Designing Photonic Integrated Circuit: Photodiode to CMOS

[BenKOTI]

I am now trying to design a photonic integrated circuit which will be used to replace the global interconnect layer of a LSI. I need to somehow take the output of a photodiode w/o TIA (10G signal perhaps) and then transmit that signal to other layer of the LSI by TSV or something. Assume a connection between photodiode and a CMOS inverter.

(1) What is the required voltage output of the photodiode. I have heard from someone that the required voltage swing for driving an inverter in 32 nm node is 200 mV, is that true? I thought it would be higher.
(2)Is it necessary to realize impedance matching in this kind of design?

 

[Twigg]

(1) Depends entirely on the supply voltage for the inverter you're using. For a 5V supply, conventionally 3.5V to 5V is recognized as "high" and 0V to 1.5V is recognized as "low". If possible, it may make sense to introduce an analog comparator before the inverter.

(2) Not unless the signal is RF. If it's short of 10MHz at the photodiode, I wouldn't bother.

 

[analogdesign]

I am now trying to design a photonic integrated circuit which will be used to replace the global interconnect layer of a LSI.

That has been one of the holy grails of the electronics industry for at least 20 years now. If you can solve it, you'll be rich beyond your wildest imagination! I'm surprised you would have to ask our opinion here...

I need to somehow take the output of a photodiode w/o TIA (10G signal perhaps) and then transmit that signal to other layer of the LSI by TSV or something. Assume a connection between photodiode and a CMOS inverter.

In your scenario you would be using the CMOS inverter as a (poor) TIA. But I'm assuming you know that. Why not use an actual TIA and get *much* better performance? You can use the output of the TIA to drive an inverter, but then you'd be using the inverter as a (poor) limiter. But I'm assuming you know that. Why not use an actual limiter circuit? You can make those (and many are made) in CMOS, you know. Do you know how analog front ends for optical communications transceivers are designed? They evolved the way they did for a reason.

(1) What is the required voltage output of the photodiode. I have heard from someone that the required voltage swing for driving an inverter in 32 nm node is 200 mV, is that true? I thought it would be higher.

Do you know how a diode works? The "required" voltage output is just the turn-on voltage of the diode, as the signal is a current.

(2)Is it necessary to realize impedance matching in this kind of design?

What?

 

[dlgoff]

That has been one of the holy grails of the electronics industry for at least 20 years now.

More like 40 years.