Jitter in DAC.


I am new to this. Anyway, I have a question in my mind.
I am going to built a audio DAC with a built in jitter attenuator. My question is, can I use a high speed FIFO and a low jitter clock to do it?

This is how I am going to implement. All the Bit, Data and L/R clks will be clocked into FIFO using recovered Master Clock [ from CS8414 ], and then the clocks will be clocked out from FIFO using a clean and low jitter oscillator. All the clocks will later feed into digital filter.

So, do you think this scheme will work?

What is the advantage of using FIFO over the convention method -> PLL?

Thank you.


2001-04-09 2:54 pm
Hi ghee0,

I think that a FIFO can be used ONLY if you can control the source of the samples:

If you use separate clocks for read and write, they will have slightly different frequency :
If the reading clock is slower than the writing clock, the FIFO will sooner or later get FULL and you will lose samples.
If the reading clock is faster than the writing clock, the FIFO will sooner or later get EMPTY and the DAC will play wrong samples.

You need a control signal to tell the source when the FIFO is FULL or EMPTY
... this is not easyly done with a commercial transport.



2001-08-06 11:53 am
Hi Oweppe,

You have a very good point here.

Actually, the read and write clocks are with the same frequency but different in timing and the amount of Jitter.

On the FIFO IC, there are "Almost Full", "Read Enable", "Empty" pins that I can use to control the flow of the clocks. In other words, the FIFO will start to clock out data when the "Almost Full" is enable until it sees a trigger on "Empty".

These pins will save me from lots of trouble, don't they?

Thank you!



2001-03-06 3:07 pm
sure thing..if i get it to work =) i have borrowed heavily from http://www.diyaudio.de (dac4) i use his "adapter cards" but i have redrawen the "motherboard" i have exchanged the 7805 to Lm317...get rid of all otherstuff i didnt use...

yes the Cs8420 uppsampler with a 2.5->3ps clock takes care of jitter perfectly well....

the dac chips is Cs43122 (the latest and greatest from crystall) for outputstage i have jet to test some design but im leaning on just a passive one with lundahl transformers...

ohh by the way it is battery powered...separate supplyes for digital and analog...and about 0,1-2F in total capacitance......

To do it, you'll need a PLL to clean up the clock. The PLL locks onto the clock output from the receiver chip, so that the two clocks are close enough in frequency that you don't get fifo over/underflow. Of course, if you use a PLL, it can't remove all the jitter because it still has to pull up and down in frequency to track the receiver's clock (to a certain extent).

Due to the need to adjust the frequency, the PLL can't be made with as high a Q factor as a fixed oscillator. This means it's short term frequency stability will be lower, and the PLL itself will have a higher inherent jitter than a fixed oscillator.

So, I would go with the ASRC approach. Right now I'm building several test DACs for my big DSP project (which keeps getting sidetracked). All of the test DACs use either AD1892 combination receiver/ASRC or the AD1896, which is a superb sample rate converter. You needn't worry about the math going on inside the chip, it won't harm the sound quality as some people claim... the same kind of math goes on inside the DF1704 chip, or the PMD-100, and even more crazy math goes on inside the CS43122 etc... I can post a more detailed explanation of what happens inside an ASRC if you like.

What I really like about the ASRC approach is that the jitter rejection is about as good as you could possibly get, and you don't have to worry about complex memory management and PLL control schemes... everything's taken care of inside one chip... just plug and play. Couldn't be easier.

Anyhow, Valpey-Fischer makes some very low jitter oscillator modules. Now, wether or not you can get your hands on any, i don't know...
Many thanks on the comment.

For AD1896, if I am not going to up/down sample my clock rates.

Do you think the input bit and L/R clock can be tied to input serial ports? If it is tied, will the jitter in data line reduces?

Or they should be running from another clock source says a divider and a oscillator?

Simpson, can you tell me which configuration you are using now? With or without the extra "oscillator"?

Lastly, what is the official website for Valpey-Fischer?


I don't know anything specifically about the Audio Alchemy v1.0, but asynchronous sample rate conversion could certainly provide some benefit by removing clock jitter. And, additionally, if the DAC doesn't support high sample rates like 96 or 192 kHz, an ASRC is a good way to bring the sample rate down to one compatable with the DAC.


The AD1896 always up/downsamples to remove jitter, because no two clocks will run at *precisely* the same speed, and the FIFO inside the AD1896 would eventually overflow or underrun. The input to the AD1896 is clocked from the receiver ic (eg CS8414). Then, the AD1896 output is clocked from a local crystal oscillator (eg. 24.576MHz). The AD1896 outputs the new data, along with new low-jitter BCLK and LRCLK signals which go to the DAC. Here are some useful links for you:


[Edited by hifiZen on 09-06-2001 at 10:42 PM]
hifiZen said:
hmmm.. well it's a bit too much to put in a posting, especially without some schematics... have you read the datasheets?

hahaha,I know its a big problem!
I have all the datasheet,but I study in material so I dont understand the electrical principle.I read the datasheet but still dont know how to do.
I got these chips by a people I dont know.
And I put them in my desk for a long time,hahaha.
Right now I am just a beginner in DIY DAC,I think maybe one day I will know how to use them.
hi folks
i am new here so sorry for the mess.

i heavyly modified my player.
better op amps,coupling caps/5 power supplies with 3 transformers (1 for audio a/1 for audio d & dac /
original transformer for cpu of player & servo.
servo section was upgraded with hi speed switching transistors (traverse/spindle/trak/fokus)& +/- 12 volts
you can hear the difference!!!!!!!!!
but but but
when i started modifing the clock, i was like in another planet wow
folks spend time modifing the clock seriously!!!!!!!!!!
if someone knows about klock reshaping upgrades or cards that you can fit into your unit pls. let me know.
i am available to trade info about upgrades

pls let me know