Interesting Thread re: Jitter

rskarvan

Disclaimer: I did not author this. - rskarvan

This has been covered before in r.a.o., r.a.h.-e., r.a.p., but it's always
useful to go over it again. First, though, I'll make two important points:

1. I make no claims about audibility.
2. I'm not talking about how a good DAC *should* behave, but about how most real
DACs *do* behave.

My assertion, then, is the following:

"The *measured* analog output for identical digital inputs can be different
depending on the digital cable used for digital signal transmission from the
digital source to the DAC."

The reason for this has little, if anything, to do with the ones and zeros being
interpreted incorrectly ("data errors"). Rather, the data encoding scheme
(S/PDIF or AES/EBU) can cause timing errors. That's because the signal
transmission is *asynchronous.* You see, the digital data are crammed down the
wire serially, one bit after the other, without an explicit clock signal to tell
the digital input receiver in the DAC when one word of data ends and the other
begins (roughly speaking). The signal transmission scheme is based on biphase
mark encoding, where the time period for each bit is divided into sub-bits; a
zero is encoded as no signal transition at sub-bit boundaries and a one is
encoded as a signal transition at the sub-bit boundary. Furthermore, there is a
forced signal transition at every bit boundary and there are occasional coding
"violations" and "preambles" to help the receiver find bit boundaries and word
boundaries. I'm blathering on, but I just want to illustrate that there's a lot
to this business of recovering an asynchronous stream in real time.

Anyway, the long and short of this is that the digital input receiver not only
has to recover the correct data from the asynchronous stream (easy), it must
also recover the clock that tells the DAC chip *when* the correct data needs to
be converted to analog (not so easy). The stream of biphase-mark-encoded ones
and zeros is actually an analog signal, with associated rise and fall times,
skew, overshoot, eye-pattern closure, etc. The more distorted this signal gets,
the harder it gets for the digital input receiver (more precisely, VCO/PLL) to
extract a good, steady clock signal from it. I assure you, different cables can
cause varying amounts of distortion of the *analog* signal going from the
transport to the DAC. Hell -- you can see it on a scope. When the zero-to-one
and one-to-zero transitions move around in time from where they're supposed to
occur, this is called jitter.

The net result is that even though we can correctly interpret the ones from the
zeros, we can end up with the correct data at the wrong time. This leads to
distortion in the analog output of the DAC. Again, I'm not saying it's audible
and I'm not saying it has to be this way. I'm just saying you can *measure* it
and it frequently *is* this way. There are well-known ways to minimize the
digital input receiver's sensitivity to jitter, but no matter how well they are
implemented, you can almost always at least measure some distortion due to
injected jitter.

Anyway, it's trickier than you might guess and once you wrap your head around
it, you might wonder why "they" don't just provide a clock line or use a
different transmission scheme. Good questions....

atrustsd

yeah what ever you say. Good job

audiobliss

Wow, I'm actually glad you pulled that back up to the top. That was a really interesting read.