Meridian Unplugged - Powered by MrTechGuy.co.uk Forums Discussion Meridian - Systems S/P-DIF format

Period is the American name for "."
Full stop is the English name for "."

In the UK, we would say "This is just the best, full stop." to mean that this is the best, and there is no doubt about it.

In America, they say "This is just the best, period.", meaning the same thing.

Martin was simply stating that he thought that the statement needed no argument, and being friendly to the English and the Americans by using both dialects.

You will notice, as you learn more about the English language that it is not one language. The language Americans call English is quite different, in many small ways, to the language the English call English.

It's all good fun.


Anyway, moving on to Jitter:

The S/P-DIF format is inherently jittery. This is because it puts data down the wire using a block type structure, with a regular pattern, every 192 samples. This patterning promotes jitter. This is made even worse by patterning that happens in the audio data caused by the waveform of the music.

Meridian goes to great lengths to minimise this jitter, but there's only so far you can go within the limits of S/P-DIF. Putting a FIFO in the speaker is the only way of removing audible the effects of jitter completely.

MHR is a scrambling system that makes the data on the wire look like random noise, therefore 'decorrelating' the data on the wire. The S/P-DIF block structure remains, but this system does remove jitter caused by patterns in the audio waveform.

Upsampling is a completely different issue. Running the system faster will push the audible effects of jitter up the audio spectrum, possibly pushing it out of audible range. I say *possibly* because I don't really understand how upsampling can significantly reduce jitter.


Simon.

Simon,

if I remember correctly the "time stamp" is removed from the digital data stream to give a purer sound. This is done before leaving the source (Meridian 596). So once the signal passed the FIFO circuit how can the FIFO do its work in the DSP speaker?

Speedy

OK, I've been meaning to respond to this, but haven't found the time to craft the lengthy response I have in mind.

However, I have to quickly chime in on a couple of things.

Simon, first off, please educate me on how SPDIF is "inherently jittery"? Being a hardware engineer designing high-speed networking equipment for the last 9 years, I don't buy this statement.

The only negative that I see to the SPDIF format on the wire is that it uses edge transitions to denote 1's and 0's, rather then voltage levels. This does complicates the construction of the clock-data-recovery logic of the receiver, but isn't, in and of itself, inherently jittery. I suppose its possible that corners have been cut on the design of SPDIF receivers for cost reasons, which may add jitter, but that's a slightly different issue.

Also, how exactly does the "block" format of SPDIF add to jitter? SPDIF organizes the audio samples into frames, each frame containing two audio samples or sub-frames, one for the left channel and one for the right. 192 frames makes a block.

With each sub-frame containing one-bit for channel-status, that's 384 bits that contain a repeating pattern, of which about 344 are defined as reserved and should be 0.

So, 384 bits of 12,888 are "repeating", which is just over 3%. So, how does this add to jitter as compared to the 6,144 bits representing "random" audio samples?

The bi-phase mark encoding of SPDIF actually does a great job at balancing the data on the interface. Regardless of the sequential number of 1's or 0's in the data stream, there is a guaranteed number of transitions, which is the ideal scenario for serial data transmission and clock/data recovery.

For those interested, the following is a link to a nice primer on SPDIF:
www.epanorama.net/documents/audio/spdif.html

Regards,

Dan W.