Correlating measurements to audibility

[LeftCoastTim]

You may find this link on topic: https://www.ecoustics.com/articles/ten-biggest-lies-audio/

Thanks for this. I love it. This needs more likes.

Audio lies:
1. Cables
2. Tubes
3. Analogue
3a. Hi-Res
4. Sighted listening tests
5. Feedback systems
6. Burn-in
7. Bi-wire
8. Power conditioner
9. CD treatments
10. Golden ears

 

[trl]

Under some circumstances (feel free to call it bad luck) few of the "lies" might be plausible:

1. Cables - really thin cables or inside oxidized contacts might cause a high resistance between amplifier's plugs and speakers. Of course, carefully choose of a regular copper thick-enough speakers cable should take care of this.
3a. Hi-Res - in all of my 16/44 vs. 16/48 measurements the 16/48 won by having a lower spuriae above 10KHz, although moving further to 24/192 didn't improved anything. I personally vote with the 48KHz sampling rate because of that, but I'm aware I can't point out which one sounds better with my ears (I bet they both sound the same in the end).
6. Burn-in - might make a point for headphones and speakers, at least 10-20 hours of burn-in might not be bad.
7. Bi-wire - like stated in item# 1., choosing a really thin cable might be problematic, so by bi-wiring we will cut in half cable resistance between amp and speakers. However, purchasing two pairs of thin cables is usually more expensive that purchasing from the beginning thick-enough regular copper cables.

 

[trl]

Slew rate is not enough. What signal do you feed the amp to measure slew rate? A step? An impulse? How is the other channel being driven when this happens? Not good enough.

I guess measuring the rise and fall of a 20KHz squarewave should be more than enough, although we can visually see the effect with sinewave signals too, because the amplitude will decrease while approaching 20KHz (of course, if the low-pass filter is not very aggressive).

Here https://training.ti.com/system/files/docs/1221 - Slew Rate 1 - slides.pdf TI have a reasonable way of calculating the slew rate in opamps, by calculating the rise-fall difference between 10% and 90% of signal's amplitude, while the formula here https://www.electronics-notes.com/a...ts/operational-amplifier-op-amp/slew-rate.php is calculating the slew-rate from 0% to 100%.

​

Based on the above, an amplifier with a bandwidth of 20KHz and an output swing of 20V RMS/ 56V peak-to-peak (100W/4Ohms) should have a minimum slew rate of 7V/uS. Although, inside an amplifier there are more than one gain stage, each with it's own internal gain, hence internally (per each individual stage) there is a lower slew rate needed (we don't really need 7V/uS opamps inside, 2-3V/uS should do perfectly).

I really don't think slew-rate is an issue anymore these days, but is anyone aware of a contemporary amplifier lacking slew rate?

 

[solderdude]

Thought experiment:
Consider a signal like 18kHz sine wave at -1dB which is alternated by an 18kHz sinewave at -60dB.
So a sudden change in level (which is dynamics)

Put this signal through an opamp and it will come out looking closely the same as the original waveform.
Run it through a minimum phase bandwidth limited system (44.1kHz) and it won't look like the original.

Run it through a power amp with a resistive load and it will also look closely the same as the original waveform
Run it through a power amp connected to an actual speaker and the waveform will not look the same (back EMF, reactance, capacitance).

This despite the slewrate of used components being well above the minimum required for 20kHz at FS output.

The above are just thought experiments with (by me) expected results.
How audible this all will be remains to be seen.
One can play with amplitude differences and frequencies.

 

[jsrtheta]

Thanks for this. I love it. This needs more likes.

Audio lies:
1. Cables
2. Tubes
3. Analogue
3a. Hi-Res
4. Sighted listening tests
5. Feedback systems
6. Burn-in
7. Bi-wire
8. Power conditioner
9. CD treatments
10. Golden ears

Yep. Still true 20 years later.

 

[Wes]

For amps, I'd like to see how they measure driving difficult loads, e.g. speakers with dips to low impedance.

For pre-amps, it should be a lot easier.

But, in practice, the speakers are what really matters and are something easily controlled by the consumer (unlike the recording).

Maybe a modern system should have 90% of the cost put into speakers? (active ones, with DSP ---> 2 models so far...)

 

[pma]

I guess measuring the rise and fall of a 20KHz squarewave should be more than enough,

Makes no difference if square wave repetition frequency is 20kHz or 200Hz, the only thing of any importance is how you set the oscilloscope time base (like 2us/div) and trigger source and type. Of course generator with short edges is supposed as a necessary condition. Rise time of the amplifier step response is same regardless square repetition frequency. With the 20kHz square, there is a big chance you will burn a resistor in the output Zobel R-C circuit, if you test a power amplifier at higher output voltage swing.

 

[trl]

Maybe a modern system should have 90% of the cost put into speakers? (active ones, with DSP ---> 2 models so far...)

Speakers are like monitors, while amps are like PCs. On a poor screen with measurable geometrical distortions and a bad colorimetry you can't possibly do any CAD or digital imaging work. Same applies to speakers as well, at least if we really want to compare the sound that audio engineer recorded in the studio with the sound our ears are hearing in our homes.