Is this how to measure THD+N when using a notch? Numbers seem too high (likely user error)

I found a couple of older, unresolved topics on the matter, but haven’t found a good answer. Essentially, I’m trying to measure a high end DAC - specifically the differential output of an ES9039SPRO @ 4Vrms. According to the datasheet

When I use the QA403 alone, I get a reading of around -115dB, which I think is limited by the QA403 itself. I’ve finally managed to get all the adapters and what not to add a notch into the loop - this particular notch has a -30dB gain at 1KHz, immediately followed by 20dB of gain.

I’m following this guide, for reference.

Firstly, I sweep the notch. Specifically, I am using a sweep at -20dBFS (to account for the notch’s built in 20dB gain), and I have set the differential output of the DUT to 4Vrms. The result is this:

I then export the notch and do another sweep with it enabled as a user weighting to prove it looks as it should do:

I’m not entirely sure whether my next steps are correct. With no stimulus and no weight, the noise is just flat across the frequency spectrum. With no stimulus and the user weighting applied, I can see the positive compensation of the weighting at 1KHz:

Now I play a 1KHz tone, with the user weighting applied, and at -20dBFS. The reported THD+N is -104dB:

Now, how do I interpret this? Do I add 20dB to account for the built in gain of the notch?

Alternatively, I play a 0dBFS 1KHz sine wave and now it’s reporting a THD+N of ~-124dB.

This can’t be right - that’s exceeding the spec of given in the datasheet. To take it to the extreme, if I repeat the steps above but add 60dB instead of 20dB of gain to the notch, I start getting crazy numbers like a THD+N of -127dB.

So could anyone advise on where I might be double-counting, or something?

Hi. It is as if you have a notch with a gain of 0 dB followed by a low-noise amplifier with a gain of 20 dB. You need to enter this gain (20 dB) in the ‘dBV’ context menu so that the values calculated and displayed on the tiles are correct. I understand from the image that places the frequency response of the Notch that it has sufficient Q so that practically the impact on the upper harmonics is negligible and thus the annoyance of exporting the Notch can be avoided. You can then use the THD dBr, THD+N dBr and SNR dBr tiles that have been set up for this purpose. Read the tread at this link to find out how to use them (this is the mode I use when I use Notch):

Ah, that link resolved the matter very nicely. Using the dBr method, I’m much more confident in my results, thanks for the pointer.