10kHz Square Wave - How To Use & Interpret It

[gadget73]

I think thats just a limitation of the measurement method. It nulls out the fundamental and anything that isn't the fundamental is lumped together into a single figure that is a % of that original signal. Not really a super easy way to extract the background noise level from it and give a pure THD value with single-function measurement equipment but I suppose with software it wouldn't be all that difficult. Measure the noise profile, measure THD+N, invert the noise to cancel it and display the resulting difference as THD but other than just showing a somewhat lower number I don't know that it would really tell you very much.

A total THD number is not the most useful thing anyway. A spectrum plot showing distribution of harmonics would tell you more, but simple numbers are easier to compare, and its a less complicated instrument to make.

 

[nerdorama]

I use ARTA for acoustic measurements but haven't used it for THD, spectrum tests, etc. Guess I should investigate. I have a Picoscope which makes freq response easy among other oscilloscope type functions. Pico is probably not the best tool for spectral info but making relative comparisons for power supply noise levels is handy. I also use a DATs for impedance scans and measuring inductors and capacitors.

I've read that REW is very good for THD, spectrum tests, as well as acoustical and impedance plots. Seems like a good all-in-one package somewhat like ARTA. Not enough time or energy to learn them all.

 

[StepheK]

I've enjoyed using the USB Analog Discovery 2 with the freeware software package Audio Analyzer Suite. Lots of "one button" testing and have been able to discover some serious flaws with some commercially sold amps using it and also use it for tuning my own stuff. On square waves, sometimes I am looking for things other than how stable an amp with global negative feedback is (especially given none of the SE amps I build use GNFB). But I'm also still learning, am not a classically trained electrical engineer and I find it interesting to play around with different tests and how the results corollate to other tests.

For example when working on the Willsenton R300 300B amp, at low power levels both the frequency response and the 10K square wave looked reasonable. If I only tested it at 1W I would never have seen this problem. At about 1/3-1/2 power the frequency response starts suddenly rolling off on the top end around 8K and at 12k has a sharp dip of negative 10db+ and at that same power level the 10K square wave goes CRAZY! I bread boarded in a different pair of output transformers, the problem disappeared. Shortly after posting the review, the manufacturer came out with a new "MKII" version with different output transformers that fixed this anomaly.

After that experience, I now do testing at various power levels to make sure something crazy like this isn't happening. Obviously YMMV.

 

[StepheK]

I've read that REW is very good for THD, spectrum tests, as well as acoustical and impedance plots. Seems like a good all-in-one package somewhat like ARTA. Not enough time or energy to learn them all.

But REW is using a microphone listening to the speakers and the room right? I've used it for testing speakers so am familiar with it, but it seems like adding speakers + room into the measurement isn't measuring the amplifier.

 

[nerdorama]

I think it can also take electrical signals from a soundcard. I haven't used it but I'm pretty sure I've read the reports of others who have done THD, FR, etc for amplifiers. Can't take my word as absolute on this.

 

[gadget73]

my understanding is that its just comparing input to output. When the mic is used, there is a correction curve applied. I don't see why you couldn't just connect the output of an amplifier to the input in place of the mic and not use a correction curve. That should still indicate the difference between output signal and input signal, though in this case the difference would be caused by the amplifier.

never fooled with it though so I may be way off on that.

 

[THD+N]

my understanding is that its just comparing input to output. When the mic is used, there is a correction curve applied. I don't see why you couldn't just connect the output of an amplifier to the input in place of the mic and not use a correction curve. That should still indicate the difference between output signal and input signal, though in this case the difference would be caused by the amplifier.

never fooled with it though so I may be way off on that.

Your understanding is correct. Again, depending on how low (THD/noise) you want to measure with accuracy, the computer soundcard/interface will be a limiting factor. REW is a great tool too and it is FREE.

 

[TriodeLuvr]

I think thats just a limitation of the measurement method. It nulls out the fundamental and anything that isn't the fundamental is lumped together into a single figure that is a % of that original signal.

It's a limitation of the older measurement method, which is now obsolete. Using a FFT or sound card, we can restrict the measurement to the harmonics above the fundamental. That removes line noise from the equation, unless we're measuring below the line, such as 20Hz. And of course, the third harmonic of 20Hz is 60Hz, so in the U.S., the resulting measurement may be somewhat incorrect, even if the THD+N function is turned off. Again, this is the shortcoming of the sound card approach.

Jack

 

[THD+N]

On a lot of distortion analyzers (HP, Sound Technology, etc) there are 400Hz and 80kHz filters that are typically selected ON to help control the bandwidth of your audio testing. The 400Hz filter keeps out low frequency junk from the power supply/line frequency and the 80kHz keeps out potential RF junk.

I have FFT on my AP, but still use the low and high filters depending on what I am measuring or looking for.

 

[TriodeLuvr]

The industry should have had a new system by now to report harmonic distortion. I'd like to see nomenclature like HD2, HD3, etc. (denoting the second and third harmonic energy). 1% of the second harmonic sounds vastly different than 1% of the third. Maybe the numbers should also be in dBC instead of percent. The legacy specification of THD (and -sorry- particularly THD+N) is really lacking when it comes to defining equipment performance.

Jack

 

[THD+N]

On DIYAudio, a lot of the die-hard amp guys use PPM (parts per million) for distortion. I don't care for PPM since I have used % for so long. It's just a conversion but still... I prefer Imperial for some stuff and metric for others as well as Fahrenheit vs Celsius.

Multi-tone testing seems to be becoming fashionable in some circles.

 

[AlexHempel]

You can get state of the art results with REW because there are standalone DACs and ADCs capable of better performance than all but the very best audio analyzers. For instance, the E1DA cosmos DAC and ADC, plus all the DACs you can browse on Audio Science Review capable of <120 dB THD+N. The real drawback of REW is the lack of any inherent calibration (because the software knows nothing about the hardware it's plugged into) and the fact standalone DACs and ADCs don't have software switchable attenuators. Also, for testing driver stages, transformers, and insensitive power amps, it's nice to have the higher output amplitude of something like the QA403, although some of the pro-tier DACs have very hot outputs.

 

[AlexHempel]

the industry should have had a new system by now to report harmonic distortion. I'd like to see nomenclature like HD2, HD3, etc. (denoting the second and third harmonic energy). 1% of the second harmonic sounds vastly different than 1% of the third. Maybe the numbers should also be in dBC instead of percent. The legacy specification of THD (and -sorry- particularly THD+N) is really lacking when it comes to defining equipment performance.

Jack

I agree it would be very nice to have a numeric representation, but I do think the common practice now of posting spectrum screenshots and multi-tone tests helps achieve this a bit.

 

[gadget73]

You can get state of the art results with REW because there are standalone DACs and ADCs capable of better performance than all but the very best audio analyzers. For instance, the E1DA cosmos DAC and ADC, plus all the DACs you can browse on Audio Science Review capable of <120 dB THD+N. The real drawback of REW is the lack of any inherent calibration (because the software knows nothing about the hardware it's plugged into) and the fact standalone DACs and ADCs don't have software switchable attenuators. Also, for testing driver stages, transformers, and insensitive power amps, it's nice to have the higher output amplitude of something like the QA403, although some of the pro-tier DACs have very hot outputs.

I don't suppose its got the ability to self-cal by doing a simple loopback connection? Output right back to input, and let it figure out any deviation from flat.

 

[zackthedog]

You can get state of the art results with REW because there are standalone DACs and ADCs capable of better performance than all but the very best audio analyzers. For instance, the E1DA cosmos DAC and ADC, plus all the DACs you can browse on Audio Science Review capable of <120 dB THD+N. The real drawback of REW is the lack of any inherent calibration (because the software knows nothing about the hardware it's plugged into) and the fact standalone DACs and ADCs don't have software switchable attenuators. Also, for testing driver stages, transformers, and insensitive power amps, it's nice to have the higher output amplitude of something like the QA403, although some of the pro-tier DACs have very hot outputs.

I bought a QA403 but haven't had time to set it up yet. Bob Cordell has provided a step-by-step guide and measurement templates.

 

[AlexHempel]

I don't suppose its got the ability to self-cal by doing a simple loopback connection? Output right back to input, and let it figure out any deviation from flat.

Oh yeah, frequency response isn't an issue -- modern DACs and ADCs are flat out of the box. But it doesn't know the amplitude of input and output signals because that depends on the specific hardware implementation of the DAC and ADC. You can set the levels in REW by generating a sine wave (for the DAC) or measuring a sine wave (for the ADC) and telling it what the RMS voltage of the wave is. It's not a serious issue by any means, but another little thing that makes the QA403 a bit more convenient to use.

 

[AlexHempel]

I bought a QA403 but haven't had time to set it up yet. Bob Cordell has provided a step-by-step guide and measurement templates.

You're going to love the QA403 one you learn the few quirks. Here's the biggest one: when you're running the spectrum analyzer, the signal generator output is pulsed, not continuous. Wasted like a day thinking a Williamson was oscillating at LF when it was really just the pulses from the generator. I was expecting a normal sine wave because that's what my DACs generate with the REW signal generator. You can get a nice continuous output from the QA403 to look at on a scope by enabling "idle" and then stopping the spectrum analyzer with the "stop" button.

 

[zackthedog]

You're going to love the QA403 one you learn the few quirks. Here's the biggest one: when you're running the spectrum analyzer, the signal generator output is pulsed, not continuous. Wasted like a day thinking a Williamson was oscillating at LF when it was really just the pulses from the generator. I was expecting a normal sine wave because that's what my DACs generate with the REW signal generator. You can get a nice continuous output from the QA403 to look at on a scope by enabling "idle" and then stopping the spectrum analyzer with the "stop" button.

Thanks for the tip!

 

[TriodeLuvr]

Oh yeah, frequency response isn't an issue -- modern DACs and ADCs are flat out of the box.

But the connecting cables and interface might not be. The ability to calibrate in loopback is a standard feature of the commercial FFTs I've used (Stanford Research). This is a necessary feature if you want to be able to ascertain response to 0.1dB or so.

Jack

 

[trobbins]

Yes REW has loopback calibration that normalises a particular test setup (excluding the DUT) so that added cables and attenuators and probes and amplifiers can be compensated for. But if the original soundcard, or probes etc have significant amplitude and/or phase change out at the frequency extremes then the automated compensation requirement can become large and hence final DUT measurement results can become tatty at frequency extremes.

For me that happens out beyond 60-80kHz with a 100:1 scope probe (I typically only use an unbalanced input setup), so the region up to 96kHz can require some interpretation unless a 10:1 or 1:1 scope probe can be tolerated by the soundcard for the particular application.

A soundcard and REW includes some functions and performance not available from the PicoScope, so I just choose the measurement platform needed to cover my bench testing. No big deal, as each test tool lives in its own self-contained aluminium BBQ tray and gets moved in or out of the test area easily.