Udo,
- the level mismatch is 0.07dB exactly,
- the 2nd file is inverted by my mistake (+IN and -IN of the soundcard balanced input swapped), it seems that I forgot to put the latest non-inverted version into the zip. However music test files were not inverted.
- the files are 5dB below FS for the reason of soundcard increasing distortion above that level,
- both files are dithered
- the source is analog generator that's why there are harmonics in the orig file, but deep below masking level. 549 file has same harmonics even if driven from digitally generated signal.
- CCIF IMD 19+20kHz or 13+14kHz produces 1kHz IM tone, however the test is dangerous to be posted, because listeners will set high gain to hear the difference tone and the 19,20kHz tones, for most people inaudible, may cause hearing damage. I will not post such test. Moreover, music files contain such frequencies at very low amplitudes so the IM product (of some 0.02% - 0.05% as here) is again inaudible with music.
BTW, do you have a positive abx result of these 1kHz files?
Last edited:
No, i have no abx result - i have no change on my laptop as the soundcard loopback SNR is less than 85 dB, even with headphones (but i can hear my laptop accessing the RAM in the headphones...).
Even in an anechoic room i doubt that i can discriminate -86 dB near the strong fundamental.
I guess than everyone who has a positve ABX result discriminates on second order effects of the recording chain, or of his own setup.
As i understand the problem with crossover distortion is that it is happening
at the zero crossings, even if the audio is played quietly.
Sorry for being a bit sloppy with the numbers. I assumed that the harmonics at -95 dB originate from non-dithering.
Even in an anechoic room i doubt that i can discriminate -86 dB near the strong fundamental.
I guess than everyone who has a positve ABX result discriminates on second order effects of the recording chain, or of his own setup.
As i understand the problem with crossover distortion is that it is happening
at the zero crossings, even if the audio is played quietly.
Sorry for being a bit sloppy with the numbers. I assumed that the harmonics at -95 dB originate from non-dithering.
Last edited:
Yes, but it is all about audibility threshold, not only about the fact of crossover distortion. Please see plots for -60dBFS, -40dBFS and -20dBFs level. Please also note how the chipamp PSR reflects in mains components increasing with level (higher PSU ripple).
Attachments
Last edited:
I think it was fairly clear from the first posts that the question was whether fairly severe cross-over distortion could actually be detected. It's been posted as the boogyman for ages, yet no-one appears to be able to detect it in music and are having a hard time even with a sine wave.
Very thought provoking. And note I have a chip amp based power amp that wipes the floor in measured performance (bar max power) with almost anything out there.
Your insistence of 16 bits bad 24 bits good for playback is highly illogical. Don't forget that the plural of anecdote is still not data!
Very thought provoking. And note I have a chip amp based power amp that wipes the floor in measured performance (bar max power) with almost anything out there.
Your insistence of 16 bits bad 24 bits good for playback is highly illogical. Don't forget that the plural of anecdote is still not data!
See for instance: http://soekris.dk/images/fft_-1db.jpg
I have such DAC and sounds good... and remember a class-B amp "inverted" (more high order harmonics with HIGH amplitude).
Of course, in this "DAM" DAC case, is a mismatch in logic gate arrays.
For now I don't read all thread, but do you have a OPA549 measurement with speaker attached? (close yous ears and make some 1mW and 1W measurements with speaker load)
A low average volume music like eg. classical maybe widens the chance of perception.
PS.: I don't able to identify sound difference of the files. pure guessing...
System: Stax SR-507 earspeaker, DIY triode amp with 5687 out, E88Cc input, DAC is a ESS9018 using its internal S-PDIF stand-alone. All balanced path.
I have such DAC and sounds good... and remember a class-B amp "inverted" (more high order harmonics with HIGH amplitude).
Of course, in this "DAM" DAC case, is a mismatch in logic gate arrays.
For now I don't read all thread, but do you have a OPA549 measurement with speaker attached? (close yous ears and make some 1mW and 1W measurements with speaker load)
A low average volume music like eg. classical maybe widens the chance of perception.
PS.: I don't able to identify sound difference of the files. pure guessing...
System: Stax SR-507 earspeaker, DIY triode amp with 5687 out, E88Cc input, DAC is a ESS9018 using its internal S-PDIF stand-alone. All balanced path.
Of course, class-C chinese el-cheapo amplifiers, so common for car, having zero bias for output, have ridiculous high crossover distortion. At low volumes, this sounds like a treble boost
mixed with some kind of noise.I've measured 16% THD at some level with "infinite harmonics" with some of these China amp.
Talking about amps, SNR, cross... My second-hand Marantz:
Marantz SR4500 A/V Receiver HT Labs Measures | Sound & Vision
In a laboratory, the measurements can be excellent but in real life, with an increasingly noisy electrical grid, the harsh reality is quite another.
Marantz SR4500 A/V Receiver HT Labs Measures | Sound & Vision
But after ALL my tweaks it sounds much better than it come home, much better. The key is to have greatly attenuated noise, RFI/EMI interference and continuous.
In a laboratory, the measurements can be excellent but in real life, with an increasingly noisy electrical grid, the harsh reality is quite another.
Last edited:
I don't claim to know how much cross over distortion is bad, but we know undithered 16-bit audio is audible to some people. For those people, such distortion could potentially be unpleasant. I don't like it when I notice it, don't know about anyone else.
Possibly, it looks like one remedy for keeping cross over distortion from being bothersome is to ABX test for it until sufficient ear fatigue is produced to make the distortion inaudible. Isn't that what these tests actually show?
Regarding 16 verses 24 bits, more can be better or not depending on source material, IMHO. Unfortunately, we don't seem to be taking data to see how the data might help correlate bit depth with what many people describe hearing. I don't understand why it would be preferable not to take take such data, but maybe someone has a thoughtful reason why. Presumably, it might depend on exactly what one would like to find out.
Possibly, it looks like one remedy for keeping cross over distortion from being bothersome is to ABX test for it until sufficient ear fatigue is produced to make the distortion inaudible. Isn't that what these tests actually show?
Regarding 16 verses 24 bits, more can be better or not depending on source material, IMHO. Unfortunately, we don't seem to be taking data to see how the data might help correlate bit depth with what many people describe hearing. I don't understand why it would be preferable not to take take such data, but maybe someone has a thoughtful reason why. Presumably, it might depend on exactly what one would like to find out.
Last edited:
Mark: Pavel did say the files are dithered. The distortion being tested for is waaaay above any level that would make 16 vs 24 bit valid and waaaay above any nasties in the converstion chain, despite your sighted dislike for the dacmagic. I cannot see your logic as you seem to be running on belief here and finding boogeymen that are just not there for this test.
Well, this is the last one I am was willing to create and post
- 440Hz sine, digitally generated
- orig recorded as a loopback
- opa549 recorded through the amp under test
- polarity is correct
- level matching within 0.02dB
- files are TPD dithered and processed in Adobe Audition (fade-in, fade-out)
- file level is about -2.5dBFS
http://pmacura.cz/440_549.zip
At 440Hz, the amp crossover distortion has fallen to 0.008%, so it will be absolutely impossible to discern files by listening, unless one has faulty D/A conversion with nonlinearities and glitches.
- 440Hz sine, digitally generated
- orig recorded as a loopback
- opa549 recorded through the amp under test
- polarity is correct
- level matching within 0.02dB
- files are TPD dithered and processed in Adobe Audition (fade-in, fade-out)
- file level is about -2.5dBFS
http://pmacura.cz/440_549.zip
At 440Hz, the amp crossover distortion has fallen to 0.008%, so it will be absolutely impossible to discern files by listening, unless one has faulty D/A conversion with nonlinearities and glitches.
Last edited:
Heck, no. Whether I could pass or not, I have no interest in playing that particular game. I think it's a poor test for low level discrimination.
However, on the last hi-res listening test thread PMA did, he did give me credit for sorting some files by ear pretty well in order of distortion, with one miss where I failed to make a direct comparison between two files. It's a better type of blind test for low level discrimination, I would suspect. However, no handy foobar plugin for doing it, so out of luck on that.
The insistence on ABX seems partly because it's what we have, and partly because the foobar version has a verification feature to help reduce cheating, although if someone really wanted to cheat they probably could. As I mentioned before, the digital stream from foobar to the DAC could be intercepted and analyzed. It wouldn't be too hard if somebody cared to bother with it, especially if using an outboard DAC, or if SPDIF out were available. However, no interest in cheating here.
Attachments
Pavel: Thank you for your patience on this, it is fascinating.
Mark: If you don't want to play then why on earth are you posting on this thread. Pavel made it very clear he wanted foobar ABX testing. If you are not willing to do ABX testing due to not feeling it is good enough then why come on and post negatively? Would it not be easier to decide that ABX was not for you and ignore the thread? If you have a better method for us then please raise a thread on it.
Mark: If you don't want to play then why on earth are you posting on this thread. Pavel made it very clear he wanted foobar ABX testing. If you are not willing to do ABX testing due to not feeling it is good enough then why come on and post negatively? Would it not be easier to decide that ABX was not for you and ignore the thread? If you have a better method for us then please raise a thread on it.
My initial intention was to leave it alone. But, people started talking about things other than ABX results. Responding to some of it was not inappropriate once it got going. I did not describe what I heard until directly asked by Mooley. Once ABX started up again and people stopped talking about what it sounded like, I stopped talking about that too. After that I did ask about why 16-bits because I would like to know why. Still wonder why. Curse of being curious I guess.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.