Improving rejection ratio of subtraction circuit

[halfgaar]

Hi,

I've build a prototype of a device in which I can plug two analog audio interlinks and the device outputs the difference between the two. The purpose of which should be clear.

I've equipped it with on-board jumpers which bypass the cable so that I can calibrate it. Calibration is done with a multi turn pot.

This is (the right channel of) the circuit:



J2 and J3 are RCA jacks. JP3 and JP4 are the jumpers.

At 1 kHz, I get a rejection ratio of 48 dB (7 Vrms in becomes 26 mVrms out, measured with digital scope). I of course want to improve upon that.

There is a region in the trimpot R26 where adjustment doesn't change the output voltage, i.e. it won't go lower than 26 mV for a few turns to goes back up again. DC offset remains the same in that zone.

I thought improving the virtual earth impedance of IC5B with a higher open-loop gain opamp would help, so I replaced it with an Burr Brown OPA2134, but that did not make one dB of difference.

I can't read any signal between the several ground points on the PCB, so ground appears to have low enough impedance.

Can anybody give advice how to improve rejection ratio?

[jcx]

look up instrumentation amp - a popular cirucit, lots of details when rolling your own

I would reduce (scale) R values by ~5-10x

the inverter and unity buffer have different feedback ratio, hence differing frequency response, phase shift - "noise gain" compensation may help

but really the simplest solution is to just buy a chip from Analog, Linear, TI, or THAT

[halfgaar]

I actually already ordered some INA128 chips from TI for a related project. It didn't occur to me to use them here too. I will look into that.

In the mean time, I will try fiddling with the resistor values a bit to see what happens.