Putting THD in the perspective

[Ron Texas]

What? Recess is over?

I don't make the rules around here.

 

[DS23MAN]

Exactly.



True, but luckilly ignore function helps adults to have a normal discussion.

Sometimes it is not wise to ignore someone:

https://www.diyaudio.com/forums/class-d/306815-hypex-fusion-plate-amps-154.html#post5913539

 

[watchnerd]

THD in the reverberation field has little meaning.

And my turntable/tonearm/cartridge sits in the reverb field, too, which adds to this.

 

[Julf]

Just subscribed to Audioxpress, by the way

I feel they have gone downhill lately, pushing a lot of woo to please their advertisers.

 

[SIY]

And that measurement microphone's distortion is part of the unknowns.

For some measurement mics, but generally good ones aren't going to be your limitation. The big issue at any reasonable and meaningful test level and distance is still noise (room, mostly), mic distortion runs a distant third except at really high SPL. That's why I usually use a PCB Piezotronics 376A32 1/2" phantom calibrated omni for 1M and room measurement, but switch to a 376A31 1/4" for Keele-style near-field.

 

[SIY]

I feel they have gone downhill lately, pushing a lot of woo to please their advertisers.

If you could PM me with some examples, I'd be curious to know. When I've criticized woo and the performance of products aimed at the woo crowd and the companies complained, my editor stood by me quite firmly.

 

[Julf]

If you could PM me with some examples, I'd be curious to know. When I've criticized woo and the performance of products aimed at the woo crowd and the companies complained, my editor stood by me quite firmly.

I don't have any recent issues handy, but look at the intro pieces by the editor (recent ones I remember have been pushing MQA and hi-res), the show reports and the vendor profiles.

 

[NTK]

The sweep's rate of frequency change has no bearing on this. What matters is the total duration of the sweep plus the duration of the capture after the sweep ends. Measurement sweeps run low frequency to high, so even very severe low frequency resonances are unlikely to be above the noise floor by the end of the captured content unless measuring a very large space with a very short sweep.

I was referring the time it takes for the physical buildup of the resonance mode in the room, and is not related to the "accuracy" of the measurement method.
I borrowed this figure from this webpage as an example. The resonance frequency is 0.398 Hz. However, if you excite it for 25 seconds and measure (~10 periods) vs 50 seconds (~20 periods) , you'll get different amplitudes because in this case it takes ~100 s to reach steady state. This gives a probable cause for the differences in the resonance peak heights measured with a sweep vs measured at a stationary frequency.

 

[SIY]

I don't have any recent issues handy, but look at the intro pieces by the editor (recent ones I remember have been pushing MQA and hi-res), the show reports and the vendor profiles.

I hope you think the equipment reviews are fair and balanced.

 

[Julf]

I hope you think the equipment reviews are fair and balanced.

Yes, still enjoy them.

 

[Krunok]

So, as a conclusion, before thread waters off, here is a full perspective which includes noise floor, 0.1% and 1% THD threshold limits of my speakers playing at my usual listening level (85dB of average SPL at 4m) and my DAC and amp THD upper limits according to independent measurements.

It is obvious that vast majority of speaker's THD components are going to be burried in the noise while amp and DAC harmonic distortion components are definitely not worth to mention in this context as their distortion levels are below listening level.

 

[JohnPM]

I was referring the time it takes for the physical buildup of the resonance mode in the room, and is not related to the "accuracy" of the measurement method.
I borrowed this figure from this webpage as an example. The resonance frequency is 0.398 Hz. However, if you excite it for 25 seconds and measure (~10 periods) vs 50 seconds (~20 periods) , you'll get different amplitudes because in this case it takes ~100 s to reach steady state. This gives a probable cause for the differences in the resonance peak heights measured with a sweep vs measured at a stationary frequency.

It doesn't. With a sweep the stimulus is the entire sweep. The response to that stimulus needs to be captured, ideally at least to the point it falls below the measurement system's noise floor. In the limit a sweep spends zero time at any specific frequency, but that is irrelevant to its purpose as a stimulus to derive the system's impulse response, which will be accurately obtained in the absence of gross truncation of the response. The impulse response will faithfully capture the time domain behaviour of the system.

 

[NTK]

It doesn't. With a sweep the stimulus is the entire sweep. The response to that stimulus needs to be captured, ideally at least to the point it falls below the measurement system's noise floor. In the limit a sweep spends zero time at any specific frequency, but that is irrelevant to its purpose as a stimulus to derive the system's impulse response, which will be accurately obtained in the absence of gross truncation of the response. The impulse response will faithfully capture the time domain behaviour of the system.

I think you are correct. The mathematics is pretty clear on this. Thanks

 

[Rja4000]

For some measurement mics, but generally good ones aren't going to be your limitation. The big issue at any reasonable and meaningful test level and distance is still noise (room, mostly), mic distortion runs a distant third except at really high SPL. That's why I usually use a PCB Piezotronics 376A32 1/2" phantom calibrated omni for 1M and room measurement, but switch to a 376A31 1/4" for Keele-style near-field.

This kind of info is not easily available.
Thanks a lot for your contribution

 

[Thomas savage]

Sense some ruffled feathers, the conversation seems to of carried on ok..

Carry on being civil... Staring Sid James ...

Cheers

 

[Thomas_A]

Here is one example measured on gated response in-room (Hamming, duration ca 7 ms, 8 averages, measured 1.2 meter from speaker at position, i.e. near damped wall). IMO a decent speaker should give below 0.5 % distortion, high-end speaker below 0.1 % at normal listening levels. My DIY speakers have small and very cheap drivers and still manage reasonably well 200 Hz-10 kHz. (The SPL not known in this case, but volume is about what I usually play as maximum.) Its the small 5 inch woofer that has some problem with 3rd harmonic 800 Hz-3000 Hz (perhaps first break-up modes, the tweeter >3 kHz seems fine). However, it is difficult to get accurate speaker distortion figures without using special and silent rooms.

 

[Rja4000]

I guess the dashed lines are 2nd and 3rd harmonics ?
Could you add the (averaged) noisefloor level ?

 

[Thomas_A]

I guess the dashed lines are 2nd and 3rd harmonics ?
Could you add the (averaged) noisefloor level ?

Yes the fat dash 2nd harmonic, the other thinner is 3rd harmonic. Not sure how to make the noise floor in Fuzzmeasure comparable to REW, but another measurement in REW shows that most except the 800 Hz-3 kHz region is below noise floor.

REW:

Tried to do something similar in Fuzzmeasure, but unsure how to do it Results regarding the 800-3000 region is basically the same dominated by 3rd harmonic, i.e. its above the noise floor. The absolute levels are not the same, and the output levels differ from the softwares (Fuzzmeasure being higher) about what I can say about these simple tries.

 

[Rja4000]

That's an interesting measurement.
How exactly are you doing it?
One sine wave per 1/12th octave?
Then you limit the measurement time to avoid reflections? That's what you mean by "gated"?
If 7ms is the gate time, how do you measure below 140Hz?
And why 1.2m?
Wouldn't that be more accurate by having the mic much closer (1cm or so)?
Of course, you'd need to combine each individual driver data then.

 

[Thomas_A]

That's an interesting measurement.
How exactly are you doing it?
One sine wave per 1/12th octave?
Then you limit the measurement time to avoid reflections? That's what you mean by "gated"?
If 7ms is the gate time, how do you measure below 140Hz?

I use sweeps. Yes, if I would have a long time window and get all of reflections, THD from the speaker would not be easily interpreted. In this case I use Hamming window narrow it down as much as possible to speaker response. It will not be possible to look below 200-300 Hz or so, then I need to measure outside or anechoic. Or near field measurements, driver by driver.