Hello, I would like to split a 50 Ohm analog signal into 2 branches, 50 Ohms each. The 2 branches will be part of a Constant Fraction Discriminator Circuit (targeting sub-100ps resolution), so jitter is very important and of course a simple Y-split in the PCB is not enough.
I have been exploring a few options for passive fanout like Wilkinson resistive divider, or chips like PS2012GT2-R50-T1 resistive divider and EP2RCW+ MMIC splitter, but perhaps an active fanout circuit using op amp might be even better?
If active fanout is the way to go, could you please suggest any suitable circuits for that (or reading material on how to do that)?
My input signal is an analog pulse with sub-ns rise time and will be fed to a pi-attenuation branch and a delay-line branch.
Any advice would be really appreciated!

 

why is a Y spliter in pcb not good enough ?
make each branch 100 ohms !
what are the details of the signal to split ?
digital or analog levels ?
constant signal or occasional pulses ?
do the two outputs need to be identical , or is it just that the phase shift between the two branches must be constant ?
do you need a constant dc level or can you be ac coupled ?
finding an "op amp" with constant over pvt phase delays over the ps range will be interesting ..
https://www.digikey.co.uk/en/articles/the-fundamentals-of-rf-power-dividers-and-combiners

 

why is a Y spliter in pcb not good enough ?
make each branch 100 ohms !
what are the details of the signal to split ?
digital or analog levels ?
constant signal or occasional pulses ?
do the two outputs need to be identical , or is it just that the phase shift between the two branches must be constant ?
do you need a constant dc level or can you be ac coupled ?
finding an "op amp" with constant over pvt phase delays over the ps range will be interesting ..
https://www.digikey.co.uk/en/articles/the-fundamentals-of-rf-power-dividers-and-combiners

Hi, thank you very much for your answer.
I believe that a Y-splitter with 100 Ohm branches is not good enough, because although the input node does see 50 Ohm at the junction, the loads are actually mismatched and not isolated and that could result in reflections, crosstalk etc, that would significantly degrade the timing performance.
A passive resistive divider on the other hand, would give true broadband match and isolation but also gives around ~6db loss per output, that's why I was thinking of active fanout.
Let me explain a bit better my circuit, the input signal is an occasional analog pulse coming from readout electronics (based on an RF amplifier) of an SiPM. It has to be split in 2 identical branches, one branch will be attenuated by a pi-network and the other one will drive a pcb delay line. These 2 branches will be fed to an ultra fast, low jitter comparator that will find the zero-crossing point for these 2 signals.
As I said, I'm targeting sub-100ps resolution that's why I need to maintain the sharp edges of the input signal, and best possible SNR/jitter.

 

A quick search without too much detail.
A gain stage of +10db before the splitter. ~ 50ps
https://www.ti.com/lit/ds/symlink/trf1305b2.pdf?ts=1757192404415

 

Hi, thank you very much for your answer.
I believe that a Y-splitter with 100 Ohm branches is not good enough, because although the input node does see 50 Ohm at the junction, the loads are actually mismatched and not isolated and that could result in reflections, crosstalk etc, that would significantly degrade the timing performance.
A passive resistive divider on the other hand, would give true broadband match and isolation but also gives around ~6db loss per output, that's why I was thinking of active fanout.
Let me explain a bit better my circuit, the input signal is an occasional analog pulse coming from readout electronics (based on an RF amplifier) of an SiPM. It has to be split in 2 identical branches, one branch will be attenuated by a pi-network and the other one will drive a pcb delay line. These 2 branches will be fed to an ultra fast, low jitter comparator that will find the zero-crossing point for these 2 signals.
As I said, I'm targeting sub-100ps resolution that's why I need to maintain the sharp edges of the input signal, and best possible SNR/jitter.

Why is 6db attenuation a problem ? One arm is going into an attenuator anyway, and the other branch is into a comparitor , so actual level is not important.

Why does 100 ohms matter ? One side goes into an attenuator, make that 100 ohms , the other goes into the competitor.input , make that 100 ohms not 50.

Reflection , why would that be a problem if your matching the impedances , there should be minimal if any reflection .if it's not matched , even with an active output , you will have to have a very well matched rf system / layout to ensure a good pulse shape / edge .

Adding an amplifier , will add an inherent delay , more than a simple Y or resistive spliter.

With the photo multiplier, do need the DC component ? It's "common" to AC couple , and use a clamping circuit in the "quiet" period of the signal.. the advantage is ac amplification is much easier than DC to the GHz you need ...

I'm thinking with the trimming of the request that your on a university placement at some level , if your in industry , unless you want to sell these things , there are off the shelf things from people like vertilon corp that cover what you have outlined .