Jump to content
IGNORED

DSD-wide (8-bit) test files


mansr

Recommended Posts

I'm toying with some software and would like to see how 8-bit data behaves. It appears to be a format used more during production than for distribution, so the usual suspects don't have any samples for download.

 

Mansr, thank you for information.

AuI ConverteR 48x44 - HD audio converter/optimizer for DAC of high resolution files

ISO, DSF, DFF (1-bit/D64/128/256/512/1024), wav, flac, aiff, alac,  safe CD ripper to PCM/DSF,

Seamless Album Conversion, AIFF, WAV, FLAC, DSF metadata editor, Mac & Windows
Offline conversion save energy and nature

Link to comment
I'm toying with some software and would like to see how 8-bit data behaves. It appears to be a format used more during production than for distribution, so the usual suspects don't have any samples for download.

 

I've seen how 5-bit delta sigma audio behaves (as 2L Records' ADC is front-ended with such a modulator) when it comes to its bandwidth, and it was very similar to today's DSD 128 & 256 formats.

Link to comment

For any audio applications we need (if talk about resolution):

 

1. 0 ... 20 kHz frequency range,

 

2. Noise floor in this range.

 

That above 20 kHz apparatus can eliminate without sound quality troubles.

 

Main part of DSD quality is noise management (push noise out 20 kHz).

 

N-bit in DSD is one of dimensions of the noise management.

 

We can reduce noise next ways:

 

1. More sample rate

 

2. More complex noise shaper in sigma delta modulator

 

3. More bits per sample

 

 

1. and 2. points both lead to more demands of calculation power.

 

1. and 3. points both lead to increasing storage area.

AuI ConverteR 48x44 - HD audio converter/optimizer for DAC of high resolution files

ISO, DSF, DFF (1-bit/D64/128/256/512/1024), wav, flac, aiff, alac,  safe CD ripper to PCM/DSF,

Seamless Album Conversion, AIFF, WAV, FLAC, DSF metadata editor, Mac & Windows
Offline conversion save energy and nature

Link to comment

I'm aware of those things. A multi-bit also format has the additional advantage of enabling some DSP operations, which is probably the main reason for using DSD-wide.

 

Now I was looking at the DSF spec and noticed it allows 8-bit data, so I wanted to find a sample to play with. If none exist, so be it.

Link to comment

1 bit DSD processed fully identically multibit DSD.

 

However multibit allow decrease sample rate with keeping of quality. It can give decreasing of demand in computing power.

AuI ConverteR 48x44 - HD audio converter/optimizer for DAC of high resolution files

ISO, DSF, DFF (1-bit/D64/128/256/512/1024), wav, flac, aiff, alac,  safe CD ripper to PCM/DSF,

Seamless Album Conversion, AIFF, WAV, FLAC, DSF metadata editor, Mac & Windows
Offline conversion save energy and nature

Link to comment

See http://dsd-guide.com/sites/default/files/white-papers/DSFFileFormatSpec_E.pdf

 

Page 3 (bottom) "Bits per sample". Can have value 1 or 8. And details in annotation 4.

AuI ConverteR 48x44 - HD audio converter/optimizer for DAC of high resolution files

ISO, DSF, DFF (1-bit/D64/128/256/512/1024), wav, flac, aiff, alac,  safe CD ripper to PCM/DSF,

Seamless Album Conversion, AIFF, WAV, FLAC, DSF metadata editor, Mac & Windows
Offline conversion save energy and nature

Link to comment

To make things perhaps more clear: mansr is adding DSD processing to Sox open source software, and wants "DSD-wide" files to check that it works properly.

One never knows, do one? - Fats Waller

The fairest thing we can experience is the mysterious. It is the fundamental emotion which stands at the cradle of true art and true science. - Einstein

Computer, Audirvana -> optical to EtherREGEN -> microRendu -> ISO Regen -> Pro-Ject Pre Box S2 DAC -> Spectral DMC-12 & DMA-150 -> Vandersteen 3A Signature.

Link to comment
Page 3 (bottom) "Bits per sample". Can have value 1 or 8. And details in annotation 4.

 

Yes, the item name is completely misleading, but the annotation is clear. There are both types of files around, with oldest bits in MSB or LSBs. It's the same stuff ASIO also has. However, ASIO has third variant which is the 8-bit sample format.

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
Yes, the item name is completely misleading, but the annotation is clear. There are both types of files around, with oldest bits in MSB or LSBs. It's the same stuff ASIO also has. However, ASIO has third variant which is the 8-bit sample format.

 

In that case, do you have any samples with oldest bit in MSB?

Link to comment
I don't have doubts on that, just the neither DSDIFF nor DSF support DSD-wide.

 

Sorry, I misunderstood you. DSDIFF definitely only supports 1-bit. If you're right about DSF not supporting 8-bit, that would obviously have to be stored in some other container. That also makes this spec a strong contender for the most misleading I've ever seen.

Link to comment
You can easily make one yourself by taking bytes as-is from DSDIFF which uses that bit order, you'll just need to rearrange to the different data layout.

 

Of course I could do that, but I prefer testing code using input files from an unrelated source. Otherwise I might end up with bugs on both ends that happen to cancel out.

Link to comment
It is from my code, just few seconds of RedBook converted to DSD. (I have support for both bit orders)

 

How did you figure out that "8 bits per sample" in the spec actually means "MSB first"? Are there other such file around?

Link to comment
How did you figure out that "8 bits per sample" in the spec actually means "MSB first"? Are there other such file around?

 

Did you check the Annotation 4 section in the spec referred to by Audiventory earlier? That describes the field in more detail.

 

[Annotation4]

For example, if dsd stream data is 0x00, 0x01, 0x02, 0x03, 0x04....., then this is the sample data in data

chunk.

If “Bits per sample” is equal to 1, then store the data as LSB(Least Significant Bit) first.

0x00, 0x80, 0x40, 0xC0, 0x20, ....

(00000000, 10000000, 01000000, 11000000, 00100000, ....)

If “Bits per sample” is equal to 8, then store the data as MSB(Most Significant Bit) first.

0x00, 0x01, 0x02, 0x03, 0x04.....

(00000000, 00000001, 00000010, 00000011, 00000100, ....)

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now


×
×
  • Create New...