Dummy question: amplifiers measurement

[DonH56]

Couple of quick comments then need to get back to the lab and swear at a SerDes for a while...

1. As has been said, a single sample, while not an ideal infinite-bandwidth impulse, represents the limit for a sampled system and is a common test for stability, frequency and time response, etc. A sampled system uses the zero-order hold math rather than analog impulse but the results are similar and the single-pulse sample tells how the system handles its maximum bandwidth input.
2. Back to the original question, I (and I think most of us including Amir) acknowledge the biggest amplifier performance metrics influencing audible changes are likely to be noise floor (hum and hiss, mainly for high-sensitivity speakers) and output impedance. The latter is IME/IMO most likely to explain audible differences among amplifiers operated below clipping (and above the noise floor, natch). That is the load issue discussed many times in many threads (including this one).

FWIWFM/HTH/IME/IMO/YMMV/my 0.000001 cent (microcent) - Don

 

[françois/ EAR]

Hi, DonH56,
Thank's for your interest and responds.
So, should I understand that noise floor and output impedance are the most important measurements to explain any audible difference between amplifiers? No other measurements required? Not completely conviced for myself to believe that, but I'm in an investigation way... (I said at once I'm not completely "objectivist", I choose audio equipement with both measure reports and Hearing them...)
Her's a thread on a French forum talking about ASR: https://www.homecinema-fr.com/forum...discussions-autour-des-mesures-t30100373.html
Welcome to all to discuss.
(I wrote without any traducter, so I hope I'm in progress in this way!!!)

Regards

 

[DonH56]

Your original question was about TIM, hysteresis (memory effect), and thermal distortion, which I think are below audibility for most amplifiers and systems. There are always exceptions but in my experience too many people tend to think they are the only ones who have exceptional systems.

"Most important" is subjective. Here are some random thoughts that may or may not be shared by anyone else:

• The noise floor you need depends upon your speakers and room. If you have very sensitive horns in a very quiet room you'll probably want higher SNR than if you have low-sensitivity speakers in a noisy room.
• Distortion (THD, IMD) is important if audible. Most of the amplifiers have low enough distortion that you won't hear it in music.
• Power is important; you generally want enough to provide adequate headroom in your system and at your listening levels. Too little and you'll get clipping, harsh sound and excessive tweeter heating. Too much and you may have excessive hiss.
• You need enough bandwidth (frequency response) to cover the audio band plus enough beyond each extreme (LF and HF) to be flat within the audio band.
• Amplifier output impedance is important as it causes speaker response to vary with frequency and volume and can be quite audible.

If you have adequate SNR and low enough (inaudible) distortion over your normal power range, then I think (my guess and perhaps not shared by others) that amplifier output impedance accounts for much of the difference, but test conditions can be much more revealing than normal listening.

There is an almost endless list of design parameters and you can probably find people who think every one of them is "night and day" important.

As an aside, tube amplifiers generally exhibit higher noise, distortion, and output impedance, so it is usually easy to pick out a tube amplifier from a solid-state (SS) amp in listening tests. But even that is hard for well-designed tube amps operated well within their limits.

FWIWFM - Don

 

[solderdude]

So, should I understand that noise floor and output impedance are the most important measurements to explain any audible difference between amplifiers? No other measurements required?

Frequency response, reaction to back EMF, distortion levels, impedance swings, capacitive and inductive currents as well as resistive, output impedance variances with frequency, noise levels... they all matter.

To characterize an amplifier one must look at all measurements, not single out 1 or 2 and take different loads into account.
An amplifier can play fine from certain speakers and be bad driving electrostats.

I would not call this a sound of an amplifier but rather an inability to properly drive a certain load.
In the end it may result to different sound.

Speaker, headphone and line-level pre-amps all differ considerably. One can used less measurements to characterize some devices.

 

[Sal1950]

There are always exceptions but in my experience too many people tend to think they are the only ones who have exceptional systems.

LOL, AMEN to that Don!

 

[BaaM]

Your original question was about TIM, hysteresis (memory effect), and thermal distortion, which I think are below audibility for most amplifiers and systems. There are always exceptions but in my experience too many people tend to think they are the only ones who have exceptional systems.

There's something I don't understand.
A brief reading of ASR will reveal that what is actually measured is also most often below the audibility threshold.
Why not measure everything, everything that is measurable to make a true and totally objective ranking, rather than choosing one parameter over another which will inevitably influence the ranking?
@amirm

 

[DonH56]

There's something I don't understand.
A brief reading of ASR will reveal that what is actually measured is also most often below the audibility threshold.
Why not measure everything, everything that is measurable to make a true and totally objective ranking, rather than choosing one parameter over another?

Ask @amirm but the usual answer is a lack of infinite time and resources. Perhaps the most valuable parts of Amir's work are showing how good many components are whilst flagging the few poorly-performing ones.

 

[amirm]

Ask @amirm but the usual answer is a lack of infinite time and resources.

That's exactly it. It is like asking a friend mechanic to help you survey a bunch of used cars for you, and expecting him to do a full teardown of every engine!

The tests are enough to give us confidence or lack thereof in an audio product. What data is leaves undiscovered, pales in comparison to having a full database of equipment to analyze.

 

[Sal1950]

There's something I don't understand.
A brief reading of ASR will reveal that what is actually measured is also most often below the audibility threshold.
Why not measure everything, everything that is measurable to make a true and totally objective ranking, rather than choosing one parameter over another which will inevitably influence the ranking?
@amirm

What parameters is he missing that you think might be audible?

 

[BaaM]

What parameters is he missing that you think might be audible?

TIM, hysteresis (memory effect), and thermal distortion as stated above.
These may be audible parameters for a few amps, as @DonH56 said.
I understand that amirm has to make a choice to preserve a good balance between quality/quantity though.
But Ive to admit that I would be curious to see some new, more exotic measures, to learn about them, know their audibility thresholds and so on

 

[Blumlein 88]

Amir has already answered, but I'll add my view of his decision to key on SINAD. If a unit has an outstanding SINAD it must have low noise, and low distortion. It also is unlikely (though possible in rare cases) to have messed up frequency response and other issues. So anything topping the SINAD charts has to be pretty darn good.

Now some devices can be low noise, and modestly good on distortion that you'll probably never hear while getting a middling SINAD result. You also could have gear with exceptionally low distortion, but enough noise to compromise the SINAD and probably be audible in some use cases. Yet both devices will have a similar SINAD. You'll need to know more, and Amir does do other tests that would let you see the difference in two such devices.
However for a single number to reliably get the very best designs to come to the top, and allow that to be discernible to a reader without them needing to be an electrical engineer the SINAD isn't a bad choice.

It can be misused like owning a device with 115 db SINAD and feeling you'll need to swap out your DAC when someone makes a 121 db SINAD DAC. The overwhelming odds are you would never be able to hear a difference.

 

[restorer-john]

Amir has already answered, but I'll add my view of his decision to key on SINAD. If a unit has an outstanding SINAD it must have low noise, and low distortion. It also is unlikely (though possible in rare cases) to have messed up frequency response and other issues. So anything topping the SINAD charts has to be pretty darn good.

That's fine and dandy for D/A converters, but for amplifiers (the subject of this thread), it's just one parameter, hardly the most important and certainly not the only one to focus laser-like attention on, especially to the detriment (or complete exclusion or summary dismissal) of, all things an amplifier is required to do. That is, it's sole purpose for existing- the performance into complicated, difficult and varying loads at high powers and all frequencies for a long period of time.

 

[amirm]

In my last amplifier review, I provided step function response (square wave). I asked what people learned from it. No one that asked for it answered.

I can run thousands of tests. If I do, I get asked what they mean. If I can't give an answer people relate to, then it becomes something I won't do.

 

[pma]

In my last amplifier review, I provided step function response (square wave). I asked what people learned from it. No one that asked for it answered.

I can run thousands of tests. If I do, I get asked what they mean. If I can't give an answer people relate to, then it becomes something I won't do.

Sorry I overlooked that and had now hard times to search which measurement you mean. My reply just posted:

https://www.audiosciencereview.com/...22nc-stereo-amplifier-review.9999/post-280025

 

[Blumlein 88]

That's fine and dandy for D/A converters, but for amplifiers (the subject of this thread), it's just one parameter, hardly the most important and certainly not the only one to focus laser-like attention on, especially to the detriment (or complete exclusion or summary dismissal) of, all things an amplifier is required to do. That is, it's sole purpose for existing- the performance into complicated, difficult and varying loads at high powers and all frequencies for a long period of time.

It would be nice if at a minimum amps were tested at 1 khz with a +45 and -45 degree phase angle. I think the power levels and distortion into those loads would separate excellent amps from not so excellent amps. Or maybe it makes more sense to do it at 100 hz.

 

[amirm]

Sorry I overlooked that and had now hard times to search which measurement you mean. My reply just posted:

https://www.audiosciencereview.com/...22nc-stereo-amplifier-review.9999/post-280025

That's a reply but not an answer...

 

[françois/ EAR]

I thank all participants, and between your forum and the one I told you, in France, I will have a lot of reading to perfect my knowledge

 

[Digital Mastering System]

It would be nice if at a minimum amps were tested at 1 khz with a +45 and -45 degree phase angle. I think the power levels and distortion into those loads would separate excellent amps from not so excellent amps. Or maybe it makes more sense to do it at 100 hz.

It seems unlikely a modern power amplifier with a respectable amount of feedback (or shall we more correctly say loop gain) would even faintly notice +/- 45 degree of load variation.
(Although it might be interesting to see how that last Audio Research amp that Amirm tested reacts, given the low feedback.)
You would notice it even less at 100Hz, given typical Miller compensation. There is so much gain here that the amp will shrug off any load phase with ease. What really matters to the amp is load. (More current, lower Z, is your real problem.)
The greater you can make your open loop gain to the highest frequency, the lower your potential distortion. Greater open loop gain and Gain Bandwidth product is the main reason modern opamps are so much better than those of the 80s and 90s.
Similarly, newer bipolar power transistors and VFETs provide ~10X increase in HF power response, in addition to greater beta flatness at high currents.
Anyway I digress - A well designed amp these days covers slew rate because we really want high feedback at full power at 20KHz. It's simple to calculate how much slew rate you need for X volts (out), which can be designed and tested for. Slew rate is a design thing, and should not need to be tested.
As for capacitor sound, DA and such, we all have to study Bateman and follow the following rules:
1. If a capacitor is used for a filter, and has voltage across it in the audible band, use a dielectric that's half decent: Polypropylene, teflon, polystyrene, or NPO. Never any electrolytic. Never. Avoid Mylar (polyester). Never X7R or anything other than NPO (or C0G)
2. Coupling caps if you need them - please try to not need them by designing them out with DC servos - if you need them, then design their cutoff <1/10 below the systems LF cutoff, defined by a good cap somewhere else. These will probably be Aluminum electrolytics. If they are, please make them Bi-polar. As Bateman says, the distortion is much lower if, God forbid, you actually have voltage across them.
3. If you follow the rules above, you don't need to bypass the electrolytics with polypropylene or whatever because they are already effectively out of the circuit ...

Anyway, enough of my blither. Point is: If THD+noise and the IMD are good - it's a pretty good bet everything else is good too.