The last point made by amir is the most crucial of all, and is the reason why the discussion of sources from the context of jitter will only cause arguments to go around and around and around in circles ...
Frank
Looking for a 5th Grader's Basic Understanding of "jitter"
- Thread starter audioguy
- Start date
Frank
It also provides greater isolation for each section, but as you say a design-implementation can be a source of noise.
As an example in Hifinews recently they reviewed the Yamaha music server, correlated jitter was nice and low but the Yamaha generated more random noise based jitter to the extent it did thicken the signal quite a lot.
The Micromega wifi airplay integrated originally also had problems caused by the internal wifi circuitry (which is meant to be resolved now).
Cheers
Orb
Music servers are counter-intuitive based on the simplest principles of digital audio. Remember, the first thing we did to improve computer audio was get the DAC outside of the computer box to avoid corrupting the analog output stage with that noisy environment. The server puts it right back in there. I still think isolation is the best path; galvanic and digital from analog. I know it has worked for me. I know my system's limitations, and understand that the power and bass of bigger systems makes them a completely different experience, but the clarity I'm getting from about 60Hz up is as good or better than I've heard anywhere. And I think a big part of that is due to a very clean, uncorrupted signal.
And of course I"m still skeptical of the discipline of the audio press. How did hifi news determine that the Yamaha server's signal was thickened quite a bit?
Tim
Because they use a process based upon the Dunn J-test and his papers and work (this is integral to the Miller Audio Research QC Suite measurement tool and jitter analysis), which means they can make a specific signal "music" file and measure the output from the music server of it being played.
This enables measuring even wireless solutions such as Sonos (John Atkinson did the same thing for the Sonos review and for music servers).
As a recent example in this months Hifinews the MF Clic was reviewed and measured.
They measured USB/SPDIF/network and could calculate both jitter and various levels of distortion.
It is possible then to show jitter-distortion anomolies (if any) along with the standard jitter measurement usually used.
This also fits into the discussion me and Amir were having in the Devialet thread, I should had responded with clear measured examples but of course me and Amir had to go geeky instead
Cheers
Orb
This enables measuring even wireless solutions such as Sonos (John Atkinson did the same thing for the Sonos review and for music servers).
As a recent example in this months Hifinews the MF Clic was reviewed and measured.
They measured USB/SPDIF/network and could calculate both jitter and various levels of distortion.
It is possible then to show jitter-distortion anomolies (if any) along with the standard jitter measurement usually used.
This also fits into the discussion me and Amir were having in the Devialet thread, I should had responded with clear measured examples but of course me and Amir had to go geeky instead
Cheers
Orb
Last edited:
Not all servers have DACs in them. The one I just purchased (The Music Vault) does not have one for the very reason you stated.
Because they use a process based upon the Dunn J-test and his papers and work (this is integral to the Miller Audio Research QC Suite measurement tool and jitter analysis), which means they can make a specific signal "music" file and measure the output from the music server of it being played.
This enables measuring even wireless solutions such as Sonos (John Atkinson did the same thing for the Sonos review and for music servers).
As a recent example in this months Hifinews the MF Clic was reviewed and measured.
They measured USB/SPDIF/network and could calculate both jitter and various levels of distortion.
It is possible then to show jitter-distortion anomolies (if any) along with the standard jitter measurement usually used.
This also fits into the discussion me and Amir were having in the Devialet thread, I should had responded with clear measured examples but of course me and Amir had to go geeky instead
Cheers
Orb
OK. Sorry to be dense, but did they come up with any specific measurements that could be correlated to "thickness?" Changes in FR? Increased noise floor? Increased distortion of any variety? It is easily possible to show jitter distortion anomalies that cannot be heard. "Thickness" is a descriptor I think I can relate to -- congestion, distortion in the midrange, excess warmth -- lots of things I've heard come to mind, but they could all be traced back to a measured inaccuracy.
Tim
Not sure how you can get measured innacuracy from such tests Tim.
The thickness is widening of the test signal at the jitter frequency where it should be very narrow with minimal sidebands, it should not happen and potentially depending upon its width-output may affect the sound (this is not the behaviour for all types of generated jitter).
In such situations this I guess will become a debating point and as you have seen those pages even here where some argue jitter cannot be heard while others state it can, and this is made complicated by what type of jitter we are discussing and its measured trait.
However experience from Paul Miller suggests such thickening measured for the Yamaha causes a subtle loss of heard detail, in this situation we are talking about uncorrelated jitter and thickening at the root of the test signal, caused by phase noise on the streamed network connection.
Does this help?
I get the feeling my use of "thickening" was taken to mean a description of the sound, when it is actually a description of the measurement picture-data, the effect on sound possibly being a subtle loss of detail.
Found the magazine review I have for the Yamaha, PM specifically says:
Orb
The thickness is widening of the test signal at the jitter frequency where it should be very narrow with minimal sidebands, it should not happen and potentially depending upon its width-output may affect the sound (this is not the behaviour for all types of generated jitter).
In such situations this I guess will become a debating point and as you have seen those pages even here where some argue jitter cannot be heard while others state it can, and this is made complicated by what type of jitter we are discussing and its measured trait.
However experience from Paul Miller suggests such thickening measured for the Yamaha causes a subtle loss of heard detail, in this situation we are talking about uncorrelated jitter and thickening at the root of the test signal, caused by phase noise on the streamed network connection.
Does this help?
I get the feeling my use of "thickening" was taken to mean a description of the sound, when it is actually a description of the measurement picture-data, the effect on sound possibly being a subtle loss of detail.
Found the magazine review I have for the Yamaha, PM specifically says:
CheersPM said:
Orb
You got exactly the right feeling. I took "thickening" to be a description of the effect the jitter was having on the sound. Regarding the audibility of jitter, I certainly believe it is possible, though I suspect that it has to be pretty bad to be audible, even to trained listeners under ideal circumstances. And I firmly believe that jitter which is audible -- resulting, as in your example, of a subtle loss of detail -- should also be measurable, and I await such evidence from the purveyors of high-end jitter reduction. I know there are many who disagree, so let me add -- MHO. YMMV.
Tim
We can easily measure Jitter Tim. This is not like difference between cables and such where measurements are not revealing of the difference. They clearly are in this case.
The thickness btw will be caused by jitter at low frequencies. Jitter distortion sidebands are equal to sum and difference of jitter frequency and your signal. If you have jitter at 100Hz, and a signal at 1Khz, you will have new sidebands at 900 and 1,100 Hz. Seen on a spectrum analyzer, they will tend to crowd the shoulders of the original 1 Khz frequency making it look "thicker." This gets worse as you have more and more jitter sidebands.
The thickness btw will be caused by jitter at low frequencies. Jitter distortion sidebands are equal to sum and difference of jitter frequency and your signal. If you have jitter at 100Hz, and a signal at 1Khz, you will have new sidebands at 900 and 1,100 Hz. Seen on a spectrum analyzer, they will tend to crowd the shoulders of the original 1 Khz frequency making it look "thicker." This gets worse as you have more and more jitter sidebands.
Ah good, the force is strong in me then
Well it is measured and shown by PM, why I chose that example.
The debate for some is does this actually translate as audible, and would feel the description giving by PM in this example even though anecdotally described by the reviewers is far from enough.
In a way they are right but then how far does one want to go for their proof, nothing will be enough until peer reviewed scientific studies back them up, which unfortunately we have seen in various papers never to be done to the correct detail (such as focus on artificial random jitter and stating that as jitter not heard - means absolutely nothing).
The closest we get to the truth is the practical real world measurements provided by such as PM and JA, utlising both Audio Precision with PM's Jitter Measurement software, but sadly only possible to correlate with anecdotal subjective comments due to never being used in the science study to compare with real world products.
Cheers
Orb
Yeah I get all that, Amir. What I'm talking about is measurements of the audible effects of jitter -- how it impacts frequency response, distortion and the noise floor. You know me, I believe if it can be heard it can probably be measured, and if it can't be measured within the audible spectrum, I'll have my doubts. Not absolutes, mind you, but doubts.
Tim
Let me recap a little, somwhat simplyfied...
Rise time of a Sqwave can be distorted at the xport by mechanical issues, noise Emi/Rfi, drivers that pass the signal onto a Bus of some type where xmission line querks can be introduced to possibly distort more, then received at a Dac where some more distortion/noise could be introduced and eventually decoded from the digital Domain, until then all this distortion is cumulative.
Rise time of a Sqwave can be distorted at the xport by mechanical issues, noise Emi/Rfi, drivers that pass the signal onto a Bus of some type where xmission line querks can be introduced to possibly distort more, then received at a Dac where some more distortion/noise could be introduced and eventually decoded from the digital Domain, until then all this distortion is cumulative.
Problem is as I said, IMO no really useful studies have been done as their focus is not comparable to real world jitter behaviour, which is further compounded by say focusing only on simulated uncorrelated jitter that may not be comparable to the specific jitter patterns that are of concern as seen with AP and Miller Audio Research Jitter Measurement software.
If you do not accept the measurements having meaning from these tools and shown by JA and Paul Miller, it will be impossible for you to ever shift position from doubt, unless very indepth future studies are done (and I doubt it for jitter as it is becoming less of an issue with modern technology and designs, although it still can be a major headache in some technology-situations such as HDMI where its performance is far from consistent between products).
Ronm1, thanks for expanding and yeah agree.
Thanks
Orb
Similar threads
- Sticky
