You're welcome!
FWIW, I just looked very closely at the offsets of your files, to +/- one sample. You did a pretty good job aligning them. If we take the Ref file as Zero delay, the others are just slightly delayed by a fraction of a millisecond.
File start delay.
Not too shabby. The biggest offset is between A&B (sorry, I lost D)
I don't see how that could affect the test as we know which file is which. I guess you might be able to hear the glitch in a blind A/B test, so someone could pick up on glitch or no glitch in the ABX tester, even subconsciously. I ignored any glitches and listened to the music.
The overall levels were very well matched, but there are definite noise floor differences and some small level differences right at the start. Not sure what that indicates. Won't have time to dig further.
FWIW, I just looked very closely at the offsets of your files, to +/- one sample. You did a pretty good job aligning them. If we take the Ref file as Zero delay, the others are just slightly delayed by a fraction of a millisecond.
File start delay.
- Ref: 0
- A: 0.11 mS
- B: 0.79 mS
- C: 0.50 mS
- E: 0.41 mS
- F: 0.27 mS
Not too shabby. The biggest offset is between A&B (sorry, I lost D)
I don't see how that could affect the test as we know which file is which. I guess you might be able to hear the glitch in a blind A/B test, so someone could pick up on glitch or no glitch in the ABX tester, even subconsciously. I ignored any glitches and listened to the music.
The overall levels were very well matched, but there are definite noise floor differences and some small level differences right at the start. Not sure what that indicates. Won't have time to dig further.
Hi Pano,
thanks for confirming diffmaker works OK, check for yourself(first parameter):
Audio diffmaker
> B-A parameters: -663,9usec, 0,211dB (L), *0,032dB (R)..Corr Depth: 32,1 dB (L), 48,4 dB (R)
> C-A parameters: -377,3usec, 0,106dB (L), *0,007dB (R)..Corr Depth: 38,3 dB (L), 65,5 dB (R)
> D-A parameters: -99,65msec, 0,433dB (L), *0,271dB (R)..Corr Depth: 26,0 dB (L), 30,4 dB (R)
> E-A parameters: -285,7usec, 0,258dB (L), *0,058dB (R)..Corr Depth: 30,4 dB (L), 43,9 dB (R)
> F-A parameters: -149usec, 0,304dB (L), *0,087dB (R)..Corr Depth: 29,0 dB (L), 40,0 dB (R)
>
Additionally, 2.& 3. parameter are level corrections.
PS please check D file mannualy
thanks for confirming diffmaker works OK, check for yourself(first parameter):
Audio diffmaker
> B-A parameters: -663,9usec, 0,211dB (L), *0,032dB (R)..Corr Depth: 32,1 dB (L), 48,4 dB (R)
> C-A parameters: -377,3usec, 0,106dB (L), *0,007dB (R)..Corr Depth: 38,3 dB (L), 65,5 dB (R)
> D-A parameters: -99,65msec, 0,433dB (L), *0,271dB (R)..Corr Depth: 26,0 dB (L), 30,4 dB (R)
> E-A parameters: -285,7usec, 0,258dB (L), *0,058dB (R)..Corr Depth: 30,4 dB (L), 43,9 dB (R)
> F-A parameters: -149usec, 0,304dB (L), *0,087dB (R)..Corr Depth: 29,0 dB (L), 40,0 dB (R)
>
Additionally, 2.& 3. parameter are level corrections.
PS please check D file mannualy
Last edited:
I beleieve this is true only when Sample rate correction is needed
("Compensate for sample rate drift" checked).
In that case program use oversampled auto correlation and goes out of memory for large samples.
If Sample rate correction is not needed you can use it for long samples.
My advice:
when making test files it is important that ADC&DAC clock is on constant temperature during all runs so:
1. Turn on all hardware and let temperature stabilise
2. Do all runs in one go
That way, very probably Sample rate correction is not needed.
Yup, I tried really hard, but nothing happens 😀
After that, I tried to measure.
Look, I really only want to help, don't have any point except that.
After that, I tried to measure.
Look, I really only want to help, don't have any point except that.
Well that is what it is all about. Listening.
If we all bought our audio gear (or anything else) purely on technical specs then we would all end up with the same stuff. The only thing that analysing these files proves is that they are all different, and which of course they are.
When all is said and done and when you sit down to listen, then no amount of someone saying 'this or that is that is better' is going to cut it if you actually prefer something else.
There is no right and wrong. Some feel they can detect a difference, and the natural progression of that would be to repeat the test again with different material and see if the same kind of pattern emerged.
All good fun and interesting 🙂
I didn't listen on my main system as I had planned to. Still curious to find out what's what, though.
TL071 sounds plain distorted, as is to be expected with 390R load.
390R is not a normal/realistic load for pretty much any usual opamp, and renders this listening test another useless waste of time.
I suggest a rerun with build out resistance (100R ?) and say 2K+ load to more accurately represent real world conditions.
The result of this test is that 5534 and 833 sound relatively clean but different, the other opamps don't cut it with 390R load.
Dan.
390R is not a normal/realistic load for pretty much any usual opamp, and renders this listening test another useless waste of time.
I suggest a rerun with build out resistance (100R ?) and say 2K+ load to more accurately represent real world conditions.
The result of this test is that 5534 and 833 sound relatively clean but different, the other opamps don't cut it with 390R load.
Dan.
Sorry Dan but I have to take you to task over this,
The signal levels presented to the circuit under test ensured that none of the opamps even came close to output stage limiting.
The 5534 wasn't tested here as it would not have been suitable for this configuration, it was the 5532, and neither was the TL071 (a typo ?).
That you could differentiate the TL072 as distorted and presumably identify the 5532 and LM833P as clean in comparison would have made for truly brilliant subjective evidence... if you had presented it before knowing what the files were.
Being wise after the event is easy. Have a read at post #49
The signal levels presented to the circuit under test ensured that none of the opamps even came close to output stage limiting.
The 5534 wasn't tested here as it would not have been suitable for this configuration, it was the 5532, and neither was the TL071 (a typo ?).
That you could differentiate the TL072 as distorted and presumably identify the 5532 and LM833P as clean in comparison would have made for truly brilliant subjective evidence... if you had presented it before knowing what the files were.
Being wise after the event is easy. Have a read at post #49
General knowledge is that TL071/072 etc WILL distort at low load resistance.
5534/5532...071/072 ok, typo, whateva.
Yes, I did arrive late to the party and with knowledge of the opamps used took a brief listen on my laptop tiny little inbuilt speakers.
However such speakers are good enough to differentiate the files and the TL072 failed miserably subjectively prior knowledge or not, as did the others with the exception of the 5532 and 833.
I did not bother to transfer the files to my main system in this case because in my view this test proves very little of worth because of unrealistic load conditions.
Please note that I intend no dissing.
Much better would be to run your multiple series stages with more realistic loadings.
Even better would be to run the chain of opamps with gains (and attenuations) to better duplicate real world usage.
The music you chose is relatively sparse, relatively low average level and relatively clean.
Try running the likes of AccaDacca (Livewire) and the sonic characters of the opamps will be even more revealed.
Dan.
Dan, I can only reiterate that had you produced these comments and observations before the revealing of the files, then they would have carried substantial weight.
We did a test along the lines of amplify/attenuate but using only two opamps quite some while ago.
http://www.diyaudio.com/forums/ever...mate-opamp-shootout-where-you-get-decide.html
Perhaps in any future test the circuit used could be put out to scrutiny first.
We did a test along the lines of amplify/attenuate but using only two opamps quite some while ago.
http://www.diyaudio.com/forums/ever...mate-opamp-shootout-where-you-get-decide.html
Perhaps in any future test the circuit used could be put out to scrutiny first.
Hi,
can you please point the post where the files are revealed. I picked B. Would be nice to know which opamp I picked. Honestly though I don't understand technicalities of the test. But still it would be good to know.
Thanks and regards.
can you please point the post where the files are revealed. I picked B. Would be nice to know which opamp I picked. Honestly though I don't understand technicalities of the test. But still it would be good to know.
Thanks and regards.
I re-examine my assesment of the opamps in Mooly's previous test (attached is the result). It was a blind test using randomized playlist of iPod. I haven't used or referred to a FoobarABX back then, and it was my first experience with music clip blind test. Regarding TL072, I found it musical in 3 musics and distorted in 1 music (opera/classic, where there was a high note shout IIRC).
I found 833 was very good objectively, and no complaint subjectively. But I subjectively prefer the TL072 for it's musicality and because it's objective weakness didn't cause fatigue at all. I believe that TL072 is more suitable in BUFFER position (in earlier test it was in 2 gain stages). But may be this 5-buffers-with-390R-load is too difficult for it, I don't know. But I expect TL072 will perform better in normal buffer.
Earlier test has 4 different music types. And it has the attenuate-amplify-attenuate-amplify stages. But like I said before, I would prefer TL072 in buffer position. Long ago, in my active crossover (I still have it), I decided to put TL072 in buffer positions, and other BJT opamps in gain/filter stages.
Attachments
Mooly--innocent question (with totally moving goalposts), might you be able to take all said 8 files and make another randomized file name blinding? I'd honestly love to try again with greater scrutiny (and seems there might be a few of us that would), even acknowledging some of the more cantankerous feedback you've received.
And, thank you nonetheless for your work here!
And, thank you nonetheless for your work here!
That should be possible but I will also tweak each file because otherwise some will be identifiable by their properties.
One for tomorrow I think 🙂
One for tomorrow I think 🙂
- Status
- Not open for further replies.