Speaker Impedance with QA403 + QA461

Hi Matt, everyone,

Well I’ve designed a simple PCB for the current sense resistor

following your tutorial. In another thread there was some discussion about benefits of class AB over class D amplifier for this setup.

Well now I’m actually considering getting QA461. It will save me loads of hassle with getting this board made and buying a chip amp.

I see the differential to single ended conversion for the current sense resistors is done internally with QA461.

Are the following connections correct for an impedance measurement?

QA403 Output L+ to QA461 Input
QA403 Output L- terminated with 50r BNC
QA403 Input R+ to QA461 Output current sense
QA403 Input R- terminated with 50r BNC.

QA461 Output- to Speaker-
QA461 Output+ to Speaker+

QA461 Output monitor to QA403 Input L+
QA403 Input L- terminated with 50r BNC

So voltage sense and current sense are single ended?

I guess there is a way of telling the software the output is attenuator by -20dB from the output monitor port of the QA461.

Hope that is close!


Hi @Dan, the QA403 unused outputs should never be terminated. It will cause the output opamp to run hot, and eventually it will go into self-protection.

On the input side, it is correct to terminate unused inputs.

Aside from the output shorted, the rest is correct!

The speaker plug-ins allow you specify a voltage gain (-20 dB for the QA461) and sense resistor value. So, if you want to roll your own you have a lot of flexibility. But I do think the QA461 use of an INA199 was a good change, because the part has outstanding CMRR. At 1 kHz, it’s still well above 100 dB. The INA199 bandwidth is about 80 kHz. I think a future QA46x would aim to buffer the the output divider at the output port (so the output divider current isn’t counted in the output current). And probably reduce the output current sense resistor to 0.01 ohms and use the gain of 100 INA199 to keep the effective output sense resistor at 1 ohm. The OPA564 power opamp used in the QA461 has about 0.1 of output Z around 1 kHz, but it climbs quickly after that. However, the way the impedance measurements are made with the QA461 means the output Z of the power opamp doesn’t matter.

The INA199 CMRR plot is below.



Hi Matt,

Thank’s for the correction - will make sure not short my output :woozy_face:

That’s great to learn a little how it works.


Note to future self (and others), here are the connections

QA403 Output L+ to QA461 Input
QA403 Output L- no connection.
QA403 Input R+ to QA461 Output current sense
QA403 Input R- terminated with 50r BNC.
QA403 Input L+ to QA461 Output monitor
QA403 Input L- terminated with 50r BNC
QA461 Output- to Speaker-
QA461 Output+ to Speaker+

Very happy with QA461 and QA403.

The first thing I did was measure a 1 Ohm resistor I think the results are great, the output level used for the measurement was -20dBV.

I’m curious about the high-end noise and slight rise and I do just mean curious - what could cause that?

I want to measure some speakers that have resonance peaks in at around 100Hz that could be as a high as 80 Ohms. The nearest resistor I had to hand was a 100 Ohm, so I swapped that in, and used an output level of -10dBV.

I think might be pushing the QA461 a bit beyond it’s design with this. I remember in your tutorial it said the current sense resistor should be no more than 100x smaller than the largest resistance you want to measure.

Again I’m just curious at what is going on. I can hear an audible chirp which surprised me, when the output level is set high. Would you expect this measurement to fall outside of scope of the driver?

I played around with different output levels and what is shown are approximate optimum values for each case.

Other settings:

192k Sample Rate
Full scale 0
Attenuator Off
Sys. FFT 512k

Hi @Dan, the thing to keep in mind is that the impedance is computed from the left and right frequency response plots. The left is the voltage across the load, and the right is the current. It’s done as a complex calculation, which means phase matters. And so, as the signal-to-noise of the sweeps degrades at the band edges, the small changes in phase and mag will translate to larger changes in impedance. So, you can often look at the frequency domain plots and see what is the cause.

There is a new automated test coming that will you sweep inductors and capacitors at increasing current levels (assuming you are driving from the QA461). And at higher current levels the parts really squeal. The phenom is is microphonics. That is, if you mechanically distort a ceramic cap, you will change its value. And similarly, as you stimulate a R or C or L, you’ll hear it as the applied voltage/current will cause the part to slightly deform and make a noise. There are lots of app notes written on how to get loud caps quieted down in circuits. It’s a very interesting problem. I first encountered it in the 90’s when rubbing a plastic case on a cell phone would induce a change in a VCO capacitor value, changing the tank frequency slightly before the PLL loop corrected it. But as the phone was FM, it would manifest as as “shhhhhh” noise in the received (and transmitted) audio.

That makes sense, so might just be at the limits. I have not yet checked in more detail as you suggest, but I will investigate those regions for noise bearing in mind the complex division.

Oh cool! Once you have the phase between voltage and current you pretty much have all info that characterises capacitors and Inductors, including all the parasitics! As you already know ;0) I will definitely be using that when it is released, let me know if you need a beta tester!

Picture of the connections might make it a bit easier for others in the future. The DUT in this case is the resistor across the QA461 output. Note the 75Ohm (or you could use 50Ohm) terminations on the unused QA403 inputs.

Release 1.178 is located at the link below. A lot of work was done on the MISC RLC Automated Test and that write-up is HERE

Wow, you have been busy. The documentation really makes all the difference. Thanks!

Reading …

Are you doing this impedance with the DUT in series with the driver? Only the + output seems to be connected.

I see with the speaker impedance measurements it is in parallel, both - and + output are connected.

In the screen above above you can see I have connected the resistors into the QA461 in this way.

Hi @Dan, sorry if not clear

In the first part of the post, the QA461 and its built-in current sense was used. For higher-current situations like speakers, that should generally be sufficient. But for small L’s and C’s, a higher sense resistor can help a lot which is what the post tried to outline (in terms of recognizing when you were at the limits). So, in that case, an external R was used. And at that point, the diagram pretty much reverts to the old speaker impedance measurement (that is the picture that features the QA401 in the post–but the connection is the same for the QA403).

Only the + output seems to be connected.

In practice, the OUTPUT - is connected to the DUT ground, it’s just not shown in the picture. The DUT ground also picks up the ground from the BNC shell connection to the current sense output.

What would be nice in a next QA461 would be the ability to switch output Z so that measurements on very large impedances can be readily made

@matt , finally got around to trying the new QA461. Trying to follow the wiki on doing RLC measurement,
I put in a 10 ohm resistor on the dut connector on the QA461. Setting the input signal at -40dbv, the graph shows the left and right channel overlaying one another at -40dbv. It doesn’t matter whether I set external gain at 0db or -20db. Same result. I have to set external gain at -20db in order for it to read 10ohms.
Any ideas? Thx

Ho Moto, are you using the front-panel Output Monitor on the QA461? Or sensing across the resistor directly? The RLC wiki page was sensing directly across the R. So, if you were using the QA461 Output Monitor (gain of -20 dB) that might explain it???

@matt I was using the output monitor connection.
Why does changing the external gain parameter not affect the graph though?

Hi @moto,

it should. I will check it here: I am using QA461, with left channel on QA403 tapped directly off of DUT, and right channel differentially off a 10 ohm sense resistor and a 10 ohm DUT. So, I’m not using the QA461 Output Monitor.

Here are settings:


Here is result, which is correct

Next, I keep everything the same and change input gain to -20 dB just to see if graph changes


And graph shows impedance as 100 ohms instead of 10 ohms. This is wrong of course, but it shows that if you change the external input gain, the graph changes.

Can you replicate?

Sorry @matt but I’m not clear on your setup here.

Hi @Moto,

It’s this setup show below. Except, instead of a 0.1 ohm sense resistor it is 10 ohm sense resistor.

The amp I’m using is a QA461. However, I’m not sensing from the Output Monitor port–I’m sensing directly across the DUT (in this case, a 10 ohm resistor). This is the same as shown in the picture below.

note that the QA461 is a single-ended output (+ swings, - is held at ground). So, the L- is tied to ground. That’s the same as what is shown in the drawing.

This setup is from the RLC page HERE

Thx @matt. I did this and the result was as yours.
User error on my part on the earlier stuff.
With a 10 ohm resistor and using the 1 ohm sense resistor in the QA461, why is the impedance on the graph showing as, in my case, 9.65 ohms at any reasonable low frequency when the dc resistance measures 9.98 ohms with a good dvm?

Hi @Moto, take a look at the noise of the left and right traces in the main display. At the band edges, they will start to get a bit fuzzy, and that will translate to uncertainty. That’s why there’s the option to limit plot to 100 to 10k for example.

For a resistor, the definitive answer would be to measure the amplitude of tones at the desired frequency and treat as normal divider circuit. That doesn’t work if the part is reactive, of course. When the part has a non-trivial reactive component you need to incorporate phase into your calcs (and with that, things get messy fast if calculating manually).

Alternately, look at the 1 kHz figure and assume a modern resistor won’t show any frequency sensitivity down to DC from the 1 kHz figure :wink:

@matt is there any way to force the output lower than 20hz which seems to be the limit when unchecking the 100-10khz box?
Also is there any reason not to leave the QA461 powered on?

Hi @Moto, are you talking about adjusting the graph limits on the graph? Or making the swept sine go lower in frequency?

The QA461 is fine to leave on all the time.