SPDIF sample rates

Hi,

I was looking into some technical details for High Definiton Audio (Azalia) and was doing some bandwidth estimates.

Azalia is based of a 48 KHz frame rate.

At 192 KHz, 8 channels, 24 bits per sample requires a bandwidth of
= 4 (multiple of base rate) * 8 (channel)* 3 (bytes per sample) * 48 (base rate) = 4.608 MB/sec

I was wondering what would the corresponding number be for SPDIF?

Thanks.

You arrived at the correct bandwith but for the wrong reason. It seems to me that you are equating 'frame rate' with packet size as if you were talking about the size of an IP frame in TCP/IP. It doesn't work that way for S/PDIF.

192 kHz sampling rate * 8 channels * 24 bits per sample = 36, 864, 000 bits per second (36 Mbps) = 4, 608, 000 bytes per second = 4.608 MB/sec but s/pdif cannot chop that up into packets of 48K bits and transport it.

S/PDIF supports a maximum of 96kHz/24 for 2 channels = 4.608 Mbps (bits not bytes). DTS 96/24 is the highest sampling rate/bit depth possible right now but remember it is NOT lossless - it is compressed - so the bandwith required is nowhere near what would be required to pass that stream if it were uncompressed LPCM.

I don't know how the specs for the Azalia are worded, but if it mentions a 48 kHz 'frame rate' I suspect what it is talking about is its ability to downsample 8 channel 192/24 to 2 channel 48/24 or 48/16 for output over s/pdif.

Thanks for the info.

One confusion which I have always had:
Can S/PDIF carry stereo, 5.1, 7.1?
And are the sample rates the same for each of these i.e. 96 KHz/2 ch/24 bit.

In the case of 5.1 and 7.1, how does it work?
Does the S/PDIF receiver convert the 2 ch stream to a 5 channel or 7 channel stream?

This could get complicated real quick, but I shall do my best...

From the perspective of the receiver, there is only 1 stream - a stream of bits. Some of the bits are control information and some of them are the actual data. Whether the data represents 1, 2, 5.1, 6.1 channels or whatever is soley dependent on how the data is coded. Some of the control information includes information on the type of data; so for example the receiver won't interpret the data as audio samples (PCM) when it is actually a compressed format like DD or DTS.

Because I don't know what a DD or DTS bitstream looks like, I'll instead use a 2 channel PCM stream (a WAV file or PCM from a CD) to describe it - the principle is the same for other formats.

A WAV file basically looks like this:
[WAV][Header - # of channels, sampling rate, bit depth, etc][Left channel sample 1][Right channel sample 1][Left channel sample 2] [Right channel sample 2] .... [Left channel sample N][Right channel sample N].

It is the samples that are put into s/pdif frames along with control information that describes what the samples represent. There are blocks, frames, and subframes. A block is a collection of frames and a frame is a collection of subframes. There will be two subframes in every frame for 2 channel audio. Actually I believe there are always two subframes regardless of the number of channels with one of the control bits indicating which 'channel' of the audio - that information is for the decoder to figure out how to turn the stream of bits into audio.

So if you were playing a wav file on your computer, the left channel sample will be put into one subframe and the right channel into the other along with control bits to indicate it is 2 channel PCM. The number of frames required and the rate at which they are sent is determined by the format of the digital audio. For 44.1/16 from a CD or WAV file, there will be 44,100 frames every second with 16 bits of data in each frame. In the case of PCM, the data bits are actual samples and can be played directly (after D/A conversion).

In the case of a compressed format like DD or DTS, the bits have to be assembled together and passed to the decoder to decode them into PCM samples. The control information indicates that the data portion of the frame is not actual audio samples. Note that this is why it sometimes takes a second for the receiver to switch from PCM to DD because it has to look at the incoming stream and determine what format the data is in and pass along the data to the proper decoder.

So s/pdif can theoretically carry any number of channels as long as the bitrate is within its limits.