Amp placement: Low-Level Differential Trace vs High-Level Single-Ended Trace

[tinfever]

I'm working on a DC/DC boost conversion board and I have current shunts monitoring the input and output current. These go to current sense amplifiers which then go to output opamps to drive a coax output to connect to a scope.

Would it be better to place the current sense amplifier next to the coax output driver so that the current signal is kept as a differential signal as long as possible? Or would it be better to place the current sense amp next to the current shunts and then run the single-ended output 50mm across the board to the coax output drivers?

If I place the current sense amplifiers next to the shunts, I'll also have to run their 2.5V voltage reference a long distance as well. These signal traces will be running perpendicular from the main DC/DC current direction, but I'm worried a single-ended trace will pick up more noise and also have a different ground potential if it is far from the coax output.

Photo of partially done layout attached. Board power input on the left, output on the right. Current shunts are circled on left and right. Current output SMA connectors, current sense amp, and voltage reference circles on the bottom. The board is a 6 layer board with three ground planes. Either signal choice will be run on an internal layer between two ground planes.

Thoughts?

[tom66]

What kind of range of signals are we talking about here?

If you're sensing 50mV across the shunt and your current sense amplifier is amplifying that to 2.5V, for example, then I'd suggest that putting the sense amplifier near to the shunts makes more sense.  Whilst the differential measurement will theoretically cancel out some of the common mode effect, they'll be extremely vulnerable to any stray fields around there.  Many current sense amplifiers have poor CMRR at high frequencies, e.g. ZXCT1009 has only 25dB CMRR at 10mV input at 1MHz. 

Can you have a dedicated voltage reference per channel, or route the reference signal on its own plane, instead of using a trace to route it?

[tinfever]

What kind of range of signals are we talking about here?

If you're sensing 50mV across the shunt and your current sense amplifier is amplifying that to 2.5V, for example, then I'd suggest that putting the sense amplifier near to the shunts makes more sense.  Whilst the differential measurement will theoretically cancel out some of the common mode effect, they'll be extremely vulnerable to any stray fields around there.  Many current sense amplifiers have poor CMRR at high frequencies, e.g. ZXCT1009 has only 25dB CMRR at 10mV input at 1MHz. 

Can you have a dedicated voltage reference per channel, or route the reference signal on its own plane, instead of using a trace to route it?

I see your point.

The input current signal will be 0-20mV amplified by 100 to 2V.
The output current signal will be 0-10mV amplified by 200 to 2V.
The current sense amplifiers are INA296A parts with a CMRR of around 90dB at 1MHz.

I could use a separate voltage reference for each channel.

What if I placed the current sense amplifier next to the shunt, and then reconfigured the opamp that drives the coax to be a difference amplifier, instead of just a buffer like it is now. Then I could run the current sense amp output along with a ground reference trace to that differential amplifier on the output? Or I could add a unity gain difference amplifier like an INA133 before the coax output buffer. I think this might be ideal. Although, now the offset voltage and CMRR comes into play. Cost isn't an issue since this is a one-off.

I've attached an image of the current sense amp and output buffer circuit.

Edit: I think I'm just going to reconfigure the coax output driver as a difference amp. Something like an INA133 won't swing close enough to the rails and something that could swing close enough and have really low input offset, like an INA592, doesn't come in unity gain, so I'd have to double the current sense amp gain and use the INA592 in 1/2 gain mode which would mean redoing the supply voltage rails. Image attached of what I'm thinking of doing.