JohnSwenson;698661 Wrote: > The recent questions in this thread are actually very important and are > at the crux of a LOT of discussion in audiophiledom these days. I'll > try and take a stab at answering some of these based on what I have > uncovered by designing my own DAC over the last 10 years or so and > listening to and trying to analyze many others "out there". > > There are really only three external world influences that can affect > the sound coming out of the DAC: > > 1) the bits! > 2) timing and noise on the input signal > 3) noise on the power supply (this includes ground noise) > > There can also be EMI from outside sources getting into the box, but > for most reasonable designs this is such a low level I'm not going to > include it. (primarily because the measures used to screen a device so > it meets regulatory emission limits, means it also screens out external > influences) > > We cannot completely ignore #1 at this point since some of the most > popular methods for decreasing jitter DO change the bits, I'll get into > that later. > > #2 is where most of the discussion centers around. #3 is usually > ignored but can be a significant contributor. > > So I'll talk about #2 first. Lets look at coax first. The signal is an > analog voltage and current, ones and zeros are a way of interpreting the > different voltage levels. The voltage on the wire takes time to change > from a zero to a one and vice-versa. When its at a high or low voltage > level there can be noise "riding on top" of the desired signal. Higher > frequency noise shows up as voltage variations (wiggles) in the "steady > state" areas between the transitions. In addition you can have low > frequency noise (usually power supply related) which slowly moves the > whole signal up and down. Then there are what are called "reflections", > when these sharp changes in voltage hit an impedance difference > (connectors, cables, components on a board) part of the signal gets > reflected with an inverted sense back down the other direction of the > wire. If there is also a discontinuity at the transmitter end the > signal gets reflected back AGAIN, so now its sitting on top of whatever > the transmitter is NOW sending. A pulse can go back and forth several > times before it damps out. > > The result of all this is that the signal at the receiver chip is far > from perfect. The total result of all of the above makes it very > difficult for the receiver chip to determine exactly when that signal > changes from a zero to a one (and vice-versa). Figuring this out is a > very important part of a receiver. There are several different methods > used to try and do this well even with all this junk on the signal > trying to make it difficult to do. Some are more successful than > others. Some cost more money than others. Some have drawbacks which > make then not very popular (like it takes a couple seconds to lock onto > a stream) > > The engineering comunity has pretty much settled on one way of doing > this which is used for at least 99% of inputs. This approach gives > somewhere between 200ps and 50ps of jitetr on the recovered clock. (ALL > methods are bit perfect) As is these receivers are fairly susceptible to > changes in the crud on the input signal. The jitter on the output clock > (and its spectrum) will vary significantly with changing conditions of > the input signal. > > There have been a number of techniques used in different DACs to try > and improve this by employing different techniques AFTER the primary > input receiver. BUT there is another process going on which has been > almost completely ignored, the amount and timing of current drawn by > that receiver chip varies significantly with whats going on with the > input signal. These current changes cause noise in the power supply > rails AND in the ground plane of the DAC. Some designers will use > separfate regulators for the receiver chip and for the rest of the DAC, > but almost everybody ignores the noise on the groundplane. This noise > can cause jitter on the clock and can get into the analog signal > directly. So even if the designers spend a ton of effort to "clean up" > the signal coming out of the receiver chip, the input signal condition > can still affect the audio output. > > There are three primary methods used to clean up the clock from the > receiver chip: > 1) separate narrow band PLL > 2) buffer with slowly adjustable low jitter clock > 3) ASRC > > #1 used to be fairly popular and a number of "reclocking boxes" of the > past were built using this technology. The results can be quite good, > but there is a tradeoff, the lower the jitter the longer it takes to > lock on to a signal. > > #2 can produce extremely good results, but is much more costly to > implement and usually takes a LOT of development time. Because of this > very few have ever been actually built. > > #3 is the ruling darling of the industry. It seems like a slam dunk, > you have a local very low jitter clock, feed it to the ASRC and out > comes a stream that is synchronized to that clock. What could be > better? Unfortuntely this process DOES change the bits! That is it's > whole purpose in life, for a given point in time it looks at a whole > bunch of samples before and after that point and takes a little of > this, a little of that, munges them around and comes up with a value > that it thinks should be the vallue for the output clock. I'm not going > into the details of how they work (its fairly complex), but > theoretically it should do a good job. But unfortunately we live in the > real world and EVERY ASRC design on the market today makes compromises > in the design. They all cut corners to save cost. Now of course every > designer will tell you "Of course it doesn't make any difference, you > can't possibly hear the results of THAT". > > From my listening to these devices I've come to the conclusion that yes > you can hear the difference. Its not an obvious distortion, its an > obscuring of detail, the infamous "veils". So why are they so popular? > Because it DOES work, its the great equalizer, you don't have to work > nearly so hard on the input circuit, just shove ti through an ASRC and > you cut down significantly on the outside world interactions. BUT then > you get that subtle degradation of sound. > > The upshot is that a DAC which does NOT use an ASRC and spends a lot of > work on the input side, AND you feed it with a good source, good > impedance match etc will sound better than the same design which used > an ASRC. > > So this is where we are in audiophiledom right now, people that have > DACs with ASRCs can pretty much ignore a lot of this tweaking sources, > cabling etc. But those whose DACs do not have ASRCs will get different > results with the tweaking etc, and if they get it right will have > somewhat better sound than those with ASRC DACs. If they DON'T get the > tweaking right it will sound much worse. > > Every user needs to decide where they want to be. Choose an ASRC DAC > and ignore all this tweaking, or get a non ASRC DAC and and sweat the > details in the hopes of someday getting it just right and getting blown > away. > > BTW what do I like? I prefere #2 in the list above, its hard to do > right but the results are astonishing if you do get it right. Of course > then you have to deal with the ground plane issue, but thats a whole > different ball game. > > John S. John, I have two questions 1) given your objection to ASRC, does it follow that not only is it necessary to use a different method of cleaning up the clock, it is necessary to use a multibit dac chip- won't a delta sigma chip require rather more drastic sample rate conversion? Is it just the asynchronousness (asynchronicity?)which you object to? I appreciate that ASRC may end up slightly changing the information in extreme cases, whereas all SRC will alter the nominal data values. 2) assuming that we use the buffer and slowly adjustable clock option, why now should jitter and/or noise in the S/PDIF signal make any difference to the output- is it just the ground plane noise? presumably you agree that it is possible to read the data perfectly and to have a clock which is not derived from the S/PDIF stream.
-- adamdea ------------------------------------------------------------------------ adamdea's Profile: http://forums.slimdevices.com/member.php?userid=37603 View this thread: http://forums.slimdevices.com/showthread.php?t=94352 _______________________________________________ audiophiles mailing list [email protected] http://lists.slimdevices.com/mailman/listinfo/audiophiles
