Gary, I still plan to do some experiments in jitter. Unfortunately my tool only goes to a hundreds picoseconds of jitter so it is not accurate enough to measure such micro level changes.
Why DIY computer-based music servers are not ready for me
- Thread starter ack
- Start date
garylkoh
WBF Technical Expert (Speakers & Audio Equipment)
It might even be instructive to learn that there is NO difference with a resolution of hundreds of picoseconds.
I'm wondering what's the limit of hearing when it comes to jitter? Is it hundreds of picoseconds, or tens or single numbers? If I change the buffer length from 800ms to 780ms and I can hear it, will the objectivists on this forum say that I'm imagining things if we don't do a double blind test and your instruments cannot measure it?
It might even be instructive to learn that there is NO difference with a resolution of hundreds of picoseconds.
I'm wondering what's the limit of hearing when it comes to jitter? Is it hundreds of picoseconds, or tens or single numbers? If I change the buffer length from 800ms to 780ms and I can hear it, will the objectivists on this forum say that I'm imagining things if we don't do a double blind test and your instruments cannot measure it?
An objectivist isn't required, just someone who understands the power of expectation bias and the fact that no one is immune. At tolerances that small, blind listening is the only listening that is reliable.
Tim
When people start referring to picoseconds I can not avoid thinking about the intrinsic jitter of recordings. How can we be sure that all recordings are jitter free and the cases where we able to listen to jitter effects we are not listening to interaction between the two jitter spectra (from the recording ADCs and the DACs)?
garylkoh
WBF Technical Expert (Speakers & Audio Equipment)
micro, I gave up that line of thinking a long time ago as the things I cannot control - at the recording side. When I was still using transports and burning CD's I was amazed that even with the state-of-the-art at that time, with the Esoteric GR08 atomic clock, we could easily hear the difference between two burned CD's using a 'normal' CDR and an excellent CDR.
Jitter and Human Hearing
Steve Nugent said in an article that he believed the theoretical limit of human hearing was 15ps, though he didn't elaborate on his source or reasoning.
I'm wondering what's the limit of hearing when it comes to jitter? Is it hundreds of picoseconds, or tens or single numbers?
Steve Nugent said in an article that he believed the theoretical limit of human hearing was 15ps, though he didn't elaborate on his source or reasoning.
The actual value is pretty complicated in that it is frequency sensitive. Our hearing is most sensitive at 1 to 3 Khz so if jitter falls in that region, its levels need to be far lower than if it occurs elsewhere. What this means is that we must know the nature of jitter in order to say what its level should be to be below threshold of audibility. Professor Hawksford and the late Julian Dunn published an AES paper where they performed such an analysis. For a rather worst case 18.5 Khz sinewave jittter, the threshold of audiblity was determined to be just 20 picoseconds.
Real audio products have jitter components at a variety of frequencies and not just one component. This all then becomes a game of probability. What jitter does the system have at any one that that is not masked by the music. We know that if we achieve the above low targets,t hen that probability shrinks down to zero. Above that, it is an unknown number.
Amir, I hadn’t heard you say much about this for a long time, so I thought you had dropped the idea. If you ever manage to get the equipment together it would be a great tool to build the ultimate music server. We could measure the effects of different power cables, isolation devices, drives, motherboards, etc. Some guys on the DIY forums have experimented with tracing circuits on their motherboards and cutting the power to circuits like the onboard audio and networking, with the assumption that lowering current requirements will lower jitter.
I currently use latency tools to improve the sound of my server and there certainly is a notable difference between an idle reading of 5 vs 25. I use DPC Checker, because it’s easier to use than the others, but it’s not at all informative in tracing the latency source.
You may have already seen this, but the CMP2 project did a measurement of their jitter. I didn’t understand much about how they did it, but it didn’t sound like they required a team of scientists or massively expensive gear, though I may be wrong.
http://www.cicsmemoryplayer.com/index.php?n=CMP.03Jitter#Conclusion
Yes, I have seen that. They use an ordinary sound card and compute the jitter by measuring the sidebands in the spectrum analyzer. I tend to be dubious of using PCs as measurement gear in general. To the extent they have created a clean PC configuration perhaps that can be trusted.
Where did the idea come from that you have to optimize a computer for audio playback? I understand the need for memory if you have a player that plays from RAM, but as for the operating system itself, do you really need to tweak anything? I don't know anything about Windows, but for OS X, the case in favor of tweaking stuff is hardly compelling. Oddly, I've gotten a reputation as someone who advocates this -- I don't.
anyone with any thoughts on a very good sounding music server. over the pond looking at 64bit the music player , anyone any thoughts on sound quality , build , back up
Or just stick with one from this side of the pond. Linn merdian & naim
Plug into my calyx femto dac & rip all my cds and store my 24/96 , 24/192khz downloads
Or just stick with one from this side of the pond. Linn merdian & naim
Plug into my calyx femto dac & rip all my cds and store my 24/96 , 24/192khz downloads
First if jitter were the only issue it would be a simple problem to resolve. Nevertheless I have measured the master clock jitter of several sound cards and wound up focusing on the ESI Juli@. I got a 25 pS cycle to cycle jitter from a direct measurement of the clock, pretty much the noise floor of the analyzer (close to the state of the art). However I think the relevant measurements would be done the same way John Atkinson does in Stereophile. Since we care about the output, which is the accumulation of all the stuff in front, either adding or reducing jitter we should measure its impact on the analog output. That is not hard to do.
You need a good capture setup, either a good soundcard (the Juli above for example) or external ADC. Then an FFT program. There are free ones and inexpensive ones that are more than good enough. Finally a test signal. The JTest is fine for this. A web search should find one you can download. Matching published results will be difficult as will verifying the setup to make sure its not adding much of its own jitter.
Even if you do get really low jitter you still need to make sure the computer is not mangling the bits when you aren't looking (e.g. resampling), pumping a ton of noise down the powerline or into the air and that it manages your collection well. Each of these is an undertaking. You will learn a lot. You should get satisfying audio after the first weekend. It will improve forever and you will get lost in the myriad of options. A pc has millions of "moving parts" most of which are essential. I think done right it can easily exceed the accuracy of any other option available.
Similar threads
