• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Measuring the "sound signature" of two different integrated amplifiers.

OP
G

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
4,047
Likes
6,336
If I summarize properly the root cause issue around 4khz, if means that the amplifier needs to be matched with the speaker impedance.
This is very difficult/impossible to verify with the available marketing information.
Imagine what is the user experience with the Onkyo RZ50 on board amps going to safety mode without and warning.
In the old time, on some amplifiers, there was a vu-meter and a Led indicating an overload.
At list nowadays the amplifier manufacturers could put back the overload Led.

We actually do NOT know the root cause. While we can speculate about what I heard and how easy it might be heard, all of the numbers and measurements are done with a resistive load.

The power is <0.2 watts RMS so we weren’t running into overload either.

All I have been able to “prove” is that the differences are real, measurable, and meet the threshold of audibility.

What I thought I heard, described as “attack” for the piano is the lack of compression of about 1 dB into a resistive load.

The difference when compared to the raw digital file does in fact suggest the better amplifier is the more accurate amplifier.
 
OP
G

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
4,047
Likes
6,336
@KSTR figured it out! The left and right channels had been swapped during recording! Presumably this must have been the case when listening too!

The calculations had been done in stereo. He also identified that the phase was inverted between the two -- this makes sense now because the PM-11s2/SA-11s2 use the older standard not the current XLR standards.


The tape loop also seems to have a bigger difference since crossing over the channels did not make a difference.

SA11s2-direct to-PA5 vs. SA11s2 direct to PM11s2
The individual channels drop to indistinguishable levels but the L+R is different

1665318357843.png

1665318096154.png

1665318249675.png


1665318286382.png



And now SA-11s2 -> PM11s2 -> tape out -> PA5 vs SA-11s2 ->PM-11s2
1665318426795.png

1665318462182.png

1665318498756.png

1665318528618.png
 

KSTR

Major Contributor
Joined
Sep 6, 2018
Messages
2,885
Likes
6,455
Location
Berlin, Germany
This is what I get after some pre-processing of the files and using (IME) adequate settings in DW. I've been using the files for System A and System B (so, tape loop being in place), kindly provided by @GXAlan.

- manual trim of start and end alignment in an audio editor
- channel swap for one of the recordings
- polarity inversion for one the recordings
- drift correction enabled in DW (even though the accumulated drift is really low with less than 3ppm)
- 20Hz and 20kHz filters enable on reference and compare, and before and after processing
- non-linear calibration turned off as the and phase differences are not that large and less than optimum settings make result often worse
- there still is a channel level difference of 0.7dB in the left channel
- using L (or R) channels in the compare, not L+R(mono mix) or stereo. Mono mix cannot work because of the level difference and Stereo analyzes only left channels but applies results to both, cannot work either, obviously.


Right channels compare:
1665321762120.png

Note the rather good RMS null.

Other data generated by DW now also is very good match

Left channel is very similar, only slightly worse and from that I would conclude that System A and System B are "close to identical" in measurements and I'm convinced the perceived difference will be absent when comparing apples to apples.. that is, one channel of System A to the other channel of System B, connecting only one (and the same) speaker, and the polarity flip and precision level matching.
 
OP
G

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
4,047
Likes
6,336
The other lesson that is newly understood is that channel imbalance can affect the PK Metric.

When testing individual channels, Deltawave can use different algorithms to match the two contents. But when using L+R, it has to apply the same corrections to both.
 

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,790
Likes
10,568
Location
North-East
The other lesson that is newly understood is that channel imbalance can affect the PK Metric.

When testing individual channels, Deltawave can use different algorithms to match the two contents. But when using L+R, it has to apply the same corrections to both.

L+R is simply the calculated mono signal from left and right channels. It's neither left nor right.
 

tw99

Senior Member
Joined
Mar 13, 2019
Messages
469
Likes
1,077
Location
West Berkshire, UK
So, as an executive summary, it seems like there were some measurement and processing issues, and when these are corrected, actually there is no measurable difference between the two amps?

So the original contention that one amp has more "attack" is now just back to being an unconfirmed impression ?
 
OP
G

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
4,047
Likes
6,336
So, as an executive summary, it seems like there were some measurement and processing issues, and when these are corrected, actually there is no measurable difference between the two amps?

So the original contention that one amp has more "attack" is now just back to being an unconfirmed impression ?

Maybe. We still see -65 dBFS for the setup with the tape loop in the chain. And we did identify phase differences.

1) we are back to confirming that two amplifiers that are in the blue zone can be matched perfectly when running at low power
2) The differences no longer cross confident thresholds for audibility
3) The channel error affects PK metric but doesn’t affect the RMS and peak volume —> so we do know there was a compression effect with the tape loop was on the system.
 

HarmonicTHD

Major Contributor
Joined
Mar 18, 2022
Messages
3,326
Likes
4,854
Maybe. We still see -65 dBFS for the setup with the tape loop in the chain. And we did identify phase differences.

1) we are back to confirming that two amplifiers that are in the blue zone can be matched perfectly when running at low power
2) The differences no longer cross confident thresholds for audibility
3) The channel error affects PK metric but doesn’t affect the RMS and peak volume —> so we do know there was a compression effect with the tape loop was on the system.
Either way. The work you did and the way you handled the discovery of KSTR speaks volumes about your integrity and scientific mindset. I can only applaud and thank you for it. (Others in your position might have defended their results to the bone, you instead kept an open mind and critically reviewed the facts once pointed out by KSTR).

I also guess it is correct to assume that now that the measured „difference do not cross confident thresholds for audibility“ only an elaborate controlled ABX could confirm or not if there is indeed a difference in audibility.
 
Last edited:
OP
G

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
4,047
Likes
6,336
Either way. The work you did and the way you handled the discovery of KSTR speaks volumes about your integrity and scientific mindset. I can only applaud and thank you for it. (Others in your position might have defended their results to the bone, you instead kept an open mind and critically reviewed the facts once pointed out by KSTR).

I also guess it is correct to assume that now that the measured „difference do not cross confident thresholds for audibility“ only an elaborate controlled ABX could confirm or not if there is indeed a difference in audibility.

Well it’s reassuring // it means I did hear a difference, but it was configuration error to show these numbers. :) which was the whole start of this system.

All of the actual listening had the tape loop in the system which still has higher difference numbers but not the easy thresholds.

I cannot say for a fact whether or not I listened in proper stereo or reversed.

Now that we also identified the inverting phase, it may be important to test that because that would correlate with “attack” potentially.
But yeah, this is the value of the scientific process. An error was made, but in the process we learned a lot and know how to better test the next round of comparisons.

Edit: What I don’t know is if the reversal was at the E1DA hookup or if it was between the source and amp. But this pushes the PK
Metric from -50 dB to -65 dB give or take. I would lean toward the E1DA hook up since switching from tape loop to no tape loop required looking at the jacks again.
 
Last edited:

KSTR

Major Contributor
Joined
Sep 6, 2018
Messages
2,885
Likes
6,455
Location
Berlin, Germany
Either way. The work you did and the way you handled the discovery of KSTR speaks volumes about your integrity and scientific mindset. I can only applaud and thank you for it. (Others in your position might have defended their results to the bone, you instead kept an open mind and critically reviewed the facts once pointed out by KSTR)
I was also very pleased to see Alan's swift reaction to my findings, very professional. To note, everyone working in a field of engineering or science knows how easily one can overlook simple issues with setups and procedures despite all efforts of being thorough, it just happens... two steps forward, one step back... or sometimes even full reset.

I may add that it took me several minutes staring on the waveforms in the files (and scratching my head about the origin of those unusual extremely large and non-systematic differences) before I finally had the enlighting moment and saw the correlation to the other channel and vice versa. It's been there from the begining but selective attention kept me from spotting it right away.

I also did not notice the channel swap in the informal listening carried out simultaneously even though it's very obvious once your attention focusses on that.
Our brains can be tricked so easily...
 

coloboxp

Member
Joined
Jul 1, 2023
Messages
9
Likes
6
I was wondering about doing the same test or something similar at least.

The difference is that I was thinking to plug the amplifier's outputs directly to a computer's microphone jack, play an audio file with a constant frequency (test tone) so I can somehow "calibrate" them to output at the same volume and then measuring the audio with 2 or 3 songs, compare them after.

Does that make sense or am I missing something important?, I know that my computer's microphone jack is not of the maximum quality and perhaps even just mono (integrated realtek from last year).

Would it work?... I know its too high level, at least for now.

Thx!
 
Last edited:

JeremyFife

Addicted to Fun and Learning
Joined
Jan 8, 2022
Messages
866
Likes
1,046
Location
Scotland
Absolutely fascinating, thank you @GXAlan and all the other contributors. The twist at the end takes nothing away from how useful and worthwhile this was.

I'm still on a heavy learning curve since joining ASR and this was a timely reminder to avoid complacency (and of the "a little knowledge is a dangerous thing" pit I was staring into). It's easy to read a bit here and take a firm view that good amplifiers are 'wires with gain' and all sound the same ... and to evangelise and (frankly) become condescending (and wrong!)
Science requires an open mind.
 
OP
G

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
4,047
Likes
6,336
I was wondering about doing the same test or something similar at least.

The difference is that I was thinking to plug the amplifier's outputs directly to a computer's microphone jack, play an audio file with a constant frequency (test tone) so I can somehow "calibrate" them to output at the same volume and then measuring the audio with 2 or 3 songs, compare them after.

Does that make sense or am I missing something important?, I know that my computer's microphone jack is not of the maximum quality and perhaps even just mono (integrated realtek from last year).

Would it work?... I know its too high level, at least for now.

Thx!
Conceptually that works. At very low voltages it would work but you can easily blow out your sound card from too much voltage.

Take a look at my Raphaelite 300B SET review. It is going to be essentially to test through a speaker load not just a simulated speaker. There, I show a volume dependent treble gain in a speaker which is not seen in a resistor.
 
OP
G

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
4,047
Likes
6,336
Absolutely fascinating, thank you @GXAlan and all the other contributors. The twist at the end takes nothing away from how useful and worthwhile this was.

I'm still on a heavy learning curve since joining ASR and this was a timely reminder to avoid complacency (and of the "a little knowledge is a dangerous thing" pit I was staring into). It's easy to read a bit here and take a firm view that good amplifiers are 'wires with gain' and all sound the same ... and to evangelise and (frankly) become condescending (and wrong!)
Science requires an open mind.

The 300B is where I am able to show how measurements can show differences in amps which don’t show up when testing under standard resistor loads.

I firmly believe that there was a difference in sound between the PA5 and the PM-11s2, and the difference in level matching hints at that it’s not 100% identical. There was a technical error when I did these measurements, and I returned the PA5 and sold my PM-11s2 for the PM-10.

The PM-11s2 measures better than the PM-10 but my back to back comparisons kept pushing me toward the PM-10. In contrast, I have bought and returned speakers like the Vestia pretty quickly and bought and adopted speakers like the Bose 901 quickly.
I later discovered that the PM-11s2 measures worse into speaker loads than the PM-10 (REW sweep) which again showed me the value of testing through speaker loads.

If you haven’t already, take a look at this post:

It shows why calibrating SPL is difficult. B is the clear loser but it is hard for people to pick out a winner among the rest.
 

JeremyFife

Addicted to Fun and Learning
Joined
Jan 8, 2022
Messages
866
Likes
1,046
Location
Scotland
The 300B is where I am able to show how measurements can show differences in amps which don’t show up when testing under standard resistor loads.

I firmly believe that there was a difference in sound between the PA5 and the PM-11s2, and the difference in level matching hints at that it’s not 100% identical. There was a technical error when I did these measurements, and I returned the PA5 and sold my PM-11s2 for the PM-10.

The PM-11s2 measures better than the PM-10 but my back to back comparisons kept pushing me toward the PM-10. In contrast, I have bought and returned speakers like the Vestia pretty quickly and bought and adopted speakers like the Bose 901 quickly.
I later discovered that the PM-11s2 measures worse into speaker loads than the PM-10 (REW sweep) which again showed me the value of testing through speaker loads.

If you haven’t already, take a look at this post:

It shows why calibrating SPL is difficult. B is the clear loser but it is hard for people to pick out a winner among the rest.
Oh ... just in case there was *any* misunderstanding ... I meant that *I* was in danger of complacency/evangelising/condescension and that *I* needed to remember to keep an open mind. Your post was a great reminder to me of that, as well as being a fascinating read.
I can see that my post could be misread - sincere apologies if you or anyone else was offended!
 
OP
G

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
4,047
Likes
6,336
Oh ... just in case there was *any* misunderstanding ... I meant that *I* was in danger of complacency/evangelising/condescension and that *I* needed to remember to keep an open mind. Your post was a great reminder to me of that, as well as being a fascinating read.
I can see that my post could be misread - sincere apologies if you or anyone else was offended!
No misunderstanding at all. I was just adding to the commentary since this thread is being bumped back up.

It is easy to get into the perspective that measurements are everything, but it’s a lot more nuanced than that. Of course, measurements are also *right* — you just need to know what it is measuring.
 

ahofer

Master Contributor
Forum Donor
Joined
Jun 3, 2019
Messages
5,161
Likes
9,387
Location
New York City
Interesting thread. Can be useful to point out the problems of casual swapping and listening as a comparative method.
 

coloboxp

Member
Joined
Jul 1, 2023
Messages
9
Likes
6
Conceptually that works. At very low voltages it would work but you can easily blow out your sound card from too much voltage.

Take a look at my Raphaelite 300B SET review. It is going to be essentially to test through a speaker load not just a simulated speaker. There, I show a volume dependent treble gain in a speaker which is not seen in a resistor.
Indeed I haven't thought about that voltage issue, it would be vastly higher than coming from a standard microphone.. :confused:

And adding a resistance or so will likely influence on the measurements vs reality? (sorry I'm not electronic-wise guy.. how many ohms suggested?); although, if using it with each amp, the same "influence" would impact them equally.. maybe, yielding yet comparable results.
 

Gringoaudio1

Addicted to Fun and Learning
Forum Donor
Joined
Sep 11, 2019
Messages
619
Likes
837
Location
Calgary Alberta Canada
I suspect some manufacturers do in the service of making better amplifiers. I don’t think reviewers ever do. Speakers are not just complex passive impedances, they are active devices that generate their own voltages that the amp output and its feedback network ”see”.
Back EMF doesn’t make a passive speaker ‘active’. But I get what you’re saying.
 
Top Bottom