Thanks. Would you actually have the tools to make optical waveguides on a substrate? I know about your maskless waver stepper, but I have no idea which materials one would use to generate a sufficient refractive index difference between the base and the waveguide.
@@DiffractionLimited The materials would not be a problem, but the resolution might not be sufficient if I wanted to use a wavelength in the visible range. It will be a bit of a challenge though.
Yes I think so. The way I understood this (which could be wrong) is that you can represent the two incoming waves as a 2D Vector of two complex numbers. Each complex number encodes the phase and amplitude of an incoming wave. The overall action of an interferometer (+phaseshifters) can be represented as a 2x2 complex unitary matrix. The overall circuit is a combination of many of such linear transformation und thus can itself be described by a larger linear transformation (a NxN complex matrix).
Absolutely correct. That is why in pratice the optical part is interleaved with electronic non linearities after the light signal is digitzed by the photodiodes. The goal is to replace the more expensive vector-matrix multiplication.
Проблема только в том, чтобы сделать фазовращатели, которые можно программировать на смещение фазы на заданную произвольную величину.. Я Мечтаю о таких чипах уже более 10 лет. Но я не инженер и не физик, я не знаю как они должны быть устроены, но точно знаю что они будут в сотни и тысячи раз быстрее кремниевых логических схем.
Very nice. Now I want to make this in hardware!
Thanks. Would you actually have the tools to make optical waveguides on a substrate? I know about your maskless waver stepper, but I have no idea which materials one would use to generate a sufficient refractive index difference between the base and the waveguide.
@@DiffractionLimited The materials would not be a problem, but the resolution might not be sufficient if I wanted to use a wavelength in the visible range. It will be a bit of a challenge though.
Every time I find a well hidden/obscure photonics niche on TH-cam..... You're always in the comments already.
@@hullinstruments Yeah, I watch way too much TH-cam 😊
wat een verrassing! 🤗
Just came across your channel, amazing stuff!
Text to speach is getting spooky good.
So the mach-zehnder interferometer has to perform some sort of matrix-vector multiplication, right?
Yes I think so. The way I understood this (which could be wrong) is that you can represent the two incoming waves as a 2D Vector of two complex numbers. Each complex number encodes the phase and amplitude of an incoming wave. The overall action of an interferometer (+phaseshifters) can be represented as a 2x2 complex unitary matrix. The overall circuit is a combination of many of such linear transformation und thus can itself be described by a larger linear transformation (a NxN complex matrix).
you need a non-linearity for it to be a neural network!
Absolutely correct. That is why in pratice the optical part is interleaved with electronic non linearities after the light signal is digitzed by the photodiodes. The goal is to replace the more expensive vector-matrix multiplication.
Blind Willie you was born to die on bgm sounds slightly better I guess!
Проблема только в том, чтобы сделать фазовращатели, которые можно программировать на смещение фазы на заданную произвольную величину.. Я Мечтаю о таких чипах уже более 10 лет. Но я не инженер и не физик, я не знаю как они должны быть устроены, но точно знаю что они будут в сотни и тысячи раз быстрее кремниевых логических схем.