Measurement and subjective hearing of uncorrelated jitter as seen in HiFi News

This may had been better put in one of the expert sections but anyway, and I want to stress this is anecdotal so please do not be upset if feel that it does match your view or experience.

For some time Paul Miller thinks-experienced that there can be a loss of focus/sharpness to the sound-image if uncorrelated jitter widens the signal as seen from his detailed jitter plot.
Unfortunately the measurement graph I am talking about is in the paper-magazine publication and not his online site.
There has been previous examples of products reviewed where the mention of the above factors have been commented on by the reviewer without them knowing about the measurement.

However in the September 2011 issue, this has been usefully emphasised by the review of the Chord Electronic QBD76 DAC.
This DAC has a setting for buffering and can be disabled or set to a maximum of around 4 seconds, Chord Electronic developed this many years ago and has fine tuned ever since.
What is interesting is that the measurement shows correlated jitter only changes marginally between the buffer disabled and set to a maximum of 4 seconds (jitter went from 15psec to 13psec), but looking at the uncorrelated graph it is possible to see a widening around the test signal when disabled, and reduced a lot when set to maximum.
The reviewer of the product was John Bamford who has a lot of experience in the audio world and in his review says:

I found the buffering improved the sound markedly, sharpening the image and increasing clarity, so left it engaged max for listening to music.

Click to expand...

Still anecdotal, but a good indication of what Paul Miller suggests are the effects of uncorrelated jitter when involved in the thickening of the signal.
Thought it worth sharing anyway.
Cheers
Orb

Yeah by thickening it is the skirts around the test signal, which is narrow at top and broadens out towards the bottom.
And yeah sorry if it was not clear but the buffering did reduce the skirting effect, although I always thought buffering also affected correlated jitter, which seems minimal to say the least; although in this case it is fair to say the correlated jitter is very small in the first place and that helps I would say.
But anyway, it definitely is interesting as we are not talking of extreme skirt but a noticeable difference in this instance, shame the graph is only available for the paper magazine.

Cheers
Orb

Sorry for the delay responding, spent a bit of time searching previous Hifi-news with other products measured so I can show example of each; before buffer and after buffer, which helps to provide perspective on what products match the before and after.

Don,
yeah I agree that this is an incredibly complex situation (given the different hardware components-standards involved) and jitter patterns are anything as simplistic as my one liner provided

Buffering does or can have a benefit for correlated jitter for audio products but it is all relative to the performance of the hardware.
An example of the correlated effect are the side bands shown in the J-test as these are odd harmonics that are excessive and related to the intersymbol interference (ISI) mechanism, which is associated to data dependant jitter.
And depending upon the quality of the product can vary greatly, but these days more FIFO buffers seem to be used to assist, albeit in most cases marginally compared to the 1-to-4 second buffer delay in Chord Electronics gear.
So there are examples out there showing some kind of buffering helped to resolve the odd-harmonics "anomaly" and as you mention also spurs/spurious tones.

But yeah I agree, this is oversimplifying the whole aspect of jitter and the complete end-to-end architecture of CD replay.
Still, with that said, with average levels of data related jitter in a product, the buffering should also improve this, just unfortunate the QBD76 only has 15ps of jitter in the 1st place.

As an example, here is one of the earlier DAC64 models (bear in mind this has been revised around 5 times I think) showing both the noise skirt improvement and also both correlated data related sidebands and other spurious tones.
Those looking at the graphs please appreciate it is also showing a faulty initial review sample as well, so need to check them carefully; Fig 12 is the correct working sample.
http://www.stereophile.com/content/chord-electronics-dac64-da-processor-measurements-part-2

Section 2.6 in the Dunn paper expands on the data-related J-test, that utilises the ISI mechanism;
http://www.nanophon.com/audio/diagnose.pdf

I think a fair few will skip most of what we are saying, so I just want to emphasise that buffering is beneficial for coping with and resolving various types of jitter and hardware related issues, whether it is done at the CD/DVD mechanism when reading, received from the source,etc.

Back on topic.
I will go through the hifinews publications hopefully later on looking at the skirts and posting ones that look comparable to the QBD76 when buffer is engaged or not, as it is interesting that the focus/image quality was subjectively identified with a change in the skirt.

Cheers
Orb