Quote:
Originally Posted by Stoney /img/forum/go_quote.gif
Not so. First, the voltage that a set of bits is converted to is done by circuitry that is analog, and is prone to hysteresis and other analog errors. So, sound out is not identical just because bits are identical.
|
You're not talking here about the same thing you originally described. You explicitly mentioned digital circuitry before. The analog conversion circuitry (i.e. the DAC) is certainly affected by analog errors since that is where the conversion to analog is done. That is because this circuit is in essence analog, not digital. A good power supply is crucial at this point and it's worth pointing out that batteries as used in a typical DAP are much better than a power socket used by a typical CD player. In the purely digital circuitry higher up the chain the fact that any single bit has a slightly different voltage is neither here nor there as long as the bit is correctly represented as a 1 or a 0 and the tolerances are very high against this happening.
Quote:
Originally Posted by Stoney /img/forum/go_quote.gif
Second, the voltage that a digital "word" converts to has to be outputted at the exactly correct moment. "Timing thrown off" as you put it. A time error is indistinguishable from a voltage error. Digital signals are reclocked with phase-lock loops, but they aren't as stable as we'd like to believe. Just like a rocket is always correcting its course, so is a loop always adjusting its frequency. This spectrum of the "course corrections" is transferred directly into the music.
Think of an x-y plot of a straight line of points. If all the Y's are the correct values but the X's are a bit off, the line is no longer straight.
Jitter is very real and very audible. So much so that, to my surprise, the same CD burnt at different speeds sounded different, and the frequency band most affected went down with the burn speed. Others report the exact same thing. Bit by bit, the disks read the same, but players can't fully reclock out the jitter also encoded on the discs.
But, with MP3 players, it is much, much worse. The music is approximated by brief wavelets summed together, and has to be decoded. There is much variation and even judgment calls as to what sounds best in the decoding. No two algorithms should be expected to sound the same. Different output chip sets and algorithms may sound noticeably different.
Digital Domain - Jitter (in CDs, never mind compressed formats!)
|
Jitter on CDs is in practice a substantially worse problem than it is on a DAP. DAPs are buffered devices and data transfer from the hard drive, to memory and out to the codec is in essence perfect and done at extremely high speeds in parallel between buffers well above the rate the DAC outputs audio. There is no need for the reclocking you mention here as data transfer will use the same clock on both ends and will end up back in a buffer so any potential variation is factored out again by the time the word reaches the destination.
Yes, there can be variation in the implementation of each codec but any individual codec will be consistent in it's implementation. Any imperfections the codec generates will always exist, and the algorithm will not change with burn-in. Additionally, the codec's decoding algorithm implementation is only a factor with lossy audio, with lossless/raw formats the data coming out should always match the original data
exactly.
The transfer from the codec to the DAC is the place where jitter can potentially occur, typically this is the weak link of a CD based system where you hook your digital output up to an external DAC. Clock information is not transfered over a typical SPDIF or AES/EBU link, instead it is implied by the transfer (this is a really bad idea). Some high end CD player/DAC combinations get around this by having a separate cable to carry the clock information purely to minimize jitter. In a DAP like the iPod, transfer from the codec to the DAC is done over
I2S. The whole purpose of I2S is to not only transfer the audio data stream but also the clock to the DAC specifically to minimize jitter. In the iPod Classic it would appear that this I2S link is an incredibly short trace internal to the codec chip so the chance of clock skew is lessened even more. Yes, it's theoretically possible to get some jitter at this stage but the chance is
vastly lower than with a CD connected to an external DAC.
None of what I am saying here contradicts the page you linked to either. There's a lot of factors involved when it comes to jitter. You will also find many people will flatly deny any effects of jitter, it's a controversial topic. For example the waveform graphs shown on the page you liked to are in effect showing a square wave around 22050hz. The reason for using that specific frequency is that the highest frequencies would be the worse cases for jitter since there are so few samples. In practice though most people can't hear frequencies this high, and most players will filter frequencies that high so that's never a situation you will see in the real world. At lower frequencies the representation of the waves would look substantially more accurate even with jitter since there are more samples.
There is of course still another way jitter can get into your DAP. If your CD ripper does not try and correct jitter right from the start, then your digital files will contain jitter before the files are transferred to your DAP and will carry it all the way through the chain to the DAC being used for playback. This is why people should rip with EAC, Max, etc