# EPIC Online Technology Meeting on Co-packaged Optics for Hyperscale Datacenters

https://www.youtube.com/watch?v=ll47TK9DOl8

[00:00] So good morning and good afternoon
[00:01] everyone. So this is uh the series of
[00:04] online technology meetings presented by
[00:05] EPIC the European Photonic Industry
[00:07] Consortium and the event today is
[00:10] dedicated to co-ackaged optics for hyper
[00:13] hypers scale data centers. So we'd like
[00:17] to remind you that this event is
[00:20] sponsored by several members of Epic
[00:23] Consortium. Those are FI Reality, FCON
[00:26] Tech, OFS, AETS, Iridian Spectral
[00:30] Technologies, Suss Microl Optics,
[00:34] Modulite, Nanoscribe and Quantify
[00:37] Photonix. And actually we have a members
[00:40] of um those um epic members in in the
[00:43] room and I would like to give them the
[00:46] chance to to present themselves. So if
[00:49] we can start with maybe quantify
[00:50] photonics that uh is presented by
[00:55] Danny.
[00:59] Thank you. I'm Dan Henley, uh, vice
[01:01] president of sales for quantified
[01:03] photonics, and our company offers a
[01:05] broad portfolio of test and measurement
[01:07] instruments, and we're very focused on
[01:09] developing instruments for very high
[01:12] density applications. At OFC, we
[01:15] released a PAM for BERT in PXI and a 288
[01:19] channel optical power meter to support
[01:22] the rollout of 800 gig co-ackage optical
[01:25] network infrastructure. Uh so if you're
[01:28] testing considerations for highdensity
[01:30] co-ackage optical devices
[01:33] uh we recently released a high a new
[01:36] white paper that you may be interested
[01:37] in reading which is available on our
[01:39] website quantified forconics.com
[01:42] or feel free to reach out to me via
[01:44] LinkedIn. Uh my colleagues and I look
[01:47] forward to hearing from the panel today.
[01:49] Thank you.
[01:51] Thank you. Thank thank you Danny. And I
[01:53] believe we also have York from
[01:54] Nanoscribe in the room. Yes, I'm here.
[01:57] Hello everybody. Uh, Nanoscribe is a 3D
[02:01] micro pending company. We have a very
[02:03] spec specific technology to do very
[02:05] precise small microoptical elements.
[02:08] I'll talk about this later and introduce
[02:10] you how we can enable a better coupling
[02:13] in uh onboard optics co-ackage optics.
[02:17] Um, looking forward to the meeting.
[02:21] Thank you. Thank you very much. York
[02:23] Model
[02:29] maybe we don't have anyone from
[02:31] mobileite. So Modelite is a corporation
[02:33] for a fully vertically integrated
[02:36] semiconductor medical
[02:39] and laser
[02:41] applications. Uh maybe S micro optics.
[02:44] Wilfred would you would you say a few
[02:47] words?
[02:48] Yes thank you Ian. Um
[02:51] yes my name is Vid Nel. I'm from Zeus
[02:53] microoptics. We are in the transceiver
[02:56] and telecom business in terms of making
[02:59] optical micro optical elements. So we
[03:02] are making um um microl lens arrays out
[03:06] of different materials. We've been doing
[03:08] this since over 20 years now. And we
[03:11] have some recent new products with um
[03:14] which um will facilitate your packaging
[03:17] with integrated prisms and I will show
[03:20] that later today in the um in this
[03:23] event. Thank you Ivan.
[03:26] Yeah, thank you Wilfred. Next one uh
[03:30] next sponsor is Aidian Spectral
[03:32] Technologies. Do we have a presenter in
[03:34] the room?
[03:35] You bet I Jason Kibwar here. Uh nice to
[03:38] meet you all. Uh good morning. Um, so
[03:41] Aridian's been uh we're we're a custom
[03:43] optical filter uh manufacturer here in
[03:46] Ottawa, Canada. We've been doing this
[03:47] since 1998. Um, initially focused
[03:50] largely on telecommunications
[03:52] components, but we are also large uh
[03:54] involved over the last decade or so with
[03:56] a lot of applications for for uh a lot
[03:58] of filters for datom applications
[04:01] including uh wavelength division
[04:02] multiplexing filters in the Cband,
[04:05] Oband, Land um very miniaturized small
[04:08] components. We can actually even pattern
[04:10] them as a single array. Also, we have uh
[04:13] Etelon quality uh polishing capabilities
[04:15] and can coat Eelons uh for applications
[04:18] such as uh integrated tunable laser
[04:20] assemblies. So, we're all about custom
[04:22] optical fil uh uh thin film filters and
[04:25] dataccom is one of the large and growing
[04:27] areas that we're happy to participate
[04:29] in. So, happy to hear about uh what Epic
[04:31] members might need in terms of optical
[04:32] filters. Thanks.
[04:35] Great. Nice to great to have you Jason
[04:38] with us. And then Accetric
[04:44] Well maybe I'll say a few words. Acetric
[04:46] is actually actually serves
[04:48] I'm here. I'm sorry I'm a little bit
[04:50] late.
[04:52] Go ahead.
[04:53] Hello everyone. Aetric is leading
[04:56] manufacturer of micro optics including
[04:58] micro lenses and lens arrays and also
[05:01] defective optical elements. As you know
[05:04] our mic optics enables all kinds of
[05:07] optical transceivers in communication
[05:09] market and also some other sensing
[05:11] devices like quantum sensings and
[05:14] automotive 3D sensing lighters etc.
[05:17] I think for today's meeting like cexed
[05:20] optics or onboard optics I guess you
[05:23] need more integration features so we can
[05:26] do additional mechanical features or
[05:29] metalization for soldering etc on wafer
[05:32] level as well on our micro optics
[05:34] products. So we can support you from
[05:37] basing micropics to sophisticated and
[05:40] complex integration feature added on mic
[05:43] optics from uh small volume or like
[05:46] prototyping to mass production over you
[05:49] know the million chips per month we can
[05:52] deliver to you so that's all I want to
[05:55] tell about accessories. Thank you Ian.
[05:57] Okay,
[05:58] thank you too. Uh actually we have uh
[06:00] John Inhard from OFS.
[06:06] Good morning, good afternoon. This is
[06:07] John Nhard. I'm director of specialty
[06:10] sales for OFS based in Connecticut in
[06:14] the USA. Uh our business is specialty
[06:18] fibers across a wide range of
[06:20] applications from telecom, datacom to
[06:24] medical, aerospace, defense, sensing,
[06:27] fiber, lasers. So a a broad range of
[06:30] markets for specialty fibers. In
[06:33] addition, we are owned by Furakawa and
[06:37] Furakawa as many of you know has the
[06:39] Fitel product line either in terms of
[06:43] splicers or lasers and I'll be talking a
[06:47] little bit later about both specialty
[06:50] fibers and uh lasers used in co-ackaged
[06:53] optics. So looking forward to this
[06:56] discussion. Thank you.
[06:59] Thank you very much Joan and of course
[07:01] we couldn't leave without Font Tech and
[07:04] presented by Morates today. Yes, good
[07:08] afternoon Ian. Thank you. Um probably
[07:11] most of you know Font already. So we
[07:13] consider ourselves as a solution
[07:15] provider for the high accuracy market.
[07:18] So mainly photonix. So photonix um
[07:21] testing and assembly and for photonics I
[07:24] mean it's a huge field. So ranging from
[07:25] the typical applications like the data
[07:28] center like we here today doing um over
[07:31] the telecom telecom and LAR business. So
[07:35] um everything which is somehow related
[07:37] to optics and the high accuracies in the
[07:39] sub micrometer range which we um need
[07:42] for the alignment that's kind of our
[07:44] core business and I'm happy in
[07:46] presenting you today a few of the um
[07:48] ideas and developments which we have
[07:50] around that field. Thanks.
[07:54] Thank you very much. And the fiber
[07:56] reality,
[07:59] the last sponsor for today.
[08:03] And maybe I can just present it myself
[08:06] in a few words. So, F reality is an
[08:09] independent research and consulting firm
[08:11] specializing in optical equipment
[08:13] markets including telecom, datacom,
[08:16] consumer, auto, augmented reality and
[08:20] and more.
[08:21] So um yeah thank thank you very much to
[08:24] all the sponsors of today's the event
[08:25] and uh let's move on with the with the
[08:27] agenda. So the we have a
[08:31] very exciting uh lineup of speakers
[08:35] today and uh
[08:37] we will will see the presentations by
[08:40] Senko Franhoff for Ism, Huawei, OFS, the
[08:46] SUS Micro Optics, Fantex Service and
[08:49] Nanoscribe.
[08:51] But uh before before um apart from
[08:55] having the speakers from those named
[08:57] companies, we also have a large audience
[08:58] in the meeting today. And here are all
[09:01] the members of Epic that are registered
[09:04] for the event today. And my uh colleague
[09:07] Panagotis
[09:09] will will maybe say a few words about
[09:11] it.
[09:13] Survivan just make me a favor please.
[09:15] Can you please share the slides?
[09:19] We don't share your screen.
[09:22] Okay.
[09:32] Uh
[09:34] so I can
[09:36] I can start speaking about it. So today
[09:38] we have uh of course copas object and
[09:41] hypers scale data centers. In the in in
[09:44] the meantime I hope Ivan will manage to
[09:46] share his screen. Uh and uh and the room
[09:48] of course is extra crowded. uh we have
[09:50] represented for all the value chain. Uh
[09:53] today the meeting is going uh uh to be
[09:55] one of the better working events since
[09:56] we will have the chance to connect
[09:58] everyone raising and there we go ranging
[10:00] from research and development up to the
[10:02] end users. Uh we have of course in the
[10:04] room photoning integrated secrets
[10:05] companies and the packaging and supply
[10:07] and assembly assembly companies. Uh they
[10:10] are the key players of course uh in this
[10:12] uh in this event today. Uh we have also
[10:15] people from optics and fiber optics of
[10:17] course and uh people from the materials
[10:19] and semiconductor manufacturing and
[10:21] software as well and uh that's from me
[10:24] now Ivan I wish you all a business rich
[10:26] and enjoyable online technology meeting.
[10:29] Thank you. Thank you panos.
[10:32] So um so let's uh move ahead with the
[10:35] first presentation for today and uh so
[10:39] we'll start with a presentation by Tiger
[10:41] Nino Mia the technology and innovation
[10:44] manager from Senko Advanced Components.
[10:47] Tiger
[10:49] right I hope uh you're seeing my screen
[10:52] now.
[10:54] Yes.
[10:55] Okay. Thank you very much. Uh thank you
[10:57] Ivan and everybody and hello everybody
[11:00] from all over the world. Uh this is Tag
[11:03] know me as um representing Senko and
[11:06] also later on I I'll be speaking a
[11:08] little bit about Coobo but the uh let me
[11:11] focus on the optical interconnect
[11:13] challenges in pop optics as Senko's
[11:16] expertise is optical uh connectors and
[11:19] solutions.
[11:21] So let's review what is new with uh
[11:25] coming optics system. Um I'll I'll touch
[11:29] base on fiber count increase that's uh
[11:32] significant impact to how to arrange the
[11:35] fibers uh in and out as well as uh one
[11:38] of the biggest changes uh we have for
[11:41] seen is external laser source. I think
[11:42] this is being um addressed by Eric later
[11:45] on and the pace density this is also
[11:49] associated with the fiber counts
[11:50] increase but um we have to reserve
[11:53] spaces for laser sources and also TRX
[11:57] channels are uh increasing so uh we need
[12:00] to probably focus on what to do about it
[12:03] and then of course fibers are inside now
[12:06] optics are inside now so you have to
[12:08] consider the way to solve the fiber
[12:11] routing inside the system as well
[12:17] and as I addressed one of the biggest uh
[12:20] changes is the fiber counts increase. So
[12:23] let's take a example of uh DR4 regular
[12:27] pluggable uh 400 gig optics uh to
[12:30] configure 12.8 eight terabs. Uh how um
[12:35] so typical 32 ports uh pluggable switch
[12:39] uh can have eight fiber each per
[12:42] transceiver module that adds up to 256
[12:45] fibers. Uh let me do the same math to
[12:48] configure 51.2
[12:50] terab switch uh with CPO. CPO I
[12:54] illustrated earlier on. There are 16 um
[12:58] modules embedded in the on the same
[13:00] substrate of ASIC and each one of them
[13:04] if it's parallel optics can have up to
[13:07] uh six uh 64 fibers that times four. So
[13:13] it adds up to 10 uh 24 fibers a thousand
[13:17] 24 fibers that's quadruple of what you
[13:20] are needing now. So what I'd like to
[13:24] address is the um potential face plate
[13:28] density is going to be an issue as uh we
[13:31] need to have more fibers and also
[13:34] addressed external laser source spaces.
[13:38] So this is a chart I made uh to
[13:41] illustrate the um the fiber density and
[13:44] also connector fiber counts. So the uh
[13:48] y-axis I plotted the fiber counts per
[13:51] connector. So LC duplex. So it's sitting
[13:53] here similar to CSNS and the Xaxis I
[13:58] plotted the um fiber density per 19in
[14:02] oneu panel a patch panel using adapters.
[14:06] Um so of course no mo has better um
[14:10] density over duplex types of uh
[14:12] connectors
[14:15] and I also categorize in these three
[14:17] ways. So LCCSSN are single fiberbased
[14:21] connectors uh although duplex but single
[14:24] feral carry single fiber. So yellow one
[14:27] here uh multif fiber connectors uh one
[14:30] row uh so mo 16 and also I address a
[14:34] little bit later but the snmm is part of
[14:37] this and multi fiber two row so two rows
[14:39] of 20 uh two rows of 12 and 16 to
[14:43] configure np 24 and 32. I believe this
[14:47] has some correlation to the optical
[14:49] fiber uh fiber optics insertion loss per
[14:53] connector and the larger the fiber
[14:56] counts it's challenging to maintain the
[14:58] lower loss. So the benefit of
[15:01] considering this kind of SNMT new
[15:03] connector is to maintain that good fiber
[15:07] density at the panel but maintaining
[15:10] also the lower loss. So this is a quick
[15:14] comparison of the connector size NO and
[15:17] SNMT. You can see uh roughly the half
[15:20] the size and the configuring with the M
[15:22] uh adapters it can be a lot better than
[15:26] NO 32 as well even though SNMT carries
[15:30] uh 16
[15:33] right let me switch to the next one. So
[15:36] other areas uh sen also supports some
[15:38] other uh technologies to save the space
[15:42] inside or behind the face plate uh fiber
[15:46] routing options. Meat meatboard
[15:48] connectors and also fiber routing uh
[15:50] shuffle box uh back plane connectors for
[15:53] probably like a machine learning AI type
[15:56] of applications also fiber to the chip
[15:59] solutions. This is more like a
[16:00] customized but the as you know there are
[16:02] more fibers to be dealt or coupled to
[16:04] the chip. Uh we probably need to come up
[16:07] with the u great solutions to carry more
[16:10] fibers at once
[16:13] and Senko is capable to do provide these
[16:16] things but we'd like to hear for uh what
[16:20] uh industries are needing to achieve the
[16:23] tropopics system.
[16:26] So let me switch a gear a little bit. So
[16:28] I'm also a board member at the
[16:30] consortium of onboard optics. So let me
[16:33] touch base on the active uh programs.
[16:37] So I'm chairing this cop optics working
[16:39] group. Uh this uh has a specific scope
[16:42] to talk about or bring some technical
[16:45] guidance and standards for CPU
[16:46] implementation uh implementations u
[16:49] focusing on optical connectivity and
[16:51] remote laser sources. So uh it's been
[16:54] delayed a lot but uh we are foreseeing
[16:57] to release the white paper on optical
[17:01] connectivity uh fairly soon probably
[17:02] within June time frame and also we are
[17:06] expecting to bring the multiple fiber
[17:09] study uh type of white paper how we can
[17:12] utilize such new technologies.
[17:18] Another working group uh started last
[17:21] year is uh multimode wavegu interconnect
[17:23] system mwis. So um this is to develop a
[17:28] technical guidance and standards for um
[17:30] um optical wave guide system replacing
[17:33] the copper trace on PCB and we uh we
[17:37] need to develop or demonstrate some kind
[17:40] of coupler also waveguide itself as well
[17:43] as the media adapter to uh achieve this.
[17:48] So if you are interested let me go back
[17:50] one slide but um you can visit the
[17:52] onboardics.org org and contact Melissa
[17:55] for any inquiries. If you are willing to
[17:58] join as a member uh we are all really
[18:01] welcome. Uh we have a lots of things to
[18:03] discuss. So um I think we we can bring
[18:06] you some a summary of what's uh going on
[18:09] within more depths than this. With that
[18:13] Ivan um I conclude my presentation.
[18:21] Thanks. Hello Tiger. Thank you for this
[18:23] very informative pres presentation. Uh
[18:26] so it's about time to have some
[18:28] questions uh for for for Tiger
[18:34] there is anyone would like to ask
[18:36] something.
[18:38] Oh let me let me maybe ask you Tiger. So
[18:41] um the the typical Epic question what uh
[18:44] what do you expect from this meeting?
[18:46] What um what the other Epic members
[18:48] could uh help you with?
[18:50] Sure. So we have uh fully focused on
[18:53] passive components to couple the fibers
[18:56] or the connected fibers and there are a
[19:00] lot of different um requirements we are
[19:02] hearing like high temperature for reflow
[19:05] compatible not only reflow but we are
[19:08] hearing the high temperature operation
[19:10] as well. So uh I'd like to hear from
[19:13] like a system uh companies if they have
[19:17] any specific targeted uh environment to
[19:21] achieve the um achieve the um copacics.
[19:27] Okay. Thank you.
[19:28] Thank you. Uh we have a question. I saw
[19:31] the hand waving before but I saw it also
[19:33] in the chat now from Wilfred from SH
[19:36] Micro Optics.
[19:38] Uh yes. Thank you. In the last slide,
[19:41] uh, Tiger, you showed, um, uh, the the
[19:44] the optical board with these red, um,
[19:49] light beams, so to say, which show 90
[19:52] degrees, um, turn. So, so how how are
[19:56] these 90 degree turns ues?
[20:00] Is there any preferred technology for
[20:01] this or is that part of the coobo um,
[20:05] evaluation? Um I think the part of the
[20:07] coobo's evaluation uh we're open to any
[20:10] technology I believe and the um the
[20:13] purpose of turning the light of course
[20:15] is to um catch the light from the big
[20:18] cell on the top uh top of the PCB and
[20:21] also uh transmit to uh uh the PD on the
[20:26] substrate. So um as long as meeting the
[20:29] requirements I think uh we're open to uh
[20:32] adopt any technologies and we we should
[20:34] be able to demonstrate a couple of
[20:36] different technologies as well.
[20:38] Okay. Thank you. Well, good.
[20:41] Thanks a lot Tiger. And uh we have one
[20:44] more question that comes from Tracy, our
[20:47] love Tracy. Hello Tracy.
[20:52] Cannot hear you.
[20:55] There we go. Hello, Panos and hello
[20:58] Tiger. It's good to see you again. Uh,
[21:01] as always enjoyed your presentation, but
[21:03] I have one question. So, you mentioned
[21:05] that you you're looking at new ways of
[21:07] doing optical waveguide coupling. Uh,
[21:10] does that include some of the newer wave
[21:12] guides such as those made by polymers?
[21:16] Yes, I think polymer is one of the good
[21:18] candidates for this one. And the um so
[21:21] to answer the uh the other question uh
[21:25] probably it has to be compatible with
[21:27] the waveguide materials or the um the
[21:30] mode field diameters and so on. So um I
[21:33] I believe the main focus is to you
[21:36] utilize the polymer wave guide to this
[21:38] one.
[21:39] I I'm sorry I missed that last bit. What
[21:41] was the last bit? Oh to utilize
[21:43] utilize the polymer waveguide for this
[21:46] demonstration most likely. That's all.
[21:49] Oh, that's wonderful. Thank you so much.
[21:52] That sounds interesting, Tracy. No. So,
[21:55] if we don't have any other question,
[21:59] if we don't have any other question, I
[22:01] would like to pass to the next speaker
[22:02] then. Uh, it's a talk I'm taking uh from
[22:05] Franhoffer.
[22:16] Can you hear me?
[22:18] Yes, we can
[22:18] and see also thank you very much for the
[22:22] invite and you can see also my screen
[22:24] hopefully in full format
[22:28] here you can
[22:29] perfect
[22:32] so uh I'm from prize yam in Berlin so
[22:37] allow me to give one slide overview what
[22:41] we have been doing in the last decade in
[22:45] the topic of next generation in
[22:47] computing and uh data center
[22:50] applications using photonic
[22:52] interconnects. We started fur foxtrot
[22:56] project which has been the flagship
[22:58] where we concentrated on
[23:02] on
[23:04] rackto-rec boardtoboard chip to chip
[23:07] optical communications. Then in AT
[23:10] matrix we have been focusing on co-
[23:12] packaging technologies as a platform and
[23:15] recently and currently we are working on
[23:17] master uh where we concentrate on mass
[23:21] manufacturing data center transceivers.
[23:24] I hope uh mids can uh talk more on that
[23:29] mass manufacturing items. So as FL offer
[23:32] we contributed most likely in the system
[23:35] design and concept uh photonic RF
[23:38] component design, signal integrity board
[23:40] design, silicon photonix interposer
[23:43] development with uh through silicon via
[23:46] technologies, 3D integration
[23:48] technologies based on flip chip assembly
[23:51] and finalized several co packages and
[23:54] then also evaluated those components and
[23:58] systems uh and benchmarked
[24:02] Tiger directly started with the topic.
[24:06] Allow me to provide some background info
[24:08] on data center network topology and
[24:11] where the limitations coming from. Uh we
[24:15] know all the structure this spine leaf
[24:19] structures in data centers 70% more than
[24:22] 70% of the traffic stays inside the data
[24:25] center. In the recent projects we have
[24:28] been focusing on for foxro in in in
[24:31] within that tracks and in matrix this
[24:35] switching area in masterard we are
[24:37] focusing on single mode links
[24:40] everywhere. So the focus should be or
[24:44] the interfaces should allow us to have
[24:48] enough uh capability and symmetry in
[24:52] order to connect connect all the all the
[24:54] nodes over there. So that means we need
[24:57] uh high Ethernet switch radics which
[25:00] enables the this kind of scaling. For
[25:03] the time being the the S is actually
[25:07] becoming kind of limiting factor uh
[25:10] within the entire scaling
[25:15] today uh 256
[25:17] elements are there but and PGA is not a
[25:22] limit but the sure the port count uh is
[25:26] becoming and the routing becoming the
[25:27] the main challenge over there.
[25:31] So as we know in in data center network
[25:34] the switch bandwidth and transceiver uh
[25:38] bandwidths are uh following to each
[25:41] other. So again here we see on the left
[25:44] hand side switch bandwidth 51 52 terab
[25:47] switches with chip predicts 500
[25:51] sus speed 100g. it can be later on 200G
[25:56] with uh 256 and the port numbers and
[26:00] port speeds. So uh what we observe is
[26:04] there the serless power consumption
[26:07] increasing with the bit rate. So that's
[26:10] again uh identified uh is is an
[26:14] identified bottleneck over there.
[26:18] If you take a look the cost
[26:19] considerations, typical transceiver
[26:22] codes,
[26:23] we aim to have $1 per gigabit per
[26:27] second. Considering single mode
[26:29] transceivers given the 32 ports and uh
[26:33] even
[26:35] 64 ports uh we can assume 10 13k
[26:42] uh basically cost on the optics and if
[26:46] you consider the entire data center data
[26:48] center the entire uh transceiver cost
[26:51] will be somehow above $50 million.
[26:56] So
[26:58] further uh why we would like to decouple
[27:01] uh the IO's from the optical logic since
[27:06] our main in int intention is to reduce
[27:09] the power consumption by decoupling the
[27:12] IO's from the switch AIC at the left
[27:16] hand side we see the switch AIC we such
[27:19] an example 25 ter
[27:22] it's consuming 600 watt and one/ird of
[27:28] those power is directly uh linked to the
[27:33] sur power consumption. If we can
[27:37] actually reduce the service power
[27:40] consumption by introducing the
[27:43] disagregation of IO's in the optical
[27:46] domain, we can reduce the service uh
[27:49] power consumption down to 40 watt. And
[27:53] by doing so we can also increase the
[27:58] link bridge. And uh the here we see 1 dB
[28:03] uh penalty over there 1 dB loss uh that
[28:08] interconnect caused instead of uh 20 uh
[28:11] 12 dB. So I think moving from mid reach
[28:17] long reach uh optical electrical
[28:21] connections we we then have ultra short
[28:24] reach and by doing so reducing the power
[28:27] consumption at the third days but in the
[28:29] other hand uh means of uh optics we
[28:33] reach uh larger distances in the L3
[28:37] matrix project we have been focusing on
[28:40] those kind of uh 25 terab switches V
[28:43] with co- packaged optics using uh 2D
[28:48] transceiver array as shown at the right
[28:51] hand side consisting of based on silicon
[28:54] photonics with integrated uh 35
[28:57] materials. Uh in the next slide here you
[29:00] can see some more details. It has been
[29:04] addressing up to 256 lanes. Uh it it
[29:09] included integrated lasers by the
[29:14] bonding 35 on silicon photonics based on
[29:18] my sender configuration
[29:20] uh to have the the right modulations
[29:23] over there. It's it was basically
[29:26] relying on AMS 35 nanometer technology
[29:30] 350 nanome technology. Here in that
[29:34] slide we see again the co- package how
[29:36] it was uh connected the fiber array to
[29:41] the fiber array coupled directly by
[29:44] using microl lens arrays. I think our
[29:48] colleague will more talk about these 2D
[29:52] lens arrays in his presentations
[29:55] combined with silicon photonix
[29:57] integrated 35s and directly attached uh
[30:00] next to the uh switch ASIC. We have been
[30:04] demonstrating our uh results at OFC and
[30:10] I think if you would like to have some
[30:11] more
[30:14] details on that my colleagues from UCL
[30:16] IBM uh
[30:19] Valencia bright photonics MS and
[30:22] Photonics will be happy to provide some
[30:24] more details.
[30:26] Thank you very much.
[30:30] Thank you. Thank you Tongas.
[30:32] Thank you for the great talk and as
[30:34] always Franhoffer sets the technology
[30:36] bar very high for all of us. Um we have
[30:40] already a question in the room from Tuan
[30:42] Synopsis.
[30:44] Hey Toa, good to see you.
[30:46] Hi Tan, my pleasure.
[30:48] It's slide six. You showed let's say
[30:50] that that power consumption difference
[30:52] between uh electrical highspeed 30s IO
[30:56] and replacing that by optical IO.
[30:59] Yes. Um, is this based on I have two
[31:03] questions. Is this based on let's say uh
[31:05] calculations or based on true
[31:07] measurements? And second, does it take
[31:09] into account the power consumption from
[31:12] the the laser?
[31:16] So the power consumption from the laser
[31:17] is not included here. It's all about uh
[31:21] this 25 terra switch ASIC power
[31:23] consumption which is about 600
[31:26] uh watt at all. Right? Cons considering
[31:29] also the 30s to 56 uh elements on that.
[31:35] And if we move again to disagregation in
[31:39] the optical domain uh since we are
[31:42] allowed to have use ultra short length
[31:46] we can reduce the third day's power
[31:48] consumption actually from 180 to 40 watt
[31:52] and by doing so saving energy which we
[31:55] can use for new actually new new
[31:59] functionalities for 52 terra switches.
[32:03] Now 52 terror switches have been
[32:05] announced right that adding some new
[32:07] functionalities actually uh this uh in
[32:11] in this note that's unfortunately the
[32:15] the the main calculation for 7 nome
[32:19] simos
[32:21] as node uh I think that's roughly the
[32:25] thumbs T rule let me say in that way uh
[32:29] sure our industrial partners have been
[32:31] calculating in more detail and providing
[32:34] us with those numbers. I think that's
[32:37] that will be that will be sure different
[32:40] in different use cases and business
[32:43] cases and so on the ex right so okay so
[32:48] I so more or less what you're saying
[32:49] here is if I replace the longer distance
[32:53] sies by short reach sis I can save
[32:58] 40 watts and those 140 watts I can now
[33:01] use for functions in the switch ASIC
[33:05] because I no longer longer. Okay. But if
[33:07] you look at the overall power
[33:09] consumption including driving let's say
[33:11] the IO then actually the uh optical IO
[33:16] chiplet also requires um power
[33:19] consumption to function.
[33:22] Uh that's for sure at the end of the day
[33:25] you we do have all the transceivers also
[33:28] right. Yeah.
[33:29] So the the transceivers's power
[33:31] consumption with respect to the entire
[33:34] data center is in the range of uh 5%
[33:39] almost right maximum.
[33:40] Yeah. Yeah.
[33:41] Not at all. So but uh what we save here
[33:44] is the power the power in the switch as
[33:48] with respect to third days.
[33:50] Yeah. Understood. Okay. No thanks for
[33:52] the clarification.
[33:55] Thank you. And uh we have also a
[33:56] question from Eric Bernier from far
[33:59] away.
[34:00] Yeah. If you go to page nine of your
[34:02] presentation and I look at your 3D
[34:05] demonstrator.
[34:06] Yes.
[34:07] I'm trying to understand where the
[34:10] switch chip is.
[34:12] So switch will be basically down here.
[34:16] Okay. On the interoser.
[34:18] Yes. I think within that kind of
[34:20] European projects the resources to
[34:22] develop switch ASIC is
[34:24] not in it. So it's that's why it's kind
[34:28] of mimicked with the interfaces based on
[34:31] the uh the the the given switch
[34:36] dimensions and uh pin locations. Thank
[34:40] you.
[34:40] Yeah, I can understood. Um, now you're
[34:43] saying you would put the basically the a
[34:46] square in the center that has detectors.
[34:50] Do you do modulators or those those are
[34:52] lasers?
[34:52] So those are actually integrated
[34:55] modulators with uh three 35 hybrid I
[34:59] bonded lasers on that.
[35:01] Yeah. Did you look at the terminals for
[35:03] that solution?
[35:04] Pardon me.
[35:05] Did you look at the terminals for that
[35:07] solution? Uh did you do thermal
[35:09] simulations to see if uh the laser would
[35:12] have good performance and yes that that
[35:17] [Music]
[35:19] the integration of lasers directly on
[35:22] silicon is the main challenge right so I
[35:25] think apart from that I think then the
[35:30] we introduce also the cooling elements
[35:34] that respect the back side of the uh
[35:37] chip and also
[35:39] decoupled thermally from the max sender.
[35:43] So since it's Max sender is less let's
[35:47] say sensitive compared to the ring
[35:49] generator based modulators right.
[35:51] Well I was more concerned by the switch
[35:53] chip itself generating a lot of heat and
[35:56] like you know so you're you're you're in
[35:59] a very warm environment in that in the
[36:01] middle of that.
[36:02] I think
[36:04] that's that's for sure. That's for sure.
[36:06] In that given constration, we have been
[36:08] focusing on the optical domain only.
[36:11] Yeah.
[36:11] We were not able to test it with the
[36:14] real uh ASIC switch.
[36:18] That's that's pretty good to have a such
[36:20] a big array connected to
[36:23] actually the intention has been to
[36:25] provide a platform which is capable to
[36:29] do the entire co- packaging and the
[36:31] disagregation in the optical domain with
[36:33] such kind of configuration. So by doing
[36:36] so we demonstrated the capability of
[36:39] such 2D transceiver arrays combined with
[36:44] 2D lens arrays and connected to the
[36:46] entire uh drivers and receivers over
[36:50] there. Thank you.
[36:54] Thank you Tolga. Um Tracy would you
[36:57] would you do you have another question?
[37:03] We cannot hear you. Tracy, you are
[37:06] muted.
[37:09] You are still mute. Yes,
[37:10] there we go. Okay. I'm so sorry. Thank
[37:12] you. Um, Toga, very good uh,
[37:15] presentation. Really appreciate it. And,
[37:18] you know, it seems to make perfect
[37:20] sense. So, initially it was the the
[37:22] density of the connectors on the face
[37:24] plate, the density of the fibers, uh,
[37:26] and then, you know, how do we connect to
[37:29] the the pick? But, but now it's the
[37:31] certis. What other problems do you think
[37:33] that we're going to see? I mean, it
[37:35] makes sense that that as we increase the
[37:37] density, we're going to see downstream
[37:40] uh changes that must occur. So, in
[37:42] addition to the certis, what else do you
[37:43] think uh is that kind of the next uh
[37:45] focus of improvement?
[37:47] I think electronics is sure working on
[37:49] that on switch technologies, new ASIC
[37:54] based on new SIMOS nodes. I think for
[37:57] the time being what we have to focus to
[38:00] identify from photonic optical domain
[38:03] point of view which type of uh
[38:07] configuration makes sense
[38:09] shall we directly integrate lasers in
[38:12] those transceivers or shall we have
[38:15] let's say of the module lasers so I
[38:19] think those are the critical questions
[38:22] and sure uh we have to really understand
[38:27] uh if if you can meet the costs by doing
[38:30] so, right?
[38:33] And reliable
[38:34] the cost power. Yes. Uh any anything new
[38:37] on the power horizon? Any u any new uh
[38:42] solutions that look like additional uh
[38:45] things that may help reduce the power in
[38:47] addition to what you've mentioned?
[38:50] As I mentioned the contribution from
[38:52] photonic transceivers in the entire data
[38:54] set the power consumption is less than
[38:57] 5%. So and even 2% what we can achieve
[39:02] is mainly by introducing new photonic
[39:06] interconnects
[39:07] we change somehow the entire topology
[39:10] and by doing so uh the also in the
[39:15] protocol domain in all domains uh
[39:19] introducing some new features by using
[39:21] WDM
[39:23] that will allow us to have more
[39:26] flexibility in the basic development. So
[39:29] it's I think that's that will be my
[39:31] understanding.
[39:34] That's a great uh answer. Thank you so
[39:36] much Toga. Appreciate it.
[39:37] Thank you. I appreciate
[39:40] Thank you Tologga. Very quick question
[39:42] maybe Antonio.
[39:45] Yeah I was curious about the the lenses
[39:47] you you show in the last slide um to
[39:50] connect the the to the light from the
[39:52] fibers. uh do you know which material do
[39:54] you use and uh
[39:55] those are those are silicon I think uh
[39:58] I'm sure uh Vimit can give some more
[40:02] details how those uh bulk optical
[40:06] elements to the uh can be fabricated and
[40:10] what kind what what are the
[40:12] possibilities over there but I think
[40:14] it's it's based on silicon sure
[40:18] thank you
[40:19] great great exciting it's always great
[40:22] when when we're on the fringe with the
[40:24] research. So the next presentation comes
[40:26] from Huawei and Eric Beigner, the leader
[40:29] of advanced photonics
[40:31] at Hway will share his thoughts.
[40:39] Okay. Hello everyone. You must be seeing
[40:42] my slides now.
[40:43] Yes, that's good.
[40:44] Perfect. Thank you.
[40:46] So um I'm just gonna I wanted to give
[40:49] you a a short introduction about the uh
[40:52] projects that are going on into a
[40:54] standard organization called IPAC
[40:56] uh about co- package optics. So that's
[40:59] what I'll I'll be talking about
[41:03] trying to change my slides. Okay.
[41:06] Perfect. Okay. So just uh for those who
[41:10] don't know high silicon high up to
[41:12] electronic is a division of Huawei. It's
[41:15] got 1,200 employees. It's a We have
[41:17] global R&D teams. We have teams in uh
[41:21] Belgium, in uh Germany, in UK, in Canada
[41:25] and Japan and also in Wuhan and Senza.
[41:30] Uh we have revenues of about 1 14 1.5
[41:34] billion US dollars every um per year
[41:38] with the patents portfolio about,300
[41:42] patents. So what we do we develop the
[41:45] components the photonic optical
[41:47] components that go into the larger
[41:48] systems. Uh so you know traditionally
[41:51] we've done the lasers, the modulators
[41:54] and the modules that would go into more
[41:57] of the Huawei and telecom equipment and
[41:59] we're the we're in the midst of
[42:02] diversifying our offering by offering
[42:04] sensors uh like LARS and other type of
[42:07] sensor like that to
[42:10] to uh the time of flight sensor and
[42:13] things like this in order to uh to
[42:16] expand our business. Um,
[42:21] one of the thing we uh we're involved in
[42:24] at uh Huawei uh auto electronics
[42:28] uh IC and auto electronics is the IPC.
[42:30] The IPC is the international photonics
[42:32] electronic company. Uh we open standard
[42:35] organization based in Switzerland. There
[42:38] is 31 member. I got the the member said
[42:41] here on the right of the slide and we
[42:44] focus on opto electronic industry like
[42:47] optical chips, optical electronic
[42:50] components, optical modules for 5G
[42:55] in terms of things and AI.
[42:59] Um I think there's a so you have to
[43:02] disregard I think there's some leftover
[43:04] from old slides at the bottom here. So
[43:07] uh the first thing I wanted to talk
[43:08] about uh one of the project at IPAC that
[43:11] we have is the uh standard research
[43:13] report on 100T optical IO standards. Uh
[43:18] and basically what it's about is we've
[43:21] been talking in in the 3.2 terabits per
[43:24] second optical elements that people are
[43:27] talking about in those forums or the for
[43:31] the E. uh it's all targeted for a switch
[43:34] chip of 50 terabs per second and we know
[43:38] that at 50 terabits per second this is a
[43:41] first introduction to the industry for
[43:44] uh the CPU the co- package optics and we
[43:47] also know that we uh the 50 terab switch
[43:50] can survive and they can work properly
[43:52] with our current model based
[43:54] architecture and all just just do the
[43:56] same but but faster. What uh the concept
[44:00] around the industry is that at 100
[44:03] terabit switch for switch like that that
[44:06] kind of of bandwidth then it's becomes
[44:09] impossible to move the data
[44:11] electronically with the cert
[44:14] without
[44:16] uh basically
[44:18] consuming the entire uh power budget for
[44:22] the switch chip in just in order to move
[44:24] the the data in and out. if you're going
[44:26] to use electronic um long reach ser.
[44:32] So this is at 100t this is where the
[44:34] copac job really becomes uh um needed
[44:39] but something you have to look at is if
[44:41] you look at it as a module perspective I
[44:44] don't think it's possible to
[44:46] double the number of module right we
[44:49] normally want to try to minimize number
[44:51] of modules that are on a single package
[44:53] because you know that's a reliability
[44:55] issue um and also becomes harder to
[44:58] package and if you're
[45:01] laying means are still very long, right?
[45:04] Then those sidas at the corner of the
[45:07] chip needs to still be very strong and
[45:09] you don't have a lot of savings. So
[45:11] going forward, right, there's going to
[45:12] be a lot of push in order to increase
[45:15] the the density in the and the uh the
[45:17] capacity for the co- package optics,
[45:21] you know, 6.4 or 12.8 terab per second,
[45:25] right? looking at 200 gigabits per
[45:28] second per fiber
[45:31] coming out of each modules minimum
[45:34] probably going WDM multiple wavelengths
[45:37] per fiber
[45:39] just to have higher density higher
[45:42] footprint and if you look at the D let's
[45:45] say here the the the photonic D or the
[45:47] photonic engine is u is a little gray
[45:51] block here right so we also have to
[45:53] increase density of bandwidth per second
[45:56] for for the same type of area of
[45:58] infopolics. So there's lots of challenge
[46:00] going forward, right? Um we'll need to
[46:03] input more optical power into those
[46:05] engine, right? So there's lots of good
[46:09] progress in our technology in the
[46:10] development discussion in the standards
[46:12] but ultimately um we need new technology
[46:16] and and that's required and that's why
[46:18] we're doing at the IPAC we're doing um a
[46:22] research report to look at those issues
[46:25] where we're going to look at network
[46:28] system architectures we're going to look
[46:29] at evolution of the electronic and the
[46:31] packaging right so the modulator
[46:34] themsself needs to be a core of the
[46:36] The packaging needs to be a lot more
[46:38] advanced, right? We need to support high
[46:40] power. We need to reduce loss.
[46:42] Ultimately,
[46:44] uh we want to look at device process
[46:46] evolution road map the the channel
[46:49] density for improvement and the
[46:50] pluggable versus nonpluggable CPO.
[46:54] All of these CPO modules that people are
[46:57] talking about today, they are assuming a
[46:59] connector. So we can basically uh
[47:02] because it's easier for our supply chain
[47:04] and assembly chains to just keep that
[47:07] connector on uh on the back of the small
[47:11] module. However, uh if we increase
[47:13] density and we want to reduce the power
[47:15] requirement on the ser we need to get
[47:16] rid of that of that connector. That's
[47:19] something we want to look at here.
[47:22] The project started um around September
[47:25] time frame of 2021 and it's going to go
[47:28] on throughout this year where we uh
[47:30] we're looking for
[47:33] uh for submissions into IPC. Everybody's
[47:37] welcome to become a member here. We are
[47:39] you know we we we are looking for for
[47:43] the industry to push forward in those in
[47:45] those uh on those issues. Um right so
[47:49] throughout this year we're looking at uh
[47:50] at u getting contribution I'm the editor
[47:54] of the white paper so so we'll put all
[47:56] the the parts together and we are um
[48:01] targeting to issue a document by the end
[48:04] of this year for ballot and having the
[48:07] full research report published in Q2 of
[48:10] 2023. So that's part of the first
[48:12] project on OIO research. Right now the
[48:16] other thing that's going on into the
[48:18] IPAC uh standard is form factor. We call
[48:22] it the form factor um standard
[48:25] standardization where we are looking at
[48:28] standardizing a specific form factor for
[48:31] the external laser source. If we look at
[48:34] the different laser source that are
[48:35] possible from the OIF documents,
[48:39] uh there is uh multiple ways to do it,
[48:42] right? So I think Coobo is looking at uh
[48:45] center board laser. Um people have
[48:48] talked about face plate pluggable but
[48:50] face face plate with a jumper cable or
[48:53] back uh blind or blind connector or even
[48:58] we're looking at uh very uh large and
[49:03] powerful lasers that could sit at the
[49:05] bottom of the rack and be water cooled
[49:08] for example, right? So we can really
[49:10] provide a lot of photons for the
[49:12] modulator ultimately.
[49:15] Um the form factor research that we're
[49:19] talking about is about the 2A type. It's
[49:22] going to be a back uh blind connector
[49:26] um and face plate pluggable laser.
[49:30] And uh one thing we're looking at in in
[49:33] more um in more details, we're looking
[49:36] at a common arrangement where we're
[49:38] going to use QFPD or or SFP uh type form
[49:43] factor so that we don't really need to
[49:45] change the face plate organization. Uh
[49:48] we're looking at using a blind mate
[49:51] connector with a single feral. It's a we
[49:54] think it's a
[49:57] a reliability issue. We want to just
[50:00] have to deal with the single um single
[50:02] feral and just two pins for that
[50:04] connections. Spring loaded fereral to
[50:07] avoid uh overloading the power on the on
[50:11] the connector as you insert and and
[50:14] remove the modules. Seams management
[50:18] conventional power and will provide high
[50:20] mobility, operational simplicity,
[50:23] maintainability, serviceability and uh
[50:26] stability.
[50:29] That's the scope of the project. Looking
[50:31] at electrical specifications, optical
[50:33] specification, wavelength, lane
[50:36] assignments, optical characteristics,
[50:38] power project, etc. Electronics and
[50:40] optical connector, mechanical seams
[50:43] interface. Um, in this case, we're
[50:46] probably going to use what's going
[50:47] what's being defined. and my my group is
[50:50] doing a big a big um push in in defining
[50:53] what's going to be done at OF for the CS
[50:55] but we're going to be using it in IPC in
[50:58] our environmental
[50:59] uh specification
[51:02] project just started a month a month ago
[51:06] uh and it's supposed to go on and having
[51:09] a full specification published by uh in
[51:12] by early next year. It's a bit faster
[51:16] because it's uh it doesn't cover as much
[51:18] ground as the uh as the um optical or
[51:22] the research paper. Uh but again here
[51:24] right we we're looking at more people to
[51:26] participate uh so that we can get
[51:29] something that everybody wants to use.
[51:32] Um and uh and that's it for for that um
[51:36] for that project. So thank you very
[51:38] much. uh if you have any questions or
[51:40] want to participate or want to
[51:42] contribute
[51:43] uh you know we can even come like we can
[51:45] invite you to uh to to to the discussion
[51:50] you can have a try before you become a
[51:52] member just give me an send me an email
[51:54] I'd be happy to include you in those
[51:57] discussions thank you
[52:00] thank you Eric thanks a lot and we
[52:02] already have some questions from you for
[52:03] you uh uh the first one comes from
[52:06] Wilfred from Suss Micro optics
[52:09] W it would be great if you ask the
[52:11] question yourself.
[52:12] Ah thank you panos. Um yeah I have
[52:15] actually a few questions. Um you saw so
[52:18] more technologically question. So did
[52:22] you decide already because you said the
[52:24] lasers will be will be offboard which is
[52:26] quite an interesting approach which
[52:29] makes a lot of things probably easier
[52:31] but less integration. So what about
[52:34] other points like is it will be pump
[52:36] four or coarse or dense wavelengths
[52:38] multiplexing? Do you know already how
[52:39] many channels and and then my other
[52:42] question is a little bit more um there
[52:45] is this 5G situation between Huawei and
[52:48] the US. So is this going to be affected
[52:50] here or somehow how is this going to be
[52:54] are you working with Kobo on this or are
[52:56] you excluded there completely or I don't
[52:58] know if you can comment on this because
[53:00] I know it's very sensitive topic. So
[53:02] okay well it's very easy. So um standard
[53:05] organizations standardiz
[53:07] standardizationization
[53:09] publications uh you know research work
[53:11] is excluded from department of commerce
[53:14] rules of the US government. So there's
[53:16] no problem we can go and because they're
[53:18] public forum right we're not doing
[53:20] anything in we're doing everything in
[53:22] the open so that's that's no problem.
[53:25] That's a good news. Yeah. Yeah.
[53:26] Yeah. and like coming here for me to
[53:28] present today I know it's a public forum
[53:30] there's no problem right so it's all
[53:32] it's all all cautious all good
[53:34] uh so that I think that's fine I mean in
[53:36] terms of business right u of course it's
[53:39] very hard for Huawei to go and purchase
[53:42] anything from the US right but China is
[53:46] a big market is growing and uh you know
[53:49] the business is just uh is just no
[53:51] flourishing that way I don't think
[53:54] that's uh that's that's my two cents on
[53:56] on that part and I don't think I can
[53:58] comment very
[53:59] much more on that. Um, in terms of PALM
[54:04] 4 or PAM 6 or PAM 8, if you follow the
[54:07] IGE um into the 3DF group that's being
[54:11] discussed right now, there's lots of
[54:13] people that discuss different PALMs and
[54:16] it doesn't look like there is if you do
[54:19] PAM 6 or PA 8, you need more fact. you
[54:22] need more fer correction because your
[54:25] eyes is smaller and therefore it's going
[54:27] to be need more closed and be harder to
[54:29] to uh to recover. You'll need more fes
[54:32] right to just reopen eyes. Um and
[54:36] doesn't look like there is a lot of gain
[54:39] because you have you know you have to
[54:42] derate so there's not a lot of gain into
[54:44] going from PAM 4 to PAM six at this
[54:46] point and so my conclusion from that is
[54:50] um
[54:51] we're going to be at PA four for a long
[54:53] time at this point right um yeah but the
[54:58] the good benefit of move and and to move
[55:00] to your other question why you want to
[55:02] move the laser outside is because you
[55:04] know as I discussed because we're trying
[55:05] to increase the bandwidth density
[55:09] of those optical engine and the laser
[55:12] themselves
[55:14] even if they integrated they take space
[55:17] and even and right now it's possible to
[55:20] use macro like if you look Intel's
[55:23] presented papers at at WFC you can go
[55:26] look it up right but uh if you just take
[55:29] a picture from their paper and you see
[55:31] how many macroing they have in the laser
[55:34] base you you know you can see that it
[55:36] can do about a four terabit per second
[55:39] modules in the same in the space that's
[55:41] required for an optical engine. Well, we
[55:43] need to double and quadruple it. So
[55:44] something's going to go and it's going
[55:46] to have to be the laser and we can move
[55:48] the laser outside.
[55:49] Penalty for moving the laser outside is
[55:51] lost. So we need very low loss coupling.
[55:55] What what is your take on coherent? Is
[55:56] it still too expensive or are the DSPs
[55:59] an issue there or or what is the uh or
[56:03] do you think it's more than palm 4
[56:06] versus coherent or
[56:10] That's a good question. Um
[56:13] it's the DSP right now. It's the cost
[56:16] per bet. How much it's pjel per bits. So
[56:20] we're trying to go to minimum pog per
[56:22] bits. We're not trying to go very far,
[56:24] right?
[56:26] um coherent right now I think for the
[56:30] next five years the break even points
[56:31] about 10 kilometers if you go beyond 10
[56:33] kmters you want to go coherent it's my
[56:36] view if you going to go below you know
[56:38] you probably want to go IMD unless we
[56:41] find a way to do coherent very cheaply
[56:43] in terms of power in terms of numerical
[56:46] you know
[56:48] um computations so that's of course this
[56:52] is an evolution right so you have to
[56:53] keep your hand on it right that
[56:55] threshold may move a bit but uh at this
[56:57] point I think the thresholds are around
[56:58] 10.
[57:00] Okay. Thank you. Yeah, that's the
[57:01] feedback I basically got at OFC also
[57:03] from the other companies. So it's pretty
[57:06] coherent.
[57:10] Okay. Thank you.
[57:11] Yeah.
[57:13] Thank you. Thanks a lot. Uh
[57:15] that sound to be a very interesting
[57:17] discussion but we still have more
[57:18] questions. Uh the next question comes
[57:21] from Nanoscribe from our epic friend
[57:25] Yorganos.
[57:28] Nice to see you.
[57:29] Uh Eric, nice talk. Um I have a basic
[57:33] questions. You showed these empty
[57:35] fereral connector. Uh you want to bring
[57:38] a lot of PM polarized light to the OE as
[57:41] far as I understand. So uh will you will
[57:44] you reduce the pitch? Will there be more
[57:46] dense fiber density in the fereral or
[57:48] what is the next step you see see in
[57:50] your standardization?
[57:53] Well,
[57:55] I think there is a there is um a metric
[57:59] right which basically constant in
[58:01] industry right now. It's like cost per
[58:03] fiber alignment. So the more fiber you
[58:05] have, the more cost you got, right? We
[58:06] we don't want to increase cost. So
[58:09] ultimately we'll need to reduce number
[58:11] of fiber, right? I don't think we're
[58:13] there yet for like that generation of
[58:14] CPO, but I think you know going WDM like
[58:18] right now we're talking four wavelengths
[58:20] but eight or 16 wavelengths I think
[58:22] that's the way to go right in order to
[58:23] reduce cost really um and you know PM
[58:28] alignment number of PM fiber alignments
[58:32] you know it's also reduced adding a lot
[58:34] of cost right and so that's what that's
[58:36] where we need to to reduce it.
[58:40] Thank you. Thanks. Thanks for the
[58:42] question. Thanks Eric for the answer. I
[58:44] have two more question for you. Uh I
[58:46] would like I would like to ask to try to
[58:47] answer a little bit quickly because we
[58:49] have to go on but I want to give the the
[58:51] time to the people to ask their
[58:52] questions. So the next question comes
[58:54] from Tiger Ninoa. I hope I'm pronouncing
[58:58] it well from from the speaker.
[59:01] Yes, Tiger.
[59:03] Thank you and thank you Eric. Great
[59:04] presentation. A quick question about the
[59:07] pluggable SAP module that's under
[59:09] development in the uh by the IPC. Is
[59:12] there any consideration of how to
[59:15] maintenance the empty feroh end phase?
[59:17] So for such high power laser source like
[59:20] inspection, cleaning and design to make
[59:23] it easier.
[59:25] Well, this is something that's being
[59:26] looked at by the group right now. I
[59:28] don't think that's that's the project
[59:30] just started in the IP. They have not
[59:33] brought in that that issue yet. So I not
[59:36] really I cannot really comment with
[59:38] respect to the IPC project but it's
[59:41] definitely something that's important
[59:42] right any we're talking you know the
[59:45] lower power laser coming into an optical
[59:49] engine would be 18 dBm right we're
[59:52] talking going about to about 24 dBm and
[59:56] I've done contribution in the that shows
[59:59] you know connecting a laser at 20 dBm
[01:00:01] can priorize
[01:00:04] just the um uh the plastic fereral or
[01:00:08] like some part of that fereral like not
[01:00:10] the fiber core and glass itself but
[01:00:12] anything that's around it you have too
[01:00:15] much fire going there right so you can
[01:00:17] imagine if you have a little piece of
[01:00:18] dust of fiber it will just pulverize and
[01:00:21] it will be um uh you know will destroy
[01:00:24] basically the connection so yes cleaning
[01:00:27] is definitely super important um and I
[01:00:31] know you guys are working on it so you
[01:00:32] know I'm waiting for your solutions. But
[01:00:36] uh yeah, I don't I I cannot recommend me
[01:00:38] on that point here.
[01:00:40] Okay,
[01:00:42] there you go. We have connected now
[01:00:44] again and uh Tracy, of course, we cannot
[01:00:48] refute your question.
[01:00:50] I'm mute and go.
[01:00:51] Thank you so much, Banos. Very
[01:00:53] interesting talk, Eric. Um so, as I was
[01:00:55] listening to you, I was and some of your
[01:00:57] comments about, you know, it's it's hard
[01:00:59] to move the data electronically.
[01:01:01] Yes, it is. And it takes a lot of power
[01:01:03] and one of the biggest problems within
[01:01:05] these data centers is power and
[01:01:07] resultantly how do you cool them? So
[01:01:10] looking ahead maybe 10 years or more do
[01:01:13] and you know with with combine that with
[01:01:15] the moving the lasers down to the bottom
[01:01:17] of the rack, do you ever think that
[01:01:19] we're going to get to the point where
[01:01:21] this is going to be pure optical where
[01:01:22] we're looking at an optical switch as
[01:01:25] opposed to I mean we've been doing
[01:01:26] optical switches for over 20 years. Do
[01:01:29] you think that we're going to start
[01:01:30] moving all people switching into the
[01:01:32] data centers?
[01:01:35] Well, um, since I I worked maybe five
[01:01:40] years of my life building switches, I
[01:01:42] probably have some still standing record
[01:01:44] on the size of highspeed switch that
[01:01:47] were that were actually working in. So,
[01:01:50] but um
[01:01:54] the
[01:01:56] problem is really in the what to like
[01:02:00] I think the problem in adopting a
[01:02:02] photonic switch is to move away from a
[01:02:05] technology that we know we can extend
[01:02:08] know the risk is lower on electronic
[01:02:11] we've built electronic switch we've got
[01:02:13] learning we've got knowledge that spans
[01:02:16] know 30 40 years right whereas Because
[01:02:19] when you go with the photonic switch,
[01:02:20] first of all, we get something that's
[01:02:22] that can handle very very large flow but
[01:02:24] not very well. The packet size, right?
[01:02:27] Um, and we also get to a brand new
[01:02:30] technology with a lot of things we don't
[01:02:33] know about. So there is a lot more risk,
[01:02:35] right? So I think people are going to
[01:02:37] stick with electronic switch for as long
[01:02:38] as they can.
[01:02:41] when we talk about so that's the first
[01:02:43] part of my question when we talk about
[01:02:45] the power and I think what's important
[01:02:47] to to realize is
[01:02:51] what takes a lot of power in the data
[01:02:53] center is the hard drives and the CPU
[01:02:56] like when we talk about those switch
[01:02:58] that need a lot of bandwidth
[01:03:01] um they don't really make a big
[01:03:03] difference right but what really makes a
[01:03:06] huge difference the the power like
[01:03:09] basically for for a switch jet
[01:03:11] we have a power budget. We cannot really
[01:03:13] u ask this chip itself to consume more
[01:03:16] than something like six let's say 600
[01:03:19] watts, right? It's it's still a very
[01:03:21] large number, right? So it's what are
[01:03:24] you going to do with that 600 watts,
[01:03:26] right? And if we're going to spend a lot
[01:03:28] of that 600 watts on siras, then we
[01:03:31] don't have a lot to do the switching
[01:03:33] itself, right? And so at this at this
[01:03:35] point, what we're trying to do is move a
[01:03:37] lot of that power outside of the switch
[01:03:38] ship, right? That's what the CPO does,
[01:03:42] right? It moves the power away from the
[01:03:44] switching. I think you know in in in a
[01:03:47] whole as an architecture as a system
[01:03:51] because we have more loss from the uh
[01:03:53] from the external laser we will consume
[01:03:56] more power in the end we will consume
[01:03:58] more power as a system but the switch
[01:04:01] chip itself is going to basically we're
[01:04:03] going to remove power from the switch
[01:04:06] chip power budget itself and that's
[01:04:08] going to allow us to attain more
[01:04:11] performance in those chips and that's
[01:04:12] what we're trying to achieve here. It's
[01:04:14] not really to save the planet, but to
[01:04:16] make our design possible. That's why
[01:04:20] we're doing it.
[01:04:21] I mean, certainly for the EO for the the
[01:04:24] individual flows, you need EO at some
[01:04:26] point. So, I was just wondering if you
[01:04:28] thought that this could I mean, and this
[01:04:30] is a discussion that we can have later,
[01:04:32] but um very interesting talk. I really
[01:04:34] appreciated your information and
[01:04:36] certainly the answers that you just
[01:04:37] gave.
[01:04:37] Yeah. You know, if somebody wants to
[01:04:39] build a photonic switch, I'm all with
[01:04:41] you. I'll go and do more more switch.
[01:04:43] you know, I'll be very happy is, you
[01:04:45] know, I'm just waiting for a customer.
[01:04:47] My design.
[01:04:48] Yeah. No, it's it's possible, but I
[01:04:49] mean, you've got to get down to the
[01:04:50] individual EO if you're looking at the
[01:04:52] packets. Uh, but if you can aggregate
[01:04:55] those packets,
[01:04:57] then you can change the system, you
[01:04:59] know.
[01:04:59] Oh, yeah. Like you we have I've got
[01:05:02] papers of architecture where we have
[01:05:04] like photonic packets or obviously a big
[01:05:06] aggregation of packets where we
[01:05:08] aggregate the flows and manage large
[01:05:10] cues and stuff like that. there's a lot
[01:05:12] of tricks you can play in but uh you
[01:05:14] know um yeah but I think there is a lot
[01:05:17] more teaching that needs to happen in as
[01:05:19] well I know I saw some paper at the UFC
[01:05:22] uh and um you know like the
[01:05:26] understanding of all the switch and what
[01:05:28] are limitation and all the scaling and
[01:05:29] what can happen to like depending on
[01:05:31] what choice you make I don't think
[01:05:33] that's well understood even with that
[01:05:35] new generation of researcher coming in
[01:05:37] and um and you know that's that's we're
[01:05:41] just starting over again. Starting over
[01:05:43] again like some point we know.
[01:05:48] Perfect. Thank you. I'm sorry. I'm
[01:05:50] sorry. I'm Thank you, Tracy and thank
[01:05:52] you Eric. I'm sorry for interrupting but
[01:05:54] we still have your discussion trigger
[01:05:56] more questions. So there is another one.
[01:05:57] I would like to give him the floor.
[01:05:59] Uh let's try to be quick and then we'll
[01:06:01] go to the next speaker. So token uh from
[01:06:04] Franho for ACM. So is sure side topic
[01:06:08] this optical switching in data center
[01:06:10] but allow me to comment on that based on
[01:06:13] our foxro project where we have been
[01:06:15] working also on optical switching
[01:06:17] functionalities inside the rack. So I
[01:06:21] think it's possible for sure but it
[01:06:23] requires the entire adaptation of
[01:06:25] protocols by uh allowing WDM
[01:06:29] functionalities introduced to the
[01:06:31] protocols. So it's not changing directly
[01:06:35] from electronic switches to optical
[01:06:37] switches. It requires more than that
[01:06:40] also in the protocol and uh the entire
[01:06:43] control domain it should be adapted.
[01:06:45] Thank you.
[01:06:48] Good. Thank you.
[01:06:51] Thanks a lot. Thanks a lot for the
[01:06:53] statement toga and thanks a lot Dick.
[01:06:55] Maybe it's a good time uh a good moment
[01:06:56] to go to our next speaker. Um, and our
[01:07:00] next speaker, just give me a moment,
[01:07:03] is John Nhard from OFS,
[01:07:08] John, are you ready to share your
[01:07:09] screen?
[01:07:13] John, perfect. My floor is yours.
[01:07:22] Just bear with me as I get set up.
[01:07:29] Yeah, you can just
[01:07:31] Hold on one minute.
[01:07:33] We did it correctly
[01:07:35] half an hour ago. We can do it again.
[01:07:37] No.
[01:07:38] Yeah,
[01:07:43] there you go. Something's moving full
[01:07:45] screen.
[01:07:45] Let me go to full screen. Excuse me.
[01:07:59] Okay,
[01:08:03] all set.
[01:08:06] Seems to me that needs a little bit more
[01:08:08] time.
[01:08:15] Not seeing it yet.
[01:08:18] We We seen it, but it's uh it's uh it's
[01:08:22] stuck.
[01:08:24] at least my screen.
[01:08:28] We can we can go to the next speaker. In
[01:08:30] the meantime, we try to solve this one.
[01:08:31] Is that okay, John?
[01:08:33] Sure, that's fine.
[01:08:34] Okay. Ivan, can you please uh introduce
[01:08:37] the next speaker?
[01:08:38] Yeah, sure. Um Wilfred, could we ask you
[01:08:41] to to start your presentation instead?
[01:08:43] So, this is Wilfred from Asus Micro
[01:08:46] Optics.
[01:08:48] Okay, I I
[01:08:51] give me a sec.
[01:08:53] I'm ready and
[01:08:57] uh go. So
[01:09:01] thank you.
[01:09:05] Thank you very much. So I um as I
[01:09:08] mentioned earlier in the introduction
[01:09:10] I'm um in the uh doing the sponsoring
[01:09:14] section um we are uh from Z micro optics
[01:09:18] and we make micro optics since 20 years
[01:09:21] and today I want to talk about
[01:09:23] particular solutions that could make
[01:09:25] your co-acking easier. So it's like we
[01:09:28] all know there is a big it's not only
[01:09:32] optics it's also packaging. So the the
[01:09:35] the 85% of a transceiver is still
[01:09:37] packaging and with co-ackaging optics I
[01:09:40] think the emphasis will be even more on
[01:09:42] packaging than just on um uh on the
[01:09:45] optics. So the optics will be just
[01:09:47] whatever you can get but the packaging
[01:09:49] is the issue. So that we also saw that
[01:09:52] very nicely in the coobo presentation
[01:09:54] and from Huawei. So um and um this this
[01:09:58] is very evident that the packaging is is
[01:10:01] important. So we are from um Nushatel in
[01:10:05] Switzerland close to the lake. Very nice
[01:10:07] location. We have several clean rooms
[01:10:10] now for 200 millimeter wafer processing.
[01:10:13] We've been doing this since 1998. We are
[01:10:15] very big in in data and telecom and also
[01:10:18] recently we moved into automotive like
[01:10:20] three years ago. So we're fully ITF
[01:10:23] compliant um and certified. That means
[01:10:26] we can also deliver to to lighter which
[01:10:29] is getting very similar to to um um so
[01:10:33] the infrastructure and lighter the
[01:10:35] technological infrastructure becomes
[01:10:36] very similar to the one of data and
[01:10:38] telecom. So we've been doing data in
[01:10:41] telecom since many years. As I said, we
[01:10:43] also are in the semiconductor and
[01:10:46] medical field with illumination optics.
[01:10:49] Um
[01:10:51] and what we do for data and telecom is
[01:10:53] mostly cumation and refocusing optics
[01:10:57] based in fus silica and silicon. So you
[01:11:01] can pick the material whatever you
[01:11:03] prefer. Some people prefer silicon
[01:11:05] because of the for high numerical
[01:11:07] apertures of of the of the light source
[01:11:10] and fus silica is nice because it's
[01:11:12] easier to package when you use a UV glue
[01:11:15] or something. So this is typically more
[01:11:18] a question about um how stable your
[01:11:22] system is or how you assemble these
[01:11:24] things which material is better. So
[01:11:25] we've learned that fus silica is easy uh
[01:11:28] for two reasons. Like I said already the
[01:11:30] the UV glue is easier to use and there
[01:11:34] the epoxies and the other reason is the
[01:11:36] um actually the the the the change of
[01:11:40] the material properties with this
[01:11:42] temperature are very small. So
[01:11:44] refractive index change and and
[01:11:46] expansion coefficients are very small.
[01:11:47] Whereas silicon it's a little bit
[01:11:49] different. Um it's it has a larger
[01:11:52] expansion coefficient and also the the n
[01:11:54] / dt. So the chains have refractive
[01:11:56] index a little larger but because it has
[01:11:58] a higher refractive index it's more
[01:11:59] suitable for high NA optics. So it has
[01:12:02] its pros its cons. And in terms of form
[01:12:05] factors we offer lenses on the front
[01:12:08] side and the back side. We um oops um we
[01:12:11] are offering lenses uh embedded or or we
[01:12:16] call them recessed. So that means you
[01:12:18] have here a flat top. I'm going to talk
[01:12:20] about this. We offer all types of
[01:12:22] fidicials obviously and um that makes
[01:12:26] your packaging much easier. There are
[01:12:27] also trenches on the back side. So so
[01:12:28] why is it coming in handy? So in
[01:12:30] particular for if you have grading
[01:12:32] couplers or if you have a little mirror
[01:12:34] down here there are different ways of
[01:12:35] getting the light out of your wave
[01:12:37] guide. Sometimes the grading is on the
[01:12:39] same side, sometimes on the opposite
[01:12:40] side of the substrate. So there there
[01:12:42] are lots of different configurations.
[01:12:43] This is just the schematic for one
[01:12:45] possible way of getting the light out of
[01:12:48] your pick. And so there's typically an
[01:12:50] air gap here. Um so that and then um
[01:12:54] order to um either let the beam expand a
[01:12:57] little bit or to to avoid an inter glue
[01:13:00] interface to the to your whatever
[01:13:02] coupling surface you have here. And
[01:13:04] especially if it's if it's not a grading
[01:13:06] coupler, it might be easier to have this
[01:13:08] air gap there. And we these trenches
[01:13:11] that are done at the same time, they can
[01:13:13] act as glue stops or glue pockets
[01:13:15] actually. And in general you have a
[01:13:17] columnated beam but you can also
[01:13:19] sometimes refocus directly to your next
[01:13:22] optical element. And uh what we've been
[01:13:26] recently developing and it's becoming uh
[01:13:29] more and more serious as a product um
[01:13:31] actually we have one client already um
[01:13:34] is these integrated mirrors um where um
[01:13:38] you can have a 90 degrees turn of your
[01:13:41] light and so you have this micro lens
[01:13:43] down here which kind of either
[01:13:45] columnates or refocus the light to the
[01:13:47] next level. And we have a more generic
[01:13:50] design now which we can also uh which we
[01:13:54] are now setting up so that we can give
[01:13:57] them to different companies for for
[01:14:00] demonstrations for for for tests. So uh
[01:14:03] it's a it's a fully integrated way of a
[01:14:07] 90 degrees uh uh step here. You see here
[01:14:10] the zoom in and um you have here then
[01:14:14] the 90 degrees turn and at the top you
[01:14:16] have your micro lens. So here's your
[01:14:17] micro lens. So that either columnates.
[01:14:19] So we made two designs more generic
[01:14:22] designs um where one is focusing the
[01:14:25] beam here outside the microl lens and
[01:14:28] the other has a more columnated
[01:14:31] approach. So it depends a little bit on
[01:14:33] on what you do. So we we had to start
[01:14:35] with some type of design. So for our
[01:14:37] generic design, so this is the one we
[01:14:39] chose and um it can be used maybe for
[01:14:43] for some applications
[01:14:45] and already and for testing definitely
[01:14:48] and our technology has the option to to
[01:14:52] to change the angles of these different
[01:14:54] facets and and also the depths that you
[01:14:57] see here of the different elements can
[01:14:58] be changed. And obviously since we do
[01:15:00] the microl lens in a separate process we
[01:15:03] can have all types of microl lenses and
[01:15:04] microl lens arrays on that side. So this
[01:15:08] is a little bit overview of all the
[01:15:10] packaging features. So as I said before
[01:15:12] we have these simple lenses. They can be
[01:15:14] just glue to fuse silica fiber or or um
[01:15:20] to to some type of um edge emitter. It
[01:15:24] can be directly attached. We can provide
[01:15:26] the corresponding air coating. uh this
[01:15:29] recess that I mentioned earlier. This is
[01:15:31] typically used to assemble an isolator
[01:15:33] or prism and I already talked about the
[01:15:36] the grading type pick and we also offer
[01:15:39] gold tin soda pads. So you can use uh
[01:15:42] reflow processes and we can also do some
[01:15:46] additional feature by deep reactive iron
[01:15:47] etching. So if you need some passive
[01:15:50] alignment or or more or less passive
[01:15:52] pre-alignment, we can add trenches or or
[01:15:55] or or circles into the into the into the
[01:16:00] silicon lenses mostly.
[01:16:02] And I'd like to thank you for your time
[01:16:05] and I'm curious and open for questions.
[01:16:10] Great. Uh great talk Wilfred as always.
[01:16:12] Thank you very much. um in this um in
[01:16:16] this topic what would the the other Epic
[01:16:19] members be be yeah who who could be your
[01:16:23] potential partners and what would you
[01:16:25] so we are really looking forward to to
[01:16:28] to integrate this somewhere so we I mean
[01:16:32] we're talking with some companies
[01:16:33] already about this but really I we need
[01:16:36] guide more guidelines I mean Eric
[01:16:38] mentioned that there are standards being
[01:16:40] setting up and also Tiger mentioned this
[01:16:42] so that's I'm very curious about these
[01:16:44] standards. We are not part of Coobo.
[01:16:46] We're not part of IPC. So, um maybe we
[01:16:49] should. So, it would be very interesting
[01:16:51] for us to know where this road is going.
[01:16:54] there's going to be a standard
[01:16:58] and um
[01:17:01] I think there's some people unmuted
[01:17:05] the background
[01:17:06] and anyway so so for us it's important
[01:17:09] to know the big epic question is there
[01:17:12] um how can we be part of the
[01:17:15] standardization with our technology so
[01:17:18] if there's a standard that is not easy
[01:17:20] to manufacture it would be difficult um
[01:17:22] we can offer of course um uh samples and
[01:17:27] to try this and we can be we we're very
[01:17:29] open. We a lot of our discussions with
[01:17:32] our clients are are to develop new
[01:17:34] features of course in in prototyping. So
[01:17:37] we like to offer prototypes for for
[01:17:39] tests
[01:17:41] but we need to know where the road is
[01:17:43] going what what features are expected
[01:17:45] from us and because this can be quite a
[01:17:48] lengthy development. So the part that
[01:17:50] you see here that took us almost a year
[01:17:52] to develop and if um uh so it's readily
[01:17:56] available now we showed it already at
[01:17:58] OFC and we had very good feedback. The
[01:18:01] question is now what's next? So is this
[01:18:04] the right form factor or are we going in
[01:18:06] the completely wrong direction? What is
[01:18:08] the
[01:18:09] the situation from the transceivers and
[01:18:12] the integrators companies?
[01:18:15] All right. All right. Thank you Wilfred.
[01:18:16] So please make sure you stay until the
[01:18:18] discussion we have the informal
[01:18:20] discussion we have after the event. I'm
[01:18:22] sure there will be people who can who
[01:18:24] can answer those questions and uh so we
[01:18:27] have a question from Costa Golf from
[01:18:29] Engage Photonics.
[01:18:33] Costa
[01:18:34] that is correct. I would like to I would
[01:18:36] like to ask are the optical shapes that
[01:18:40] you can fabricate on the substrate uh
[01:18:44] how what kind of limitations are there?
[01:18:47] Can you fabricate aphheres or
[01:18:50] asymmetrical
[01:18:51] shapes and can they be on both sides of
[01:18:54] the substrate?
[01:18:56] Uh yes. So we can make um aphheres on
[01:19:00] both sides front and back side. We can
[01:19:03] have trenches. can combine trenches with
[01:19:05] a spheres. We can even uh I didn't show
[01:19:08] that today. We can even do head um
[01:19:13] ratings or defractive optical elements.
[01:19:15] So this is all of this is no problem for
[01:19:18] us. We've been doing this since many
[01:19:19] years uh in in in volume products. Not I
[01:19:23] think I have a slide on the defractive
[01:19:26] optics. No, I I took it out. Um so we
[01:19:30] also do defractive optics. Um
[01:19:33] aphheres we do every day we also do
[01:19:35] apheres with very high conic values. So
[01:19:38] minus 10 or minus 12 or or this we can
[01:19:41] do this it's there's no limitation and
[01:19:45] we can combine all these different
[01:19:46] features that you see here in a single
[01:19:48] element front and back side.
[01:19:50] Very nice. What is the typical surface
[01:19:52] quality of the optical surfaces
[01:19:55] surface roughness? So there's the rough
[01:19:59] there's the RMS error of the shape
[01:20:00] deviation which can be anything between
[01:20:03] let's say five and 100 nanometers
[01:20:05] depends on the lens height. So typically
[01:20:08] it's it's much less than 50 nanometers
[01:20:10] between 10 and 30 nanometers that's for
[01:20:12] the shape deviation for roughness we are
[01:20:14] in the single digit nanometers. So this
[01:20:16] is pretty standard for this technology
[01:20:19] of our our process.
[01:20:21] So this is not
[01:20:22] understood. There is no uh diamond
[01:20:25] machining involved or anything which
[01:20:27] which which has good quality but not the
[01:20:29] same quality as our process. So we are
[01:20:31] using we are semiconductor fab we use
[01:20:34] semiconductor processes for etching and
[01:20:36] machining lithography and etching. So
[01:20:39] you have very very very high accuracy of
[01:20:43] the elements and very low roughness.
[01:20:47] Thank you.
[01:20:49] Yes. Um thank you. The second question
[01:20:51] from Moritz Falcon Tech Morris, please
[01:20:55] go ahead.
[01:20:56] Um, I have two technical questions with
[01:20:58] so the first one is I mean is there a
[01:21:01] difference from the material system if
[01:21:03] you're using silica or fuse silicon from
[01:21:05] from the um processing? Can you do
[01:21:08] different more complex shapes in the one
[01:21:09] versus the other or does
[01:21:12] similar?
[01:21:13] That's a very good question. So for well
[01:21:16] um
[01:21:18] so the lens shapes are the same but
[01:21:20] since the refractive index of silicon is
[01:21:22] is a factor 2.2 higher then for fus
[01:21:25] silica you can go high to higher
[01:21:27] numerical apertures of your of your
[01:21:29] laser source. So that means if you have
[01:21:31] a 0.6 NA you can handle this with
[01:21:34] silicon but it's difficult to do with
[01:21:36] fusilica but just because of the
[01:21:38] refractive index not because of the lens
[01:21:40] shape. Um then these these mirrors that
[01:21:43] you see in the top right corner. Uh so
[01:21:45] these are easier to be done in silicon
[01:21:48] but we also have a process now to do
[01:21:50] them in fused silica
[01:21:52] and other than that we use the same
[01:21:55] processing the same uh technology for
[01:21:58] both materials. So there's pretty much
[01:22:00] the same.
[01:22:02] Okay. Second quick question. I mean you
[01:22:04] have also shown structures which have um
[01:22:07] structure on both sides of the of the
[01:22:09] material. So how well can you align the
[01:22:12] top to the bottom surface or the
[01:22:14] structures of both of them?
[01:22:15] But typically we specify better than
[01:22:17] three microns plus - 3 microns. It can
[01:22:20] be better than this. It can be like two
[01:22:23] microns. Um but we always specify better
[01:22:26] than three microns plus - 3 microns. So
[01:22:28] in X and Y. So that means um yeah so it
[01:22:33] it it for on a on a 200 micrometer lens
[01:22:36] or so this is it's more than enough
[01:22:39] typically accurate enough.
[01:22:42] Thanks.
[01:22:46] All right. All right. So let's speak uh
[01:22:48] let's thank our speaker again. Thank you
[01:22:50] very much Wilfred. Let's go back to the
[01:22:52] presentation of John Inhard from OFS.
[01:22:58] Let's try this again.
[01:23:00] Okay.
[01:23:05] Seeing anything yet?
[01:23:09] [Music]
[01:23:22] Make sure so you push the button to
[01:23:24] share the screen at the bottom of the
[01:23:26] Zoom.
[01:23:27] I can do that. Huh?
[01:23:30] Okay,
[01:23:30] I can see. Ah, okay. It's working now.
[01:23:32] Perfect.
[01:23:37] Trying to go to full screen. Just bear
[01:23:39] with me.
[01:23:41] Just push F5.
[01:23:42] Yeah, that's what I did.
[01:23:47] Let's try this.
[01:23:52] Maybe you have multiple screens.
[01:23:57] No, it's just stuck. But but but there's
[01:23:59] no problem. I can start my screen on.
[01:24:01] Okay, let's do that. Yep.
[01:24:04] Plan B.
[01:24:06] Okay. A moment. There you go.
[01:24:20] There we go.
[01:24:22] All right. Uh so again, John Earnhard uh
[01:24:25] with OFS specialty fiber and you'll see
[01:24:29] in the lower right hand corner uh we've
[01:24:31] actually had two logos on each of the
[01:24:33] slides and that's important for this uh
[01:24:36] presentation. So we are uh aptic company
[01:24:40] as OFS and we are owned by Furakawa
[01:24:44] which also is active in lasers. So we
[01:24:48] cover both the active side and the
[01:24:50] passive side uh which makes it life
[01:24:53] rather interesting when we're providing
[01:24:55] the uh the passive part to the laser
[01:24:58] part. So often the fingers are pointing
[01:25:02] different directions. Next slide please.
[01:25:16] And I'm not in the next slide. I'm in
[01:25:18] the next slide.
[01:25:22] Yep. Next slide.
[01:25:24] I am already. You cannot see it.
[01:25:26] Oh, I don't see it. Okay. Yeah. I must
[01:25:28] have a
[01:25:29] No, it's not seeable. Doesn't see it
[01:25:31] kind of.
[01:25:34] Okay,
[01:25:37] just understand everybody else can see
[01:25:39] my slides but it's not on it's only John
[01:25:41] who cannot see it.
[01:25:42] No, we can also don't only see the first
[01:25:43] slides.
[01:25:44] Yeah, we see the first slide on us.
[01:25:46] Okay. So, it's my mistake then
[01:26:01] again.
[01:26:12] Can you see it now?
[01:26:16] coming up. Not yet.
[01:26:22] It's a little bit slow.
[01:26:39] Still not yet.
[01:26:51] Thank you. You co you co- packaged too
[01:26:53] many slides in that presentation, John.
[01:26:56] Exactly. I've got a conflict between
[01:26:59] lasers and fibers.
[01:27:00] How many videos?
[01:27:03] [Music]
[01:27:09] Ianos, why don't you let John try to
[01:27:10] share his slides again?
[01:27:11] Yeah. Now, you know what I'm going to
[01:27:13] do? I'm going to try to use
[01:27:15] I'm going to go to Adobe and hope that
[01:27:17] that helps out.
[01:27:24] Okay, we can see
[01:27:25] that was faster.
[01:27:26] That was faster.
[01:27:27] How about now?
[01:27:29] We can see your cover.
[01:27:31] Let's see if we can see that.
[01:27:33] Yes, it works. Works. Works. You have
[01:27:35] just have the blue thing on the screen
[01:27:36] which you need to click away on the
[01:27:38] upper right from reopen close PDFs.
[01:27:41] Thanks.
[01:27:42] Yeah, now it's perfect.
[01:27:43] Perfect.
[01:27:45] All right. Well, thank you for your
[01:27:46] patience. So, we've been involved across
[01:27:49] uh many different specialty fiber
[01:27:51] applications and markets for the better
[01:27:54] part of 30 years. Um, but specifically
[01:27:58] in the telecom and dataccom uh space, we
[01:28:01] work mainly with OEMs uh and some of
[01:28:05] whom are actually on this uh this call
[01:28:08] now. And uh we've supplied herbium dope
[01:28:12] fibers, polarization maintaining fibers,
[01:28:16] uh low bin loss fibers. Now there's some
[01:28:19] emerging uh new fibers like holo core
[01:28:22] that we're very active in. Tiger
[01:28:25] mentioned multicore fibers. Uh so
[01:28:28] there's really a a broad range of
[01:28:31] products that we're supplying into the
[01:28:33] industry. Um on the other side for Aawa
[01:28:38] uh so we also have been very activists
[01:28:41] for Aawa for many years in uh telecom
[01:28:44] type lasers. Uh we have a chip fab in
[01:28:48] Chiba, Japan. Uh we also have a laser
[01:28:52] assembly and packaging facility
[01:28:56] owned in Thailand. Um and most of what
[01:29:01] we do in terms of lasers is DFB based uh
[01:29:05] technology. Uh historically that's been
[01:29:08] package lasers like uh microITLA tunable
[01:29:13] lasers. Uh the trend recently has also
[01:29:17] been to uh provide uh chips uh for
[01:29:22] integration with silicon photonics. Um
[01:29:26] but then uh in the future we are also
[01:29:29] looking at external laser source or ELS
[01:29:32] modules and we'll actually be shipping
[01:29:34] our first samples uh in June.
[01:29:39] Uh Tiger showed this earlier but uh this
[01:29:42] is the transition going from uh copper
[01:29:46] dominant to fiber dominant on the right.
[01:29:52] And uh just an exploded view of the
[01:29:55] fiber routing. And as Tiger had pointed
[01:29:57] out uh we have a large number of both PM
[01:30:01] fibers and single mode fibers in the
[01:30:04] architecture highlighted above on the
[01:30:06] right.
[01:30:08] So when we look at this from a fiber
[01:30:10] optic perspective,
[01:30:13] of course there's single mode and PM
[01:30:15] fiber. Uh and so what we have to worry
[01:30:19] about from an optical standpoint is what
[01:30:22] is the mode field diameter. Uh so in
[01:30:25] some cases uh we've got customers who
[01:30:28] want a conventional sort of nine micron
[01:30:30] type mode field diameter. Um and then in
[01:30:34] other cases a customer wants something
[01:30:37] that maybe uh interfaces directly with a
[01:30:39] chip uh that might be in a three micron
[01:30:43] mode field and a couple of options in
[01:30:46] between. Um low bin loss uh is becoming
[01:30:51] increasingly important as transceiver
[01:30:54] modules shrink in size and go to QSFPD
[01:30:57] and OSFP.
[01:30:59] uh and there will likely be pressure on
[01:31:03] uh low bend loss as well in the CTO uh
[01:31:06] area. Um PM fibers uh we need to worry
[01:31:10] about traditional PM properties like
[01:31:13] beat length and per and so on. And uh
[01:31:19] another area that we have to watch out
[01:31:21] for is path interference. uh
[01:31:24] particularly as you go to these low bin
[01:31:26] loss fibers, it can become more of a
[01:31:29] problem um and and looking at also
[01:31:32] matching the mode field diameters. So,
[01:31:35] so there's an optical side to this and
[01:31:38] then there's also a mechanical side and
[01:31:42] uh we will see increasing pressure as a
[01:31:44] fiber optics supplier um to uh improve
[01:31:49] our tolerances. Um, so those could be
[01:31:53] core clad offset,
[01:31:56] uh, core diameter, clad diameter. You
[01:31:59] know, on the laser side, we do a lot of
[01:32:01] things to, uh, improve our power and
[01:32:04] improve our efficiency. If we don't do
[01:32:06] things right on the fiber side, we can
[01:32:10] essentially negate or throw away a lot
[01:32:12] of that effort. Um so over time as the
[01:32:16] volume increase there will be more
[01:32:18] pressure to to improve the uh the
[01:32:20] mechanical tolerances.
[01:32:23] Um in addition uh the question is what
[01:32:26] clad diameter uh so 80 micron or 125
[01:32:30] micron the initial ELS samples we'll
[01:32:33] ship next month will be based on a 125
[01:32:36] uh cladding diameter. uh but we are uh
[01:32:40] certainly seeing pressure for more 80
[01:32:43] micron fibers both PM and SMF that we're
[01:32:46] we're currently shipping for transceiver
[01:32:49] and other types of applications.
[01:32:51] Um in addition uh the coding we're
[01:32:55] seeing uh couple of emerging trends on
[01:32:58] coding. One is smaller diameter. Uh so
[01:33:01] on a 80 micron uh cladding the typical
[01:33:07] uh diameter is approximately one uh 165.
[01:33:11] Uh we have begun shipping some samples
[01:33:14] at 135 micron. Again that's a coating
[01:33:18] diameter. Um there was some discussion
[01:33:21] about thermal issues and solder reflow
[01:33:24] earlier. So uh we've also seen more
[01:33:28] pressure to uh provide alternate
[01:33:30] coatings that will survive uh at least
[01:33:34] shortterm duration of higher temperature
[01:33:36] with solder refflow and possibly survive
[01:33:40] longer term duration with some hot spots
[01:33:43] uh in some of the tight packaging.
[01:33:48] So on the uh the ELS front uh as I
[01:33:52] mentioned we'll be shipping uh next
[01:33:54] month in June. Uh the in this particular
[01:33:58] case the ELS is pigtailed. We did see a
[01:34:01] back plane discussion earlier. Uh that
[01:34:04] is something that we will make available
[01:34:06] at a later time. Um, we're looking at
[01:34:09] either a a CWDM style format or 1311
[01:34:15] uh style. Um, it is uh it is PM uh
[01:34:20] fiberbased using OFS uh PM fiber. Um and
[01:34:26] the the current uh output power that
[01:34:29] we're uh and this would be power out of
[01:34:31] fiber that we're focusing on is uh 100
[01:34:34] mills.
[01:34:40] So this is my closing slide. uh you
[01:34:43] know, this really is all about a new
[01:34:46] ecosystem and and I really appreciate
[01:34:49] that Epic is pulling together uh all of
[01:34:51] our uh different speakers today. Uh you
[01:34:55] know, it's interesting. We're
[01:34:56] approaching this more from a component
[01:34:59] level, but we've heard uh people with
[01:35:02] higher level uh system integration and
[01:35:06] and it's going to require a network of
[01:35:08] all of us uh to develop the right
[01:35:11] standards uh to design and integrate
[01:35:13] these new components and to scale
[01:35:16] production. And we think we're in a very
[01:35:19] interesting position as what we call one
[01:35:22] for Aawa. So the combination of Furakawa
[01:35:26] active devices uh along with the OFS uh
[01:35:30] passives and u really appreciate the
[01:35:34] opportunity to speak to all of you today
[01:35:38] and thank you for the patience and
[01:35:40] getting the slides set up.
[01:35:43] Okay, thank you very much John. Um yes
[01:35:46] indeed you're right. So the ecosystem is
[01:35:48] very important uh especially at this
[01:35:50] stage and um actually Tiger from Senko
[01:35:55] has already a few questions for you.
[01:35:58] Great. Hey Tiger.
[01:35:59] Hi John. Um thank you very much for a
[01:36:01] great presentation. A couple of
[01:36:03] questions but one is u is Furukawa or
[01:36:06] OFS working on the pre-aligned and
[01:36:08] ribbon fiber for easier termination?
[01:36:12] That is a great question and that is
[01:36:14] very much uh the question of the day. Uh
[01:36:17] yes uh but as you can imagine it is very
[01:36:20] much non-trivial. Um so we can talk some
[01:36:24] some more offline about that but uh you
[01:36:27] know that has not come up so far and uh
[01:36:30] the what I'll call the packaging of the
[01:36:33] fibers uh is definitely a critical part
[01:36:37] of all of this. Uh you know we're we're
[01:36:39] dealing with a a large number of fibers
[01:36:42] that presents a bit of a rat's nest. So
[01:36:44] having uh the right packaging of those
[01:36:47] fibers and then a pre-alignment of the
[01:36:50] PMF. Uh so it's the desire is to have a
[01:36:54] drop in solution that will drop into an
[01:36:57] array or into a an NO style connector.
[01:37:01] So great question and uh if anybody's
[01:37:03] got great ideas on that uh we're all
[01:37:05] ears.
[01:37:07] Maybe by the next event there's some
[01:37:09] news hopefully.
[01:37:10] Yeah. Okay. And the sec second question
[01:37:13] was about relative to my comments made
[01:37:16] earlier about the um there are some
[01:37:18] changes to the process and operation
[01:37:21] when it comes to kovash optics of course
[01:37:22] the reflow to 260 degrees u celsius and
[01:37:27] also potentially depending on the
[01:37:28] applications uh you may have to operate
[01:37:31] at the higher temperature beyond 100 uh
[01:37:34] and I think it primarily it's about the
[01:37:37] coating of the fibers but is there any
[01:37:39] specific materials that you are uh
[01:37:42] selling now or be in development of
[01:37:45] such?
[01:37:46] Yeah, we we actually are selling uh
[01:37:48] currently at least at a prototype level
[01:37:51] uh a essentially a mid-temper dual coat.
[01:37:55] Uh we've provided papers on this at OFC
[01:38:00] in the past and uh so we are shipping
[01:38:03] some commercial quantities both of PM
[01:38:06] fiber and of single mode fiber. So thank
[01:38:09] you for that question.
[01:38:10] Thank you very much John. Yeah.
[01:38:14] Yes. Thank you John for the very nice
[01:38:16] presentation. And we go to our next
[01:38:19] speaker mod safe right and the R&D
[01:38:22] manager from Python tech service
[01:38:27] mod. The floor is yours.
[01:38:29] Perfect. Can you see my screen um
[01:38:32] yes it's all good.
[01:38:33] Perfect. Great. So yeah um welcome and
[01:38:37] thank you for the opportunity to present
[01:38:39] um a few slides and also two video
[01:38:42] snippets on what we as a um solution
[01:38:45] provider for the testing and assembly um
[01:38:48] are facing with when it comes to co-
[01:38:50] package optics. Um first of all I mean
[01:38:53] the picture you have seen multiple times
[01:38:54] already. Um so we're having a lot of
[01:38:58] fibers entering the the main um yeah
[01:39:01] main main P core and the connection of
[01:39:04] the fibers um we do mainly see two
[01:39:06] different options. The one is having a
[01:39:08] passive fiber alignment in case there's
[01:39:10] a um silicotonic chip with weak groups.
[01:39:14] So the passive um placement of the
[01:39:16] fibers is doable. And the second option
[01:39:19] would be um in case we have a silicon
[01:39:21] photonic chip which edge edge coupling
[01:39:24] um that we use an active fiber alignment
[01:39:27] then either with or without additional
[01:39:29] optics in between. So these are the two
[01:39:31] options I will um have a closer look in
[01:39:34] in the next slides. However, before
[01:39:37] looking on the assembly, um you may want
[01:39:39] to have a first check on your ideal and
[01:39:42] wifer level already about the
[01:39:45] functionality of your photonic
[01:39:46] integrated circuits. So this is where we
[01:39:49] offering so-called electrooptical wafer
[01:39:51] level testings by which we can um either
[01:39:54] using the grading couples from the top
[01:39:56] or also edge coupled um um capabilities
[01:40:00] on wer level for coupling the light in
[01:40:03] and out of the chip as well as then for
[01:40:05] sure the standard um DC and RF
[01:40:08] electrical um contacting. So in the case
[01:40:11] you have verified the functionality of
[01:40:13] your um pick device with that one um you
[01:40:17] can do the next step towards the um
[01:40:21] integration of laser diets. So we
[01:40:23] learned that either um I mean a lot of
[01:40:25] trends having
[01:40:27] the laser source external installed.
[01:40:29] However, if you looking for some heat
[01:40:31] with genius integration of this five D
[01:40:35] into the silicon pick, um we have
[01:40:38] developed a quite a nice technique based
[01:40:40] on um um let me move to the pointer.
[01:40:46] So based on an infrared imaging or
[01:40:50] optics using an um yeah microscope
[01:40:53] objective. So we are able to um look
[01:40:56] really through the silicon because the
[01:40:57] silicon is as you know transparent um in
[01:41:00] the infrared region by that we can
[01:41:03] really um monitor and look at the
[01:41:05] interface between the 35 day and the
[01:41:08] silicon substrate and you see kind of
[01:41:10] two pictures here to the right where we
[01:41:12] really looking through the silicon die
[01:41:15] and then um when having the 35 on an six
[01:41:19] axis um alignment stage we can lively
[01:41:23] monitor the position and then align
[01:41:25] until we have achieved the final
[01:41:28] position.
[01:41:30] um we have been um yeah able to really
[01:41:34] achieve an accuracy of the die tracking
[01:41:36] not of the bonding but of the die
[01:41:38] tracking through the silicon in the
[01:41:40] range of 70 nanometers of one sigma
[01:41:44] bringing us into the region where really
[01:41:46] um the direct but coupling of the 35 by
[01:41:51] intelliging photonic wave guides is
[01:41:53] possible and um combining that with
[01:41:57] localized laser soldering meaning um not
[01:42:00] only doing imaging from the bottom but
[01:42:02] also using laser soldering or the
[01:42:04] fixation of the two components. It's
[01:42:06] quite a nice combination.
[01:42:09] So um after this fixation of the 355 to
[01:42:13] the big you may want to do a second step
[01:42:15] of functional testing on W level or on D
[01:42:20] level depending on what you have done
[01:42:22] before.
[01:42:24] But then finally we come to the assembly
[01:42:27] of the fiber and for the fiber
[01:42:29] preparation we have a so-called fiber
[01:42:31] line machine series which is also using
[01:42:35] or making use of a CO2 laser to remove
[01:42:38] the coating from the fiber arrays fibers
[01:42:42] or fiber arrays as well as then using a
[01:42:44] higher focused um beam to really cut the
[01:42:47] fibers and that's what you will see in a
[01:42:51] quick video snippet here. Um really you
[01:42:55] see a CO2 laser coming from top and
[01:42:57] bottom. So from both sides at the same
[01:42:59] time removing the coating as well as
[01:43:01] then being slightly refocused and um
[01:43:04] cutting the fiber or fiber way into
[01:43:07] different components.
[01:43:09] And one example could be you could use
[01:43:10] that fiber way to bring it into the
[01:43:13] connectors. Then fix it. This is um UV
[01:43:18] epoxy and you have the connector already
[01:43:20] ready on one of the ends of the fiber
[01:43:23] array.
[01:43:24] The second side you can prepare in a
[01:43:27] similar way and then for example use for
[01:43:30] the um passive fiber to recroof assembly
[01:43:33] process which is shown here on a picture
[01:43:35] just taken from Google. So you have the
[01:43:38] fiber arrays and um when placing them
[01:43:40] into the weak grooves and they are
[01:43:43] automatically um passively aligned so
[01:43:46] that you only need to fix them um with
[01:43:49] some kind of um coverlet to keep it
[01:43:51] mechanically in position and that is
[01:43:54] exactly um the process you see on this
[01:43:57] small video which you can find also on
[01:43:59] um YouTube. So definitely there's um
[01:44:02] only a part of
[01:44:06] um right now.
[01:44:08] Can you still hear me?
[01:44:13] Yes, we're here. Well,
[01:44:15] okay. So um because there's a music to
[01:44:17] the video. Sorry for that. So you can
[01:44:19] see the fiber array which is um yeah
[01:44:23] aligned from only from T cameras from
[01:44:25] top and bottom. Oh, sorry from the top.
[01:44:28] So the fiber way and then the pick
[01:44:30] itself and then um based on machine
[01:44:32] vision the fiber way is placed to the
[01:44:36] into the V grooves so that the yeah um
[01:44:40] alignment is done in that case the
[01:44:42] fibers removed to bring in the epoxy
[01:44:45] before bringing the fiber back to the
[01:44:47] safe position.
[01:44:49] So this is the method using the reg. The
[01:44:53] second approach which I discussed
[01:44:54] earlier is um the use of um lenses or
[01:44:59] lens arrays directly um aligned and
[01:45:02] fixed to the edge of the photonic
[01:45:05] integrated circuits like I'm showing you
[01:45:07] in a picture from Tindle and then I'm
[01:45:10] using a single mode fiber connector
[01:45:14] which also has a lens at the end a lens
[01:45:16] array so that we are widening the beam
[01:45:18] to have um slightly less or looser
[01:45:21] alignment. tolerances and by that
[01:45:24] allowing for pluggable connection of the
[01:45:26] system and that is something which we
[01:45:29] also um developing together with tindle
[01:45:31] at the moment in our labs. So really
[01:45:33] using um the fiber ar away including
[01:45:36] lens away at the end on the one side.
[01:45:39] The cripper here is holding a small lens
[01:45:42] away and this is one of the pix chips um
[01:45:44] which we're using for the tests. So to
[01:45:47] couple light in and then in a loop back
[01:45:49] also out of the chip and then we can fix
[01:45:52] the lens the lens away to the chip and
[01:45:55] having a loose alignment tolerance for
[01:45:57] the pluggable for the connector which
[01:45:59] can be made pluggable at the end.
[01:46:03] Okay, so far these have been all
[01:46:05] individual um machines. So different
[01:46:08] machines for different um purposes.
[01:46:11] However, looking into really um high
[01:46:14] support concepts, definitely you want to
[01:46:17] implement everything in a more automated
[01:46:20] assembly line. And this is a picture of
[01:46:22] an assembly line for the automotive
[01:46:24] industry which we have installed in
[01:46:26] Germany a few years ago. So this is not
[01:46:29] kind of a fake image. is really um a
[01:46:31] large assembly line including multiple
[01:46:33] input and output feeder as well as
[01:46:36] different stations of the assembly
[01:46:39] and this is only the hardware so far but
[01:46:42] there's also um quite a lot of software
[01:46:44] involved one side the software for the
[01:46:47] process implementation but we also
[01:46:50] setting up um um we have a fcon
[01:46:53] software part which is allows us to
[01:46:56] really do a nice interface with KPI
[01:46:58] tracking as well there's um yeah AI or
[01:47:02] ML um algorithms processes really for
[01:47:06] process optimization in that whole
[01:47:08] assembly line. So to increase yield, UPH
[01:47:12] and all of that.
[01:47:15] Okay, thanks. That was a very quick
[01:47:17] overview about um what different aspects
[01:47:20] we could offer for the yeah copo
[01:47:24] assemblies and I'm for sure happy for
[01:47:26] any questions.
[01:47:28] Thanks a lot Maurice. Thanks a lot for
[01:47:30] this informative presentation. Uh
[01:47:32] amazing job really from Phom Font Tech.
[01:47:34] Um uh takes it I would like to take the
[01:47:38] chance to apologize from before because
[01:47:39] it might be the case that I had the
[01:47:41] technical problem and not John. And uh
[01:47:44] do we have any question for uh for
[01:47:46] Morice? It seems that we have I hope
[01:47:49] it's it's John. So John again I'm sorry
[01:47:53] before I think it was my mistake.
[01:47:57] So Moritz thank you uh very interesting
[01:48:00] and to Tiger's question earlier uh have
[01:48:03] you been doing any work with uh PM
[01:48:06] alignment in an array format so taking
[01:48:09] PM fiber and then put blocking it uh
[01:48:13] across say eight fibers and then fixing
[01:48:15] those fibers in place
[01:48:18] and so we have been working with PM
[01:48:20] fibers but I think only with um with not
[01:48:22] eight in a row so that's definitely
[01:48:24] something which is um also on on the
[01:48:26] list because that's becoming more and
[01:48:28] more important also having really um not
[01:48:30] only one or two fibers but really a huge
[01:48:32] array or multiple fibers um on one chip.
[01:48:35] So I also see that what you mentioned
[01:48:37] earlier that the interface between fiber
[01:48:40] and chip that's kind of um alo becoming
[01:48:42] a limit from the um footprint. So really
[01:48:45] having more power or more channels into
[01:48:47] one fiber and definitely be a
[01:48:50] good option to go.
[01:48:52] Thank you.
[01:48:57] Do you have any other question for uh uh
[01:49:00] well let me
[01:49:02] Yes please.
[01:49:04] Yeah I had a quick question. So you show
[01:49:06] different alignment procedures Morris
[01:49:09] and um giving the talk before from from
[01:49:12] OFS from John. So when do you um when do
[01:49:16] you need to couple light into the fiber
[01:49:18] and when can you use other features for
[01:49:21] alignment? So I mean there's there's
[01:49:23] active alignment and there's active
[01:49:25] alignment. So and then there's passive
[01:49:26] alignment. So when I talk so and uh so
[01:49:29] for active alignment sometimes you can
[01:49:31] use features on the either on the on for
[01:49:34] on the for rules or on the micro optics
[01:49:36] itself and sometimes you have to couple
[01:49:38] light into it. So is there like a rule
[01:49:42] or like do you like to have a standard
[01:49:43] for this to go even to the next step?
[01:49:46] Um
[01:49:48] um that's a good point. I mean if you
[01:49:50] really come to fiber to pick coupling I
[01:49:53] think um really coupling light so doing
[01:49:54] it actively is quite a preferred way and
[01:49:58] it also it's um from the
[01:50:01] speed point of your up I think it's not
[01:50:03] necessarily a bad thing I mean if you
[01:50:05] have the rec that's rather simple go in
[01:50:08] and um push yeah p push to the end
[01:50:11] without for sure having contact
[01:50:14] um but for for most of the other cases I
[01:50:16] definitely would prefer having some some
[01:50:17] light in the fibers at least in the
[01:50:19] moment moment um because that also can
[01:50:22] be um if it's prepared well quite um
[01:50:26] fine from from the cycle time. So if
[01:50:28] you're going to submicro accuracies um
[01:50:31] it's definitely also taking time from
[01:50:33] the vision and all of that.
[01:50:38] Okay,
[01:50:39] thank you Marit
[01:50:41] please go for it with me. Now because we
[01:50:44] we've been so our clients regular ask us
[01:50:47] for adding features and I think uh other
[01:50:51] other people in the in the in the plenum
[01:50:53] here in the in the panelist have the
[01:50:55] same request so to do fully passive
[01:50:57] alignment just with features on the
[01:50:59] micro optics. So some type of ledges,
[01:51:02] some type of of uh features that are So
[01:51:06] I just wonder how realistic you see this
[01:51:08] morets or is is this just um out of the
[01:51:12] question?
[01:51:14] I mean if it's really the silicon wave
[01:51:17] cards themselves I mean they are what a
[01:51:18] few hund nanometers in dimension. So if
[01:51:22] as soon as we going to some beam
[01:51:24] expansion kind of more size converters I
[01:51:26] think that's becoming more realistic. So
[01:51:30] good. I mean I guess that also your
[01:51:32] factors are in the
[01:51:34] micrometer range.
[01:51:37] So I think that that's somehow where we
[01:51:39] need need to be one to micrometer but
[01:51:42] not much more than that and it's um it
[01:51:45] also depends on the power loss which you
[01:51:47] have. So we are facing quite
[01:51:50] for for the quantum applications for
[01:51:51] example I mean this is becoming even
[01:51:53] more critical to have a very high
[01:51:56] coupling yield. So I think that could be
[01:52:00] difficult.
[01:52:02] Okay. Yeah, I I agree.
[01:52:06] Thanks a lot. Thanks a lot. So maybe
[01:52:09] it's better now to go to the next
[01:52:10] speaker and if you want to have uh to
[01:52:12] ask more questions to Morice, you can
[01:52:14] stay here in the informal session and
[01:52:16] keep asking questions. So the next
[01:52:18] speaker for today will be Yorg Moleski
[01:52:20] from Nanascribe. Hello Yorg.
[01:52:23] And you're ready as always. The floor is
[01:52:26] yours.
[01:52:29] You're only muted.
[01:52:31] Yes.
[01:52:32] Perfect.
[01:52:33] Okay. Can everybody hear me? Nice to see
[01:52:36] everybody or nearly. Uh I'm Yolinski
[01:52:39] from Nanosk. I'm the business
[01:52:40] development development manager. Uh
[01:52:43] first time introductory slides. So uh we
[01:52:46] are uh now more than 100 employees since
[01:52:49] this month. Uh I'm not sure if it's
[01:52:52] going to stay like this till next month.
[01:52:53] We are hiring again and again. So if
[01:52:55] anybody is interested a side note uh we
[01:52:58] have more than 40 years in the in the
[01:53:01] business and we provide 3D printing
[01:53:03] equipment which then can enable uh
[01:53:05] reduced losses optical losses for fiber
[01:53:08] to chip uh coupling. Um let's click here
[01:53:15] so I will not talk about our key
[01:53:17] technology uh the printing technology
[01:53:19] you can refer to our website and then
[01:53:21] you can have a look at it. So with
[01:53:23] technology enables basically that we can
[01:53:25] connect from fiber to chip. Um so you
[01:53:29] see here the basic scheme the idea is
[01:53:31] that we uh enlarge a mode field on both
[01:53:35] side with uh cumation lenses printed on
[01:53:38] the chip or on the fiber. Then we have a
[01:53:40] larger beam area. We can overlap and
[01:53:42] reduce by this losses and the uh the
[01:53:45] coupling losses and the alignment uh can
[01:53:47] be much more reduced in the order of two
[01:53:49] to five micron uh plus minus. So uh one
[01:53:53] thing we do is this fiber printing uh we
[01:53:56] can do uh cimation lenses uh focusing
[01:54:01] lenses tapers periscopes and we have a
[01:54:04] partnership with fix in uh Netherlands
[01:54:07] for these kinds of lens fiber arrays uh
[01:54:09] so we supply them through through fix if
[01:54:12] you have if you're interested. Uh here's
[01:54:14] an example of the alignment of the fiber
[01:54:17] arrays. So uh we had somebody uh doing
[01:54:20] some vapor in the lab and then put a
[01:54:22] green light through the through the
[01:54:23] fiber array. So you can see they're very
[01:54:25] perfectly aligned. Um we had very good
[01:54:28] experience at least in fiber fiber
[01:54:30] coupling. We also have done some fiber
[01:54:32] to chip coupling with focusing limit
[01:54:34] elements but I will show a few results
[01:54:35] later. Uh on the other side also we can
[01:54:38] print on chips. Um this will be uh is
[01:54:41] already available uh for first tests. So
[01:54:44] we are looking for people going further
[01:54:45] with this. Um so how is going with
[01:54:48] printing on fibers? Um basically again
[01:54:51] we want to have a wider beam to have
[01:54:53] relaxed tolerances in packaging or we
[01:54:56] can do a small focusing beam to couple
[01:54:59] into the wave guide order of three
[01:55:02] microns or two microns for example
[01:55:04] focusing uh spot side of the modefield
[01:55:07] diameter. Um
[01:55:10] there's something
[01:55:12] that was too fast. I wanted to show a
[01:55:15] short a small video. So if this is
[01:55:18] starting, let me quickly start the
[01:55:20] video. So here you can see a video
[01:55:22] shortly of the fiber printing process.
[01:55:25] On the left side you see the scheme. So
[01:55:26] that's was for a a green fiber array.
[01:55:29] You have the eight lenses. We first put
[01:55:31] in the uh first and last position uh
[01:55:34] costly and then the printing process
[01:55:37] starts automatically.
[01:55:38] That's the last position. Um and now we
[01:55:42] do the alignment. Basically, we look for
[01:55:44] tilt and flatness and set position. And
[01:55:48] additionally, we also look how is the
[01:55:50] core light emitting. So we have an
[01:55:51] illumination unit. Uh so we directly
[01:55:54] print in the direction of light that
[01:55:55] enables these precise parallel beams
[01:55:58] coming from the fiber array. We do this
[01:56:00] on each core. Um we could speed up the
[01:56:03] process doing it on every second but we
[01:56:05] have seen that the tolerances of the
[01:56:08] error is required to have a good quality
[01:56:10] to do it on every core.
[01:56:14] Similar thing for printing on chips. So
[01:56:16] the other side basically so again this
[01:56:19] is from the OCT project. This is a
[01:56:20] handheld uh OCT project with together
[01:56:23] with size and some other partners.
[01:56:25] Here's again we do a different
[01:56:27] technology. We use a com focal module.
[01:56:29] We can detect precisely why with Z and
[01:56:32] XY can scan the position of the wave
[01:56:35] guides and then we can print uh the uh
[01:56:39] optical elements can be here again
[01:56:42] focusing columnating lens can be tier
[01:56:45] mirrors different optical elements free
[01:56:47] completely free form.
[01:56:50] This is what it looks like. Um we have
[01:56:53] very good experience. We can also do
[01:56:55] simulations uh for customers for the
[01:56:58] coupling losses. We have a wavefront
[01:56:59] propagation inhouse software for doing
[01:57:03] uh tests and visibility studies. Uh
[01:57:05] yeah, in general this enables uh reduced
[01:57:09] losses. So but so where do they occur?
[01:57:12] So here's an example of a fiber to fiber
[01:57:14] coupling just a scheme. I will not go
[01:57:16] further in this in the PDF I supplied
[01:57:18] there are some more information on this.
[01:57:20] So mainly we see that you have frenal
[01:57:22] losses which is only can only become by
[01:57:25] coating. We have bulk absorption. This
[01:57:27] is very low with our resins we have. You
[01:57:30] can have a misalignment of the core on
[01:57:32] the fiber or on the on the V-groove. On
[01:57:34] the fiber we are at 500 nometer precise
[01:57:36] to the core. So precise more precise
[01:57:39] than the mechanical precision and then
[01:57:41] you can maybe have a mode field mismatch
[01:57:44] uh but we have seen that uh with 10% on
[01:57:47] 25 micrometer we have still we are still
[01:57:49] in the 1dB range. This is what is shown
[01:57:52] here in the next image. So we have done
[01:57:54] here green light fiber coupling uh tests
[01:57:58] in house fiber to fiber 25 micrometer
[01:58:01] cimated over 1.2 2 mm can be smaller can
[01:58:04] also be less than a millimeter can be a
[01:58:06] few hundred micrometer. Um so we had one
[01:58:09] five microme for 1 dB lateral alignment
[01:58:11] to uh tolerances and coupling was 0.7
[01:58:15] lens and 0.3 for the astral reflection.
[01:58:19] Um and here is just a coupling. So you
[01:58:21] have the fiber array let me quickly
[01:58:23] check. So this is the wave guides and
[01:58:25] you have the fiber area on this side
[01:58:27] coupling into the 1550 nanometer and
[01:58:30] there we had a 1.7 dB coupling per
[01:58:32] interface loss.
[01:58:34] So this is all done with our quantum XLI
[01:58:37] we introduced at the beginning of the
[01:58:38] year. So for printing on PM or SM fibers
[01:58:42] photonic ships and also on 3D
[01:58:44] photographies.
[01:58:46] Um now just some images to see what is
[01:58:48] possible. You have seen this already a
[01:58:50] little bit. These are the geometries we
[01:58:52] can do. It's really nice free form again
[01:58:55] the aligned beams.
[01:58:58] See the lenses in this time this were
[01:59:00] five here seven lenses printed.
[01:59:04] This is again a special one. We can do
[01:59:05] really preform do a fast axis slow
[01:59:07] access columator directly printed on a
[01:59:09] facet of edge emitting laser.
[01:59:13] And now it's coming to the end. So I'm
[01:59:15] think I'm nearly in time right now. Um
[01:59:18] what can you do for us? Come to us with
[01:59:20] your coupling challenge. we can help you
[01:59:22] or at least try to cannot promise 100%
[01:59:25] success rate but we have very good
[01:59:27] results uh and feedback from our
[01:59:29] customers
[01:59:30] um and uh what can we do uh we can do
[01:59:34] the automatic printing uh fiber arrays
[01:59:36] via fix and chips we can discuss this
[01:59:38] also um and we can validate hope to
[01:59:42] validate uh our solution for your
[01:59:44] reduced maybe reduced fiber pitch
[01:59:46] because we don't care about the pitch we
[01:59:48] align where the fiber
[01:59:51] Thank you.
[01:59:56] Okay. Thank you, York. Uh that was a
[01:59:58] great presentation. Do we have any
[02:00:00] questions for York in the in the
[02:00:03] audience?
[02:00:07] Okay. Uh Murit, please go ahead.
[02:00:11] Yes. And one question, you mentioned the
[02:00:13] um tool specifically also for quantum
[02:00:16] applications. It's called Quantum. Um so
[02:00:18] what is experience um with the 4K range
[02:00:22] and the um I mean if you're cooling it
[02:00:24] down your devices we have some of our
[02:00:28] yeah some of our residents have been
[02:00:30] cooled down to 4K without any major
[02:00:32] issues
[02:00:34] but we have not the data for all resins
[02:00:36] for some resins we have the values it's
[02:00:38] it has been proven um and for others we
[02:00:41] don't miss the information
[02:00:43] if I would imagine a system or whatever
[02:00:45] laser diet your coupling um fiber into a
[02:00:49] p kind of a different material system.
[02:00:53] So you could imagine that it survives
[02:00:55] the cooling down to 4 Kelvin.
[02:00:57] Yes, this is what customer from us has
[02:01:00] done. Yeah.
[02:01:00] Okay. Thanks.
[02:01:05] Okay, thank you. And the second question
[02:01:07] comes from the European Space Agency, L
[02:01:09] McKenzie. Please go ahead.
[02:01:12] Hi Jük, thank you very much for the
[02:01:14] presentation. Just a quick question.
[02:01:16] Have you done any kind of um mechanical
[02:01:19] assessment of um how well your lenses
[02:01:22] adhere to the surface to which you um
[02:01:26] you mount them?
[02:01:28] Uh we have not done according to any
[02:01:30] space related standards. We have done
[02:01:32] internally in the lab and with customers
[02:01:34] but not to space related standards. We
[02:01:36] are we would like to so we can talk
[02:01:38] about that. I mean if we talk about say
[02:01:40] teleordia standards have you done any
[02:01:43] assessment to those standards or
[02:01:45] I need to check with our customers. I
[02:01:47] have no information on that.
[02:01:49] Okay. Thank you.
[02:01:54] Okay. Uh thank you Jorg. I'm wondering
[02:01:56] if if um nonscribe is developing a new
[02:01:59] tools also like an advanced versions of
[02:02:02] this uh of the alignments.
[02:02:06] What?
[02:02:06] We can talk about this. We can talk
[02:02:08] about this in a year.
[02:02:10] All right. All right. Then then then
[02:02:12] make sure you you still join our
[02:02:14] seminars in a year.
[02:02:15] No worries. I love these seminars. I
[02:02:17] will join them. No worries.
[02:02:19] Great. We love you too, York.
[02:02:22] See you.
[02:02:24] Okay. Thank you very much. It was a
[02:02:25] great presentation and uh yes, let's
[02:02:28] thank uh all the speakers uh of the day.
[02:02:31] So this actually concludes our seminar
[02:02:34] for this time. Uh let me just share a
[02:02:38] few thoughts with you right now.
[02:02:58] And
[02:03:00] that's frozen.
[02:03:18] Okay. and seems to have again issues
[02:03:20] today with Zoom. Uh you can
[02:03:24] Yes. Hello. I'm I'm back. But yes, there
[02:03:26] is something going on with the screen
[02:03:28] sharing. Um let me try again. Sorry for
[02:03:32] this.
[02:03:33] Can you see it? Well, now
[02:03:36] perfect. Full screen mode.
[02:03:39] Oh, perfect. It is in a full screen
[02:03:42] mode.
[02:03:45] It's not.
[02:03:47] We can see it. We can see. Don't touch
[02:03:49] it. We see it.
[02:03:51] Okay. Good. So, uh yes, regardless of
[02:03:54] this technical issues, so uh we had a
[02:03:56] great discussion I believe today. So, um
[02:03:59] we have joined in the room wonderful
[02:04:01] speakers from Senko, Fran from Hoffer,
[02:04:04] Huawei, Nanoscribe,
[02:04:06] Source, Micro Optics, Fant and OFC. So,
[02:04:09] we have heard about the upcoming
[02:04:11] challenges in the density increase in
[02:04:13] the in the fiber connectors. We have
[02:04:16] heard about uh the frownhoffers
[02:04:19] technology for the low heat well low um
[02:04:24] energy consumption and low price
[02:04:28] uh technology uh upcoming. We have heard
[02:04:31] about the 100T cop package optics
[02:04:34] challenge from Huawei. We have uh
[02:04:37] overseen the the external laser sources
[02:04:40] and uh and the technologies related uh
[02:04:44] developed by OFS. We have heard today
[02:04:47] the talk from Fant about the
[02:04:50] electrooptical wafer level testing and
[02:04:53] also from nanoscribe about the quantum
[02:04:56] XLIN and the 3D printing and what can 3D
[02:04:59] printing can do for for the co-acked
[02:05:02] optics. So I'd like to just remind you
[02:05:05] that u so we are close to the end of
[02:05:09] this season in the online technology
[02:05:11] meetings. So so we are here on the 23rd
[02:05:14] of May and um you might be interested in
[02:05:16] joining us for the two upcoming OTMs on
[02:05:19] June 13 on the Vixel technology and
[02:05:23] applications and also on the June 27 for
[02:05:26] the quantum communication and the QKD
[02:05:29] technologies. After this, we're we jump
[02:05:32] already in the in the next season that
[02:05:34] will start in September and we'll have
[02:05:36] even more exciting events for you. Uh,
[02:05:40] another reminder, so this this um this
[02:05:44] seminar was uh possible thanks to all
[02:05:47] the team of Epic and Epic, the European
[02:05:50] Photonix Industry Consortium supports
[02:05:53] you in technology, in market reports,
[02:05:55] provides you with a networking
[02:05:57] opportunities and access to new markets.
[02:05:59] We also provide mentorship, HR support
[02:06:03] and help with investment.
[02:06:06] Uh these are the speakers of the event
[02:06:08] today. Let's uh we thank them again for
[02:06:10] supporting us and uh epic members drive
[02:06:14] the photonic revolution. Thank you very
[02:06:17] much for joining us.