Your result is not totally invalid. Your wireless headphones are doing something that makes the very low level distortion audible. That's an interesting result.
I think that the different data content of the test files can change behavior of those wireless things.
My experience is - the worse the DAC and soundcard, the more audible are the "differences", that in fact may not exist. It may be just behavior, technical flaw of the cheap equipment. These issues are quite complex.
I tried the fading part - no success.
It is most probably exaggeration of some data levels. Not necessarily connected to signal distortion. Similar to undithered signal and poor SFDR.
Two different DACs, one in the headphones in wireless mode, and the Realtek DAC in the PC in wired mode. How about the wireless driver, is it bluetooth?
Yes, Bluetooth. AptX specifically, which is normally good enough that I forget it's wireless, but with those test tones it makes the distortion much more obvious.
Now with my cheap (and tweaked+recabled) Takstar ts-671 heapdhones.
* With Lacinato ABX, impossible.
* With JRiver, I can hear a little differentiate: 1khz_opa549.wav sound more sharper, shrill. Still it would be very difficult to overcome the blind test.
* With Lacinato ABX, impossible.
* With JRiver, I can hear a little differentiate: 1khz_opa549.wav sound more sharper, shrill. Still it would be very difficult to overcome the blind test.
Maybe the DAC in the headphones is more accurate than the Realtek DAC in the PC? Some of those Realtek DACs are pretty bad, IMHO. Like maybe around -60dB THD+N for the headphone out. Line level out, if available, may be somewhat better.
Edit: When connected to the analog out on the PC, is there a volume control on the headphones, or only in the PC?
Last edited:
If it was audible during fade-out, it must have been DAC error (glitch, wrong converted value, error code). I have checked the distortion components during fade-out and they are deep deep deep below any threshold of audibility. It is always much more difficult to protect correct and educated approach than to fabricate irrational constructions. Much more difficult.
Mr. Evil said he heard something at the beginning at fade-in, not at fade-out. Some buzzing at first. Is there any buzzing at first way in the background?
Edit: they sound a little different at the beginning to me on my laptop. Also, on the laptop the level isn't perfectly steady, there is a little buzzing on and off over time. I don't suppose the level is high enough for intersample overs?
Edit: they sound a little different at the beginning to me on my laptop. Also, on the laptop the level isn't perfectly steady, there is a little buzzing on and off over time. I don't suppose the level is high enough for intersample overs?
Last edited:
The human ear is most sensitive to distortion somewhere between 2 - 3 kHz.
I must say very clearly, we are very deep under all known masking curves and thresholds of hearing with this sine test and its distortion components. So there may not be a positive ABX result, in case that the listening HW is not defective.
About the same problem for me.
I tried with some good headphones and was impossible to notice any difference. Then I tried with a good amp and speakers and no luck too.
But with an old amplifier and any speaker is easy to discern it.
Attachments
Yes, that's the problem of poor equipment in web tests. You can often read that people write: "I have used xyz cheap notebook with cheap onboard chip and clearly heard the difference with plastic PC speakers ..." and it is just sensitivity of poor DACs and electronics to any signal change.
To conclude, I am posting the non-linear difference between the last 1kHz test tones (1khz_orig and 1khz_opa549), small interchannel phase shift corrected.
http://pmacura.cz/1khz_549diff.zip
Try to listen to it at the same volume level as you did in the 1khz listening ...
Attached are as well spectra of the 1khz_orig and 1khz_opa549, with distortion number shown. The test tones were recorded simultaneously from analog generator to prevent phase and pitch differences. One path was direct from generator to A/D, second path went through OPA549 loaded with 6.8 ohm. Levels matched in analog domain by a pot.
In the situation when single sine tone distortion is not discernable, other than by playback equipment error or poor parameters, it is useless to continue with music samples.
To conclude, I am posting the non-linear difference between the last 1kHz test tones (1khz_orig and 1khz_opa549), small interchannel phase shift corrected.
http://pmacura.cz/1khz_549diff.zip
Try to listen to it at the same volume level as you did in the 1khz listening ...
Attached are as well spectra of the 1khz_orig and 1khz_opa549, with distortion number shown. The test tones were recorded simultaneously from analog generator to prevent phase and pitch differences. One path was direct from generator to A/D, second path went through OPA549 loaded with 6.8 ohm. Levels matched in analog domain by a pot.
In the situation when single sine tone distortion is not discernable, other than by playback equipment error or poor parameters, it is useless to continue with music samples.
Attachments
"1khz_opa549.wav sound more sharper, shrill"
By th way, the amplifier is the same, my second hand but very well cheap tweaked: AV Marantz SR4500
Above the amps: DC & Ripple Blocker x4 ME (aka Maty Edition), made by Aleksandar (ATL-Hifi).
Above the amps: DC & Ripple Blocker x4 ME (aka Maty Edition), made by Aleksandar (ATL-Hifi).
Last edited:
Pavel, the new data files have problems too.
The level is off by about 0.1% and the second file is inverted.
The test files are 16 bit / 48kHz and are on the limit of a 16 bit system.
You loose 6 dB because they are not at 100% amplitude.
The first harmonic is at -88dB.
There is no dithering added and a lot of harmonics at -95dB.
Please start with easier training files with 1% distortion at 24 bit / 96 kHz.
If they work and we have learned to discriminate better,
you can make the test files harder.
The worst test case are two high frequency tones, where
the intermodulation products map to 2 kHz?
The level is off by about 0.1% and the second file is inverted.
The test files are 16 bit / 48kHz and are on the limit of a 16 bit system.
You loose 6 dB because they are not at 100% amplitude.
The first harmonic is at -88dB.
There is no dithering added and a lot of harmonics at -95dB.
Please start with easier training files with 1% distortion at 24 bit / 96 kHz.
If they work and we have learned to discriminate better,
you can make the test files harder.
The worst test case are two high frequency tones, where
the intermodulation products map to 2 kHz?
I have had several attempts using headphone and speakers (laptop) and get inconclusive results.
EXPOSURE 3010S2D
This graph remember the EXPOSURE 3010S2D. Well, without the odd armonics.
[Polish] Exposure 3010S2D Wzmacniacze stereo - testy
This graph remember the EXPOSURE 3010S2D. Well, without the odd armonics.
[Polish] Exposure 3010S2D Wzmacniacze stereo - testy
Last edited:
- Status
- Not open for further replies.