S11-E3_Optical Photonic Packaging

Full Transcript

https://www.youtube.com/watch?v=uDhbVT2D_hg

[00:03] Hello everyone and uh welcome to our webinar series titled introduction to platonics packaging.
[00:10] My name is Mark Brenzing and again I'm joined by my colleague Dr. Camila Gratowski and just a brief recap over the last two uh webinars.
[00:19] So in our first webinar we provided a general introduction to platonics packaging package design and last week we looked at electrical packaging terminal management.
[00:27] So in today's webinar, Camil is going to discuss optical packaging and specifically two of the main optical coupling schemes.
[00:36] Again, this webinar will last about 45 minutes and we've allocated 50 minutes at the end for Q&A.
[00:41] So following the same pattern as the last two webinars, please feel free to use the Q&A tab for any questions that you might have and we'll address them at the end of the talk.
[00:52] So without further ado, I'd hand over to Camil to begin today's webinar.
[00:58] Well, everybody just I want to do a mic check to make sure that uh you can hear us.
[01:00] If you can indicate in the chat or
[01:05] us. If you can indicate in the chat or any other methods that uh we're coming on.
[01:07] Yeah, perfect. Okay. So, this is uh today is essentially the key of the entire photonic packaging.
[01:15] Uh uh essentially all roads lead to optical packaging because this is what differentiates us from the other uh types of packaging.
[01:24] So in electronics packaging you essentially deal with electrical signals and the way to deliver them and also perm management.
[01:32] However, as I mentioned two weeks ago already, optical packaging also has this domain of optics which we have to take care of.
[01:42] So this is what uh separates uh us. Oh sorry too laser po. Perfect.
[01:47] Okay. So optical packaging is the is the key to to what we do and uh essentially the basic principle is very simple.
[01:56] Uh what I'm going to be presenting for most of the time is uh the edge coupler paradigm and then briefly at the end I'll also touch on the grating couple paradigm but uh I'll be using edge
[02:07] paradigm but uh I'll be using edge couplers because they're conceptually a little bit easier to understand.
[02:11] Uh so that's why I'm going to concentrate most of my presentation on this specific one.
[02:16] So um in any case when you're uh packaging a photonic chip what you essentially need to do is to deliver optical signal between a fiber and the waveguide.
[02:28] So the waveguides uh for example on silicon can be very small.
[02:37] Uh their typical size on in silicon as you can see from the image here on the bottom right there's like 220x 400 nanometers.
[02:40] So the mode is very tightly squeezed into that into that wave guide in order to uh uh save on space and also have a lower loss propagation.
[02:51] However, optical fibers are typically much bigger especially when we're working with the uh let's say O and Cbank domain.
[03:00] So between 1300 and 150050 nanometers in those cases the uh the fiber modes are typically much
[03:08] The fiber modes are typically much larger.
[03:10] So essentially photonix packaging is about mediating this optical link and making sure that we can deliver light from the chip and to the chip uh in as lossless method as possible.
[03:21] So there are various methods of actually doing that.
[03:23] However uh before we we get to that we have to just uh provide few definitions of what we're working on.
[03:28] So uh typically when you're working with photonics and photonic integrated circuits uh in most cases I don't know of any cases but there might be where where the signal propagates in a single gausian mode.
[03:44] So uh that's essentially what we'll be narrowing down our scope to uh just a single mode gausian uh that's propagating for the waveguide.
[03:51] And uh as far as definitions that I'll be using the most critical one is the mode field diameter which if you take the the gausian function and uh you let's say put the intensity to unity and then you take as the uh the central
[04:10] then you take as the uh the central position also put to zero and if you
[04:13] position also put to zero and if you instead of x here put the this parameter
[04:16] instead of x here put the this parameter w then what you arrive here that the
[04:18] w then what you arrive here that the intensity at this uh parameter w which
[04:22] intensity at this uh parameter w which is called width is equal to 1 / e^ 2
[04:25] is called width is equal to 1 / e^ 2 which is about 13 1.5% of the intensity
[04:28] which is about 13 1.5% of the intensity at peak.
[04:32] So uh from that we defined our mode for diameter as two widths which is
[04:34] mode for diameter as two widths which is which is over here and uh historically
[04:37] which is over here and uh historically people were mostly using full width of
[04:39] people were mostly using full width of max half maximum which is similar but
[04:42] max half maximum which is similar but exactly it's at you know at half
[04:43] exactly it's at you know at half maximum.
[04:46] Uh and there's a very simple relationship between the two.
[04:47] However, for our purposes, we're concentrating on
[04:49] for our purposes, we're concentrating on mold vidameters and what's what you're
[04:51] mold vidameters and what's what you're going to find in most of the literature.
[04:55] going to find in most of the literature.
[04:57] And as I mentioned previously, uh essentially what we are trying to do is
[04:58] essentially what we are trying to do is to match the modes between the
[05:01] to match the modes between the waveguides.
[05:04] Uh one waveguide is our pick wavegui.
[05:07] The other uh wavegu is our fiber.
[05:09] Now how it's being done the technical tech technicalities of that
[05:12] technical tech technicalities of that those are better left to you know to.
[05:13] those are better left to you know to pick designers and to foundaries because.
[05:17] pick designers and to foundaries because u as I mentioned uh previously uh at.
[05:20] u as I mentioned uh previously uh at photonix packaging we mostly deal with interfaces.
[05:21] So we have fiber on one side and the other side provides a chip with.
[05:24] and the other side provides a chip with a spot what's called a spot size converter.
[05:26] a spot what's called a spot size converter which converts a very uh small.
[05:28] converter which converts a very uh small mode that's propagating in the silicon.
[05:31] mode that's propagating in the silicon waveguide to something that's a little.
[05:32] waveguide to something that's a little bit bigger and approximates uh what uh.
[05:34] bit bigger and approximates uh what uh what the what's the mode size of our fiber.
[05:37] what the what's the mode size of our fiber.
[05:39] Uh however if they're not matched and they are typically not matched then.
[05:42] and they are typically not matched then we start incurring losses.
[05:44] we start incurring losses.
[05:47] So these are essentially the overlap integral is the way that you calculate the uh uh in.
[05:49] way that you calculate the uh uh in general the uh the coupling efficiency.
[05:52] general the uh the coupling efficiency.
[05:55] However, for two gausian beams it can be somewhat simplified.
[05:58] So if uh in the x and y direction your uh your dimensions are different then you use this uh neat.
[06:01] and y direction your uh your dimensions are different then you use this uh neat equation which can be simplified to a.
[06:04] are different then you use this uh neat equation which can be simplified to a much simpler form.
[06:07] equation which can be simplified to a much simpler form.
[06:09] If your mode sizes on
[06:13] much simpler form.
[06:16] If your mode sizes on both sides are well if the mode sides both sides are well if the mode sides are symmetrical.
[06:17] So for example you have are symmetrical.
[06:20] So for example you have a fiber on one side uh which are a fiber on one side uh which are cylindrically symmetrical.
[06:22] So they're cylindrically symmetrical.
[06:23] So they're typically symmetrical in the x and y typically symmetrical in the x and y direction and if by any happy accident
[06:26] direction and if by any happy accident on the opic side you also have a on the opic side you also have a symmetrical mode then you can use this
[06:28] symmetrical mode then you can use this uh neat formula.
[06:31] uh neat formula.
[06:34] And of course if they don't match then
[06:36] And of course if they don't match then our theta which is our ATA sorry which our theta which is our ATA sorry which is our losses are less you know our
[06:39] is our losses are less you know our coupling efficiency is less than unity.
[06:42] coupling efficiency is less than unity.
[06:45] So uh what I'll be using throughout this So uh what I'll be using throughout this talk uh I won't be using natural units.
[06:47] talk uh I won't be using natural units.
[06:49] I'll be mostly using decibels which is a I'll be mostly using decibels which is a logarithmic unit and the definitions are
[06:52] logarithmic unit and the definitions are given here.
[06:54] given here. So if you to talk about So if you to talk about coupling efficiency that's 10 log of A
[06:57] coupling efficiency that's 10 log of A and it's typically less well it is less
[06:59] and it's typically less well it is less than zero.
[07:02] can have coupling efficiency greater than zero by by definition.
[07:03] greater than zero by by definition.
[07:07] Uh in practical terms the coupling Uh in practical terms the coupling efficiency is calculated when you
[07:09] efficiency is calculated when you perform experiment as essentially a uh
[07:14] perform experiment as essentially a uh 10 logs of uh the power that's transmitted through to the reference.
[07:19] transmitted through to the reference power that we know that's injected into fiber and if that power is equal to 1 millatt then their units become dbm.
[07:27] millatt then their units become dbm.
[07:32] useful values to uh to to know about is that zero dB coupling efficiency is equal 100%.
[07:38] uh coupling efficiency is equal 100% 1 dB which is going to be used mostly throughout this talk is about 80% it's actually around 70 79.4% 4% and 3dB is very close to 50%.
[07:51] very close to 50%. And 3dB is if you're working with micro electronics, then that's uh that's the standard uh standard notion.
[07:56] standard notion. Um so as far as fibers, I'm just not going to explain all of them.
[08:02] However, uh typically in our work, we deal with a single mode fibers because that uh as the uh photonic chip is single mode, then we also want to couple the mode uh
[08:14] then we also want to couple the mode uh into that pick using single mode fibers.
[08:17] into that pick using single mode fibers.
[08:21] Uh so for example at uh 1550 uh the single mode fibers operating at C band have a core diameter of 8.2 microns.
[08:27] have a core diameter of 8.2 microns.
[08:30] uh they're they're actually very very easy to distinguish from the multim mode fibers which course starts around 50 microns and can get pretty pretty large after that.
[08:31] easy to distinguish from the multim mode fibers which course starts around 50 microns and can get pretty pretty large after that.
[08:33] fibers which course starts around 50 microns and can get pretty pretty large after that.
[08:36] microns and can get pretty pretty large after that.
[08:38] after that. For other applications, for example, uh you can use ultra high numerical aperture fibers which have a smaller core size um and a little bit of different composition which means that the mode size a little bit different from the ones that you would have from SMF28 and uh in many applications which require uh polarization which are position sensitive you also will encounter polarization maintaining fibers.
[08:41] example, uh you can use ultra high numerical aperture fibers which have a smaller core size um and a little bit of different composition which means that the mode size a little bit different from the ones that you would have from SMF28 and uh in many applications which require uh polarization which are position sensitive you also will encounter polarization maintaining fibers.
[08:42] numerical aperture fibers which have a smaller core size um and a little bit of different composition which means that the mode size a little bit different from the ones that you would have from SMF28 and uh in many applications which require uh polarization which are position sensitive you also will encounter polarization maintaining fibers.
[08:45] smaller core size um and a little bit of different composition which means that the mode size a little bit different from the ones that you would have from SMF28 and uh in many applications which require uh polarization which are position sensitive you also will encounter polarization maintaining fibers.
[08:47] um and a little bit of different composition which means that the mode size a little bit different from the ones that you would have from SMF28 and uh in many applications which require uh polarization which are position sensitive you also will encounter polarization maintaining fibers.
[08:48] composition which means that the mode size a little bit different from the ones that you would have from SMF28 and uh in many applications which require uh polarization which are position sensitive you also will encounter polarization maintaining fibers.
[08:50] size a little bit different from the ones that you would have from SMF28 and uh in many applications which require uh polarization which are position sensitive you also will encounter polarization maintaining fibers.
[08:52] ones that you would have from SMF28 and uh in many applications which require uh polarization which are position sensitive you also will encounter polarization maintaining fibers.
[08:55] and uh in many applications which require uh polarization which are position sensitive you also will encounter polarization maintaining fibers.
[08:57] require uh polarization which are position sensitive you also will encounter polarization maintaining fibers.
[08:59] position sensitive you also will encounter polarization maintaining fibers.
[09:01] encounter polarization maintaining fibers.
[09:03] fibers. This is just one of the designs.
[09:05] This the the simplest and the most common.
[09:07] It's called a panda because of the two ears on the sides of the core which keep the polarization on the same level.
[09:09] the two ears on the sides of the core which keep the polarization on the same level.
[09:11] which keep the polarization on the same level.
[09:13] level. So if you want to match the mode between
[09:16] So if you want to match the mode between your let's say system uh which is your your let's say system uh which is your fiber or the something that's external fiber or the something that's external to your package and your photonic chip to your package and your photonic chip then you need to match those modes.
[09:25] So for example at 1550 the mode for for example at 1550 the mode for diameter in your uh single mode fiber is diameter in your uh single mode fiber is around 10.4 four microns.
[09:32] But if you have a pick, which is not uncommon to have picss with motor diameters of three three and a half microns, then if you take the uh the equations, then you find out that you have about 5 dB losses per coupling, which might be a little bit too much for for you.
[09:49] So there are many ways to match the modes.
[09:52] Uh historically the earliest and the uh some still sometimes used is to use a lens fiber which essentially has as the name suggests a lens at the tip which uh focuses the beam down to the small let's say between three and four microns in order to match the mode.
[10:10] However, this is uh uh this is uh mostly used for uh for applications which which have just
[10:16] for applications which which have just have single channels and you actually have single channels and you actually necessarily have to do that.
[10:21] Then what you can do you can do an adabetic taper splice.
[10:23] Uh it's relatively new technique where you take a single mode fiber of the standard variety and you take a high numerical aperture fiber which has different mode size and then you do a adabatic taper splice uh in between and the losses that uh we we get from this uh method are between 0.1 to 0.2 on the slice the splice itself.
[10:48] Then you can get what's called waveguide array to fiber interposers.
[10:53] Uh these are actually the technology developed by t photonics.
[10:57] Uh where you can actually use silver ion exchange in order to modify locally the refractive index of glass and uh essentially create waveguides and write them in the lithographic process.
[11:12] So uh because you can do that you can actually take um one mode on one mode
[11:17] actually take um one mode on one mode size on one side and taper it down to size on one side and taper it down to have a different mode size on the other side.
[11:22] have a different mode size on the other side. You can do that little graphically in a similar manner.
[11:25] You can do that little graphically in a similar manner. You can do the same with photonic integrated circuits which have uh which are made for this for this kind of uh tapering.
[11:30] with photonic integrated circuits which have uh which are made for this for this kind of uh tapering.
[11:35] kind of uh tapering. You can have uh a chip which has a output copper that uh has a mode at the edge which is varying with distance.
[11:39] You can have uh a chip which has a output copper that uh has a mode at the edge which is varying with distance.
[11:42] has a mode at the edge which is varying with distance. You can see here the veneer which tells you you know at the position at which the mode is approximately which side.
[11:45] with distance. You can see here the veneer which tells you you know at the position at which the mode is approximately which side.
[11:47] You can see here the veneer which tells you you know at the position at which the mode is approximately which side.
[11:50] position at which the mode is approximately which side. So here at this distance it's three microns four microns and so on and you polish it down until you arrive at the value that's uh that's of interest.
[11:52] approximately which side. So here at this distance it's three microns four microns and so on and you polish it down until you arrive at the value that's uh that's of interest.
[11:54] So here at this distance it's three microns four microns and so on and you polish it down until you arrive at the value that's uh that's of interest.
[11:56] this distance it's three microns four microns and so on and you polish it down until you arrive at the value that's uh that's of interest.
[11:58] microns and so on and you polish it down until you arrive at the value that's uh that's of interest.
[12:01] until you arrive at the value that's uh that's of interest. That's this technology was developed by fix.
[12:03] that's of interest. That's this technology was developed by fix.
[12:06] technology was developed by fix. So uh unfortunately if that was uh the end of it we would this wouldn't be so difficult.
[12:09] So uh unfortunately if that was uh the end of it we would this wouldn't be so difficult.
[12:12] end of it we would this wouldn't be so difficult. Um however unfortunately in many applications you also have to worry about relative position of the modes uh
[12:15] difficult. Um however unfortunately in many applications you also have to worry about relative position of the modes uh
[12:17] many applications you also have to worry about relative position of the modes uh
[12:20] about relative position of the modes uh between the fiber and the chip.
[12:22] As between the fiber and the chip.
[12:24] As you'll see those tolerances as we call them are quite small and the theory
[12:26] them are quite small and the theory behind is quite simple.
[12:28] as one of the mode that's displaced uh you know
[12:30] whether in vertical or horizontal
[12:32] direction then uh the modes shift
[12:35] relative to one another and you start
[12:37] incurring losses in which case your
[12:39] coupling efficiency becomes less than
[12:41] one
[12:44] uh if you perform a very simple zax
[12:46] simulations where you take uh where you
[12:48] assume that uh your mode fuel diameter
[12:51] on both sides are identical so let's say
[12:53] your chip is 10 microns mode fuel diameter and your fibers micron motor
[12:56] diameter and your fibers micron motor diameter as well and you plot it as a
[12:59] function of that mode diameter then you
[13:02] arrive at a very interesting solution
[13:04] that the slope of this line is ex equal
[13:06] 0.24.
[13:10] So in this case it means that for
[13:11] symmetrical couplers for identical
[13:14] symmetrical couplers you can expect your
[13:17] alignment tolerance to be around quarter
[13:21] alignment tolerance to be around quarter of the mold field diameter.
[13:22] This is actually a useful thing to to keep in mind.
[13:28] Uh of course uh this is a simulation that doesn't take into consideration the wave dependence but
[13:31] the mode for diameter within your fiber or the pick will be well will the size of the mode will be with dependent.
[13:39] So u uh in what you would want in order to satisfy this the fiber uh and the chip follow the same slope but uh of course that rarely happens.
[13:51] In any case this uh value of 0.24 24.
[13:54] It can be actually uh demonstrated from first principles and uh this is actually a graph from a paper that was just accepted to applied optics and I hope it will be published soon where you can actually demonstrate that your alignment tolerance at 1 dB is approximately nearly identical to 0.24.
[14:17] This parameter here square root of minus log natural logarithm of ATA at some specific uh uh
[14:25] specific uh uh uh threshold.
[14:27] uh threshold uh is is what defines you your alignment.
[14:30] uh is is what defines you your alignment tolerances.
[14:32] For example, here you can see on this graph that uh at 1 dB coupling efficiency you get your uh uh.
[14:35] coupling efficiency you get your uh uh you.
[14:40] you uh your benchmark distance your.
[14:41] uh your benchmark distance your displacement distance is 0.24.
[14:43] However, if you select your benchmark at 3dB,
[14:47] then this value then becomes 0.415.
[14:50] Uh so you'll get 3dB losses when you are 41% off from the size of your mode.
[14:54] So if your mode size is for example 10 microns then your 1 dB losses will be at 2.4 microns uh but your 3dB losses will be at 4.1 microns.
[14:58] Conversely, as you go to much uh smaller uh tolerant uh thresholds, for example, 0.1 dB, your uh you're now entering into very very tight.
[15:01] if your mode size is for example 10.
[15:03] microns then your 1 dB losses will be at.
[15:07] 2.4 microns uh but your 3dB losses will.
[15:11] be at 4.1 microns.
[15:14] Conversely, as you go.
[15:18] to much uh smaller uh tolerant uh.
[15:23] thresholds, for example, 0.1 dB, your uh.
[15:25] you're now entering into very very tight uh tolerance regime where for example,
[15:27] uh tolerance regime where for example, if your mode size is 10 microns, your
[15:30] if your mode size is 10 microns, your tolerance for to achieve that goal is
[15:33] tolerance for to achieve that goal is 0.7 microns.
[15:37] However, with this graph essentially you can choose any tolerance
[15:39] cutff point and the scaling law of this type will still apply.
[15:41] You can actually demonstrate that uh of course reality is
[15:44] not uh is such that we very rarely have identical couplers uh and symmetrical
[15:47] clappers on both sides.
[15:50] So what we mostly deal with is some sort of a symmetry.
[15:53] uh in fibers uh they typically of course have uh because of their inter
[15:55] inherent symmetry they have uh the mode size is symmetrical.
[15:58] However, picss are not.
[16:01] So this is quite a very simple simulation that kind of uh demonstrates the the principle in general where we
[16:03] have uh where we take our mode size in the fiber to be equal to 10.4 and uh we
[16:06] assume that mode fometer in the pick is
[16:26] assume that mode fometer in the pick is as well equal 10.4 four and what we get.
[16:29] as well equal 10.4 four and what we get there is that regardless of the mode.
[16:31] there is that regardless of the mode size of the pick in the other direction.
[16:34] size of the pick in the other direction the tolerance is constant is completely.
[16:38] the tolerance is constant is completely independent on the on the mode size in.
[16:40] independent on the on the mode size in the vertical direction uh the black line.
[16:43] the vertical direction uh the black line is the limiting value that's the slope.
[16:45] is the limiting value that's the slope of 0.24 24. But what you have is that in.
[16:49] of 0.24 24. But what you have is that in the other direction the vertical.
[16:50] the other direction the vertical direction the uh the coupling efficient.
[16:53] direction the uh the coupling efficient doesn't follow either this or this line.
[16:55] doesn't follow either this or this line. However, these lines are let's say.
[16:57] However, these lines are let's say independent and again the paper that I.
[17:00] independent and again the paper that I mentioned there's actually derivation of.
[17:02] mentioned there's actually derivation of the formula that shows this as well.
[17:05] the formula that shows this as well to give you a little bit of a.
[17:07] to give you a little bit of a perspective of how difficult it is to.
[17:09] perspective of how difficult it is to align the fiber to the waveguide. um.
[17:12] align the fiber to the waveguide. um what you saw here the values that you.
[17:14] what you saw here the values that you see here are of the order of 2.5 microns.
[17:17] see here are of the order of 2.5 microns for 1 dB losses and if you compare that.
[17:20] for 1 dB losses and if you compare that to for example a human red blood dot.
[17:22] to for example a human red blood dot cell you can see that the thickness of.
[17:23] cell you can see that the thickness of the blood dot cell is around 2 microns.
[17:26] the blood dot cell is around 2 microns so this is the uh tolerance this is our.
[17:30] so this is the uh tolerance this is our accuracy that we have to achieve in.
[17:32] accuracy that we have to achieve in order to uh align the fiber to the chip.
[17:36] order to uh align the fiber to the chip and be still within more or less 1db uh.
[17:38] and be still within more or less 1db uh coupling efficiency.
[17:42] coupling efficiency So that's for the displacement.
[17:44] Uh there are also other uh other axes that we have to typically deal with.
[17:48] other axes that we have to typically deal with.
[17:50] One of them is rotation.
[17:53] So if we perform a similar analysis where if we for starters assume that all mode field diameters are exactly the same.
[17:55] if we for starters assume that all mode field diameters are exactly the same.
[17:59] field diameters are exactly the same across the entire system and we plot our 1dB element tolerance uh against the mode field.
[18:00] across the entire system and we plot our 1dB element tolerance uh against the mode field.
[18:03] and they select the modeful diameter then we get this curve which doesn't seem to look like anything uh until you actually plot it in the reciprocal space.
[18:06] 1dB element tolerance uh against the mode field and they select the modeful diameter then we get this curve which doesn't seem to look like anything uh until you actually plot it in the.
[18:08] mode field and they select the modeful diameter then we get this curve which doesn't seem to look like anything uh until you actually plot it in the reciprocal space.
[18:11] diameter then we get this curve which doesn't seem to look like anything uh until you actually plot it in the reciprocal space.
[18:14] doesn't seem to look like anything uh until you actually plot it in the reciprocal space in which you get a straight line.
[18:16] until you actually plot it in the reciprocal space in which you get a straight line and you can actually show that this is uh this line is because it's tied to the the divergence angle of the gausian beam which is uh defined.
[18:18] reciprocal space in which you get a straight line and you can actually show that this is uh this line is because it's tied to the the divergence angle of the gausian beam which is uh defined.
[18:21] straight line and you can actually show that this is uh this line is because it's tied to the the divergence angle of the gausian beam which is uh defined.
[18:25] that this is uh this line is because it's tied to the the divergence angle of the gausian beam which is uh defined.
[18:27] it's tied to the the divergence angle of the gausian beam which is uh defined.
[18:30] the gausian beam which is uh defined using this formula and uh again this is using this formula and uh again this is derivation of this is in the paper uh
[18:37] derivation of this is in the paper uh and if you also add another magisterium of the wavelength you also see similar
[18:42] uh similar performance that as your wavelength increases so does your alignment tolerance increases however
[18:49] alignment tolerance increases however what's what's important to understand from to remember from these is that as your mode field diameter uh increases
[18:59] your mode field diameter uh increases your angular alignment tolerance decreases.
[19:04] That's uh that's unfortunately that's the thing what what happens.
[19:07] So as you increase your multidameter you get better at aligning in the x and y directions for example but you lose your angular tolerance.
[19:16] However if you look at these numbers so for example for 10.4 four micron mode diameter.
[19:21] Uh our alignment tolerance in the angle is around three is just shy of 3° and 3° is uh is very large for photonic packaging machines.
[19:29] So we
[19:31] photonic packaging machines.
[19:31] So we rarely have to worry about it at all.
[19:34] rarely have to worry about it at all.
[19:34] And uh just coming back a little bit uh
[19:37] And uh just coming back a little bit uh again you can show that uh uh the
[19:40] again you can show that uh uh the element torance of angle is essentially
[19:42] element torance of angle is essentially tied to your uh to the divergence angle
[19:45] tied to your uh to the divergence angle and which also includes the wavelength
[19:47] and which also includes the wavelength which is why that we have this
[19:49] which is why that we have this wavelength dependence.
[19:51] wavelength dependence.
[19:51] Of course this is kind of idealistic simulation because we
[19:53] kind of idealistic simulation because we assume that the modem doesn't change as
[19:56] assume that the modem doesn't change as a function of wavelength and it does.
[19:59] a function of wavelength and it does.
[19:59] So this is not say idealized case but it
[20:01] this is not say idealized case but it gives you some sort of a uh
[20:03] gives you some sort of a uh understanding.
[20:05] understanding.
[20:05] Of course if you start inputting all your realistic numbers
[20:07] inputting all your realistic numbers then these uh straight lines become a
[20:10] then these uh straight lines become a little bit more curved.
[20:12] little bit more curved.
[20:12] Uh so that's as far as uh let's say theory goes.
[20:15] Uh so that's as far as uh let's say theory goes.
[20:15] Uh however uh when you go to the the experimental data due to most
[20:18] theory goes.
[20:18] Uh however uh when you go to the the experimental data due to most likely to mode mismatch and some other
[20:21] to the the experimental data due to most likely to mode mismatch and some other effects what you see is uh uh what you
[20:23] likely to mode mismatch and some other effects what you see is uh uh what you can expect ballpark uh is that your
[20:27] effects what you see is uh uh what you can expect ballpark uh is that your alignment tolerance will end up to being
[20:30] can expect ballpark uh is that your alignment tolerance will end up to being
[20:32] alignment tolerance will end up to being about a third of your multiple diameter
[20:35] about a third of your multiple diameter that you have in the system. So if you
[20:37] that you have in the system. So if you have everything perfectly you know
[20:39] have everything perfectly you know matched then you'll get 1/4 as uh
[20:42] matched then you'll get 1/4 as uh predicted by theory but in practice it
[20:45] predicted by theory but in practice it comes closer to 1/3
[20:49] uh and that's only if you are dealing
[20:52] uh and that's only if you are dealing with a single fiber which is uh not so
[20:56] with a single fiber which is uh not so difficult to align uh and perform
[20:58] difficult to align uh and perform optical packaging but uh in modern
[21:01] optical packaging but uh in modern photonics we also uh rarely have such a
[21:04] photonics we also uh rarely have such a comfortable situation and typically what
[21:06] comfortable situation and typically what we have are multiple channels. Uh the
[21:09] we have are multiple channels. Uh the channels right now typically are between
[21:12] channels right now typically are between two and let's say 16 and uh there are
[21:16] two and let's say 16 and uh there are some works in the pipeline which are
[21:20] some works in the pipeline which are which are tend to increase this number
[21:21] which are tend to increase this number all the way to more than
[21:25] all the way to more than 62. I guess I think the biggest fiber
[21:28] 62. I guess I think the biggest fiber rate that we personally dealt with was
[21:30] rate that we personally dealt with was 148 fibers. But uh as photonic uh pack
[21:34] 148 fibers. But uh as photonic uh pack as photonics progresses we expect these
[21:36] as photonics progresses we expect these numbers of optical charge going to be
[21:38] numbers of optical charge going to be significant
[21:40] significant uh significantly larger than what they
[21:41] uh significantly larger than what they are currently and right now there are uh
[21:45] are currently and right now there are uh in fiber arrays there are two pitch
[21:47] in fiber arrays there are two pitch standards one is 127 microns that's
[21:51] standards one is 127 microns that's because the size of the cladding of your
[21:54] because the size of the cladding of your standard single mode fiber is 125
[21:56] standard single mode fiber is 125 microns which means if you leave one
[21:59] microns which means if you leave one micron each side from each fiber That's
[22:01] micron each side from each fiber That's why you get the 127 and the other pitch
[22:04] why you get the 127 and the other pitch standard is 250 microns. Uh and of
[22:07] standard is 250 microns. Uh and of course you can derivatives of those. So
[22:09] course you can derivatives of those. So you can get 500 micron fiber ray, 750
[22:12] you can get 500 micron fiber ray, 750 micron fiber array and and such. Uh
[22:15] micron fiber array and and such. Uh other pitch standards are also available
[22:17] other pitch standards are also available on demand but it's currently it's a
[22:19] on demand but it's currently it's a little bit more difficult to get
[22:20] little bit more difficult to get something which is pitched at less at
[22:22] something which is pitched at less at 127. Although there are some
[22:24] 127. Although there are some experimental fiber arrays that you can
[22:26] experimental fiber arrays that you can get that use fibers which have a
[22:28] get that use fibers which have a cladding of diameter of just 80 microns.
[22:32] cladding of diameter of just 80 microns. So uh but I don't know if we'll be able
[22:35] So uh but I don't know if we'll be able to get much more down with that
[22:37] to get much more down with that diameter. It's it's possible but the
[22:39] diameter. It's it's possible but the reliability might suffer because there
[22:41] reliability might suffer because there also manufacturing tolerances of the
[22:44] also manufacturing tolerances of the fibers themselves and also of the fiber
[22:47] fibers themselves and also of the fiber uh of the you know of the array as it is
[22:50] uh of the you know of the array as it is constructed. So for example, core
[22:52] constructed. So for example, core concentricity or how well the core is
[22:54] concentricity or how well the core is aligned to the outside cladding is
[22:56] aligned to the outside cladding is better than half a micron. That's the
[22:58] better than half a micron. That's the typical uh quotation you'll find. Then
[23:02] typical uh quotation you'll find. Then there's the pitch tolerance between you
[23:05] there's the pitch tolerance between you know between different or adjacent uh uh
[23:09] know between different or adjacent uh uh channels. So depending on uh the number
[23:12] channels. So depending on uh the number of channels this can be from plus or
[23:14] of channels this can be from plus or minus.7 microns all the way to 1.5
[23:17] minus.7 microns all the way to 1.5 microns if your fiber is sufficiently
[23:20] microns if your fiber is sufficiently large. And if you remember that we have
[23:22] large. And if you remember that we have around 2.4 microns for coupling to a
[23:26] around 2.4 microns for coupling to a chip which has a at 1550. This tells you
[23:30] chip which has a at 1550. This tells you that you will incur some losses but
[23:32] that you will incur some losses but there might not be egregious. And of
[23:35] there might not be egregious. And of course uh final tolerance is the
[23:37] course uh final tolerance is the polishing angle tolerance at the angle
[23:40] polishing angle tolerance at the angle at which you polish the fibers. that
[23:42] at which you polish the fibers. that tolerance of that can be between 3 and
[23:45] tolerance of that can be between 3 and 0.5°
[23:48] and uh and the more you there's there's
[23:52] and uh and the more you there's there's an issue in photonix packaging that's
[23:54] an issue in photonix packaging that's been recognized for several years now
[23:56] been recognized for several years now and that's that of short density the
[23:58] and that's that of short density the more you um
[24:01] more you um the more channels you have the more
[24:03] the more channels you have the more let's say the length of your chip you
[24:06] let's say the length of your chip you need to have for example if you are
[24:08] need to have for example if you are pitched at 250 microns and you have uh
[24:11] pitched at 250 microns and you have uh let's say let's take a small number 20
[24:14] let's say let's take a small number 20 uh optical channels then your base and
[24:17] uh optical channels then your base and the length of your chip has to be 5 mm
[24:20] the length of your chip has to be 5 mm at the very least in order to to to
[24:22] at the very least in order to to to support all that. If you have let's say
[24:25] support all that. If you have let's say you're going towards 120 microns then
[24:27] you're going towards 120 microns then you need 30 mm and uh in photonics you
[24:30] you need 30 mm and uh in photonics you pay very dearly for real estate. So you
[24:32] pay very dearly for real estate. So you have to be very conscious of you know
[24:34] have to be very conscious of you know how big of a chip do you want to have in
[24:37] how big of a chip do you want to have in order to support that many of channels.
[24:39] order to support that many of channels. And of course you can reduce the pitch
[24:41] And of course you can reduce the pitch uh to 127 essentially cutting this in
[24:44] uh to 127 essentially cutting this in half. Uh but you can also keep reducing
[24:46] half. Uh but you can also keep reducing the pitch using some other uh methods.
[24:49] the pitch using some other uh methods. Uh uh the interpose that I mentioned
[24:52] Uh uh the interpose that I mentioned previously it's for example the chip
[24:54] previously it's for example the chip interposer but also the glass uh
[24:57] interposer but also the glass uh interposer. They can be used not only to
[24:59] interposer. They can be used not only to reduce or modify your the size of your
[25:02] reduce or modify your the size of your mode but they can also be used to reduce
[25:05] mode but they can also be used to reduce the pitch. So we did a project some time
[25:08] the pitch. So we did a project some time ago about se six or seven years ago
[25:11] ago about se six or seven years ago where we used such an interoser which
[25:13] where we used such an interoser which had which was going from 127 to 63 and
[25:16] had which was going from 127 to 63 and 1/2 micron pitch so that we could
[25:19] 1/2 micron pitch so that we could squeeze a lot of channels into a very
[25:22] squeeze a lot of channels into a very small space. So for 120 uh uh optical
[25:26] small space. So for 120 uh uh optical channels you're already down to let's
[25:28] channels you're already down to let's say about 7 mm which is more manageable.
[25:33] say about 7 mm which is more manageable. Uh so this is uh how such an optical
[25:36] Uh so this is uh how such an optical interposer works. Uh this case this is
[25:38] interposer works. Uh this case this is the waft interposer where you have your
[25:40] the waft interposer where you have your optical channels on one side pitched at
[25:42] optical channels on one side pitched at standard fiber uh standard fiber array
[25:45] standard fiber uh standard fiber array scale and then it's being pitched down
[25:48] scale and then it's being pitched down to match not only the the the mode size
[25:51] to match not only the the the mode size but also the pitch to match that on your
[25:53] but also the pitch to match that on your pick. So you can do many many uh
[25:56] pick. So you can do many many uh interesting things with your interosers.
[25:58] interesting things with your interosers. Last year, last week I spoke about uh
[26:01] Last year, last week I spoke about uh electrical condos that can also be used
[26:03] electrical condos that can also be used to reduce the pitch of the op electro
[26:06] to reduce the pitch of the op electro connections and this is the optical
[26:08] connections and this is the optical equivalent of that. The next topic that
[26:11] equivalent of that. The next topic that has to be uh highlighted is when you're
[26:14] has to be uh highlighted is when you're talking about edge coppers is how do you
[26:16] talking about edge coppers is how do you prepare the facets
[26:18] prepare the facets because if you don't prepare the facets
[26:20] because if you don't prepare the facets properly or or well then you'll start
[26:23] properly or or well then you'll start incurring losses due to scattering. So
[26:26] incurring losses due to scattering. So the simplest way is just to use a dicing
[26:28] the simplest way is just to use a dicing saw or a laser to cut through the wafer.
[26:30] saw or a laser to cut through the wafer. It's uh fast, scalable uh because you
[26:33] It's uh fast, scalable uh because you can you can just program your dicing saw
[26:35] can you can just program your dicing saw on your wafer and just
[26:38] on your wafer and just having run wild. However, as you can see
[26:41] having run wild. However, as you can see from the image here, uh there's a lot of
[26:44] from the image here, uh there's a lot of uh roughness to the edge uh which lead
[26:48] uh roughness to the edge uh which lead uh which leads to high scattering and
[26:51] uh which leads to high scattering and also especially with laser you can have
[26:53] also especially with laser you can have a lot of debris on surface as well.
[26:55] a lot of debris on surface as well. uh then you can also use
[26:57] uh then you can also use crystalallographic planes to break the
[26:59] crystalallographic planes to break the wafer which is quite fast. However, if
[27:02] wafer which is quite fast. However, if you have a very large wafer, it's
[27:03] you have a very large wafer, it's actually quite difficult to do uh to
[27:07] actually quite difficult to do uh to have a neat cleave on the entire wafer.
[27:09] have a neat cleave on the entire wafer. So, I believe it's best suited just for
[27:12] So, I believe it's best suited just for uh for laser chips and not so much for
[27:15] uh for laser chips and not so much for photonic integrated circuits. Uh then if
[27:19] photonic integrated circuits. Uh then if for example you dice your wafer you can
[27:21] for example you dice your wafer you can then polish it afterwards which leaves
[27:23] then polish it afterwards which leaves you with very smooth facets. However
[27:25] you with very smooth facets. However it's a slow process because essentially
[27:27] it's a slow process because essentially you have to take your chip one by one
[27:29] you have to take your chip one by one putting in a polishing machine and then
[27:31] putting in a polishing machine and then grind it down using different grits
[27:33] grind it down using different grits until you get the quality of the facet
[27:35] until you get the quality of the facet that you need. So it's quite difficult
[27:37] that you need. So it's quite difficult to execute on a large wafer and large
[27:39] to execute on a large wafer and large scale. And it's quite difficult to see
[27:42] scale. And it's quite difficult to see on this image but this is the image I
[27:44] on this image but this is the image I took several years ago. And if you
[27:47] took several years ago. And if you compare this line which is the reference
[27:49] compare this line which is the reference line uh on the chip that should be
[27:52] line uh on the chip that should be parall to the surface you can see that
[27:54] parall to the surface you can see that the polishing is actually about in this
[27:56] the polishing is actually about in this case one degree off. It's actually very
[27:58] case one degree off. It's actually very easy to polish at uh at angles which are
[28:02] easy to polish at uh at angles which are not perpendicular to your copper. So
[28:04] not perpendicular to your copper. So there's there's a little bit of a
[28:06] there's there's a little bit of a problem when you when you're trying to
[28:08] problem when you when you're trying to polish. Then uh recently the most common
[28:12] polish. Then uh recently the most common method of uh getting facets which is
[28:14] method of uh getting facets which is actually with this you can get
[28:17] actually with this you can get exceptionally good facets is to perform
[28:20] exceptionally good facets is to perform a chemical wet etch of the face. Uh so
[28:23] a chemical wet etch of the face. Uh so what you have is your wafer. You take
[28:26] what you have is your wafer. You take your wafer and then you just uh etch it
[28:29] your wafer and then you just uh etch it down just where your uh edge coppers
[28:32] down just where your uh edge coppers are. Of course you have to pattern it uh
[28:34] are. Of course you have to pattern it uh in such a way that you etch in the
[28:36] in such a way that you etch in the correct place. And you actually etch it
[28:38] correct place. And you actually etch it using two steps. You can see here
[28:40] using two steps. You can see here perhaps from this image that there is a
[28:42] perhaps from this image that there is a very very nice surface which extends
[28:46] very very nice surface which extends perhaps 20 microns down and then the
[28:48] perhaps 20 microns down and then the rest of the uh of the face is a little
[28:51] rest of the uh of the face is a little bit more rugged. That's because the two
[28:53] bit more rugged. That's because the two different etching solutions are used to
[28:55] different etching solutions are used to first to create a very nice and smooth
[28:57] first to create a very nice and smooth facet and then the rest of it to uh to
[29:00] facet and then the rest of it to uh to just etch it down to a specific depth.
[29:04] just etch it down to a specific depth. Uh and afterwards what you do, you
[29:07] Uh and afterwards what you do, you essentially have these trenches and you
[29:09] essentially have these trenches and you use your dicing saw to just dice right
[29:11] use your dicing saw to just dice right in between to have these uh facets
[29:14] in between to have these uh facets exposed. However, uh you can see from
[29:17] exposed. However, uh you can see from this SCM image that what this creates is
[29:19] this SCM image that what this creates is this L-shaped uh cavity which is not
[29:22] this L-shaped uh cavity which is not very good for packaging because uh what
[29:26] very good for packaging because uh what you need to do in that case is you
[29:29] you need to do in that case is you cannot have a standard fiber ra with a
[29:31] cannot have a standard fiber ra with a lid which is uh flushed with your
[29:33] lid which is uh flushed with your fibers. You actually have to get a
[29:35] fibers. You actually have to get a special fiber made which has a lid
[29:37] special fiber made which has a lid retracted by half a millimeter to a
[29:39] retracted by half a millimeter to a millimeter and with that you have enough
[29:41] millimeter and with that you have enough space to actually approach with the
[29:43] space to actually approach with the fibers to your pick. So this is what you
[29:45] fibers to your pick. So this is what you have. Uh if you look from the side of
[29:47] have. Uh if you look from the side of such a pick, you have your pick
[29:49] such a pick, you have your pick substrate. This is the L-shaped ledge
[29:52] substrate. This is the L-shaped ledge that was created using etching process.
[29:53] that was created using etching process. You can see the two steps here. This is
[29:55] You can see the two steps here. This is the fast step. This is the slow step.
[29:57] the fast step. This is the slow step. And uh this has to be etched down by
[29:59] And uh this has to be etched down by more than at least more than half of the
[30:01] more than at least more than half of the diameter of the fiber which is 125
[30:04] diameter of the fiber which is 125 micron. So you have to etch down at
[30:06] micron. So you have to etch down at least 65 microns if not more.
[30:09] least 65 microns if not more. Uh another way of dealing with that uh
[30:12] Uh another way of dealing with that uh edge issue is undercutting. What you can
[30:15] edge issue is undercutting. What you can do is uh if you're quite good at it uh
[30:18] do is uh if you're quite good at it uh and we have here specialists that that
[30:20] and we have here specialists that that can do this uh uh with exceptional
[30:23] can do this uh uh with exceptional accuracy is you can take your dicing saw
[30:26] accuracy is you can take your dicing saw and you can dice
[30:29] and you can dice up until this moment. You just position
[30:31] up until this moment. You just position your plate so that the uh it stops over
[30:35] your plate so that the uh it stops over here and you can see these these grooves
[30:39] here and you can see these these grooves are made using the dicing. So coming
[30:42] are made using the dicing. So coming from the bottom of the chip and this
[30:44] from the bottom of the chip and this essentially exposes your facet. So so
[30:46] essentially exposes your facet. So so you can approach with a standard fiber.
[30:49] you can approach with a standard fiber. This is a tricky process. Uh that's uh I
[30:52] This is a tricky process. Uh that's uh I wouldn't perhaps recommend a large scale
[30:53] wouldn't perhaps recommend a large scale but you can deal with uh uh if you need
[30:56] but you can deal with uh uh if you need to have standard fiber rays of auto
[30:59] to have standard fiber rays of auto tractor that this is another way of of
[31:01] tractor that this is another way of of dealing with that. Uh another dicing
[31:04] dealing with that. Uh another dicing method which is perhaps uh for in my
[31:07] method which is perhaps uh for in my experience gives the best facets is
[31:10] experience gives the best facets is stealth dicing where you take you focus
[31:13] stealth dicing where you take you focus a laser not on the top surface where it
[31:15] a laser not on the top surface where it would leave debris and jagged lines
[31:19] would leave debris and jagged lines but you actually focus it uh on the
[31:23] but you actually focus it uh on the sacrificial layer some somewhere within
[31:25] sacrificial layer some somewhere within the body of the silicon wafer and then
[31:28] the body of the silicon wafer and then you just you know you just scratch it
[31:31] you just you know you just scratch it essentially inside And then you can uh
[31:33] essentially inside And then you can uh separate it and break it off. And this
[31:35] separate it and break it off. And this method uh is actually quite good in
[31:38] method uh is actually quite good in creating very smooth as you can see very
[31:41] creating very smooth as you can see very smooth fastens very smooth surfaces. And
[31:43] smooth fastens very smooth surfaces. And when you compare using the same chip uh
[31:46] when you compare using the same chip uh design uh the difference between a saw
[31:49] design uh the difference between a saw dicing uh and sticing is about in this
[31:52] dicing uh and sticing is about in this case 2db. So with saw dicing we get 3dB
[31:55] case 2db. So with saw dicing we get 3dB losses per facet at 1550 and with slow
[31:58] losses per facet at 1550 and with slow dicing you get uh around one and u if
[32:02] dicing you get uh around one and u if you additionally fill that gap with uh
[32:05] you additionally fill that gap with uh index matching liquid then your losses
[32:08] index matching liquid then your losses drop to half a dB.
[32:13] Uh before I finish with uh edge facets
[32:16] Uh before I finish with uh edge facets uh I need to talk about the special case
[32:19] uh I need to talk about the special case which is indium facet indium phosphide
[32:21] which is indium facet indium phosphide chips which uh uh since they have
[32:24] chips which uh uh since they have monolithic integrated lasers and
[32:26] monolithic integrated lasers and detectors typically what you will find
[32:28] detectors typically what you will find with them is that at the edge the uh
[32:32] with them is that at the edge the uh couplers are angled at this typically
[32:35] couplers are angled at this typically it's 7°. This is to stop reflection
[32:38] it's 7°. This is to stop reflection feeding back into laser because these
[32:40] feeding back into laser because these lasers are modulated as you can see
[32:41] lasers are modulated as you can see here. If you remember how to recognize
[32:44] here. If you remember how to recognize the RF traces, these are modulated at
[32:45] the RF traces, these are modulated at high frequencies. So in order to stop
[32:48] high frequencies. So in order to stop reflections from interfering with the
[32:49] reflections from interfering with the operational laser, these uh uh
[32:52] operational laser, these uh uh waveguides are angled at uh 7°. However,
[32:56] waveguides are angled at uh 7°. However, this presents a problem for packaging uh
[32:59] this presents a problem for packaging uh at least twofold. One is uh you either
[33:02] at least twofold. One is uh you either have to polish your uh your fiber array
[33:05] have to polish your uh your fiber array at a specific angle. But not only that,
[33:07] at a specific angle. But not only that, you also have to adjust the pitch of
[33:10] you also have to adjust the pitch of your fiber arrays. You no longer can use
[33:13] your fiber arrays. You no longer can use your standard 127 to 50 microns because
[33:17] your standard 127 to 50 microns because of the uh laws of refraction. So if you
[33:21] of the uh laws of refraction. So if you crunch the numbers, you'll find out that
[33:23] crunch the numbers, you'll find out that if uh for example your waveguide pitch
[33:25] if uh for example your waveguide pitch on the chip is 250 microns, then your
[33:29] on the chip is 250 microns, then your actual fiber array pitch that matches it
[33:32] actual fiber array pitch that matches it optically is smaller 241 microns. So you
[33:36] optically is smaller 241 microns. So you have to have a very special uh fiber
[33:38] have to have a very special uh fiber array made to match this. And if your
[33:41] array made to match this. And if your waveguide pitch would be smaller then
[33:44] waveguide pitch would be smaller then without interposer you cannot get fiber
[33:46] without interposer you cannot get fiber array with a standard fiber that would
[33:49] array with a standard fiber that would match this pitch because this is smaller
[33:50] match this pitch because this is smaller than the the size of your fiber.
[33:53] than the the size of your fiber. Alternatively what you could think about
[33:55] Alternatively what you could think about is to reverse the problem where you
[33:57] is to reverse the problem where you start with a standard fiber a pitch and
[33:58] start with a standard fiber a pitch and then you just adjust your waveguide
[34:00] then you just adjust your waveguide pitch. However, if I understand
[34:02] pitch. However, if I understand correctly that's uh uh foundry pdks do
[34:05] correctly that's uh uh foundry pdks do not really allow to have custom
[34:07] not really allow to have custom waveguide pitch. So that's uh for now
[34:09] waveguide pitch. So that's uh for now that's a no-go. Uh but you can uh come
[34:13] that's a no-go. Uh but you can uh come back to the interposer idea for example
[34:15] back to the interposer idea for example and have the interposer do all the uh
[34:18] and have the interposer do all the uh dirty work for you where you have your
[34:20] dirty work for you where you have your standard fiber array which is attached
[34:22] standard fiber array which is attached to the outside and the deposer not only
[34:24] to the outside and the deposer not only matches the pitch but also matches the
[34:26] matches the pitch but also matches the angle and you still have your normal
[34:29] angle and you still have your normal plane of coupling preserved. So this is
[34:31] plane of coupling preserved. So this is one of the uh one of the better ways of
[34:33] one of the uh one of the better ways of dealing with the indium phosphide angled
[34:35] dealing with the indium phosphide angled facets.
[34:37] facets. So I'm just going to briefly touch on
[34:40] So I'm just going to briefly touch on the grating coppers because from the
[34:41] the grating coppers because from the packaging perspective whether coupling
[34:43] packaging perspective whether coupling to the edge or to the grating there is
[34:46] to the edge or to the grating there is very little difference uh uh let's say
[34:48] very little difference uh uh let's say theory wise you're just matching the
[34:50] theory wise you're just matching the modes matching the angles and everything
[34:53] modes matching the angles and everything has to be uh you know has to be well
[34:55] has to be uh you know has to be well positioned. uh with grating copper
[34:57] positioned. uh with grating copper actually a little bit easier because you
[34:59] actually a little bit easier because you can design them to match the mode size
[35:02] can design them to match the mode size of your fiber a little bit better than
[35:04] of your fiber a little bit better than you can with edge copters. With edge
[35:06] you can with edge copters. With edge cutters you're dealing with some farm
[35:09] cutters you're dealing with some farm restrictions and how far you can expand
[35:11] restrictions and how far you can expand the mode using any sort of spot size
[35:13] the mode using any sort of spot size converter. So it's a little bit more
[35:15] converter. So it's a little bit more difficult although technology is getting
[35:17] difficult although technology is getting better on that front and we're getting
[35:19] better on that front and we're getting mode field diameters which are already
[35:22] mode field diameters which are already approaching let's say 10 microns at
[35:23] approaching let's say 10 microns at 1550. So we're we're getting very close
[35:26] 1550. So we're we're getting very close with grating coupers. It's a little bit
[35:28] with grating coupers. It's a little bit easier uh because the principle is a
[35:30] easier uh because the principle is a little bit different instead of just
[35:32] little bit different instead of just directly matching the mode. So you're
[35:34] directly matching the mode. So you're using grating using gratings essentially
[35:37] using grating using gratings essentially to scatter your light and to um uh to
[35:40] to scatter your light and to um uh to interfere in such a way that the light
[35:42] interfere in such a way that the light that is let's say goes in from the
[35:46] that is let's say goes in from the waveguide is coherently uh scatter in
[35:49] waveguide is coherently uh scatter in such a way that it leaves the plane of
[35:52] such a way that it leaves the plane of your chip at at a specific angle and
[35:55] your chip at at a specific angle and much of the power is maintained within
[35:57] much of the power is maintained within this beam. Of course, as you can see
[35:58] this beam. Of course, as you can see from the simulation, there are also
[36:01] from the simulation, there are also losses associated with uh uh you know
[36:03] losses associated with uh uh you know with uh transmission into the uh into
[36:07] with uh transmission into the uh into the body of the chip. So they're a
[36:09] the body of the chip. So they're a little bit inherently more lossy than
[36:11] little bit inherently more lossy than your edge copters. So in historically
[36:15] your edge copters. So in historically grating copters were better solution
[36:17] grating copters were better solution because they had a smaller losses than
[36:18] because they had a smaller losses than edge copters. But as the latter became
[36:21] edge copters. But as the latter became more more researched and better
[36:25] more more researched and better technologies were invented, now it's
[36:27] technologies were invented, now it's reversed. Now edge coppers have much
[36:29] reversed. Now edge coppers have much better coupling efficiency than grating
[36:31] better coupling efficiency than grating coppers. So here are some uh videos
[36:34] coppers. So here are some uh videos taken of numerical simulations where we
[36:36] taken of numerical simulations where we can see the beam from the fiber entering
[36:38] can see the beam from the fiber entering the grating copper at a specific angle
[36:40] the grating copper at a specific angle and being coupled in. I'll run that
[36:43] and being coupled in. I'll run that again. You see? Yes. So you also what
[36:47] again. You see? Yes. So you also what you saw very briefly though is some of
[36:49] you saw very briefly though is some of the beam was transmitted through and
[36:51] the beam was transmitted through and that's that's the major source of your
[36:53] that's that's the major source of your losses in the system. The element
[36:56] losses in the system. The element tolerances of uh fiber to gring coppers
[37:00] tolerances of uh fiber to gring coppers are because you can match the mode a
[37:02] are because you can match the mode a little bit better. They're also quite
[37:03] little bit better. They're also quite more relaxed. This is one of the
[37:05] more relaxed. This is one of the measurements I performed. Uh so what you
[37:07] measurements I performed. Uh so what you see here is your alam torance is around
[37:10] see here is your alam torance is around 3 microns which is congruent with 10
[37:12] 3 microns which is congruent with 10 micron mode size. Um however the shape
[37:15] micron mode size. Um however the shape is a little bit asymmetric because in
[37:17] is a little bit asymmetric because in this case we had a focusing graing
[37:18] this case we had a focusing graing coupler. So the shape of this tolerance
[37:21] coupler. So the shape of this tolerance curve is also asymmetric. We also have
[37:23] curve is also asymmetric. We also have this uh little bump here in the Z
[37:26] this uh little bump here in the Z direction which is this one. It occurs
[37:28] direction which is this one. It occurs over here when your fiber is positioned
[37:29] over here when your fiber is positioned over here. And this is actually a
[37:33] over here. And this is actually a reflection from light traveling. If you
[37:37] reflection from light traveling. If you So what this what this point shows is
[37:40] So what this what this point shows is when you retract your fiber towards this
[37:43] when you retract your fiber towards this side and what happens there is the light
[37:46] side and what happens there is the light gets uh transmitted through this is the
[37:49] gets uh transmitted through this is the silicon oxide silicon boundary which can
[37:53] silicon oxide silicon boundary which can reflect part of the light and it's
[37:54] reflect part of the light and it's getting essentially coupled from the
[37:56] getting essentially coupled from the bottom. So this is a little bit of a
[37:59] bottom. So this is a little bit of a spirious result but uh so when you're
[38:02] spirious result but uh so when you're doing optical packaging you have to be
[38:04] doing optical packaging you have to be kind of careful in order not to couple
[38:07] kind of careful in order not to couple uh you know from the wrong side. Uh the
[38:10] uh you know from the wrong side. Uh the coupling efficiency as I mentioned u is
[38:12] coupling efficiency as I mentioned u is uh is one thing. Um right now the best
[38:17] uh is one thing. Um right now the best uh
[38:19] uh the best grating coupers will have
[38:21] the best grating coupers will have around just north of 1 dB losses I guess
[38:24] around just north of 1 dB losses I guess but uh in comparison the edge coppers
[38:27] but uh in comparison the edge coppers they have much limited bandwidth. So the
[38:29] they have much limited bandwidth. So the edge coppers can be made they have
[38:30] edge coppers can be made they have bandwidth of more than 200 nanometers.
[38:33] bandwidth of more than 200 nanometers. Now it's uh pretty easy to do. However,
[38:36] Now it's uh pretty easy to do. However, grating coppers have due to the uh
[38:39] grating coppers have due to the uh coherent nature of uh of their
[38:41] coherent nature of uh of their operation, they have much narrower
[38:43] operation, they have much narrower bandwidth. So for this specific example,
[38:45] bandwidth. So for this specific example, for example, we have one bandwidth of 20
[38:48] for example, we have one bandwidth of 20 nanometers and 3D bandwidth of about 39
[38:52] nanometers and 3D bandwidth of about 39 nometers. But uh for each gritting
[38:54] nometers. But uh for each gritting coupler from different foundaries,
[38:57] coupler from different foundaries, different materials, you will see
[38:59] different materials, you will see different mileages. Your your bandwidths
[39:02] different mileages. Your your bandwidths and also losses will vary greatly.
[39:06] and also losses will vary greatly. Also uh rating coppers are not only
[39:10] Also uh rating coppers are not only sensitive to to wavelength but also
[39:12] sensitive to to wavelength but also sensitive to angle. So as you saw from
[39:14] sensitive to angle. So as you saw from the simulations uh they all showed that
[39:16] the simulations uh they all showed that the beam was uh uh
[39:20] the beam was uh uh u was not perpendicular to the great cup
[39:24] u was not perpendicular to the great cup was actually coming at a specific angle
[39:26] was actually coming at a specific angle of incidence and in standard uh PDKs
[39:30] of incidence and in standard uh PDKs you'll find either 8° or 10° standard
[39:33] you'll find either 8° or 10° standard angle of incidence. So that's the angle
[39:34] angle of incidence. So that's the angle at which you must hit the grating copper
[39:36] at which you must hit the grating copper to get the best losses. And this graph
[39:38] to get the best losses. And this graph just shows a simple measurement where it
[39:40] just shows a simple measurement where it was just the angle of the fiber was uh
[39:43] was just the angle of the fiber was uh modified and uh the you coupling
[39:46] modified and uh the you coupling efficiency is of course for this
[39:48] efficiency is of course for this specific coupler is best at 10° because
[39:51] specific coupler is best at 10° because that's the design uh AOI but uh your 1dB
[39:55] that's the design uh AOI but uh your 1dB coupling efficient your 1dB angle of
[39:58] coupling efficient your 1dB angle of incidence tolerance is around 3° which
[40:02] incidence tolerance is around 3° which is kind of congruent with if you
[40:04] is kind of congruent with if you remember the uh graph I showed before
[40:08] remember the uh graph I showed before where at 10° we had about 2.5
[40:11] where at 10° we had about 2.5 uh degrees uh tolerance from theory
[40:14] uh degrees uh tolerance from theory that's kind of it's ballpark there.
[40:19] that's kind of it's ballpark there. So if you want to couple your fiber
[40:21] So if you want to couple your fiber array to the grating coppers there's two
[40:24] array to the grating coppers there's two ways in which you can do that. One is to
[40:25] ways in which you can do that. One is to go from the top. So you polish your
[40:28] go from the top. So you polish your fiber array at the angle which is
[40:30] fiber array at the angle which is exactly equal to the ay and then just
[40:32] exactly equal to the ay and then just attach it from the top. This is quite
[40:35] attach it from the top. This is quite simple thing to do. um quite simple
[40:37] simple thing to do. um quite simple thing to align and uh and package.
[40:39] thing to align and uh and package. However, uh as you might imagine, this
[40:41] However, uh as you might imagine, this requires quite a tall package because
[40:43] requires quite a tall package because the fiber just fiber just sticks out
[40:45] the fiber just fiber just sticks out vertically. So, if you want to create a
[40:47] vertically. So, if you want to create a low profile package, then this is going
[40:50] low profile package, then this is going to be essentially something that's
[40:51] to be essentially something that's sticking up at you know uh vertically
[40:54] sticking up at you know uh vertically and uh there's a cable management issues
[40:56] and uh there's a cable management issues that might arise. Alternatively, you can
[40:59] that might arise. Alternatively, you can use total internal reflection
[41:02] use total internal reflection at which you polish your uh uh your
[41:06] at which you polish your uh uh your fiber at a very specific angle which is
[41:08] fiber at a very specific angle which is given by this formula.
[41:11] given by this formula. And then what you will have is the u
[41:14] And then what you will have is the u light travels through the core and is
[41:16] light travels through the core and is reflected internally from the facet and
[41:19] reflected internally from the facet and then is injected into your rating
[41:21] then is injected into your rating coupler. However, you have to take into
[41:23] coupler. However, you have to take into consideration that there is a minimum a
[41:25] consideration that there is a minimum a angle of insertion that you can satisfy
[41:27] angle of insertion that you can satisfy due to the loss of total internal
[41:29] due to the loss of total internal reflection at the at the uh uh glass to
[41:33] reflection at the at the uh uh glass to air interface. This also means that uh
[41:36] air interface. This also means that uh when you're packaging such a device, you
[41:40] when you're packaging such a device, you have to put adhesive underneath over
[41:42] have to put adhesive underneath over here. But if you have any of that
[41:45] here. But if you have any of that material encroaching on your facet, then
[41:49] material encroaching on your facet, then you will completely lose your uh total
[41:52] you will completely lose your uh total reflection effect. So uh when you're
[41:54] reflection effect. So uh when you're packaging, be careful not to put any
[41:56] packaging, be careful not to put any adhesive on the front facet or or your
[41:58] adhesive on the front facet or or your package is not going to work at all.
[42:01] package is not going to work at all. Also keep in mind that this fiber array
[42:05] Also keep in mind that this fiber array is essentially a lidless package uh a
[42:07] is essentially a lidless package uh a lidless fiber because this is the
[42:09] lidless fiber because this is the V-groove block uh schematically. is a
[42:11] V-groove block uh schematically. is a V-groove block and the fiber is just
[42:13] V-groove block and the fiber is just laid in the V-groove block. If there was
[42:15] laid in the V-groove block. If there was also a lid on top of it, then the
[42:16] also a lid on top of it, then the distance between your uh your fiber and
[42:20] distance between your uh your fiber and your uh chip would be 1 mm which means
[42:24] your uh chip would be 1 mm which means that at that distance your beam would
[42:26] that at that distance your beam would spread out uh quite quite far and you
[42:28] spread out uh quite quite far and you would have need to have a very large uh
[42:31] would have need to have a very large uh grating copper in order to capture that
[42:33] grating copper in order to capture that beam. So uh for uh coupling into grating
[42:37] beam. So uh for uh coupling into grating culprits in the horizontal or planer
[42:40] culprits in the horizontal or planer scheme you need to have lidless fibers.
[42:44] scheme you need to have lidless fibers. And this is just an example of such a
[42:46] And this is just an example of such a package where you see we have a photonic
[42:49] package where you see we have a photonic integrated circuit with a electronic AC
[42:51] integrated circuit with a electronic AC integrated on top. And since we wanted
[42:53] integrated on top. And since we wanted this to be quite low profile package, we
[42:56] this to be quite low profile package, we have this fiber array with a facet
[42:59] have this fiber array with a facet polished at uh in this case it was I
[43:01] polished at uh in this case it was I believe uh 40° to match the angle of
[43:04] believe uh 40° to match the angle of incidence. So these packages are quite
[43:06] incidence. So these packages are quite simple andite quite convenient to
[43:09] simple andite quite convenient to design.
[43:10] design. So u to to summarize the the webinar
[43:14] So u to to summarize the the webinar optical packaging it differs from the
[43:17] optical packaging it differs from the electronic packaging that require great
[43:19] electronic packaging that require great precision at least as far as uh
[43:22] precision at least as far as uh positioning in the x and y direction is
[43:24] positioning in the x and y direction is typically it will be of the or of single
[43:26] typically it will be of the or of single microns if if not less. Uh there are
[43:29] microns if if not less. Uh there are benefits to using both grating coppers
[43:31] benefits to using both grating coppers and edge coppers and I touched on some
[43:33] and edge coppers and I touched on some of those uh already. uh the right now
[43:38] of those uh already. uh the right now the way that things are moving the edge
[43:39] the way that things are moving the edge coppers are going to be the the uh
[43:43] coppers are going to be the the uh better for the future. However, rating
[43:45] better for the future. However, rating coppers have one um advantage in that
[43:48] coppers have one um advantage in that they are surface features. So you can
[43:50] they are surface features. So you can position them um so by positioning them
[43:53] position them um so by positioning them on the surface you can actually see them
[43:55] on the surface you can actually see them better during the alignment process and
[43:57] better during the alignment process and also you can utilize surface to have
[43:59] also you can utilize surface to have more of your coppers per unit area.
[44:02] more of your coppers per unit area. While with edge coppers you just have
[44:03] While with edge coppers you just have your edge which uh you know scales your
[44:07] your edge which uh you know scales your number of uh copper scales only linearly
[44:10] number of uh copper scales only linearly instead of uh as a square
[44:13] instead of uh as a square and uh going u to another conclusion
[44:16] and uh going u to another conclusion from this is uh the best way to minimize
[44:19] from this is uh the best way to minimize the losses is to match the motor
[44:21] the losses is to match the motor diameter of the copper and the fiber. uh
[44:23] diameter of the copper and the fiber. uh that's mostly on the on the foundry side
[44:27] that's mostly on the on the foundry side but you can mitigate any mismatches
[44:29] but you can mitigate any mismatches using several techniques which I touched
[44:31] using several techniques which I touched on briefly from uh uh uh fibers
[44:35] on briefly from uh uh uh fibers interposer and such. Uh and we also uh
[44:39] interposer and such. Uh and we also uh saw some quite interesting coupler
[44:41] saw some quite interesting coupler designs mostly grating couplers which uh
[44:44] designs mostly grating couplers which uh perform exceptionally well in lab
[44:46] perform exceptionally well in lab conditions. However uh sometimes uh we
[44:48] conditions. However uh sometimes uh we are uh we're not sure if they're
[44:51] are uh we're not sure if they're actually a very good solution for
[44:52] actually a very good solution for packaging. You can get very good
[44:54] packaging. You can get very good coupling for some of the designs but
[44:56] coupling for some of the designs but they might not be very easy to package
[44:59] they might not be very easy to package or you might not be able to package them
[45:01] or you might not be able to package them at all using at least using the current
[45:02] at all using at least using the current technology. So that is uh it for this
[45:07] technology. So that is uh it for this part of the webinar.
[45:09] part of the webinar. Thank you for your attention.

https://www.youtube.com/watch?v=uDhbVT2D_hg

Summary

TL;DR — This webinar introduces optical packaging, focusing on the critical challenge of coupling light between optical fibers and waveguides. It details the principles behind optical coupling, including mode field diameter matching, and explores various methods like edge and grating couplers, along with the complexities of facet preparation and alignment tolerances.

Key points

• Optical packaging is crucial for photonic devices, differing from electronic packaging by managing optical signals in addition to electrical ones.

• The primary challenge is efficiently coupling light between small waveguides (e.g., 220x400 nm on silicon) and larger optical fibers, minimizing signal loss.

• Key coupling methods discussed are edge couplers and grating couplers, with edge couplers being conceptually simpler and currently favored for higher efficiency.

• Precise alignment is critical, with tolerances often in the single-digit micron range, comparable to the size of a red blood cell.

• Facet preparation methods, including dicing, polishing, and wet etching, significantly impact coupling efficiency by affecting surface roughness and scattering.

• Indium phosphide chips present unique packaging challenges due to angled facets, often requiring specialized fiber arrays or interposers.

Takeaway — Effective optical packaging hinges on precise mode matching and alignment between fibers and waveguides, with ongoing advancements in coupler technology and fabrication methods aiming to reduce losses and improve integration.