Jitter is easy to measure and compare. Audibility of jitter is another matter - most tests show that jitter has to be pretty severe to be audible. In any case, jitter is mostly caused by noise, and not by the transmission characteristics of the cable.
With jitter on the DAC you get converted the correct value at the wrong time, and this could be audible.
And I have to counter with another question: "which jitter"?
We have two very different things - we have audio clock jitter, caused by instability of the outgoing audio clock (this causes small differences in bit timings that appear as weak intermodulation noise in the output) on one hand, and data transmission jitter (caused by the sending transport, computer and possibly cable) that has no effect as such, as long as the data is received correctly, on the other hand
The problem is that many DACs derive their audio clock from the incoming data clock to ensure they stay in synch with the sender. Even then the jitter in the incoming data is usually filtered by a phase-locked loop.
Let as consider that we have a jitter level of 1μsec (huge because usualy it is under 1nsec).
This corespond of a frequency of 1MHz.
Who hears such a frequency?
How many amps can reproduce this frequency?
What speakers can reproduce this frequency?
If we consider only the phase deviation, a jitter of 1μsec will generate a phase jitter of 3.6° at 20KHz and 2.72° at 15KHz.
How many people can notify such a variation?
But a variation of 1000 times smaller as it really is for 1nsec jitter?
You seem to be confused about the definition of jitter... to complicate the matters, there are several possible definitions.
Some copypaste from wikipedo :
"Period jitter (aka cycle jitter) is the difference between any one clock period and the ideal/average clock period. "
"Cycle-to-cycle jitter is the difference in length/duration of any two adjacent clock periods."
For audio, period jitter is more relevant.
Peak period jitter, is about guarantee that the clock period will not drift more than +/- x% , or x nanoseconds, during a specific observation interval.
As the observation interval increases, it is called "drift" instead of "jitter", caused by temperature, aging, etc...
Now, if you examine the phase difference between a real clock and a perfect clock of the same average frequency, you will see the phase has fluctuations. This is called phase noise, another way of looking at the same thing.
So, "a jitter level of 1µs" does not mean anything.
If you add the words "period jitter 1µs RMS white gaussian", then you will know that each period will have a gaussian white distribution error of 1µs RMS. In this case, your DAC running at 44.1 kHz would have a SNR corresponding to about 6-8 bits.
Why ?
Suppose you have two consecutive samples of different value.
Each sample will last 1/44100 seconds, or 22 µs.
Suppose first sample has value 0, next sample has value 1.
Now, if the transition between those two samples is shifted by 1µs, as per your example, that is 4% error, one of them will be 4% longer, the other will be 4% shorter. It is equivalent to both samples having perfect timing but 4% amplitude error...
Last edited:
Thanks Julf, I'm still reading up on eye diagrams, and agree it is more informative for sure. But if the receiver can repair timing error, it is not so critical that the pattern be perfect or as close to as possible?
Measuring the DAC output would also be fun to do anyhow, do you mean from the analog output stage, i.e. RCA sockets, or from the chip itself? (chip is sealed and hide from view anyway). I do know that is a rare 80's Philips 16 bit chip, and I know it is not 1541..
Guilty as charged.
See, DAC is not a problem, bigger problem is wife. I have been using the same wife for more than 2 decades
Lol, it must be time for an upgrade then! You can get a very obedient sample from the far east via the internet these days, quite economical too I hear..
Exactly. But all depends on the DAC and how good it is in dealing with input variations.
The RCA's would be fine - that is the signal you are interested in, after all.
Hmm... That would suggest the digital input circuitry probably isn't very sophisticated...
There is more here than meets the eye. First of all, the samples at 44100 rate consists of many bits, 16 or even 24 data bits.
So the bit time is much smaller than 1/44100.
Secondly, it is physically impossible that two consecutive samples are 0 and 1. That would mean that the original is a jump of 0 volt analog till max volt analog in 2 microseconds or so. Even if there exists an audio signal that does this (it does not exist) there is still the ADC antialias filter that will prevent it.
If you have a digital audio data stream, the slight movement of the zero crossings of the bit edges is the jitter. Yoiu can express jitter in nanoseconds or picoseconds, but also in 'unit intervals'. One UI is the time between two edges, so a jitter of 0.01Ui is a movement of the zero crossing of 1% of the bit time.
Jan
Thanks very much Marce, I did begin reading your links which are very useful and in depth, I do need to have more time to sit and read (in peace and quiet!) to get my head round this stuff, see, I'm not trained in EE or anything like it, just very interested and willing to learn, so your input is valuable and appreciated. It sounds like you do a lot with this stuff, shame I'm not up your way but in the soggy south!
Jan, yes I did, thanks. The DAC chip I'm certain is very good, and it's implementation is purist; i.e., no over sampling, no LP filters, hence the requirement for a good transport and connection cable.
I assume you are aware of the fact that the low pass filter is an essential part of how the DAC operates? It is not called a "reconstruction filter" for nothing.
You might choose to use the rest of your chain as a LP filter, but a filter that is actually designed for the job might be better optimized.
Julf, I am aware that LP filters are normal, but am also aware that they are not entirely necessary, and removing them from the signal path does have a sonic benefit. But, the DAC is what it is, and fed the best signal possible, it makes a very realistic reproduction of the recorded event, the only other digital I have spent time with that rivals it (and maybe betters it in truth) is Dcs Vivaldi.
But, the DAC is not what I want to get bogged down with here, it is getting the best signal possible to it.
For measuring purposes, would you recommend a scope of say 100MHz or higher, or is that a bit excessive?
Again, many thanks!
I would have phrased this like:
"The DAC chip I'm certain is very good, and it's implementation is purist; i.e., no over sampling, no LP filters, hence it is very insensitive to transport and connection cable"
Jan
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.