I think you need to connect the QA403 differentially like this.
I think I saw in a photo above the shield being used as Driver- like this,
If you correct as shown things will start to work, or at least not be so wrong 
(things are always a little wrong!).
That gets it closer but still no where near accurate. For example try a 27k Rdut and 2k Rsense. With the box checked in the new software and connected just as you have shown, I get Rdut=24k. I did connect the dac output to your pdb differentially, but why should this be necessary?
Using my formula and measuring the voltages off your pcb, I get Rdut=27088. This is with dac output not connected differentially.
Thanks for trying that with a differential connection. I think I understand.
If I let Rd = 27k and Rs = 2k and put into the apparent resistance expression,
It gives 24k Ohms which is what the QA403 should measure without correction applied and what you found.
Inside the Speaker Impedance plugin, @Matt should take this 24k value and correct it using this expression,
If you do that you recover Rd = 27k.
If the new checkbox does not recover the 27k then there could be a bug in the source code, because it should be capable of backing-out the correct DUT impedance given the above example.
Right now the QA40X software requires differential connection for the correction factor to have a hope of working. The reason is you get two very different circuits.
Differential connection
Single-ended connection
This will have a different set of correction equations. Maybe you have found them already and this is why it is working for you in single-ended mode?
Ok. Thx @Dan.
I used the same simplified diagram that @matt did and derived the equation the same way he did.
This ignores Rg1(because it is small I think) and R1 because it is large compared to Rg(I think thats why he did it) and ignores R4( because it is large compared to Rg2).
It seems to get very close but suffers when Rs is too small compared to Rd.
I’ll use your equations on those “bad” cases and see if that cures them.
I’m sorry if I’m wasting too much time on this. I’m 75 and have nothing else to do.
Yeah I think you are right. A small series resistance and a large shunt resistance taken to the limit remove themselves from the circuit. The load in the circuit needs to be in between those limits for it to be a good approximation.
OK! We have a plan 
I will calculate the expressions for the single ended case and see how we do.
Well I’m 42 and have 10 month old baby crawling around and a full time job and I’m working on it for fun too. I’m really enjoying the challenges you keep on throwing up, let’s get to the bottom of it. 
More mistakes. Its too late for me to be doing math at midnight.
Ok. New day.
I validated that actual measurements on the pcb match the TINA simulation of the differential input schematic.
I used the equation in the pic of chatgpt. Vd= voltage across the dut. Vs= voltage across the sense resistor.
V5=voltage from the junction of the sense and dut resistors to ground. The accuracy wanes when the sense resistor is say smaller than 5-10% of the dut.
In the pic below the dut is 9955 ohms.
Here is a much higher dut Rd=74510.
@Dan I tried using your correction formula on a variety of cases. All works fine. Both the formula I use and your correction one end up from say 1 to a few 10ths of a percent off. Interesting they seem to differ from each other in different directions in different cases. Not sure I understand the pattern. Anyway its unimportant. If @matt adds your correction to the check box result we should be good to go.
It just needs to be clear about the differential source.
I know my previous equation works with a single sided source. Maybe there needs to be a check box for that as well.
Great, thanks for coming back to confirm. Yes I agree! I did a little bit more work too and here are the final correction expressions for both differential and single-ended connection schemes.
Differential connection (the DUT is being drive by outputs L+ and L-)
Single-ended connection (the DUT is being driving by output L+)
Same definitions as before,
These are calculated from circuits show here.
You can access these circuits using the Falstad simulator here https://tinyurl.com/yndnx86z
Glad we got to the bottom of it. Do you get lower noise when you connect differentially rather than single-ended?
I love the falstad applet! Funny but I forgot about the noise! I’ll check.
So the UI needs two checkboxes: One to decide if user wants to compensate for analyzer impedances and another where the user specifies if balanced or single-ended, is that right?
Your call! I think that’s what @moto wanted. I tend to connect to the DUT with the QA461 or a rack power amplifier for higher power stuff. I just got involved to help Moto because he was using the PCB and it turned into an epic thread!
Thx to both of you. @matt the 2 checkbox alternatives would be great!
OK, got it. I think this can get done today.
Release is here. Note there is a new plug-in called MISC Input Impedance so it can be better tailored to the task. There are two settings that allow you apply input Z compensation. The first must be checked if you want compensation applied. The second checked determines the type of compensation that is applied.
I checked using 10 ohm sense into a 10K load (unbalanced), and without comp it showed 8.44K, and with comp is showed 10.1k. Since you have the setups on your desk and are knee deep in this, please let me know what you see.
Thx @matt. I just tried it. This is dut 26954 and sense 3000.
The SE measurement is great. Leaving everything exactly the same except adding the check to differential and rewiring the input to differential instead of SE we get the blue trace. So that somehow is not getting adjusted.
Also @matt my measurements seem to be sensitive to the ratio of sense to dut. Here is a 10.17 ohm sense and 9985 dut. This is se with output set to 0dbv and 0 attenuation. BH is used. This is on @dan pcb with l+ l- r+ r- connected. L+ output us used.
Here is the same measurement switching the sense to 1000.1
This may be more informative. Sense from 10-2000 ohms SE.
Sorry about the photos. Something is going on with my screen capture.
I’m using,
I get better results with balanced connection than single-ended (same scale, both plots). It seems I have a systematic offset compared to @matt results on the single-ended case. The waveforms are a little noise today, I think because the windowing function is Flat Top.
Balanced case looks great!
@Moto are you interested in reproducing these plots? I’m surprised by the single-ended results I got, they are not that good.
Ha! I would do anything to figure this out!
How about first I send pics of my setup to see if they match yours. Then a measurement and we can diagnose from there.
@dan , @matt think I have a clue. Here is the graph from the above pic yesterday but in red. 0db out and 18db atten. I redid it this morning with 24 db atten ( blue) and got correct results. Playing around with varying output and atten levels you can get different results.
Oh that’s really weird. The impedance depends on if the attenuator is set?! Does it change the input impedance? I will try and reproduce that. It’s a good point, we should synchronise on measurements settings as well as resistances. I see you connected a load box! I just have 10r and a 10k load soldered in.
I can’t wait to see your test on those different attenuations. I was thinking that it was the effect on qa input impedance. Also in general, when you have those high impedance dut’s and low impedance sense resistors, at 0db signal strength those voltages across the sense after 24db attenuation are getting pretty small.
Hi @Moto, the impedance changes you are seeing should be easy to see by looking at the amplitudes of 1 kHz tones when input ranges are changed. Can you see reported input levels change as you change full scale input ranges change??
