Effect of 440kHz switching frequency residuals on tweeter distortion?

[pma]

Effect of 440kHz garbage on tweeter distortion?

Following my measurements on Hypex NC400 modules, I was curious if permanent output residuals of class D switching frequency, which I measured 943mVpp at 446kHz, would have any effect on measurable tweeter distortion. Especially taking into account that level of useful audio signal during listening at moderate or low volume would be well below the HF switching frequency residual.

So I prepared a test, where the power amp is driven from a sum of two signals, one comes from the sound card in audio band, the second one from sine generator GAG-810. The level of HF signal was set to get about 960mVpp/440kHz at power amplifier output, and audio signal level was set to get 220mVpp in 1st round and 440mVpp in the 2nd round of tests. The 440kHz HF frequency was switched on and off and measurements taken in both cases. The output was sent to Beyma T2030 tweeter and measured by Behringer ECM8000 microphone, in near field.

Conclusions

No significant effect of 440kHz HF frequency with 940mVpp amplitude was observed in acoustical measurements. Neither in frequency response, nor in distortion plots. My previous suspicion was not confirmed by measurements. Very small differences in harmonics plots are at the limit of repeatability of acoustical measurement.

As I have the setup still connected, please feel free to tell me your suggestions what more you would like to measure re this topic, if you have any.

Images

Measurement setup

Tweeter + microphone

Frequency responses from 5 measurements. Lower, at 220mVpp signal level, HF on/off. Higher, at 440mVpp signal level, HF on/off. HF on/off curves overlap in both cases. Y-axis in dB, but level not calibrated.

Distortion, signal level 440mVpp, HF off.

Distortion, signal level 440mVpp, HF on.

 

[Soniclife]

How much do you have to lower the frequency before it shows up? Some class D use much lower frequencies.

 

[JohnYang1997]

How much do you have to lower the frequency before it shows up? Some class D use much lower frequencies.

This is a great suggestion.

 

[pma]

How much do you have to lower the frequency before it shows up? Some class D use much lower frequencies.

This is a good question, thank you for that. I made a new set of measurements with HF frequency from 200kHz decreasing down to 30kHz. Please see it below. Though higher frequencies seem to change H3 a bit, but leaving overall THD unaffected, 30kHz and 40kHz show clear intermodulations with the audio sine sweep. This is a new set of measurements and my microphone has moved a bit from #1 post, so please the 1st plot in this post (HF = off) as a reference (repeatable - tested).

 

[Soniclife]

Weird how the 3rd harmonic drops lower as the frequency lowers, before it all goes wrong.

So simple conclusion, good class D good, bad class D bad.

 

[pma]

Weird how the 3rd harmonic drops lower as the frequency lowers, before it all goes wrong.

So simple conclusion, good class D good, bad class D bad.

Yeah it is interesting, with the H3 "smoothing". The amp output voltage distortion remains unchanged with HF injection (I checked it for all frequencies used), so it seems something is happening in the driver when additional HF power is added, and it seems to linearize symmetrical distortion. However at inaudible level. Maybe some speaker/driver expert might help us with an answer?

 

[solderdude]

The H2 distortion is (with an almost 100% certainty) caused by the used condenser microphone.

 

[pma]

The H2 distortion is (with an almost 100% certainty) caused by the used condenser microphone.

This is highly questionable. Vout(rms) is only 150mV, SPL on acoustical side is low even in the proximity and I have the microphone measured to have distortion of -70dB and better at that low SPL. Another check is to go more distant, at the expense of higher noise, so the higher harmonics become unmeasurable. The setup chosen is quite optimal and distortion is that of the tweeter. As a conclusion, I do not agree with your statement.

 

[SIY]

Again, my thanks for you showing test setups in detail so that they can be replicated and/or critiqued.

I wouldn't necessarily discount the mike as a source of the 2nd, especially with the nearfield placement. It's a significant mechanical perturbation to the acoustic area. I use a smaller diameter (1/8") mike there, but that might even be insufficient- how do you see the 2HD vary with mike distance?

 

[mjwin]

[...] it seems something is happening in the driver when additional HF power is added, and it seems to linearize symmetrical distortion. However at inaudible level. Maybe some speaker/driver expert might help us with an answer?

I'm not an "expert", but my hunch would be that the increased linearity has something to do with biasing the magnetic core (as is done intentionally with the ultrasonic bias on magnetic audio tape). Magnetic materials can be terribly non-linear and, although in a speaker, they are present in the form of a permanent magnet, the material is still within the magnetic circuit of the voice coil.

Interesting tests though. I think the real problems occur when the sum & difference intermodulation tones approach audio frequencies - as in the early class D amps & switching power supplies which ran at frequencies only just above audibility.

 

[JohnYang1997]

Hmm. If one want to dig out distortion, get a proper piston type. You can get one pretty cheap from China. Less than 200 bucks. Without calibration tho.

 

[keantoken]

Cool. I wonder what the effect would be from using highpassed noise instead of a sine wave?

Also, what about using a triangular pilot tone instead of sine? The reason being a sine wave lingers at peaks whereas a triangle or sawtooth wave is a linear sweep. One step further would be a square wave, which would result in a triangular excursion of the diaphragm. Might help narrow down the source of the distortion.

 

[Vini darko]

If I'm reading it right. More than 100k switching frequency =good enough. Less = problems in the audible band?

 

[SIY]

If I'm reading it right. More than 100k switching frequency =good enough. Less = problems in the audible band?

If one is to believe a measurement using a very artificial stimulus and a toy microphone. But whether or not it's correct, the relevance is questionable- are there any modern Class D amps of any pretension to even middling quality with switching frequencies that low?

 

[Vini darko]

If one is to believe a measurement using a very artificial stimulus and a toy microphone. But whether or not it's correct, the relevance is questionable- are there any modern Class D amps of any pretension to even middling quality with switching frequencies that low?

I'm not aware of any stooping so low in frequency. And your right this test only objectively applies to this setup. Still interesting though.

 

[Vini darko]

Does make me wonder about my old 1990's class D stuff. Especially the 1991 pioneer I'm using for main speakers currently.

 

[keantoken]

Making a low-frequency class D amp to create this effect is the Rube Goldberg way of reducing tweeter distortion. Any home theater device with a microcontroller could be programmed to play a pilot tone. Anyone with a DAC could leave a 48KHz sine playing in the background and do a listening test. Reducing distortion with a pilot tone has already been discussed off and on for decades I believe.

I don't think we need to raise the frequency to avoid overload, just reduce the level. Lower frequency means more diaphragm movement and more core magnetization, one of these is probably the source of distortion reduction. Lower frequencies at lower levels are less likely to be problematic for the electronics, so the pilot tone should probably be >40KHz, high enough not to intermodulate into the audio band but low enough to be effective at a low signal level.

The latest REW has great distortion measurement features that would answer a lot of questions about this effect. Someone could do distortion vs power plots for each pilot frequency. Maybe if you hit a tweeter resonance you would need even less pilot tone to reduce distortion.

 

[Scytales]

I just stumbled across this interesting thread.

Many thanks to PMA for having done these tests. They are most enlightening.

Some thought-inspiring observations made by mjwin and keantoken, also.

 

[AnalogSteph]

It's wild to think (but certainly plausible) that high-frequency AC bias may help linearize the electrical side of speaker drivers as well. Someone should try that with the woofer in an active (DSP) 2-way, I imagine about 20 kHz should easily do in this case so no outlandish sample rates required (but don't let it leak into the tweeter circuitry for obvious reasons, and stay away from breakup modes).

 

[Heinrich]

It's wild to think (but certainly plausible) that ...

I arguably don't get what you mean ... If there's some noise on the line, then this is going to be taken up by the transducer, according to its frequency response. There might be a resonance not shown in the graphs due to cut off at 12kHz. (Why?!)

The microphone may sense the noise up there, at the resonance. By what should the mike discriminate that HF noise from distortion?

It might be as easy as that: the HF noise is plotted as distortion once the HF comes low enough to be converted to sound and also lies in the range of the mike.

All speculations would then be based on a faulty interpretation fueled by cutting off the high frequency part from the display.

Don't worry, I did it myself professionally ...