JohnSwenson;216719 Wrote: > Bring out the shotguns, here is a hopefully not too boring attempt at > going over jitter, interfaces, audio, USB and anything else I can think > of to throw in here. (BTW non of this is new or uniquely my own, its > based on reading a lot of stuff and my own experiments, measurements > and listening) > > Jitter, what is it? In a nutshell its the variation in timing between > signal edges. In particular we are dealing with the jitter on a clock > used to clock out data in a DAC. The clock is supposed to be converting > the data at precise intervals (44.1 KHz, 96KHz etc) If the time from one > edge to another varies from one clock pulse to the next, there WILL be > distortion in the analog waveform. > > Jitter is usually refered to in the hifi press in terms of a single > number, such as 250 pico seconds. Unfortunately that does not > characterize jitter very well. You find out that the one clock might be > off from perfect by 50ps, the next by 135ps, the next by 600ps etc, it > varies all over the place. So if you want to use time numbers you have > to use averages, or other statistical means to try and come up with > something meaningful. What this means is that "500ps" from one device > can have very different actual jitter than "500ps" from another device. > > > Most technical people use the term "phase noise" to describe jitter, > this is usually displayed as a frequency spectrum of the clock. This > seems to have much better correlation to audibility than just the > single time number, but its much harder to understand and compare, so I > understand why the press doesn't use it. Why is this important? Because > the frequency of the jitter relates directly to sidebands in the > recovered audio. If you have jitter that is varying in a nice > repeatable pattern that shows up in a spectrum, that will produce > sidebands in the audio signal. (its not quite that simple, but I really > don't want to get into that here!) These sidebands are NOT harmonically > related to the "signal" so they are not easily masked. > > Listening seems to indicate that "random jitter" which looks like noise > on the phase noise spectrum is much less audible than phase noise at > specific frequencies. The S/PDIF receivers in common use are famous for > their narrow frequency jitter spectrum. Very narrow specific frequencies > but at fairly high levels. Other systems such as some types of USB > receivers LOOK much worse, but they are much more random with the > jitter spread out over broad frequency ranges. > > Since I'm on S/PDIF lets go breifly into that. As has been mentioned by > others, the issue is NOT data integrity, its the jitter in the recovered > clock coming from the S/PDIF receiver. As I mentioned earlier the > spectrum of the recovered clock contains some strong "spikes" which > seem to be audible under some circumstances. This is using the > traditional "transport is in control" methodology, the DAC has to > synchronize itself to what is happening in the transport. It is > possible to break this relationship,but no tvery many DACs actually do > that. One way is to have a separate low jitter clock which "clocks out > the data", the problem is its not synchronized with the input stream, > you will either drop samples on the floor or duplicate samples, this of > course DOES cause waveform distortion. > > You can stick a buffer in, but in order to handle any situation the > latency of the buffer gets to be somewhat long, people don't like > pushing play and waiting several seconds for the sound or having a lag > time when pushing buttons on the transport. The latency also makes it > really difficult to use such a DAC with video! There are some very > fancy buffer management techniques that can alleviate some of these > issues, but very few companies have tried to deal with that. > > There are a whole host of other methods to try and make a DAC less > sensitive to jitter coming from the receiver. Some more effective than > others. > > I've done an experiment where I look at the phase noise of a recovered > S/PDIF clock and tried different cables, connectors, moving the cable > etc and all this things DO make changes in the phase noise of the > recovered clock coming out of the S/PDIF receiver. It was really > interesting to watch the spectrum change as I moved the cable around! > (this means that you might actually have vibrations of the cable > changing the clock jitter!) > > Now on to USB, there are three official USB audio "modes", all of these > are called "isochronous" this means the bus reserves space for these > packets, nobody else using the bus can interrupt them. There is > synchrounous, adaptive and asynchronous. Synchronous sticks a PLL > directly on the USB data stream, the jitter is terrible and extremely > sensitive to what happens on the bus. It has the same problems as > S/PDIF, the source is in control. Very few modern devices use this. > Some early ones did which is where you here about USB having 1500ps > jitter or some such, again this is not the norm today. > > Almost everything today uses adaptive mode, again the Source is in > control, but the reciver has its own clock generator which is slowly > changed to match the average data rate coming across the bus. These are > MUCH better than synchronous, but there is still a variable frequency > clock generator in there. The different implementations vary quite a > bit with the best having quite low jitter without large spikes in the > spectrum. All of these are quite sensitive to implementation, board > layout, supply bypassing, power supply noise, jitetr of the clock > feeding the frequency synthesizer all can make huge differences in the > jitter even with the same chip using the same value components! > Implementation makes BIG difference. Again there is NO feedback to the > host it just sends out packets at what it thinks is the right speed and > the DAC has to synch with it. > > There is also asynchronous mode, here the data is clocked out by the > local oscillator when the buffer starts to over or underflow it tells > the host to slow down or speed up the data stream. This is the only > mode where the DAC is in control. The host is still sending the packets > isochronously, there is no packet by packet handshake or anything like > that, but there IS a path back from the reciver to the host. This is > almost never done. There are only one or two USB chips that support > this mode and ALL of them require re-writting the firmware inside the > chip, not an easy task. I only know of one person who has ever done > that. I tried for a long time and gave up. (thats a whole nother > story). > > There a few devices that DO have the DAC in control (such as the EMU > 0404 USB) but they do not use the official USB audio spec, they came up > with their own packet protocol which means they have to write their own > drivers. > > Back on S/PDIF and external DACs, the jitter and power supply noise of > the signal driving the S/PDIF transmitter CAN show up as increased > jitter in the recovered clock, this is why various "digital mods" > actually can make a difference. BUT you still have the jitter from the > reciver circuit, these two (source jitter and receiver jitter) interact > in interesting ways. It is definately possible that the changes in the > source jitter can be swamped by the receiver jitter, in such a case the > "digital mod" would hardly make a difference, in a different reciver > circuit the source jitter can be much more prominant, and in other > receivers both types can be diminished significantly. Thus if you don't > hear it in your system, doesn't mean it won't be audible in someone > elses. If you can' hear it doesn't mean that every one that can is > deluding themselves, and if you CAN hear it, it deosn't mean that > everyone that can't is deaf. There are so many different combinations > of devices and interactions between them that its impossible to make > blanket statements about jitter audibility based on your system. > > OK I hope I didn't bore too many people. I get carried away sometimes > and this thread was too big an invitation. > > John S.
Gulp! . -- haunyack Transporter -> B&K Reference 200.2 -> Vandersteen 3A Signature. RWA (Analog) SB3 -> Rotel RB 1070 -> B&W Matrix 805. Fridgidare -> Mirror Pond pale ale -> easy chair w/remote -> irritated neighbors. ------------------------------------------------------------------------ haunyack's Profile: http://forums.slimdevices.com/member.php?userid=9721 View this thread: http://forums.slimdevices.com/showthread.php?t=37044 _______________________________________________ audiophiles mailing list [email protected] http://lists.slimdevices.com/lists/listinfo/audiophiles
