Trouble measuring impedance with REW

[tucura]

Hello to all,
I am using REW 1.8.0_332 64bit running in MacBook Air, OSX 11.6.4
Measuring input impedance of a diy 3-way speaker cabinet found results differ from predicted values. That motivated to test an LC tank build with a 1.2mHy coil and a 25uF capacitor that were previously measured.
For impedance measurement I use the line input of a Creative Extigy USB sound card, signal output from Topping E30 and power amplifier, and suggested impedance measurement jig (60ohms sense resistor).
Measuring a 10ohms resistor give a flat amplitude plot at the correct impedance value.
The LC tank impedance measurement result is in the attached plot from REW. The resonant frequency is near 6kHz (instead of 918Hz).
Measuring the resonant frequency using the REW signal generator, manual frequency sweep, and measuring max voltage with a voltmeter results in 927Hz.
I cannot explain the disparity of results.

 

[John Mulcahy]

Using different interfaces for input and output isn't ideal, since there will be small clock rate differences between them and the latency from output of one device to input of another can be variable. They need to be operating at the same sample rate of course, selected in Audio Midi Setup to match the rate chosen in REW. The different devices shouldn't have a big impact though, so there could be an issue elsewhere in the setup. You say the results "differ from predicted", but by what margin? If the impedance measurement is set up and calibrated correctly it provides very precise measurements of components and complete systems, should probably start by attaching an mdat file with the results of the open, short and reference calibration measurements.

 

[tucura]

Thank you for your prompt answer. The predicted (and also measured) resonant frequency is around 900Hz. The REW impedance plot shows resonance near 6kHz. But it is a very clear resonance shape. Then is difficult that the such a frequency shift is due to using two different interfaces. But please understand that I attribute an error of such magnitude to my method/setup and not to REW. I will send you the calibration plot early tomorrow.

 

[tucura]

Here are the results of a calibration round after 1/6 octave smoothing (very noisy otherwise) using the previous jig with a 10ohms test resistor.
Not sure if the near 100 ohms trace is correct given that the jig uses a 60 ohms sense resistor. Anyway the real measurement discrepancy relates to resonant frequency detection, shown as 5.8kHz on the REW impedance plot, and 820Hz yield both by hand calculation and separat measurement as described before.

 

[tucura]

I made additional measurements using a 1.2mHy inductor with a 3.3uF capacitor. Resonant frequency is ~2.5kHz . REW impedance plot shows resonance at near15kHz (as seen in attached plot). Interestingly the ratio of resonant frequencies (REW/measured) is very similar in both cases:

1.2mHy//25uF. 5.8kHz/920Hz = 6.3
1.2mHy//3.3uF. 15kHz/2.5kHz = 6.0

 

[John Mulcahy]

If you need so much smoothing something is wrong. The open circuit check is for gain mismatches between the input channels, the 'impedance' result it shows is really a percentage channel match. Attach mdat files rather than images, please, there is much more info in the file. As I mentioned above, make sure the input and output devices are operating at the same sample rate by checking the rate selected for each of them in Audio Midi Setup and make sure that matches the sample rate chosen in REW.

 

[tucura]

I just noticed you are the author of REW. It is the most outstanding piece of software I've seen put in the public domain (and I have seen many!). Every audio passionate should be thankful to you and your effort to produce and release to the public domain such a well done product.

 

[tucura]

Both dac and adc are using 48kHz sampling rate. Anyway, I will try to use only the Creative Extigy card to eliminate that variable.

 

[tucura]

I managed to use the Creative Extigy card only, plus a power amplifier driving the impedance measurement jig. The results are the same. I have further made some additional tests using a 10ohms resistor in parallel with 25uF capacitor. The attached results using 1/6 smoothing suggest some differences difficult to explain by me, the significant difference in the low frequency region. I have to mention that the test with parallel RC is order of magnitude noisier before filtering than the single R. The sound card works all right reproducing music, but it is true that the single loop-back configuration (while calibrating) is noisier than expected.
John, I attached the mdat files as you requested.

 

[John Mulcahy]

There is something very broken in that setup. I also see the measurement labelled with "96 kHz" has the sample rate set to 48 kHz in REW? There also does not appear to be any selection made in the soundcard preferences for the input and output of the Extigy interface?

Please go through the impedance calibration steps again from the start and attach an mdat file with the 3 calibration measurement results, it might have some clues about where things are going wrong.

 

[tucura]

The sampling rate can be set only on REW. I repeated all measurements after setting the sampling rate to 44.1, 48 and 96kHz. Results look all about the same. It was my fault not to include the calibration files in a previous message. This plot here corresponds to 48kHz sampling rate, 60ohms sense resistor and 10ohms test.

 

[tucura]

I will test again and attach the mdat file later today.

 

[John Mulcahy]

Open the Audio Midi Setup app on your mac (in utilities), select the device and look at the Format setting to view/change the sample rate.

 

[tucura]

Please find attached the calibration run using Extigy card only:
Rsense = 60 ohm, Rtest = 10 ohm, Fsamp=48kHz

 

[tucura]

Additional test using 10ohm only and 10ohm//25uF with Extigy card only and 48kHz sampling rate.
No filtering. Test with resistor only shows low noise. Test with resistor//capacitor very noisy. Test repeated several times with same result.

 

[tucura]

Open the Audio Midi Setup app on your mac (in utilities), select the device and look at the Format setting to view/change the sample rate.

The Midi Setup shows sampling rate for Extigy card as 48kHz, both line input and output, same as used on REW. I changed to 96kHz on some of the tests to see if it had any impact on noise, but most tests were done on 48kHz.

 

[John Mulcahy]

The cal results look fine, other than the variation between 5 and 20 Hz, but the interface roll-off is probably the cause of that.
The 10R // 25uF is crazily noisy, though not as bad as the earlier measurement with the amp. Difficult to know what could be causing that.
Does your Mac have built-in audio, headphone out and line in? If so, worth trying those.

 

[tucura]

The Mac doesn't have stereo inputs. At this point I am not interested in measuring impedance with any precision, only care about frequency dependent issues. And a 6x error in the evaluation of the resonant frequency of a simple LC tank make this task impossible.
My next step is to evaluate the tank using the SPL measurement setup using the Mac built-in audio and a power amplifier using a jig with a 100 ohms resistor to detect resonance. I need a good dose of good luck!!!

 

[John Mulcahy]

It is difficult to fathom what is causing the odd measurements. I don't have many components around me, but here is a measurement of a series LC, 9.97 uF and 1.03 mH - would expect the impedance min to be at 1.57 kHz and the resistance there to be around 14.48 ohms (from the series R of the inductor and the ESR of the cap, both also measured in REW) and that's what I get.

 

[tucura]

I am amazed of how clean the curves are: I have never, under any circumstance, managed to get such a clean plot. Before I attributed my noisy plots to measure the speakers in a noisy room, now I understand there are other source of problems.

 

[tucura]

Surprisingly (for me!) the Macbook Air doesn't seem to have an external microphone input.
Using a simple USB headphone/mic interface and a 100ohm resistor driving the LC parallel tank I managed to verify resonance using the SPL mode on REW. Resonant frequency shows now located at ~900Hz as expected. You can verify the result on the first data file (1.2mHy_25uF_USB-DAC_JUL19.mdat)
Moving back the same setup to the Extigy resonant frequency looks now like being near 6kHz (the original problem that motivated my request for help). The result is in the second data file.
It is now clear to me that in order to make impedance measurements with REW I have to invest on a new sound card with stereo inputs and retire the Extigy. But It is hard for me to recognize that I cannot understand what can be wrong with the Extigy that I've been using for 20 years for recording and playing audio. It is difficult to justify the origin of a 6x frequency shift on the frequency spectrum.
I am very grateful for all the attention and help I received.

 

[John Mulcahy]

Those old Soundblaster cards used to have their own control console for configuration (on Windows at least) and including a suite of "EAX" effects. There is a button on the card to turn the CMSS off, but I think the control console is needed to disable all EAX processing. If you take a look at the impulse responses you'll see the Extigy result is a little...different to what one usually sees. Perhaps the remote gives you some additional control over it to make sure all processing is off and the input signal is not being sent back to the output?

 

[tucura]

Good point! I will try to look into your suggestion. I will try with Windows 98 running inside a VirtualBox.

 

[tucura]

It was impossible for me to get a copy of the Extigy original CD. I managed to measure with sufficient accuracy (not much needed) the input impedance of the crossover using the mono input of a USB audio card, a power amplifier and a 100 ohm resistor. Also REW in the SPL measurement mode. Calibration done with a 10ohm resistor. All very crude but enough for my current needs. I appreciate all the help received here.

 

[PMA]

REW suddenly started to give me strange plots when measuring impedance. It is a comb filter like plot (below). With all sound cards I have. SPL measurement is OK. Impedance measurement with LIMP is OK as well. I have no idea what has happened, any advice would be highly appreciated.

Edit: fixed, "use loopback as timing and cal reference" must not be used, only "use as timing reference".