How much jitter...

[Bazukaz]

The bits are not the problem. Timing is the problem. Most DACs derive their output data timing from a PLL locked to the incoming data timing. The PLL filters out high frequency jitter, but can't (in fact, mustn't) fiddle with low frequency jitter. It appears that basic optical transmission systems introduce more jitter than basic coaxial cable systems. Hence for short distance cable is better. For long distances optical is better, but if you want to preserve timing then you need something better than the basic setup.

Hmm so it means that in addition to audio data pulse width is used to determine the frequency of transmitter timing ? And digital receiver will adjust it's own clock to match it; low frequency variation in this timing causes what we call "jitter" ?

I have bought a dac based on CS8416 chip. The following is stated in the datasheet(Page 53) :

"There are some applications where low jitter in the recovered clock, presented on the RMCK pin, is important. For this reason, the PLL has been designed to have good jitter attenuation characteristics, as shown in Figure 25. In addition, the PLL has been designed to only use the preambles (PDUR=0) of the bi-phase encoded stream to provide lock update information to the PLL. This results in the PLL being immune to data dependent jitter affects because the preambles do not vary with the data."

Does this imply that this DAC is mostly immune in the jitter we are talking about? I have a bit trouble understanding some phrases, like "preambles" and similar.

Regards,
Lukas.

 

[DF96]

The preamble is the stuff which comes before actual data in a data packet. As the preamble does not change it is a better place to derive timing, as it reduces the risk of signal-related jitter - which can be worse than random jitter.

The issue is not the data pulse width, but the timing of the data transition. This feeds through to the data transition on the DAC output.

Yes, the DAC is largely immune to jitter caused by the transmission method. The receiver chips are designed to make this so. The remaining jitter comes from two sources:
the receiver chip and DAC themselves - nothing is perfect!
the data source, where it is largely decided by the jitter of the crystal oscillator

In my opinion people can get too concerned about cable or fibre jitter, which is largely eliminated by the DAC receiver chip. A low jitter oscillator in the source is more important.

 

[Bazukaz]

The preamble is the stuff which comes before actual data in a data packet. As the preamble does not change it is a better place to derive timing, as it reduces the risk of signal-related jitter - which can be worse than random jitter.

The issue is not the data pulse width, but the timing of the data transition. This feeds through to the data transition on the DAC output.

Yes, the DAC is largely immune to jitter caused by the transmission method. The receiver chips are designed to make this so. The remaining jitter comes from two sources:
the receiver chip and DAC themselves - nothing is perfect!
the data source, where it is largely decided by the jitter of the crystal oscillator

In my opinion people can get too concerned about cable or fibre jitter, which is largely eliminated by the DAC receiver chip. A low jitter oscillator in the source is more important.

Thanks for explanations, it's a bit more clear but also rises more questions.
My search for better DAC started some time ago when I noticed my integrated sound card(laptop) producing some unintended noises in high frequency part of audio spectrum. The noises are tied to the signal and depend on it's level. It was even recorded by measurement microphone, so the problem seems to be not imaginary. From your explanation it looks like I can still be getting excess jitter if I use optical out from digital source on my PC, assuming integrated card has a poor quality crystal. I have also had to buy an optical->coaxial converted to feed CS8416 board; so the solution turns out to be quite expensive and not necessary any better.
But why the receiving DAC can reject jitter caused by cables/interconnects and cannot reject one that comes from drifting frequency of the crystal in the digital audio source?

Regards,
Lukas.

 

[kenpeter]

I am following along because I am tesitng jitter cleaner EVMs at work this week.

I notice none can be any better over the long time than the crystal that drives them.
And none can be any better over the short time than the VCO locked to that crystal.
What gives best results for phase drift over periods inbetween these extremes?

Even if you use a VCO to get rid the "jitter", reasonable priced ref crystals are all off
frequency and drift with temperature a lot more that I previously realized. Maybe its
just cause I never drilled down to look so closely before? Was a real shocker how
bad they drift, compare to rubidium. They are going to take that nice clean VCO on
a slow wander wherever they drift, and thats a jitter of sorts too.

I sat there in amazement watching a Transko 40Mhz crystal drift 2Hz/Sec.
Don't know if I could hear that, but the imaginary horror alone is enough....

 

[DF96]

The receiver needs to keep roughly in step with the incoming data stream. To do that it has to follow long-term (i.e. slow) variations in the clock rate, including low frequency jitter. To avoid this problem you would need a long FIFO register so you can reclock the data, or a shorter FIFO with some clock feedback to the data source. Some people do this, but few domestic sources have the ability to receive a clock.

In practice, given a good clock in the source and a good receiver before the DAC it is not really necessary to do anything more complicated.

Very low frequency 'jitter' (i.e. long-term frequency stability) does not matter at all, as even the cheapest quartz crystal will be much more stable than any musical instrument. Absolutely no need for rubidium locking etc. Very stable oscillators may have worse jitter, as all circuits are a compromise so improving one thing may degrade something else.

 

[Bazukaz]

The receiver needs to keep roughly in step with the incoming data stream. To do that it has to follow long-term (i.e. slow) variations in the clock rate, including low frequency jitter. To avoid this problem you would need a long FIFO register so you can reclock the data, or a shorter FIFO with some clock feedback to the data source. Some people do this, but few domestic sources have the ability to receive a clock.

In practice, given a good clock in the source and a good receiver before the DAC it is not really necessary to do anything more complicated.

Very low frequency 'jitter' (i.e. long-term frequency stability) does not matter at all, as even the cheapest quartz crystal will be much more stable than any musical instrument. Absolutely no need for rubidium locking etc. Very stable oscillators may have worse jitter, as all circuits are a compromise so improving one thing may degrade something else.

Hi,

Thanks for explanations. It's a bit more clear how it works. Currently I have quite a chain of audio devices (PC->CS8416 based DAC->MiniDSP(ADC->DSP->DAC)). I am a bit concerned about quality of ADC/DACs and board layout in miniDSP as there is quite a lot of background noise(sounds like white noise), significantly more than a cheap integrated audio card in my laptop, and is quite audible from even two meters with more sensitive speakers.
Are there ways to evaluate if jitter is a problem without sophisticated equipment? All I have is a PC and a 20 MHz analog scope.

Regards,
Lukas.

 

[DF96]

I think jitter has to be quite bad before you can measure it with simple equipment.