Do you like tube distortion? - listening test

Which of the files you prefer by listening?

  • I prefer ella1 but I do not have an ABX result

    Votes: 7 53.8%
  • I prefer ella2 but I do not have an ABX result

    Votes: 5 38.5%
  • I prefer ella1 and I do have an ABX result

    Votes: 1 7.7%
  • I prefer ella2 and I do have an ABX result

    Votes: 0 0.0%

  • Total voters
    13
  • Poll closed .
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I observed on some speaker vs resistor load that my 6bl7 tube amp which is triode and low feedback has a 0.7 db gain at 50hz vs 100 and 1000 hz.

However this amplifier has a poor frequency response -3db at 8khz but it is very enjoyable.

6bl7 triode with almost 4 db of feedback is not a lot in push pull, however the THD is very low!

The repeated impulse at -3db shows at 2V peak-peak a 20us rise time and 69us fall time.
This means that in theory 11khz is the true cut off point.
This response is bad under 0.5 watts and gets better over 2 watts because transformer losses I believe.

I don't think it has anything to do with THD since the listening levels are typically low.

This makes me realize a lot of things for tube amplifiers especially.

Since the bigger transformers have usually more loss and coercion their bandwidth is especially limited for very low signals such as 0.2 watts but improves as the power is augmented because the loss becomes more averaged.

Getting a small output transformer would actually be the best solution for low feedback amplifiers.

If you add feedback the feedback will restore the bandwidth.

I can easily boost the feedback to 10 db in my 6bl7 amp and restore the high frequency response.

There is absolutely no ringing or any artifacts on the scope.

We really need to come up with something better than spectrogram and thd figures to gauge the quality of the amplifiers....

What we need is an amp that can control a difficult load at low power (0.5watt) with flat response, and multiple signals without any ringing and as low IMD as possible and THD.
 
Last edited:
If you are calling for rise time measurement, it is the easiest thing to measure and it is about 1us for this circuit, which reflects in >300kHz -3dB corner. Everything seen in time domain has its image in frequency domain. However, non-linear distortion is not easy to see in the scope screen, unless you use a notch filter. IMO everything is well known and there is no need to "investigate" some different methods.
 

Attachments

  • lampy 15kHz sq 30Vpp rise.PNG
    lampy 15kHz sq 30Vpp rise.PNG
    21.3 KB · Views: 402
Pavel,
Have you done a control experiment where you compare the original wav file to the recorded file you made with a "straight wire" to see if you can tell a difference?
I ask because this would inform the sensitivity of your test.
IOW the sampling/ADC process has to be of sufficient accuracy to allow a certain amount of wav file difference to be distinguishable.
Brian
 
Last edited:
If you are calling for rise time measurement, it is the easiest thing to measure and it is about 1us for this circuit, which reflects in >300kHz -3dB corner. Everything seen in time domain has its image in frequency domain. However, non-linear distortion is not easy to see in the scope screen, unless you use a notch filter. IMO everything is well known and there is no need to "investigate" some different methods.

I see more like 3.5us , and what was the frequency for the square wave or was it only a rise time test?

Can you do the same test with 10000hz square wave with 95% duty time at only 2v peak-peak into a speaker or resistor?
 
Pavel,
Have you done a control experiment where you compare the original wav file to the recorded file you made with a "straight wire" to see if you can tell a difference?
I ask because this would inform the sensitivity of your test.
IOW the sampling/ADC process has to be of sufficient accuracy to allow a certain amount of wav file difference to be distinguishable.
Brian

Yes of course. Since some time, I always do it. I was NOT able to tell the difference in a DBT. However, I am not going to post original file at the moment, as it always invokes sighted impressions. It may happen, if there will be participants interested, that after the poll expiration I may prepare 3 files: original, wire and tube. I need then all of them to have exactly same size. The equal size is one of the reasons I post wire x tube and not original x tube. People would take any measures to guess which is which.
 
I see more like 3.5us , and what was the frequency for the square wave or was it only a rise time test?

Can you do the same test with 10000hz square wave with 95% duty time at only 2v peak-peak into a speaker or resistor?

No, rise time is defined as the time between 10% and 90% of the total step swing level. We need to keep standard definitions and we need to have some basic knowledge to lead such discussion. BW is then 0.35/Tr and it fits quite perfectly.

10kHz square or single step makes absolutely no difference in Tr (10%-90%). It (rise time) is always the same. And I cannot load a link preamplifier stage with a speaker. This is a task for the power amplifier, I hope you understand the difference.

I am sorry but we need to keep some knowledge level. I am not going to debate the elementary definitions and principles, as it would start a never ending circle of off topic questions.
 
Ella1 has much higher high frequency distortion, especially from brass section, and is unpleasant to hear. I noticed the difference in a couple of seconds. I first heard ella1 and than ella2. It was not necessary to wait till brass section from ella2 started to blast, I noticed the difference from the first seconds of ella2. And all that from the crappy miniature loudspeaker on my Lenovo notebook!
 
Administrator
Joined 2007
Paid Member
This one:

foobar2000: Components Repository

Select (highlight) the tracks you want to ABX and do as Pano says and right click the highlighted selection.

When you have done, highlight the tracks again and right click and 'Remove' them (removes them from Foobar only).

(I got 100% on this, good eh ;))
 

Attachments

  • Annotation.png
    Annotation.png
    223.2 KB · Views: 263
  • Annotation2.png
    Annotation2.png
    90.5 KB · Views: 262
  • Annotation3.png
    Annotation3.png
    41.6 KB · Views: 253
  • Annotation4.png
    Annotation4.png
    43.5 KB · Views: 255
I downloaded the files, and played each file one time through the computer speakers at my desk. A couple of seconds into #2, I was certain it sounded "better."

I waited an hour, did it again - played each file one time, and can't say one sounded any different from the other.

Win W5JAG
 
Yes of course. Since some time, I always do it. I was NOT able to tell the difference in a DBT. However, I am not going to post original file at the moment, as it always invokes sighted impressions. It may happen, if there will be participants interested, that after the poll expiration I may prepare 3 files: original, wire and tube. I need then all of them to have exactly same size. The equal size is one of the reasons I post wire x tube and not original x tube. People would take any measures to guess which is which.
Is the recording you used the same as the one I posted in #2?
If not then I'll wait until/if you post the one you used as input. I just wanted to explain why I can hear a difference.
 
Harmonics monotonic decrease with moderate THD = "natural" sound

Hugh (AKSA):

Aksa Lender P-mos Hybrid Aleph (ALPHA) Amplifier

Lastly, the harmonic profile has been thrown into prominence in the last ten years, but in fact has been pivotal for good sound and Jean Hiraga demonstrated this in Le Revue du Son in Paris in the mid-sixties. He said that if you can arrange the first five or six harmonics off the fundamental along a straight, descending line, with at least 10dB reduction with successive higher harmonic, the sound will be perceived as 'natural'. H2 should start no higher than 68dB down from the fundamental.

This was a controversial statement at the time, and even now, because it pays little regard for THD, which is a rms summation of the harmonic percentages of the fundamental, with no information about each actual harmonic.

I have been designing amps using this philosophy for twenty years, and if you give regard to the monotonic decrease of the harmonics and a moderate THD (no more than 0.03%) you can create a good sounding amplifier.

There are other issues, of course, but it certainly explains why many amps with very low THD still do not sound good. The fact that this is a subjective assessment, and humans enjoy squabbling about the subjectives has only clouded high end. During the eighties the Japanese opened up the 'THD Wars', to very little benefit. Still a lot of people like single ended triode amplifiers, and they deliver outrageous THD figures...
 
Last edited:
Administrator
Joined 2004
Paid Member
Yes, the Aska quote above nicely describes harmonic masking. of which Jean Hiraga has ling been an advocate. E.G. Hiraga loads the 300B at 2.5K not because the distortion is lower than a higher impedance load, but because the harmonic structure is more favorable and less audible.

The interesting thing about the tube buffer distortion added to the test file is that it appears to consist of only H2 and H3, nothing more. And yet no one has reliably heard it on this re-recording. It would still be interesting to test it with a clean violin or cello recording.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.