Can you tell original file from tube amp record? - test

Which file is the original and which do you prefer

  • Apricot is the original file

    Votes: 7 46.7%
  • Avocado is the original file

    Votes: 5 33.3%
  • I prefer Apricot by listening

    Votes: 7 46.7%
  • I prefer Avocado by listening

    Votes: 7 46.7%

  • Total voters
    15
  • Poll closed .
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
AX tech editor
Joined 2002
Paid Member
It is quite easy to make recommendations, right? I understand that the forum is primarily about debates.

Sorry Pavel, no intention to irritate. I just get these ideas building on what I read from you and others, trying to contribute in my feeble way to getting a better understanding.

I no longer have the ears nor the equipment for such tests, but can't stop my brain ;-)

Jan
 
Hi Jan,

my point was that we usually know how to make the test, but it takes a lot of time to prepare usable music samples (to avoid level and time differences etc.) and also to perform the listening tests then (50 trials would be really exhausting).

I wanted to show that even a seemingly huge difference between files as here (original vs. digital chain with inserted tube-SS amplifier with considerable distortion) is not easily audible. If there was a "big sonic difference", people would come with plenty of 16/16 ABX results. As we can see, this is not the case and the audible difference, if any, is just at the edge.

I am doing these tests for the reason that we often read, from highly regarded persons, claims about audibility of resistors (that have nonlinearity like -140dBFS), quantum purifiers, interconnect wires inside the amp, PCB materials etc. These claims are, IMO, in a huge contradiction with real thresholds of human hearing abilities, and I am trying to give an idea of the hearing resolution abilities in a real world.
 
Administrator
Joined 2007
Paid Member
I have voted :)

I listened on the PC with Sony MDR V-7 headphones and a flat response using Foobar and could not statistically pick a preference. I tried several times.

Next I converted the files to high bitrate MP3 so that I could play them on my main system via speakers and using the USB/DAC option on my Marantz SA-CD player.

Now I felt there was a very subtle difference in the extreme lower bass. I felt I preferred Avocado by just listening to it and suspect this could be the original file. That in itself surprised me because normally I seem to go for the more coloured option. I put this down to the fact my main amp does nice things to music and was designed as a 'whole' taking pre and power characteristics into account. So perhaps the file I didn't prefer just pushed things a little to far in that direction.

So that's my thoughts fwiw :)

I shall await to have that theory blown out of the water ;) and also now having felt I could hear a difference via speakers may encourage me to go and do another Foobar test again.
 
What kinds of systems are people using to make their comparisons? If listening in headphones from a computer means using a pair of decent headphones (Grado, Sony MDR-V6, etc.) into the headphones jack socket on a laptop running Windows 10, then there are all sorts of ways the sound could be messed with by Windows (everything resampled to 16 bit/48kHz using whatever processing Microsoft threw in there, for example). Another variable is the type of audio interface you might be using connecting to your computer. For instance, if you're using a cheap 'n cheerful Behringer UCA202 soundcard's headphone output, you'll want to make sure your headphones have a high enough impedance not to be loading down the headphone amp IC (which has Zout of 47 ohms, according to an NwAvGuy review).

I'll be giving this a try with a pair of Sennheiser HD650 headphones (300 ohm) driven from the output of a UCA202 soundcard's headphone out. I have other headphones but they're all lower impedance (32 to 50 ohms), so the UCA202 will color those, especially in the bass (because of high Zout causing lack of damping).
--
 
So I tried doing the ABX test, using foobar ABX Comparator.
After a while, I became so wracked by indecision that I couldn't go on. At least through my listening apparatus for this test (Windows 10 PC > foobar 2000 > Behringer UCA202 > Sennheiser HD650) I could not tell the difference with any kind of certainty.

It will be interesting to find out what level of THD the tube voltage amp was adding to the original signal, and what the harmonic spectrum of that THD was (All 2nd harmonic with nothing higher? Descending levels of 2nd, 3rd, 4th?).

It will be interesting to try this test with a well-recorded excerpt of orchestral music that includes concerted strings, some high pitched percussion, and some good bass content. I can make suggestions for a piece to try, if you like.

Good test.
--
 
AX tech editor
Joined 2002
Paid Member
Hi Jan,

my point was that we usually know how to make the test, but it takes a lot of time to prepare usable music samples (to avoid level and time differences etc.) and also to perform the listening tests then (50 trials would be really exhausting).

I wanted to show that even a seemingly huge difference between files as here (original vs. digital chain with inserted tube-SS amplifier with considerable distortion) is not easily audible. If there was a "big sonic difference", people would come with plenty of 16/16 ABX results. As we can see, this is not the case and the audible difference, if any, is just at the edge.

I am doing these tests for the reason that we often read, from highly regarded persons, claims about audibility of resistors (that have nonlinearity like -140dBFS), quantum purifiers, interconnect wires inside the amp, PCB materials etc. These claims are, IMO, in a huge contradiction with real thresholds of human hearing abilities, and I am trying to give an idea of the hearing resolution abilities in a real world.

Pavel I am in full agreement with your position on audibility.
There is a good reason why audiophools generally hate controlled tests! Nobody wants to be the emperor with no clothes on ;-)

Jan
 
I persevered, and here's what I got:

foo_abx 2.0.5 report
foobar2000 v1.3.10
2018-11-22 10:43:33

File A: apricot.wav
SHA1: 366faaef08134ef83262c95ee8a5ccbdeedb7395
File B: avocado.wav
SHA1: b404fd5526030ace6f56bead6dd1f780cab0a533

Output:
WASAPI (event) : Speakers (3- USB Audio CODEC ), 16-bit
Crossfading: NO

10:43:33 : Test started.
10:51:02 : 01/01
10:52:57 : 01/02
10:54:46 : 01/03
11:02:26 : 02/04
11:05:14 : 03/05
11:24:09 : 04/06
11:24:42 : 05/07
11:25:22 : 06/08
11:26:15 : 06/09
11:26:55 : 07/10
11:27:31 : 07/11
11:28:32 : 07/12
11:30:55 : 07/13
11:44:27 : 08/14
11:46:00 : 09/15
11:47:28 : 10/16
11:47:28 : Test finished.

----------
Total: 10/16
Probability that you were guessing: 22.7%


It looks like I could tell the difference. ??

--
Edit to add:
- uploaded txt file output from ABX Comparator
- I only concentrated on telling the difference between the two files. After an initial listening, I was skeptical that I'd do any better than 8/16. I thought I'd be guessing. So I didn't keep track of which track was apricot vs. avocado. Sorry...
- I can say that one track had slight 'stronger' sounding bass than the other. Ever so very slight difference. I could guess that the 'tube-ified' track has artificially puffed up bass, or I could guess that the high impedance of the tube stage makes the bass a little weaker. So I'd only be guessing which file is which.
--
 

Attachments

  • ABX_result_rongon.txt
    765 bytes · Views: 30
Last edited:
It will be interesting to find out what level of THD the tube voltage amp was adding to the original signal, and what the harmonic spectrum of that THD was (All 2nd harmonic with nothing higher? Descending levels of 2nd, 3rd, 4th?).


--

You can see some measurements in this post
Can you tell original file from tube amp record? - test

THD vs. amplitude plot is up to full amplitude in the test file. Distortion spectrum at full amplitude is now attached. You can see that H4 is below 0.001%, so in fact we have only H2 and H3 that might affect the sound. As the signal amplitude is lowered, higher order components completely disappear and there is only H2 then.

Thank you for the ABX report as well. I think that 10/16 is better than just guessing, if you were able to repeat the result, it would bring even more confidence, as now we have 22% probability of guessing. But I take your result and impressions serious.

You asked what we use for testing. I took a Roland external USB card and Vivanco SR-750 and Sennheiser HD598 headphones in this test.
 

Attachments

  • tubedist.png
    tubedist.png
    75.3 KB · Views: 217
Last edited:
I have voted :)

I listened on the PC with Sony MDR V-7 headphones and a flat response using Foobar and could not statistically pick a preference. I tried several times.

Next I converted the files to high bitrate MP3 so that I could play them on my main system via speakers and using the USB/DAC option on my Marantz SA-CD player.

Now I felt there was a very subtle difference in the extreme lower bass. I felt I preferred Avocado by just listening to it and suspect this could be the original file. That in itself surprised me because normally I seem to go for the more coloured option. I put this down to the fact my main amp does nice things to music and was designed as a 'whole' taking pre and power characteristics into account. So perhaps the file I didn't prefer just pushed things a little to far in that direction.

So that's my thoughts fwiw :)

I shall await to have that theory blown out of the water ;) and also now having felt I could hear a difference via speakers may encourage me to go and do another Foobar test again.

Thank you for giving it a try, Karl. We shall see more after 4th Dec.
 
One of the most interesting measurements on the amplifier under test is a multitone measurement (with wideband multitone). I am attaching the plot where left channel is measured through the amplifier under test and right channel is a wire loopback (i.e. test setup limit). We can see that the distortion components in the amplifier channel produce apparent noise floor elevation of some 30dB. Still, the dynamic range probably remains high enough for this to be inaudible or just at audibility limit.
 

Attachments

  • hybrid_multitone.png
    hybrid_multitone.png
    135.3 KB · Views: 151
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.