• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

ABX distortion test on a real-world tube preamplifier, about 2.7% THD

pma

Major Contributor
Forum Donor
Joined
Feb 23, 2019
Messages
2,289
Likes
4,057
Location
Prague
After a short rest, I have a new listening test. Now, the original file is tested against the file that was played and recorded through the real tube preamplifier and of course there was a DAC/ADC path to record the file. The link to test files is

https://www.dropbox.com/s/1nji8wcczathq3b/Quintet_tubetest.zip?dl=0

If you would like to participate in the test, please download the test file (you do not need to have a dropbox account. Click somewhere outside log-in window and then click to download the quintet_test.zip file. Do no try to play the files directly from dropbox, it is impossible.)

In case you want to participate, I would ask you to make an ABX test in Foobar 2000

http://www.foobar2000.org/components/view/foo_abx

and then to post the full protocol of the test, which we are able to verify.

Test description
The output of the DAC was amplified by the tube preamp to some 11Vrms at full scale. Then it was divided by a stepped attenuator to the same amplitude as was the DAC output. This attenuated signal was connected to ADC input. The original file and the recorded file were time aligned as the last step.

tubepre_test_THD_s.png

Distortion of the tube preamp close to full scale of the ADC


tubetest_setup.JPG

Photo of the test setup

However, as the test file peak amplitude is about -8dBFS, the distortion near peaks is lower than at FS, and is about 1%
tubepre_test_THD_-9dB_s.png

THD and THD+N at -8.7dBFS, which is close to peaks in the music sample

The tube preamp used is the SRPP+ by Glassware and it was described here earlier
https://www.audiosciencereview.com/...ube-preamp-kit-review-and-measurements.15883/

Again, have fun listening the samples!
 
Last edited:
OP
pma

pma

Major Contributor
Forum Donor
Joined
Feb 23, 2019
Messages
2,289
Likes
4,057
Location
Prague
My result - no chance

Code:
foo_abx 2.0.2 report
foobar2000 v1.4.8
2020-12-14 15:46:28

File A: Quintet3.flac
SHA1: 5e123ca2ae69c238cc9bc38a7bbd8f258c3a329a
File B: Quintet4.flac
SHA1: 8d68ff31f23796ffca3b769ac147a247b0de3b37

Output:
ASIO : Creative Sound Blaster ASIO
Crossfading: NO

15:46:28 : Test started.
15:47:38 : 01/01
15:47:52 : 01/02
15:48:13 : 02/03
15:48:27 : 02/04
15:48:40 : 03/05
15:49:00 : 04/06
15:49:20 : 04/07
15:49:40 : 04/08
15:49:54 : 04/09
15:50:13 : 04/10
15:50:13 : Test finished.

 ----------
Total: 4/10
Probability that you were guessing: 82.8%

 -- signature --
042d1be6cca132e4845254e628fb7134a5b4b763
 

flipflop

Addicted to Fun and Learning
Joined
Feb 22, 2018
Messages
806
Likes
1,051
foo_abx 2.0.6c report
foobar2000 v1.3.16
2020-12-14 15:59:06

File A: Quintet3.flac
SHA1: 5e123ca2ae69c238cc9bc38a7bbd8f258c3a329a
File B: Quintet4.flac
SHA1: 8d68ff31f23796ffca3b769ac147a247b0de3b37

Output:
DS : Primær lyddriver
Crossfading: NO

15:59:06 : Test started.
16:03:05 : 00/01
16:03:28 : 00/02
16:03:39 : 00/03
16:04:40 : 01/04
16:05:08 : 01/05
16:05:21 : 02/06
16:05:51 : 03/07
16:06:16 : 03/08
16:06:35 : 04/09
16:06:48 : 04/10
16:07:22 : 04/11
16:07:33 : 04/12
16:07:33 : Test finished.

----------
Total: 4/12
p-value: 0.927 (92.7%)

-- signature --
40054c681b286eb3158dcd25581134b658a73e9b
It's a very busy track. Too many instruments playing at the same time. There's 2-3 solo piano notes at the start of it, but piano is no good for revealing harmonic distortion.
 
OP
pma

pma

Major Contributor
Forum Donor
Joined
Feb 23, 2019
Messages
2,289
Likes
4,057
Location
Prague
foo_abx 2.0.6c report
foobar2000 v1.3.16
2020-12-14 15:59:06

File A: Quintet3.flac
SHA1: 5e123ca2ae69c238cc9bc38a7bbd8f258c3a329a
File B: Quintet4.flac
SHA1: 8d68ff31f23796ffca3b769ac147a247b0de3b37

Output:
DS : Primær lyddriver
Crossfading: NO

15:59:06 : Test started.
16:03:05 : 00/01
16:03:28 : 00/02
16:03:39 : 00/03
16:04:40 : 01/04
16:05:08 : 01/05
16:05:21 : 02/06
16:05:51 : 03/07
16:06:16 : 03/08
16:06:35 : 04/09
16:06:48 : 04/10
16:07:22 : 04/11
16:07:33 : 04/12
16:07:33 : Test finished.

----------
Total: 4/12
p-value: 0.927 (92.7%)

-- signature --
40054c681b286eb3158dcd25581134b658a73e9b
It's a very busy track. Too many instruments playing at the same time. There's 2-3 solo piano notes at the start of it, but piano is no good for revealing harmonic distortion.

Thank you for participating!

There are not too many instruments. It is a small quintet. We need more than one instruments to hear intermodulations. Like in the previous test

https://www.audiosciencereview.com/...tortion-test-about-5-on-a-music-sample.18217/

where many people were successful. It is all about distortion audibility thresholds. This test has lower distortion, though it is "real distortion" and not mathematically added, like in the previous test. However, the tube creates almost ideal polynomial distortion which is independent on frequency.

My goal is to show that the race for lowest SINAD (THD+N) numbers is pointless. It is 100x more important to get the good S/N after all the components of the audio chain have been interconnected. Test the complete chains rather than isolated components.

tubetest_thdfreq_10V_s.png

THD vs. frequency of the tube pre under test, at 10V output
 
Last edited:

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
10,095
Likes
22,070
Location
The Neverlands
I remember the good old days of cassette decks. It really made one aware of noise levels. Who cared about distortion... 1% was already great.
Depends a bit on the type of distortion though.
Some folks even preferred to hear 'pumping' of noise reduction over continuous tape hiss.

With the current high efficiency ear phones S/N ratio is yet again more important than distortion or tons of output power.

Fortunately ASR measurements show us S/N ratio, even at 2 'reference levels', as well as distortion (various types), FR and whatnot so each reader can look at what they find important.

It is 100x more important to get the good S/N
100x = 40dB :)
 
Last edited:

Veri

Master Contributor
Joined
Feb 6, 2018
Messages
8,042
Likes
9,796
My result - no chance

Code:
foo_abx 2.0.2 report
foobar2000 v1.4.8
2020-12-14 15:46:28

File A: Quintet3.flac
SHA1: 5e123ca2ae69c238cc9bc38a7bbd8f258c3a329a
File B: Quintet4.flac
SHA1: 8d68ff31f23796ffca3b769ac147a247b0de3b37

Output:
ASIO : Creative Sound Blaster ASIO
Crossfading: NO

15:46:28 : Test started.
15:47:38 : 01/01
15:47:52 : 01/02
15:48:13 : 02/03
15:48:27 : 02/04
15:48:40 : 03/05
15:49:00 : 04/06
15:49:20 : 04/07
15:49:40 : 04/08
15:49:54 : 04/09
15:50:13 : 04/10
15:50:13 : Test finished.

----------
Total: 4/10
Probability that you were guessing: 82.8%

-- signature --
042d1be6cca132e4845254e628fb7134a5b4b763
Hey, at least I scored 1 more ;) ha. Was doing pretty good up to the 4/5...
Code:
foo_abx 2.0.6d report
foobar2000 v1.4
2020-12-14 21:09:55

File A: Quintet3.flac
SHA1: 5e123ca2ae69c238cc9bc38a7bbd8f258c3a329a
File B: Quintet4.flac
SHA1: 8d68ff31f23796ffca3b769ac147a247b0de3b37

Output:
DS : Primary Sound Driver
Crossfading: NO

21:09:55 : Test started.
21:10:31 : 01/01
21:11:24 : 01/02
21:11:46 : 02/03
21:12:15 : 03/04
21:12:32 : 04/05
21:12:53 : 04/06
21:13:23 : 05/07
21:13:31 : 05/08
21:13:43 : 05/09
21:14:43 : 05/10
21:14:43 : Test finished.

----------
Total: 5/10
p-value: 0.623 (62.3%)

-- signature --
5aeb949c089e337338b1395e36ff5fb4b9b797b5
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
12,814
Likes
19,468
My results. Absolutely perfect.

Perfect for someone guessing.
foo_abx 2.0.6d report
foobar2000 v1.5.3
2020-12-14 22:36:02

File A: Quintet3.flac
SHA1: 5e123ca2ae69c238cc9bc38a7bbd8f258c3a329a
File B: Quintet4.flac
SHA1: 8d68ff31f23796ffca3b769ac147a247b0de3b37

Output:
DS : marantz-AVR (Intel(R) Display Audio)
Crossfading: NO

22:36:02 : Test started.
22:39:04 : 01/01
22:39:50 : 01/02
22:40:07 : 02/03
22:40:33 : 03/04
22:40:59 : 04/05
22:41:53 : 05/06
22:42:45 : 05/07
22:43:01 : 05/08
22:43:25 : 06/09
22:43:56 : 06/10
22:44:14 : 07/11
22:44:38 : 08/12
22:45:08 : 08/13
22:45:19 : 08/14
22:45:34 : 08/15
22:46:02 : 08/16
22:46:02 : Test finished.

----------
Total: 8/16
p-value: 0.5982 (59.82%)

-- signature --
428fa9c2ddd82ec7619bcae3da873934d6df9abb
 
OP
pma

pma

Major Contributor
Forum Donor
Joined
Feb 23, 2019
Messages
2,289
Likes
4,057
Location
Prague
Thanks @Veri and @Blumlein 88 for trying the test. I think we may say that 1 - 2% THD is inaudible on this music sample. THD+N is the same and converted to the popular SINAD it tells 34 - 40dB.

We also spoke about S/N. Measured S/N here is 110dBA to FS 11V. The peaks here are about -8dBFS, still 102dBA. This tells how important it is to make a difference between the 2 components of that one parameter.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
12,814
Likes
19,468
You didn't have to convince me. I've never liked THD+N or SiNAD. You need both bits of info separately. I think SFDR (spurious free dynamic range) is actually better.

I understand why Amir uses it like he does. If something has great SINAD numbers it is both low distortion and low noise and has good design for both. But it can give the wrong idea about whether a given device really sounds better than another device. You have to explain a bit more to someone who doesn't know this stuff and it looks more complicated.

I could have something with THD+N of 1% and it may never be audible on music if the one percent is all 2nd and 3rd harmonic with a low noise floor. Or I could have something with THD+N of 1% which has little distortion but lots of noise or maybe 60 hz hum. Such a device rather than being audibly transparent for music is in fact unusable for any hifi purposes. It could be a complete non-starter. Anytime a spec for device A can be great hifi for music, and device B with the same spec is uselessly broken then you don't have a good spec. The saving grace with THD+N is you won't get good to great results without it being okay in both noise and distortion.
 
OP
pma

pma

Major Contributor
Forum Donor
Joined
Feb 23, 2019
Messages
2,289
Likes
4,057
Location
Prague
I could have something with THD+N of 1% and it may never be audible on music if the one percent is all 2nd and 3rd harmonic with a low noise floor. Or I could have something with THD+N of 1% which has little distortion but lots of noise or maybe 60 hz hum. Such a device rather than being audibly transparent for music is in fact unusable for any hifi purposes.

That's correct. And you have exactly described why I do not like simplifications to just one number. That's how it goes. If we know, we do not need explanations. If we do not know, we cannot understand and tend to accept some simple universal belief. Like one number specification. We already were there, decades ago.
It is necessary to explain. But you know how frustrating it is. If you cut off one hydra head, two more would grow back in its place.
 

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,137
Failed (9/16), log attached.

@pma, would you be able to report the distortion % and spectrum at whatever the average level of that recording is?
 

Attachments

  • ABX Result Quintet.txt
    722 bytes · Views: 57
OP
pma

pma

Major Contributor
Forum Donor
Joined
Feb 23, 2019
Messages
2,289
Likes
4,057
Location
Prague
I'm not surprised that's not audible given that the third harmonic is already roughly 100dB down from the fundamental.

Not exactly. The question was to average level and the measurement was at average level. At peak level of the music sample, the distortion is about 1%
dist_peak_level.png


The tube distortion is twice higher at twice the amplitude. Quite perfectly linear plot. At FS, it is about 2.7%

thd_ampl_srpp.png


So it goes, unless we speak about crossover distortion. Single numbers are pretty misleading.
 

playmusic

Member
Joined
Dec 14, 2020
Messages
10
Likes
6
Just for completeness: I made three attempts but was not able to discriminate the two tracks.

It was interesting, though: Several times I was sure to hear some differences between the tracks.
But then the test result proved the opposite.

I have given up for the moment.
Nevertheless, I will ABX the two tracks again, sometime.
 
OP
pma

pma

Major Contributor
Forum Donor
Joined
Feb 23, 2019
Messages
2,289
Likes
4,057
Location
Prague
Just for completeness: I made three attempts but was not able to discriminate the two tracks.

It was interesting, though: Several times I was sure to hear some differences between the tracks.
But then the test result proved the opposite.

I have given up for the moment.
Nevertheless, I will ABX the two tracks again, sometime.

Thank you for your participation!
 

orangejello

Active Member
Joined
Apr 18, 2019
Messages
197
Likes
239
That's correct. And you have exactly described why I do not like simplifications to just one number. That's how it goes. If we know, we do not need explanations. If we do not know, we cannot understand and tend to accept some simple universal belief. Like one number specification. We already were there, decades ago.
It is necessary to explain. But you know how frustrating it is. If you cut off one hydra head, two more would grow back in its place.
This is also the likely answer to why tube recordings in the fifties and early sixties could sound so good despite having inferior THD specs compared to what is possible today. It just doesn’t matter that much in the end result. We were discussing this in another forum...
 

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,137
This is a fair question. Please find it attached.

View attachment 99944


There is also fairly good crosstalk
View attachment 99952

Thanks :)

In general, of all forms of distortion, it is clipping that is the least audible, partly because our masking thresholds rise as level increases, and partly because our auditory systems tend to average inputs over a period of approximately 200-250ms.

I therefore find it relatively unsurprising that the test has defeated all of us (so far)....
 

XpanD

Member
Forum Donor
Joined
Apr 7, 2018
Messages
97
Likes
92
Location
Netherlands
Bit late, but here's mine:

foo_abx 2.0.6d report
foobar2000 v1.6.2
2020-12-27 13:37:55

File A: Quintet3.flac
SHA1: 5e123ca2ae69c238cc9bc38a7bbd8f258c3a329a
File B: Quintet4.flac
SHA1: 8d68ff31f23796ffca3b769ac147a247b0de3b37

Output:
Default : Primary Sound Driver
Crossfading: NO

13:37:55 : Test started.
13:38:38 : 00/01
13:39:21 : 01/02
13:39:59 : 02/03
13:40:47 : 02/04
13:41:26 : 02/05
13:41:50 : 03/06
13:42:14 : 03/07
13:42:49 : 04/08
13:42:49 : Test finished.

----------
Total: 4/8
p-value: 0.6367 (63.67%)

-- signature --
b112803faf9f2048c5c40071b69cdb185dbccf11

The two differences I occasionally felt like I noticed were slightly nasal highs, as well as a difference in clarity in the bass. I could only "hear" this during some of the tests, not all of them. I did not read the full test details beforehand, just the title and the setup steps. None of this ended up bearing out in the results, though!

Thanks for giving me an excuse to get this set up, good to be able to do some of these comparisons now.
 
Last edited:

Pdxwayne

Major Contributor
Joined
Sep 15, 2020
Messages
2,673
Likes
904
I did the tests, might as well share it here.

I was told cumulative results of 30/48 is about 3% to 5% chance of guessing. So, I might be able to sense distortions using my chain a little more often than not?

My chain is laptop USB to Gustard X16 to Topping L30 to AKG K371.

Test results, in order of being taken:

first_ABX_7_out_of_10.PNG


second_ABX_6_out_of_10.PNG


third_ABX_7_out_of_10.PNG


4th_ABX_10_out_of_16.PNG
 
Last edited:
Top Bottom