What is the THD measurement limit with the built-in generators?

I am trying to characterize an opamp circuit with respect to THD and THD-N using a QA400. It occurred to me that when I directly feed the output into the input and measure a 1kHz 0dBV signal I already get a THD of 0.0098%.
Does this mean that the lowest THD that can still be measured using the generators accurately is about 0.01%? The FFT resolution was set to 65536 and Fs:48kHz (seems my computer cannot do 192kHz). Thanks much in advance for advice!

Hi @filter7,

You are correct–the loopback measurement you made effectively determines the lower limit of what can be measured in a DUT. With THDN, larger FFTs don’t really buy you much. With THD, a larger FFT can indeed buy you a lot. Play around with sweeps to study that more.

The most expensive analyzers in the world cannot usually measure THD, THDN, etc of a simple op-amp circuit, because the performance is so good in modern opamps. For example, an opamp like the opa1612 has THDN around -136 dB, which means you’d need an analyzer with -146 dB of THDN to measure it accurately. The best test equipment today can deliver around -122 dB using very sophisticated analog electronics (tracking oscillators and notches).

For this reason, you usually need to rely on “distortion magnifier” of some kind to force the opamp to distort by a known amount. You then measure that, and subtract the magnifier amount. See section 8.3 in the OPA1612 spec linked below to where TI explains the process.

There is also some exceptional work from a fellow named Samuel Groner. He didn’t stop at an aggregate for opamp distortion, but instead delved into input/output/loading. It’s a fantastic piece of work.

Now, the good news is that if you have a fairly simple opamp circuit and you have screwed it up somehow (using X7R caps instead of C0G), then you will see that mistake on a QA401 in the form of higher harmonics. But in general, well-designed simple opamp circuits are very difficult to measure accurately even on very expensive equipment.

http://www.nanovolt.ch/resources/ic_opamps/index.html

Hi @Matt,

Thanks much for your quick and thorough response!

I studied chapter 2 of the Groner paper. So the basic idea is to run the opamp at a high (60 dB) gain and use the same 60 dB voltage divider to reduce the input signal from the generator running at full output signal. Thereby the generator runs at best SN and therefore most of the measured THD+N at the output of the opamp is generated by the opamp itself?

Can one then simply subtract the 60dB ’noise gain’ from the measured THD+N to get the ‘opamp contribution to THD+N’ at a 0dBV output signal level?

Thanks much for answering newbie questions!

Hi @filter7, I think the genesis for the Groner and TI techniques come from the engineers at Burr-Brown way back when. The link below has an EDN article written by a guy named Jerald G Graeme. That has as much as you’d want to know on the topic.

As Graeme outlines, with the addition of that one special resistor, you gain control over the signal gain and distortion gain independently. So, you want your signal gain to be 1, and your distortion gain to be 40 dB (in the OPA1612 spec) or 60 dB (in the Groner paper). Groner calls the distortion gain “noise gain” in his paper. I’m not sure why he calls it that. Note he does have a large cap in there, which kills the distortion gain as you get closer to DC.

In any case, it’s a pretty potent technique for application-specific questions you might have on an opamp.

Can one then simply subtract the 60dB ’noise gain’ from the measured THD+N to get the ‘opamp contribution to THD+N’ at a 0dBV output signal level?

I’ve never done it, but I think you’d want to first determine the equivalent input noise of the opamp. That’s usually done by shorting the input and then dialing in 60 dB of signal gain and measuring the noise. And then revert to signal gain of 1, and dial in 60 dB of THD gain. Then you know the THD and N, and from that you can arrive at the THD+N. I think as Groner points out, it’s not an easy automatic process and it’s quite intensive.