# Instrumentation amplifier with frequency-dependent gain?

I have an AD8421 (instrumentation amplifier); it has two pins to connect a resistor Rg that sets the gain --- G = 1 + 9.9k / Rg. For example, a 100 ohms resistor produces a gain G = 1 + (9900 ohms / 100 ohms) = 1 + 99 = 100.

I need the amplifier to have a lower gain at DC; to avoid additional stages or even a passive filter after the output, I wonder if I could use an R-C network where Rg is supposed to go, so that the "gain resistor" varies with frequency. In particular, what about placing the following:

simulate this circuit – Schematic created using CircuitLab

With this, I get a gain of 10 at DC, and 100 at high frequencies, with transition being from approx. 1.6Hz to 16Hz.

Sounds like in theory it should work; is there any reason why it might not work in practice? Any experience out there with this or similar setups?

Thanks,

Cal-linux

Can't think of a reason it won't work. The AD8421 uses a standard three op amp topology. It isn't one of those wacky ones designed to have a fixed gain when you leave RG open and another fixed gain when you short it. It should just work, and it shouldn't hurt the CMRR of the output stage of the IA. Give it a try, and kindly report back!

• It might be tempting to split the 100 into two 50's, one on either side of the RC, just to maintain symmetry, but I wouldn't think that you need to. Mar 30, 2016 at 22:47
• Probably won't give it a try; I noticed the "region of operation" in the datasheet (the Vout vs. Vcommon charts) and that suggested that it may be harder to tell whether the circuit is operating safely when you introduce any non-resistive components (mostly because I'm close to saturation due to DC). Also, that made me think that I could simply try increasing the supply voltages (that was sort of a "duh! moment" --- I had it at +/- 9V. I just tried +/- 15V and it looks like with my current setup it is ok; if needed, I could lower the gain to half what I have now and I should be safe) Apr 1, 2016 at 12:59

The gain will vary from 10 at DC to 100 at about 1kHz, but below 1kHz the circuit is unusable because of the change in gain with frequency. So long as the frequency of the signal is well above 1kHz then it may be usable. But most 100uF capacitors are very imperfect components and are likely to give other frequency related errors.

I presume that you are using an instrumentation amplifier to get a high CMRR. In the datasheet there is a comment regarding Rg.

Parasitic capacitance at the gain setting pins (RG) can also affect CMRR over frequency. If the board design has a component at the gain setting pins (for example, a switch or jumper), choose a component such that the parasitic capacitance is as small as possible.

In other words, if you add capacitance to the Rg pins you will degrade the CMRR. Unfortunately the datasheet doesn't say by how much, but if AD are concerned about a few pF of parasitic capacitance then 100uF is surely going to be a big problem.

If adding capacitance to the Rg pins was a viable way to tailor the frequency response I suspect that Analog Devices would have mentioned it. I would look at adding a traditional high pass filter after the AD8421.

If you're not convinced then it may be worth while trying to simulate the effects of adding the capacitor to the AD8421.

• I can see how unequal capacitances at each RG pin can be bad for cmrr, but not capacitance between the pins Mar 31, 2016 at 0:36
• Good catch, I overlooked this comment in the datasheet. However, that sounds to me like it would affect high frequencies (the AD8421 is supposed to go up to many MHz --- 2MHz at the gain that I'm using it). As Scott pointed out, adding explicit capacitance (in a network with a resistor in series) in the Rg path should be different than unbalanced parasitic capacitance, no? I'm planning to go with a SMD ceramic capacitor (digikey part number 399-11270-1-ND) --- quite surprising that they pack 100uF into a 1210 size! I wonder whether it may have nasty artifacts within my frequency range? Mar 31, 2016 at 13:13
• Actually, the AD8421 datasheet does not have that comment --- you may have been looking at the 8429 or 8226/8228. I wonder whether that same comment is not really applicable to the AD8421? (seems strange that it wouldn't, right? but it also seems strange that they would omit it if it is indeed applicable) Mar 31, 2016 at 14:34
• @Cal-linux This is the datasheet that I'm using - analog.com/media/en/technical-documentation/data-sheets/…. See page 21, under the heading 'Common-Mode Rejection Ratio over Frequency'. Mar 31, 2016 at 15:59
• Huh --- turns out that I was using the AD8421-EP datasheet, where -EP apparently stands for "Enhanced Product" (curious that with a Google search for AD8421 datasheet, the only PDF from analog.com is the -EP). It is 20 pages (instead of 28 for the standard one), and it does not mention this issue of parasitic capacitance. I'm pretty sure that the one I got from Digikey is the standard one; either way it's worth being careful about that detail. Mar 31, 2016 at 17:16