I have a high power 115 VAC, 400 Hz sine wave output, 20 KHZ PWM inverter (90 kVA) that runs from a nominal 450 VDC Bus. I'm using TI AMC1100 isolation amplifiers for the sine wave output sensing and also for the DC Bus sensing. My control is via a TMS320F2812 DSP. I have two main choices for design and I need to make the choice that will minimize the impact of picked-up switching noise of the sensed analog signals. The input side of the ACM1100 iso amps is the same for both choices and only the secondary-side is different. Note that I have my DSP control board with the F2812 and a separate Interface PC Board with the AMC1100s and associated circuitry on it. These two boards are mounted side-by-side on a metal panel with short standoffs (1/4" long) and are about 8" to 12" apart and will be connected together largely with ribbon cable and IDC box connectors. My two design choices are as follows:
1) Run the AMC1100 output side from 5VDC. Connect the AMC1100 iso amp outputs to a standard differential op-amp circuit (also on a single 5V supply) using possibly a TLV2322 op-amp. The single-ended op-amp output then goes to the DSP board thru the ribbon cable to a single DSP channel that has proper anti-aliasing filters (on the DSP board). These anti-aliasing filters on the DSP board also will help mitigate any noise picked-up between the differential amp outputs and the DSP board A/D inputs.
2) Run the AMC1100 output side from 3.3VDC. Connect the AMC1100 iso amp outputs directly to the DSP PC Board thru the ribbon cable and each AMC1100 output would connect to two A/D channels on the DSP board for differential processing. There would be no anti-aliasing filter between the op-amp output and the A/D inputs, however, an anti-aliasing filter will be contained in the AMC1100 and/or the differential op-amp circuit. I would have no real way to mitigate possible severe noise pickup between the AMC1100 outputs and the DSP A/D inputs.
Please note that I have no choice in this design with respect to having a separate DSP board and Interface board. I'm stick with that initial condition. Although I don't have every detail worked out yet, I expect the cable lengths from the DSP to Interface boards that carry the isolated analog signals to be about 1' to 1.5' long and likely be ribbon cable. You may wonder why I can't use the same anti-aliasing filter I spoke of for choice-1 (single-ended to DSP BD) in choice-2 (differential to DSP BD). The reason is that the DSP board is a mature board and I have to use it the way it is. It only has these anti-aliasing filters on a few A/D channels and I would either have to remove them from all channels or use the differential A/D with half having a filter and half not having a filter. This is probably a bad thing to do.
There may be other things I can do to mitigate possible noise pickup in the ribbon cable that I haven't thought of. I hope you can help.