Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Amplifier sound quality test - method and samples (Read 15762 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Amplifier sound quality test - method and samples

I present a method for testing amplifiers, which let you compare original sound samples to different amplifiers by comparing test sample files.
This allows for doing the listening evaluation on any equipment, anytime, anywhere, and you can use abx to verify.

I am not a regular reader on Hydrogenaudio, so please forgive my ignorance if something very similar to this has already been discussed.
I just wanted to present this to an audience that are familiar with and have an actual interest in listening test evaluation, and also tries to have a scientific approach and the knowledge for it.

I had hoped to be able to verify sonic differences in power amplifiers, the intention was not to present yet another "all sounds the same".
So far, I have failed to find any audible differences between original sound samples and amplifiers, when playing music signals.
I found audible differences between original and also between amplifiers using multitone test signals.
But that's not what we listen to, is it?

I describe the method, what I tested for and my findings and conclusions so far here:

Article:
http://kvalsvoll.blogspot.no/2015/03/testi...nd-quality.html


Test sample files:
http://www.kvalsvoll.com/Articles/abx/


I have only posted this on data-bass and a local hifi-forum.
I expected some interest on the hifi-forum, considering the energy they put into discussions about cables and other snakeoil, but there is little or no interest.
Actually I don't think they understand the test method and the implications of the results.

So it is no surprise that so far:
- The method has not been challenged
- No reports received from positive tests that have been able to verify differences

Note the 5x loop through the instrumentation test - since the loop contain a consumer av-receiver, and this indicates that it is completely audibly transparent, many audiophiles could save a lot of money and potentially get improved sound at the same time.
Just saying, though I suspect very few are listening.

It would be interesting to see what you make out of the sample files - can you verify differences that I myself fail to notice, do you see obvious flaws or shortcomings to the method?

And if the test sample files can entertain some of you through the weekend, I personally find this experiment a success regardless of what conclusions comes out of it.

Amplifier sound quality test - method and samples

Reply #1 –
Were the amplifiers loaded (i.e. had speakers connected), or just connected to a line-level recorder?

Cheers,
David.

Amplifier sound quality test - method and samples

Reply #2 –
Yes, I'm not quite clear how one would hear anything other than a transduced version of the electrical output
Loudspeaker manufacturer

Amplifier sound quality test - method and samples

Reply #3 –
Quote from: 2Bdecided on
Were the amplifiers loaded (i.e. had speakers connected), or just connected to a line-level recorder?


Good point.

In two instances of people who tried this sort of thing before they loaded the UUT's with carefully engineered dummy loads that simulated loudspeaker loads reasonably well.

Swedish Amplifier Tests  (not really DBTs I don't think.)

PCABX power amp tests on the wayback machine

Ethan Weiner has similar files related to audio interfaces: http://ethanwiner.com/aes/

Amplifier sound quality test - method and samples

Reply #4 –
Quote from: 2Bdecided on
Were the amplifiers loaded (i.e. had speakers connected), or just connected to a line-level recorder?

Cheers,
David.


All music samples were recorded with real loudspeaker load, at decent output signal level.

Level was adjusted to max peak (0dBFS digital from the source material) -3dB below clipping of the small amplifier.
In this room on those speakers that equals -6dB on the master volume, where 0dB is calibrated to 85dB (standard cinema reference).

Multitone tests are awful to listen to, so I just recorded them with a 8ohm resistive load.

Amplifier sound quality test - method and samples

Reply #5 –
Some flac files are corrupted.
"I hear it when I see it."

Amplifier sound quality test - method and samples

Reply #6 –
I see my article is a bit short of information, as it says nothing about output level or loudspeaker load (see my reply above for info).

I tried to present this as simple as possible, and seemingly some important details got lost in the process.

Amplifier sound quality test - method and samples

Reply #7 –
Quote from: xnor on
Some flac files are corrupted.


To be sure I went through the list of files and downloaded them, they all seem fine.

Maybe you got a partial download, or the files simply don't play on all players.

Amplifier sound quality test - method and samples

Reply #8 –
What device was used to bring the speaker level signal down to line level to record it? A pad? A voltage divider?

Last I looked into them, the ones I would have deemed "affordable" (in my book) had some minor frequency response errors which looked to be potentially audible (not that I bought any or tried it),  which might be dependent on the differing output impedance of these various amplifiers being tested.


Amplifier sound quality test - method and samples

Reply #9 –
Quote from: xnor on
Some flac files are corrupted.


When I try to load them into Cool Edit Pro 2.1 with the FLAC filter, the file open locks up at the end, and I have to close the program with the Task manager.

Happens with about half the files.

Never happened to me before.

Amplifier sound quality test - method and samples

Reply #10 –
Quote from: Kvalsvoll on
To be sure I went through the list of files and downloaded them, they all seem fine.

Maybe you got a partial download, or the files simply don't play on all players.


foo_verifier plugin reports:
Code: [Select]
Item: "D:\temp\I_Can_Feel_It_Humming_amp1.flac"
MD5: 7AA6272CD1C3774EA1CF03823B7DA007
CRC32: 08EBE15F
Warning: Reported length is inaccurate : 1:19.400000 vs 1:19.307120 decoded
Error: Corrupted FLAC stream
Error: MD5 mismatch


1 item could not be correctly decoded.

List of undecodable items:
"D:\temp\I_Can_Feel_It_Humming_amp1.flac"


The file itself has
File size : 13.7 MB (14 468 253 bytes)
MD5: f7e92217d4649596d243abb6abe31f2d
SHA-1: 6d20213dbcb6927a65fe8b22840d19fecb71c121
"I hear it when I see it."

Amplifier sound quality test - method and samples

Reply #11 –
Thanks for the info about the files, xnor and Arnold B. Krueger, with this information I will be able to fix it.
I will post here when I have found the problem, and updated the files.

Until then, we will have to do with what we have, the files should still play just fine, at least they do on foobar and xbmc here.

Amplifier sound quality test - method and samples

Reply #12 –
Quote from: mzil on
What device was used to bring the speaker level signal down to line level to record it? A pad? A voltage divider?

Last I looked into them, the ones I would have deemed "affordable" (in my book) had some minor frequency response errors which looked to be potentially audible (not that I bought any or tried it),  which might be dependent on the differing output impedance of these various amplifiers being tested.


Level attenuation is considered a part of the recorder, and is very important for getting good results.
The one I use here is designed specially for this purpose, it is balanced, it tries to avoid adding noise, and attenuation is chosen so that the level is poperly gain matched to the recording device to avoid clipping and noise.

Before I record a sample, I monitor frequency response and noise.
It is very easy to make a mistake, such as having a faulty setting on the player device, and to prevent that, the best way is to actually verify performance before each recording is done.

For the record, the loop is flat through the audio range (falls off below around 10hz, which makes it easy to identify amplifier samples using a spectrum analysis), and noise levels are low enough to distinguish clearly between 16bit and 24bit samples.

WARNING to anyone trying this by making a simple attenuator - the amplifier output terminals may be live, even the one marked "-" may be live, and if you connect this to ground you can destroy the amplifier.
 

Amplifier sound quality test - method and samples

Reply #13 –
Sample files that reported faulty are now fixed and uploaded.

Never occured to me that it should be necessary to verify flac-encoding of a file.

Amplifier sound quality test - method and samples

Reply #14 –
You mean a bridged amplifier.
Could you draw up a quick schematic that shows one amplifier channel, the speakers/load, your attenuator and A/D converter?

edit:
Quote from: Kvalsvoll on
Sample files that reported faulty are now fixed and uploaded.

Never occured to me that it should be necessary to verify flac-encoding of a file.

Thanks. I noticed it because Adobe Audition fails to open the file with a decoding error.
"I hear it when I see it."

Amplifier sound quality test - method and samples

Reply #15 –
Well, after a quick look at the im2 and im3 test files there indeed does not seem to be a large difference. I wonder if the similar levels of intermodulation distortion is coincidence, or a "problem" in your measurement setup.

Besides the original input files, do you also have recorded these files without an amplifier? So we could see the baseline performance of the recording device.
"I hear it when I see it."

Amplifier sound quality test - method and samples

Reply #16 –
I hope the files are fine now, and for the more detailed description of the instrumentation, it is quite simple:

Original sample file -> Playback device (Computer+DAC+pre) -> Power amplifier under test -> Loudspeaker and attenuator

attenuator -> Recorder with ADC -> Computer -> sample file


The complete playback device is a computer running Kodi(XBMC) into a commercial AV-receiver.
I suspected this playback would be sufficient, i.e. audibly transparent, signal processing analysis indicates so.
The only thing I was worried about was noise, and if this should prove to be audible in the loop, a better DAC+pre with lower noise and less distortion could be used.
However, for convenience, this is how I started out.

The recorder consist of an attenuator, AD-converter, and computer.

The whole instrumentation loop is tested in the tuttabella_test sample files, and the 5x sample is a 5-times loop-through done by playing back the previously recorded samle, then playing it again and repeat.
The attenuator is not part of this test loop, but all faults from this is easy to verify because it will be mainly noise induced due to the reduction in source impedance from 0 (amplifier output) to around 200ohm, the attenuator is simply a passive resistor network.


The attenuator is just connected to the speaker terminals, you can connect it to the terminals on the amplifier or on the speaker, just make sure there are no significant hf losses through the speaker wires due to capacitance and inductance in the wire if connecting to the speaker.

A bridged amplifier is one of the configurations only waiting for the disaster when you ground one of the live output terminals.

Anyway, all discussion about potential sound quality issues in the instrumentation loop - computers, dac/adc, attenuator - is moot if you can not hear the difference between the original and the amplifier test samples.

Amplifier sound quality test - method and samples

Reply #17 –
Quote from: xnor on
Well, after a quick look at the im2 and im3 test files there indeed does not seem to be a large difference. I wonder if the similar levels of intermodulation distortion is coincidence, or a "problem" in your measurement setup.

Besides the original input files, do you also have recorded these files without an amplifier? So we could see the baseline performance of the recording device.


I did not record the im2, im3 multitone tests through the loop without amplifier.
Maybe I should do that.

If you do a spectrum analysis of the samples you will see differences between the amplifiers, and it should also be possible to hear differences, provided you abuse your ears a little - you need quite loud levels.
I also noticed the differences are not equal on different playback systems - an indication that the playback amplifier and headphones/speakers actually mask some of the differences one is looking for.

Amplifier sound quality test - method and samples

Reply #18 –
Yes, the files I've tried were fine so far.

Quote
The whole instrumentation loop is tested in the tuttabella_test sample files

So in that test loop there is no power amp or load? That would be the baseline performance of your whole measurement chain.
Could you repeat just that one test with the 19+20 kHz tones? The reason I'm asking is because both the playback and recording devices add noise and distortion and it would be easy to see in a spectrum of just 2 tones.


Quote
The attenuator is just connected to the speaker terminals

So it is connected in parallel to the load. What is it, something like that: http://en.wikipedia.org/wiki/Attenuator_%2...nuator_circuits? Or a simple voltage divider?


Quote
Anyway, all discussion about potential sound quality issues in the instrumentation loop - computers, dac/adc, attenuator - is moot if you can not hear the difference between the original and the amplifier test samples.

That's right. Still haven't had the time to do a critical listen.
"I hear it when I see it."

Amplifier sound quality test - method and samples

Reply #19 –
Quote from: xnor on
Yes, the files I've tried were fine so far.

Quote
The whole instrumentation loop is tested in the tuttabella_test sample files

So in that test loop there is no power amp or load? That would be the baseline performance of your whole measurement chain.
Could you repeat just that one test with the 19+20 kHz tones? The reason I'm asking is because both the playback and recording devices add noise and distortion and it would be easy to see in a spectrum of just 2 tones.


Quote
The attenuator is just connected to the speaker terminals

So it is connected in parallel to the load. What is it, something like that: http://en.wikipedia.org/wiki/Attenuator_%2...nuator_circuits? Or a simple voltage divider?


Quote
Anyway, all discussion about potential sound quality issues in the instrumentation loop - computers, dac/adc, attenuator - is moot if you can not hear the difference between the original and the amplifier test samples.

That's right. Still haven't had the time to do a critical listen.


And that is why I decided to post here - maybe someone with more experience in abx testing could reveal the differences, and tell how to hear it.
Maybe we need new samples with different music, maybe we should try with a 4x looped amplifier test.

I can do a sample of the test loop with 19k+20k tones.

Yes, the attenuator is a simple divider, it does not add a significant load on the amplifier.

Amplifier sound quality test - method and samples

Reply #20 –
Quote from: Kvalsvoll on
Sample files that reported faulty are now fixed and uploaded.

Never occured to me that it should be necessary to verify flac-encoding of a file.


It is tech, everything can go wrong, even the simplest thing. Murphy's law. ;-)

The samples seem OK for me now.

I'm also missing files for just the test setup - IOW just rerecord the output of the DAC.

BTW  funny story - on the IM2 files I could clearly hear the 1 KHz spurious response, even with the .org file that obviously had none of any significance. The problem was with my monitoring system - it appears its the headphones.

Another test file to try to re-record is this one: http://www.hydrogenaud.io/forums/index.php...st&p=893070. It is for making very quick FR tests.

I just updated that post to include a bucket-aligned multitone.

Two hints about testing your rig with test tones - one approach is to use the Audio Rightmark program, and another is to built your own test file with multiple signals in it. a 1 KHz tone, some simple multitones, a swish, etc.

Amplifier sound quality test - method and samples

Reply #21 –
Quote from: Kvalsvoll on
Yes, the attenuator is a simple divider, it does not add a significant load on the amplifier.


My favorite attenuator is a 5 K ohm 2 watt (RV4) potentiometer built into the dummy load, whether resistive or reactive. With reasonable length cables its response is flat enough.

ebay link

Jameco link

Since my measurement rig (M-Audio AP 24192) has balanced inputs, I also use standard balanced mic attenuators:

Audio Technica Balanced Attenuator

One of the  problems with using sound cards for test equipment is that they aren't really sensitive enough to use with portable audio gear. I'm experimenting with using a modified mic preamp for that purpose and having some success.  So far I haven't found a mic preamp with spurious responses > 100 dB down, despite numerous spec sheets claiming otherwise, but 90 dB is good enough for most work.

Amplifier sound quality test - method and samples

Reply #22 –
Kvalsvoll, one thing you might want to do is match the volume of the recorded files using something like the ReplayGain scanner in foobar2000. Otherwise you might hear volume differences in an ABX test.
"I hear it when I see it."

Amplifier sound quality test - method and samples

Reply #23 –
Quote from: xnor on
Kvalsvoll, one thing you might want to do is match the volume of the recorded files using something like the ReplayGain scanner in foobar2000. Otherwise you might hear volume differences in an ABX test.


As long as he is dealing with good amps which seems to be the case from my limited investigations, matching levels by simple means such as putting test tones into the files, or just looking at average levels after time-matching will suffice.

While I don't see the details of his tests, they might for example provide counter-evidence to the audiophile misapprehensions about using amps that are not vastly overpowered for the application.

If he were working real junk equipment, or with analog media - vinyl or mag tape then the more complex means mentioned above might be warranted.

Amplifier sound quality test - method and samples

Reply #24 –
While we are talking about attenuators, attenuators that are used to tap down the outputs of power amps need to be able to handle power and this impacts the choice of alternatives.

The power paramers are set by the maximum voltage that the attenuator might see - probably on the order of 50 volts RMS unless one is going after really power amps, and the need for relatively low impedances to preserve high frequency response.  I've found that attenuators based on a 5 K potentiometer can work out well within these limits.

A 5 K potentiometer with 50 volts RMS across it has 10 milliamps flowing through it for a total dissipation of 500 milliwatts or 0.5 watts. 

Using a 2 watt linear pot has not been a problem, but it needs to be remembered that resistor power ratings are based on relatively high operating temperatures on the order of 200 degrees C.

If the parts do not have low temperature coefficients you can end up with a setup that is changing its parameters on its own while you use it.  Metal film resistors with low temperature coefficients are a really good idea for this application.

This immediately eliminates one class of potentially attractive options being  Alps RK27 and RK271 style conductive plastic attenuators that seem to be typically rated at 10 milliwatts.

This probably also eliminates rotary switch based RK27 equivalents that are based on SMT resistors.

The next step up power wise are the rotary switch based attenuators that are based on 0.1 watt leaded resistors. The lower valued resistors should be no problem, but the resistors for the upper few steps may need to be upgraded.

Logarithmic potentiometers need to be looked at carefully because their power dissipation is not uniformly distributed over the part, but tends to concentrate in the upper ranges.