Data Jitter Versus Electrical Jitter?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi,

Normally I don't post much.. but today I saw an interesting thread on Head-fi.org about the E-MU 0404 soundcard. It wasn't the soundcard that was particularly interesting in the thread, but the topic of jitter. Ian, one of the persons in the thread, stated that toslink was indepedent of the computer power supply ground. I asked how this was possible, in which another person named Solude responded stating the following:

"The key here is optical. Though the transmitter itself is being fed a less than rock solid power feed, the LIGHT it transmits has no electrical component. The received bits from light have no electrical jitter. When electrical jitter is not an issue then coax is usually preferred for it's on paper performance benefits over optical. But the PC is an electrical jitter nightmare. A few of ther Stereophile reviews show this pretty well where the jitter on an RME coax side is great data jitter wise but junk for electrical jitter. The optical on the same card had the same data jitter but ZERO electrical jitter resulting in superior results compared to coax on the same card. Pretty sure it was the RME PST?"

Can anyone give me their thoughts on the subject? Is toslink superior than Coax in a soundcard situation?

Thanks for your help,
300_baud

Edit: For those of you interesting in viewing the thread, go here:
http://www4.head-fi.org/forums/showthread.php?t=78053&page=3
 
I used to use an optical from my soundcard for this very reason (now computer is in a different room to the hifi though), assuming...

"electrical jitter".... Not sure what this is really meant to mean. I assume you mean RFI? Jitter is clock variations, what you are referring to as data jitter. RFI is any old random junk (which if severe enough can cause false triggering), spewed out by clocks, switch mode power supplies, motors, etc. You can imagine that with all these things in abundance a computer is an RF nightmare. It's worth noting that no personal computer would pass the applicable standards for electromagnetic compliance (FCC 15, CISPR 13, EN55013) were it to be tested on the basis of being consumer electronics.
 
Hi Isaac,

Well a response was made to the head-fi.org thread which said:

"Data jitter is where there are slight delays in the data's arrival or if the data arrives in the wrong order and electical jitter is noise in the signal caused by RF interference and dirty power."

So you would recommend to use toslink in my application? (soundcard use)

Anyone else have their thoughts on this topic?

Regards,
300_baud
 
While an optical interface isolates electrically one component from the other, including ground (and earth) problems, the optical transtmitters/receivers also benefit from a clean and stable psu.
Because they need to transform a good electrical signal into good light.:D

But yes, I also think it's better to use Toslink on internal soundcards.
 
I still don't know the audible effects of jitter. Only thing I can think of is With my Pioneer GEX-FM903XM XM radio receiver. The STA450, which is the XM service layer decoder, has i2s output. There is a data line, serial or binary clock, LR clock, and an M clock or the master clock to run the DAC. I have these 4 lines connected to a DIT4096. The DIT4096 is set up to output a TTL level S/PDIF signal that is connected to a toshiba toslink transmitter. The optical cable is connected to a sharp toslink reciever which is connected to a conversion circuit which uses a 74HC04 CMOS inverter, two 680 ohm resistors, a 100 ohm resistor and a 1:1 pulse transformer. The transformer secondary is the S/PDIF output, which is connected to my DAC which was designed my jwb (very good design, uses a CS8420 and a CS43122, sounds really really good too). I get distortion and poping, it seems the samples are not being decoded properly. When I play around with the i2s wires or the toslink connector, I can fix it and make it free of distortion. On my Sony XM reciever connected the same way with the same circuit I get no distortion at all. Is this distortion I hear because of jitter because the pioneer XM radio is so noisy or is it something else? The pioneer's STA450 also has an S/PDIF outpit on pin78 which is TTL level. Simply connect a toslink transmitter to it and you have good quality distortion free S/PDIF output on the pioneer. Unfortunately, this S/PDIF output on pin 78 is controlled by a register, and is not enabled on the generation 2 sony XM receiver (model number DRN-XM01MK2), but is enabled on the first generation receiver (model number DRN-XM01). Even so, the pioneer has this distortion with the i2s output, the sony does not. The CS8420 has good jitter rejection, but I imagine the jitter is so bad that it can't even lock on. Anyone have any ideas?
 
Hi,
jitter is a very slippery beast... I never feel safe when addressing it... however, I try to give a tentative and incomplete answer to the rigina question. Hope someone else can help in checking it...

Jitter has different components with different causes. The easiest way to measure it seems to use a jitter test signal with a high level tone at 1/4 the sampling frequency Fc and some low level signal, typically a toggling of the LSB every 192 samples, and get an FFT spectrum of the signal around Fc/4 of the resulting analog output.

This setup is also the one Stereophile used for RME card.

What happened in the test was that when the electrical connection was used there was a very high jitter, while with SPDIF this was very low. This is not surprising, as the grounds inside a PC ar not that clean, as you can imagine: in this case the SPDIF is probably causing a big ground loop.

The idea at the basis of the specific declaration that the data related jitter is low is that in the unjittered spectrum of the test signal there are some low level (-110/-120 dB) lines; these lines are related to the LSB bit toggling.

When these lines are enhanced, it is said this is jitter related data. Here is where I am not completely sure...

The standpoints I know about are the following.

1) The jitter theory says that if you have a pure tone signal which is converted with a jittered clock, then jitter

appears as spurious lines spaced around the pure tone. This can be checked/simulated also using the Excel spreadsheet FFT function, if you want. It is also possible to deduct jitter amount from the side bands level.

2) One of the problems with separate dacs is spdif. SPDIF is imperfect, even from the theoretical standpoint, as the digital code used causes symbol interference, which in the end means that the clock transmitted through SPDIF is necessarily jitter affected (in a variable extent, however), and the jitter depends on the transmitted data!

Note that digital audio receivers like CS8412/14 etc do not solve completely the problem. This can be very easily detected by observing the mess any data different from all zero causes to the master clock of an 8412 through a high frequency spectrum analyzer. Fortunately, all digital audio receivers apply rather effective counter measures to avoid ransferring such a messy clock to the DAC...

In my view, however, this is data related jitter.

Going back to the test signal, the Fc/192 LSB toggling causes a periodic modification of the DIGITAL signal, which, based on the symbol interference problem described above, causes an amount of clock jitter with a frequency equal to Fc/192/2 and higher odd harmonics. This jitter can be detected as an increase of the Fc/192/2 LSB toggling related lines which are present also in the unjittered signal.

So the data related jitter is the one appearing around the main tone at a distance of Fc/192 and multiples, while what was called "electrical jitter " is randomly distributed around the main high level tone. Note that most of this is composed by 50Hz and multiples (in Europe, 60Hz in US) and is therefore power supply jitter.

The attached spectrum is affected by jitter. Data jitter is shown by the yellow arrows: in the unjittered signal, the , lines should all be approximatively equal to their values at the extreme of the spectra portion (7.525KHz-14.525KHz) shown. Non-data related jitter is shown by the blue lines that in the unjittered signal are completely absent.

300_baud, hope this can help.

And all of you guys, please tell me if what I wrote is wrong....

Giorgio
 

Attachments

  • presentazione2rid.jpg
    presentazione2rid.jpg
    60.4 KB · Views: 169
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.