Bit perfection

[Theo]

The concept behind "bit perfect" is that it is a guarantee that the digital data has not been corrupted. So a lot of software, equipment, cable... makers claim to be "bit perfect" so music will sound a lot better, veil lifted and all... However, a proper interpolation or resampling algorithm working in 32 or 64 bit depth shouldn't distort the signal above the calculation error, which is quite small if you do it properly. So, is bit perfection a necessity or just another audiofool concept? Are there measurements showing a comparison between BP and not BP?

 

[SIY]

Having electrons all be the same charge is vitally important as well.

That's my sarcastic way of saying that the VAST majority of equipment is, from the standpoint of digital transmission, essentially bit perfect. Ditto things like noise shaping and similar resampling. You have to work hard not to make it so. When digital data are lost, the audible consequences are obvious- pops, clicks, silence, stutters. The idea that imaging, tone, or mysterious qualities of "musicality" or "PRaT" are somehow affected is indeed audiophoolery.

Now, what's NOT bit perfect? Lossy data compression (e.g., MP3) and systems like MQA. But these are deliberate add-ons. Using lossless source files (e.g., FLAC, WAV) will assure that you are working "bit perfect."

 

[miero]

Some DACs have a built-in test for bit perfect transfers. This is good to find if a playing system is "incorrectly" configured, for example a low quality resampling is used.

And it's a bit ironic that the MQA (lossy data compresion) requires the bit perfect data transmission to the DAC.

 

[solderdude]

MP3 may not be perfect but has nothing to do with bit-perfect.
For MQA things are different as that may require software decoding and in the DAC the last bit of decoding.
Arguably the file and what goes to the DAC is not the same, but should be closer to the signal present during the encoding.

AFAIK Bit-perfect means (or should mean) the bits that describe the audio signal (sample values) from the music file is arriving not altered in any way at the input of the DAC (in this case).

Add digital volume control or other processing and the data arriving at the DAC differs from the original file.
So not bit-perfect by definition.
Of course when only the volume is changed, in a good way, the final analog waveform is not different other than in amplitude and (IRL) the signal to noise ratio is smaller (due to noise in the analog domain having a constant level).
The S/N ratio of the digital signal may well be unaltered.
The same goes for DSP which is intended to alter the original waveform.

The term bit-perfect is often misused to describe other things IMO.

All DACs that oversample or have digital filtering inside (all SD) are altering the applied sample values by definition anyway so the whole 'bit perfect' stuff is nonsense anyway, when viewed from the position that the sample values at the output have the exact same value as the discribed bit value at that time. True for all DACs including R2R.
This is what most people think it means... the actual output voltage of the DAC is exactly the same as the value described at that moment.
It is always a very, very, very close approximation. Less so for R2R as they are all plagued by linearity issues.

 

[sergeauckland]

Having electrons all be the same charge is vitally important as well.

That's my sarcastic way of saying that the VAST majority of equipment is, from the standpoint of digital transmission, essentially bit perfect. Ditto things like noise shaping and similar resampling. You have to work hard not to make it so. When digital data are lost, the audible consequences are obvious- pops, clicks, silence, stutters. The idea that imaging, tone, or mysterious qualities of "musicality" or "PRaT" are somehow affected is indeed audiophoolery.

Now, what's NOT bit perfect? Lossy data compression (e.g., MP3) and systems like MQA. But these are deliberate add-ons. Using lossless source files (e.g., FLAC, WAV) will assure that you are working "bit perfect."

This should be writ large at the heading of every audio forum. (except this one, we know that!)

S.

 

[Soniclife]

This is what most people think it means... the actual output voltage of the DAC is exactly the same as the value described at that moment.

Is it? To me it's the right data being delivered to the DAC.

It's an important concept when we live in a world where people like Google and Apple are likely to end up doing the bulk of the data delivery, and they don't care about such things. Being able to shame them with a popular term is a good thing, Google did fix the CCA eventually.

 

[Vincent Kars]

IMHO this has a lot to do with using a PC as a source or as a DAW.
Depending on the OS, resampling could be disastrous (the Win XP Kmixer) or showing artifacts (Win Visa and higher).
Almost all OS do mix, hence convert to float, dither and convert back to integer.
Totally unnecessary if only 1 stream is playing. If the whole path is 16 bit, this might have a slight negative outcome.
Badly implemented audio software can also cause distortion

It is measurable.
Couple of examples can be found here:

http://www.thewelltemperedcomputer.com/Intro/SQ/SampleRateConversion.htm

http://www.thewelltemperedcomputer.com/KB/SRC.htm

 

[miero]

Comparison of resample algorithms in many audio tools:
- http://src.infinitewave.ca/

 

[Veri]

MP3 may not be perfect but has nothing to do with bit-perfect.
For MQA things are different as that may require software decoding and in the DAC the last bit of decoding.
Arguably the file and what goes to the DAC is not the same, but should be closer to the signal present during the encoding.

AFAIK Bit-perfect means (or should mean) the bits that describe the audio signal (sample values) from the music file is arriving not altered in any way at the input of the DAC (in this case).

Add digital volume control or other processing and the data arriving at the DAC differs from the original file.
So not bit-perfect by definition.
Of course when only the volume is changed, in a good way, the final analog waveform is not different other than in amplitude and (IRL) the signal to noise ratio is smaller (due to noise in the analog domain having a constant level).
The S/N ratio of the digital signal may well be unaltered.
The same goes for DSP which is intended to alter the original waveform.

The term bit-perfect is often misused to describe other things IMO.

All DACs that oversample or have digital filtering inside (all SD) are altering the applied sample values by definition anyway so the whole 'bit perfect' stuff is nonsense anyway, when viewed from the position that the sample values at the output have the exact same value as the discribed bit value at that time. True for all DACs including R2R.
This is what most people think it means... the actual output voltage of the DAC is exactly the same as the value described at that moment.
It is always a very, very, very close approximation. Less so for R2R as they are all plagued by linearity issues.

solderdude for president

Comparison of resample algorithms in many audio tools:
- http://src.infinitewave.ca/

isn't this site pretty old though? Still a valuable resource for SRC freaks I guess

 

[SIY]

If you're still using XP, "bit perfect" is the least of your worries!

 

[mitchco]

Fun testing bitperfect

https://audiophilestyle.com/ca/bits...gital-Audio-–-Bit-Perfect-Audibility-Testing/

https://audiophilestyle.com/ca/ca-academy/JRiver-Mac-vs-JRiver-Windows-Sound-Quality-Comparison/

https://audiophilestyle.com/blogs/entry/161-1644-vs-24192-experiment/

https://audiophilestyle.com/blogs/entry/154-jriver-vs-jplay-test-results/

https://audiophilestyle.com/blogs/entry/153-flac-vs-wav-part-2-final-results/

http://archimago.blogspot.com/search?q=bitperfect

 

[Blumlein 88]

https://wiki.jriver.com/index.php/Audio_Bitdepth

Look at the section Bit Perfect on this page. It describes how 64 bit volume control is not a problem. The errors with 64 bit volume processing are so small even 100 million conversions will not change the waveform enough to alter the 24 th bit in a 24 bit file.

Otherwise bit perfect is important for reasons like the OS doing poor conversions which happens way more often than it should at this date in time. Quality DAWs or playback software normally won't be a problem even when doing volume control.

 

[tmtomh]

I of course agree that poor and unnecessary conversions should be avoided whenever possible. For example, if you have a music library with files in a bunch of different sample rates that are not even multiples of each other - for example 44.1k and 96k - then it makes sense to have a playback system that can do on-the-fly resolution switching so that you are not stuck with a single, brute-force downsampling of 96k to 44.1 or upsampling of 44.1 to 96.

But that aside, I do find the obsession with bit-perfect output slightly amusing, since as far as I know things get distinctly non-bit-perfect inside almost every modern delta-sigma DAC ever made. In most cases, your PCM or DSD signal gets turned into some kind of 4 or 5-bit PCM/PWM hybrid before being spit out the other end. And the result is still functionally bit-perfect (if not technically so) despite that.

So I wouldn't say bit-perfect is audiophoolery. but I would say that as with many audiophile articles of faith, people tend to overgeneralize about it, turn it into an absolute requirement, and not be terribly discerning about when it's important and when it's not.

 

[LF78]

In my opinion expecting a bit-perfect path from source file to DAC is totally reasonable, given the fact that a non bit-perfect path is most probably caused by bugs or bad software/hardware design.

From DAC to analog output: what does it even mean? DAC's output is an analog signal, not bits.

 

[Vincent Kars]

given the fact that a non bit-perfect path is most probably caused by bugs or bad software/hardware design.

No, you will have a non-bit perfect path the moment you apply any kind of DSP in your media player, be it volume control, EQ, DRC, etc.

 

[LF78]

No, you will have a non-bit perfect path the moment you apply any kind of DSP in your media player, be it volume control, EQ, DRC, etc.

Of course, but here we are talking about intentional filters/modifications, not unwanted/unexpected data corruption. Let me rephrase: "In my opinion expecting a bit-perfect path from source file to DAC (or DSP/digital EQ/digital xover/volume control) is totally reasonable, given the fact that a non bit-perfect path is most probably caused by bugs or bad software/hardware design."

 

[mansr]

A bit-perfect path from storage to DAC chip is essential as a starting point. It means you have full control and can add quality processing steps if desired. If you can't get a bit-perfect stream, something outside your control is messing with the data. Often, this is OS level resampling and mixing, which can be quite poor. It's about avoiding unwanted processing, nothing else, no phoolery.

 

[tlr]

I agree with @mansr and others. For me, verifying a bit-perfect signal path is about the peace of mind that nothing in the chain is mixing, downsampling, reducing bit depth, etc. Maybe it makes an audible difference, maybe it doesn't, but if you can easily remove that variable, why not do it?

 

[HammerSandwich]

Serious answer:
Digital audio would never have been developed without a solid understanding of the relevant math. The obvious conclusion is that DSP volume/EQ/etc., done properly, isn't an issue. Verify that the signal path works, then worry about your acoustics & speakers. Those things definitely affect more than the LSB.

Smartass answer (which is no less true!):
What percentage of the people who obsess over bitperfect playback store their files on ZFS/BTRFS? QED.

 

[garbulky]

I guess if you are talking about "bit perfect" you could talk about multibit NOS DACs. Schiit claims that their multibit filter is also bit perfect though it does perform oversampling. However you still have pre and post ringing with Schiit's filter. I'm not sure of NOS DAcs though.