We have an AD621 measuring the voltage drop across a 2.5R sense resistor.We are measuring 78uV of noise at the output of the AD621 with x10 gain. I have taken the inputs 2,3 out of circuit and directly shorted them. The noise is still there.The data sheet does not provide a Power supply rejection value or graph.Do you have such data? also is there any advise you can offer to track down the noise?
Please refer to the AD620 datasheet. The AD620 is very similar to the AD621, except that the gain setting resistor is outside rather than inside the part. The AD620 does have PSRR graphs.
Note that the noise peak to peak noise in the datasheet is referred to input and restricted to the range of 0.1 Hz to 10 Hz. i.e. if your measurement system restricts the bandwidth to this range, you should see about 5.5 uV on the output. If your bandwidth is larger, you need to use the 13 nV/rt(Hz) number. Assuming you have a one pole roll off in his filter, the way to calculate the rms noise would be:
Gain * input referred noise * sqrt(BW * PI/2), which is 10 * 13 * sqrt(BW * 1.571)
To convert from rms to peak to peak noise, it depends on whose definition you use. (Since noise is random, theoretically if you wait long enough you can get whatever peak you want.) However typically a factor of 6 is used. So to get peak to peak:6 * Gain * input referred noise * sqrt(BW * PI/2), or 6 * 10 * 13 * sqrt(BW * 1.571)
In summary, depending on the bandwidth being used, your circuit may be just fine and 78 uV is the inherent noise of the part. If this ends up the case, we have a newer part, the AD8421, which has less noise at gain of 10 (~7nV/rt(Hz).
We also have some older parts, the AD624 and AD625, which will have a noise of 8.5 nV/rt(Hz) at a gain of 10. All of these parts will require anexternal resistor.