SoundAndMotion
100+ Head-Fier
- Joined
- Jul 15, 2015
- Posts
- 353
- Likes
- 229
Well, this is a strange thread.
Some, mostly haiku and bilboda, state that they believe playback of FLAC files "on-the-fly" could sound different from WAV or pre-decompressed FLACs. They never state that they think the PCM data going to the DAC or bits are different or corrupted. Yet nearly all responses to them center on the fact that the bits are the same! I don't think this is a deliberate straw man, rather a simple misunderstanding... or at least I hope that's the case.
I can think of 3 possible ways that the sound can be different.
Please don't ignore my use of the word ”possible”. I am answering those who say "impossible", "never" or "always". I don't think anyone here can speak with authority about the likelihood or probabilities of certain problems w.r.t. all devices that may be currently in use. I know I can't.
First, problems in the speed of the algorithms, whether hardware, software or both. The fact that your desktop quad i7 can solve complex systems of equations in real time does not make a <$10 FLAC player run faster. Granted, most weak processors either won't play a FLAC -8 file or will cause audible clicks or stuttering, and some here have labeled that “broken”, but they’re not illegal, so they exist. But if longer gaps or delays (>10ms) create such artifacts, what about shorter gaps or delays that occur periodically?
I was curious, so I created a program that inserts a selectable number of zeroes or fixed values (zero order hold) every so often (set or variable period of time). Quick summary: some combinations were clearly audible, while others were inaudible to me. Intermediate values produce varying degrees of degradation. I wonder whether a degradation that isn’t obvious, can perhaps be ABX-able and simply sound “better”. Does this mimic a too-slow process? I don’t know. Do you?
Second, simple software/firmware bugs, other than time-inefficiency. They exist and when a clear problem presents (e.g. doesn’t play or no sound), they get fixed. But if a small, simple, non-fatal math error occurs, is it guaranteed that it will be found?
Before you claim these don’t exist, or at least not any more, google: flac playback problems. Recent biggies include Android 4.4 (Kit Kat - 2013) and VLC 2.2.0 (2015).
And finally, analog noise, dependent on processor “effort”. Assuming that this noise is not severe enough to disturb the digital bits in the signal path (good assumption, I think), the idea is additional current draw from an active vs. nearly idle CPU could find its way to the post-D/A-conversion side of the path. This idea is the most speculative. Although analog noise from USB data has been shown on DAC outputs (mostly the 8kHz microframe noise), it has been very low level. But it demonstrates the concept, and makes the claims of "impossible", "never" or "always" unnecessarily risky. I think “unusual”, “rare” and “typically” would be better choices.
I suspect that reports of hearing differences often occur when there actually is none, but to state that it cannot occur is not true.
Some, mostly haiku and bilboda, state that they believe playback of FLAC files "on-the-fly" could sound different from WAV or pre-decompressed FLACs. They never state that they think the PCM data going to the DAC or bits are different or corrupted. Yet nearly all responses to them center on the fact that the bits are the same! I don't think this is a deliberate straw man, rather a simple misunderstanding... or at least I hope that's the case.
I can think of 3 possible ways that the sound can be different.
Please don't ignore my use of the word ”possible”. I am answering those who say "impossible", "never" or "always". I don't think anyone here can speak with authority about the likelihood or probabilities of certain problems w.r.t. all devices that may be currently in use. I know I can't.
First, problems in the speed of the algorithms, whether hardware, software or both. The fact that your desktop quad i7 can solve complex systems of equations in real time does not make a <$10 FLAC player run faster. Granted, most weak processors either won't play a FLAC -8 file or will cause audible clicks or stuttering, and some here have labeled that “broken”, but they’re not illegal, so they exist. But if longer gaps or delays (>10ms) create such artifacts, what about shorter gaps or delays that occur periodically?
I was curious, so I created a program that inserts a selectable number of zeroes or fixed values (zero order hold) every so often (set or variable period of time). Quick summary: some combinations were clearly audible, while others were inaudible to me. Intermediate values produce varying degrees of degradation. I wonder whether a degradation that isn’t obvious, can perhaps be ABX-able and simply sound “better”. Does this mimic a too-slow process? I don’t know. Do you?
Second, simple software/firmware bugs, other than time-inefficiency. They exist and when a clear problem presents (e.g. doesn’t play or no sound), they get fixed. But if a small, simple, non-fatal math error occurs, is it guaranteed that it will be found?
Before you claim these don’t exist, or at least not any more, google: flac playback problems. Recent biggies include Android 4.4 (Kit Kat - 2013) and VLC 2.2.0 (2015).
And finally, analog noise, dependent on processor “effort”. Assuming that this noise is not severe enough to disturb the digital bits in the signal path (good assumption, I think), the idea is additional current draw from an active vs. nearly idle CPU could find its way to the post-D/A-conversion side of the path. This idea is the most speculative. Although analog noise from USB data has been shown on DAC outputs (mostly the 8kHz microframe noise), it has been very low level. But it demonstrates the concept, and makes the claims of "impossible", "never" or "always" unnecessarily risky. I think “unusual”, “rare” and “typically” would be better choices.
I suspect that reports of hearing differences often occur when there actually is none, but to state that it cannot occur is not true.