A proper blind test will tell you for certain whether a DAC or other device is audibly transparent. No measuring needed!
I have to admit I don't think peering in very deeply has much value. I think with FFT/IFT it's paramount to remember the inherent time-frequency tradeoff with various widowing sizes and types. And that they all end up being a snapshot averaged over some time period that has to mathematically describe the signal as if it were steady state. Which is seldom, if ever, the case.
I don't think it pays to assume any one windowing choice is the correct one, so if follows logically not to get too serious with any of them. I think trying to chose the window size and type wisely for the measurement's intended purpose, is about the best can be done.
IIUC, the files are the same in the frequency domain spectral view (within the limits of 16-bit dither), as they were synthesized by MATLAB scripts (which were vetted by 3rd parties). Most likely any differences in measurements are artifacts of dither differences and or of FFT calculation artifacts.
Of course, the time domain is a very different situation. In the time domain, phase of frequencies can matter a lot. In a frequency domain spectral view plot, phase information has been discarded.
Moreover, phase rotation is a well known method of amplitude modulation or frequency modulation. IOW, it is well known that only changing the phase of certain frequency components will change the modulation type.
Last edited:
There you go. None of them are audibly transparent!
More seriously, today's professional perceptual scientists know how complicated it is to do blind testing well. EE's with who think its all a matter of common sense have fooled many people into thinking they can't hear what they actually can hear. Its a very unfortunate situation.
More information in another thread:
https://www.diyaudio.com/community/...ble-difference-whatsoever.314762/post-8039411
https://www.diyaudio.com/community/...ble-difference-whatsoever.314762/post-8039416
https://www.diyaudio.com/community/...ble-difference-whatsoever.314762/post-8039399
Last edited:
Look at the wave forms! One stays at the same level and the other varies. So again they do not measure the same.
LOL, except you haven't done the tests! You're just making claims with nothing to back then up. Of course some devices are clean enough to not damage the sound audibly. This has been proved again and again, even going back decades to 1980s DACs:
http://www.bostonaudiosociety.org/bas_speaker/abx_testing2.htm
Blind tests are not easy to do correctly:
https://www.prosoundweb.com/discerning-differences-how-to-conduct-proper-useful-listening-tests/
In the time domain or a frequency domain spectral view? You have to be specific.
Actually, I have done more than you know about. You appear to be jumping to conclusions again.
Based on your posts to this thread so far, I won't believe it until I witness you do a blind test in person. 😗
Where are you located? I am in Auburn, CA. Since I have ultra low distortion Sound Lab electrostatics this is where I think any serious listening tests would have to be done. You would be welcome to bring any dac you like for comparison. We can find a mutually agreeable disinterested 3rd party to do the blind device swapping.
I'm in Connecticut, but at age 76 my traveling days are over. I hate airports! So I won't be coming out there. But it doesn't have to be a DAC that is tested. You said "none of them" (DACs or other devices) are transparent, so the tests could be a unity gain preamp inserted into the signal path, or anything else with active electronics. But let's stick with DACs for now. I'm surprised you actually believe that your hearing is so good that you can identify artifacts more than 100 dB below the music while the music plays. That's quite a claim!
I have several "blind tests" on my web site that you could do in the comfort of your home with your own loudspeakers. Then you'll email me your choices and I'll tell you if you're right or not. If you're up for that, this first A/D/A test asks you to identify an original recording, then an A/D/A loop-back of one generation, then five generations, then ten generations. Your mission is to identify the files in order of degradation.
https://ethanwiner.com/loop-back.htm
This next test asks you to identify a live music source recorded simultaneously through three converters ranging in price from a $25 SoundBlaster X-Fi sound card through to a $2,000 Lavry Blue converter. You have to pick which of the three files was recorded through which converter:
https://ethanwiner.com/converters.html
This last test has you identify two 24/96 hi-res sources with copies reduced to 16/44. You haven't claimed that you can identify hi-res versus CD quality, so this may not be a relevant test for you.
https://ethanwiner.com/hd-audio.htm
I have several "blind tests" on my web site that you could do in the comfort of your home with your own loudspeakers. Then you'll email me your choices and I'll tell you if you're right or not. If you're up for that, this first A/D/A test asks you to identify an original recording, then an A/D/A loop-back of one generation, then five generations, then ten generations. Your mission is to identify the files in order of degradation.
https://ethanwiner.com/loop-back.htm
This next test asks you to identify a live music source recorded simultaneously through three converters ranging in price from a $25 SoundBlaster X-Fi sound card through to a $2,000 Lavry Blue converter. You have to pick which of the three files was recorded through which converter:
https://ethanwiner.com/converters.html
This last test has you identify two 24/96 hi-res sources with copies reduced to 16/44. You haven't claimed that you can identify hi-res versus CD quality, so this may not be a relevant test for you.
https://ethanwiner.com/hd-audio.htm
Ah, but that's not my claim. Where I will find differences in sound will probably be more like at -60dB to -90dB, and or in localization precision. The differences won't be in terms of measured nonlinear distortion nor of fixed noise floor.
As I said before, the things we typically measure tend to get fixed. Its the things that we don't usually or don't ever measure which are the things that still need fixing.
Nope. Don't trust your ability to produce valid blind tests. Maybe you can or maybe you can't, but its not scientifically proven you can. After all, I have seen your paper on isolators, and read your arguments with @gedlee. I have good reasons to have doubts.
Why would we need to do that? As I see it, it would be only of we wanted to cheat... like .mp3 etc. But we don't - so if in=out , we are good - no?
//