Jon Risch's Web Site


JITTER
By Jon M. Risch
Last Updated 3-24-02

A very common misconception about digital signal
transmission with respect to audio is that if the signal does
not get corrupted to the point of losing or changing the 1's
and 0's, that nothing else can go wrong. If the transmission
system had been designed with cost no object, and by
engineers familiar with all the known foibles and problems
of digital transmission of audio signals, then this might
be subtantially true. No differences could rear their ugly
head.

Unfortunately, the systems we ended up with DO NOT remain
unaffected by such things as jitter, where the transistion
from a 1 to a 0 is modulated with respect to time. There are
many ways that jitter can affect the final digital to analog
conversion at the DAC. Jitter on the transmitted signal can
bleed or feed through the input reciever, and affect the DAC.
How? Current drain on the power supplies due to the changing
signal content and the varying demands made on the power
supply to the logic chips and the DAC. Modulate the power
supply rails, and the DAC will convert at slightly different
times. HOWEVER the power supply gets modulated, it will
affect the DAC. One version of this has been popularly
refered to as LIM or Logic Induced Modulation by the
audiophile press. See:
"Time Distortions Within Digital Audio Equipment Due to
Integrated Circuit Logic Induced Modulation Products"
AES Preprint Number: 3105 Convention: 91 1991-10
Authors: Edmund Meitner & Robert Gendron

Many of the logic chips in a digital audio system behave
very poorly with respect to dumping garbage onto the rails
and even worse, onto the ground reference point. Even as I
post, logic manufacturers such as TI are advertising the
benefits of their latest generation of logic chips that
reduce ground bounce. The circuitry itself generates it's
own interference, and this can be modulated by almost
anything that also affects the power supply or ground.

Who cares what the power supply rails or the ground is doing?
The DAC cares, beacuse it is told to convert a digital signal
value at a certain time. This time is determined by the master
clocking oscillator in some designs, and in others by the
digital data stream itself by deriving a clock from the clock
data embedded in the data stream, and when the DAC
has determined that a
transistion from logical one to a zero, or a logical zero to a
one, has in fact occured. The point at which the DAC decides
this has occured, depends on the absolute value of the power
supply rails near the moment of detection/conversion. The purity
of the master oscillator signal is also affected by PS and ground
variations, as well as sound vibrations, and the activity of the
various subsystems within the CD player/DAC box. If this master
oscillator signal is not perfectly pure, and free from noise, phase
jitter, and other artifacts, then even if the DAC was totally
unaffected by PS perturbations (virtually impossible to accomplish),
then the master oscillator signal itself would cause jitter.

The amount of jitter that it takes to affect the analog
output of the signal used to be thought of as fairly high,
somewhere on the order of 1,000 to 500 pS worth. Now, the
engineers on the cutting edge claim that in order for jitter
to be inaudible and not affect the sound of the signal, it
may have to be as low as 10 to 20 pS. That's for 16 bit
digital audio. That's a very tiny amount of jitter, and
easily below what most all current equipment is capable of.

Computer systems never convert the 1's and 0's to time
sensitive analog data, they only need to recover the 1's
or 0's, any timing accuracy only has to preserve the bits,
not how accurately they arrive or are delivered. So in this
regard, computer systems ARE completely different than
digital audio systems.

Look into digital audio more thouroughly, and realize that
the implementations are not perfect or ideal, and are
sensitive to outside influences. Just because they could
have been and should have been done better or more nearly
perfect does not mean they were! People are not hearing
things, they are experiencing the result of products designed
to a cost point that perform the way they do in a real
world because of design limitations imposed by the consumer
market price conciousness all the mid-fi companies live and
die by.

Jitter read from a CD will affect how well the read servo
stays locked, and how much the read servo has irregular power
supply demands. Just about everything and anything affect the
power supply, so reduce jitter read from the disc, and it will
affect the accuracy of the playback event.

With digital cables, there are three things that are paramount:
proper impedance, proper cable termination, and wide bandwidth.
It may be that a particular cable more nearly matches a systems
actual impedance. The other factor, proper termination includes,
but is not limited to the actual electrical termination inside
the components, as well as the connector on the end of the
cable. If the connector is NOT a perfect 75 ohm, 110 ohm, or
whatever, it will cause minor reflections in the cable, which
makes our old friend JITTER raise it's ugly head again.

The third factor, bandwidth, is only an issue because both the
AES/EBU and the SP/DIF interface formats were designed before
Sony/Phillips knew all there was to know about digital problems, and they
require PERFECT unlimited bandwidth cables in order for the
transimission systems to be free of jitter. The more you limit
the bandwidth, the more jitter. This is a known engineering
fact, and an AES paper was given about this very subject not
too long ago.
"Is the AES/EBU/SPDIF Digital Audio Interface Flawed?"
Preprint Number: 3360
Author: Chris Dunn
Author: Malcolm O. J. Hawksford

The effective data rate of SP/DIF is about 3 Mhz, and the
design of the transmitters and recievers is abysmal. Maybe
if everything else was done right, then cables, etc. wouldn't
matter. So much was done wrong or cost cut till it screwed
up that they do come into the picture.

A good web source for info on jitter is located at:
http://www.digido.com/jitteressay.html
AND
http://www.digido.com/wegetletters.html#anchor2484124

http://www.audioprecision.com/publications//audiotst/jan96/jan961.shtml

http://www.audiotest.com/publications/audiotst/dec99/jitter_theory.html

http://www.jitter.de/english/engc_navfr.html
(This site has many links to good references, as well as decent explanations of what jitter is, etc.)

Replication News Article on Jitter:
http://www.AudioAsylum.com/audio/general/messages/977.html

http://www.elantec.com/pages/apppdf/d40954.pdf
(Impedance variations of a short cable)

http://www.toshiba.com/taec/components/Datasheet/TOTX173.pdf
(One of the more popular TOSLINK optical components)

http://www.axon.nl/axon/axon_comp.nsf/whitepapers/whitepapers/$file/jitter.pdf

http://www.galstar.com/~ntracy/acg/AandE/npt.on.jitter2.htm

http://www.stereophile.com/fullarchives.cgi?280
(An archived Stereophile review, that shows measured jitter levels.)

http://www.dcsltd.co.uk/papers/jitter.pdf
(Talks about optical based jitter problems toward the end)

http://www.nanophon.com/audio/jitter92.pdf
http://www.nanophon.com/audio/diagnose.pdf
http://www.nanophon.com/audio/towards.pdf

http://www.homecinemachoice.com/frame.html?http://www.homecinemachoice.com/testbench/DVDPlayers/DVDSound.shtml

http://www.benchmarkmedia.com/appnotes-d/jittercu.asp

http://www.music.mcgill.ca/~martin/bibliography/digital/d_conversion.html
**********************************


HOW JITTER GETS GENERATED BY ALMOST ANY STAGE IN A DIGITAL AUDIO DEVICE
As a further explanation fo how jitter gets into everything, here
is another stab at it:

The critical thing to realize is that jitter can only occur at the DAC, whether that is in a CD player, or a separate
box connected with a SPDIF digital interconnect.

However, just about anything that causes power supply variations (henceforth PS = Power Supply) WILL also
affect the DAC. It does seem that data buffers would eliminate any problem, but the whole thing arises because
the DAC can be affected by PS variations.

Let's go into this, so there is no doubt of the root cause of jitter in digital audio systems. How does a logic chip,
whether it is a DAC, a hex buffer, or a complex digital filter, detect the presence of a one or zero? The digital
data stream is buzzing along, and the signal traverses from ground to a full logic +5 volts. Somewhere along
the way from ground to +5, or from +5 to ground, a one or a zero is detected. Exactly when this detection takes
place is a function of a ratio of the power supply. Most all modern logic chips have hysterysis built in, that is, the
point that a one is detected is difernet than the point a zero is detected. With the power supply siting at exactly
5 volts, for detection of a one the signal level may have to go up past 3.33 volts, and to detect a zero, the signal
would have to swing below 1.67 volts. This is for CMOS, the old TTL type logic chips, which I do not believe is
used in any modern CD players, used 2.8 and 0.7 volts respectively.

Further, the digital data stream does not have instantaneous logic transistions, there is a finite time that the
data will translate from ground to +5 volts, or back down (and not necessarily the same amount of time for
each). The transition time from a total ground level up to a full +5 volts takes a certain amount of time. Hence,
when the transition from a logical one to a zero occurs, or vice versa, has the potential to shift around a bit.

Here is the big deal part: since this detection point is a ratio of the PS voltage at any given moment IF THE PS
RAILS DEVIATE FROM 5 VOLTS, THEN THE POINT AT WHICH THE DETECTION OCCURS IS MOVED.
Once again, let's take it slow, and flesh this out carefully. For instance, if the PS were to deviate down to 4.8
volts, now the trigger thresholds would be 3.20 and 1.60 volts, instead of 3.33 and 1.67. So even if a perfectly
regular and precise data stream were being passed to the DAC, if the PS to the DAC varies during the various
data logic transitions, then a timing error is introduced (jitter). If we follow a series of 4 transitions, let's see what
happens.

First transition, a zero to one, occurs while the PS is exactly 5 volts, detection occurs at the expected transition
voltage. Lets say for convenience, that this occured at the 'normal' timing interval, and reference to it for the next
three transitions. Now the next one is coming along, but wait! The PS droops during the time when the signal is
in transition, and the timing is altered. If it was a one to a zero, then it would be later, if it was a zero to a one, it
would be earlier. In our example, it is a one to a zero, so the signal is later than normal. The next transistion
comes along, and the power supply is steady again at +5 volts, a normal detection occurs, and the signal
transition is back to normal timing. Then the final transition we are looking at in our example, another one to
zero transition, but wait, the PS has had a load removed (some gates just got done switching OFF
somewhere), and low and behold, the PS voltage rises above normal during the transition period. Now the
detection of the transition has been made earlier than normal.

Once you get the basic idea, then it becomes clear that ANYTHING that affcts the PS voltage will affect the
DAC timing/detection process of the digital data, and hence affect the output signal. Not only does the PS rail
(the +5 volts) affect this, but the ground does as well. The end result of a lot of logic gates switching on and off
is the dumping of a lot of current into the local grounds, and as a result of the ground reference points real world
inductance (not even a so-called ground plane is without some inductance, or local current densities), the
ground reference point itself will vary from absolute 0 volts, either going temporarily below or above 0 volts.

Now the argument has already been made before that the power supply does not vary like this. In the long term
average, this is kind of true, as the feedback loop of the common Integrated Circuit voltage regulator will
maintain the set voltage to the best of it's abilities.
However, most of these IC voltage regualtors have what amounts to a 741 op-amp (or worse) in control of the
regulation, and the transient response of the regulator IC is not even as good as that of a simple 4558 op-amp.
Sudden transient loads, say several logic gates firing at the same time, load the supply, and for a brief
moment, the PS voltage goes down. Then the regulator senses this, and jacks the output up, often overshooting
in the process, and an entire PS transient event has been put in motion. A SINGLE LOAD TRANSIENT
COULD TAKE MANY HUNDREDS OF CLOCK CYCLES TO FULLY SETTLE OUT. One popular IC voltage
regulator is speced at over 100uS to settle from a load transient, one single isolated event. Again, once you
realize that these load transients keep coming, in a cachaphony of digital sing-song inside the player, then you
can begin to appreciate that the PS is never really stable in terms of moment to moment, only for the long term
average.

Web reference URL's:
Digital Logic chip output waveforms and ground bounce info:
TI
http://www-s.ti.com/sc/psheets/sdya010/sdya010.pdf
page29-32,  waveform views of various devices, shows great shots of non-square square waves

Fairchild
http://www.fairchildsemi.com/ms/MS/MS-541-MISC.pdf
Pages 10-16, digital logic waveforms, showing less than perfect logic signal transmission, ringing, ground bounce, etc. Esp p. 15-16
http://www.fairchildsemi.com/ms/MS/MS-539-MISC.pdf
Shows good waveforms on page 4, also t-line effects for 3  foot cables.
http://www.fairchildsemi.com/an/AN/AN-831.pdf
Shows RF output of logic chips.
http://www.fairchildsemi.com/an/AN/AN-754.pdf
Shows mis-termination results, and (best case) real world waveforms.
http://www.fairchildsemi.com/an/AN/AN-375.pdf
Clearly shows waveform abberations due to the real world.

info on voltage regulator transients, re digital audio and jitter.
http://www.cherrysemiconductor.com/product/PDF/CS-5233-3PDF.pdf page5
http://www.linear-tech.com/pdf/lt0117.pdf, page 5
http://www.national.com/ads-cgi/viewer.pl/ds/LM/LM117.pdf, page 4
http://www-s.ti.com/sc/psheets/slvs297a/slvs297a.pdf
Fig. 12 and Fig. 13

Ground Bounce Info:

Good basic tutorial.
http://www.ultracad.com/g_bounce.pdf

Shows actual measurements, scope pics, etc.
http://www.fairchildsemi.com/an/AN/AN-640.pdf

About grounds.
http://www.signalintegrity.com/news/2_12.htm

No DAC chip currently made is immune to this problem, as it is itself usually an LSI IC, and the local circuits
inside of it all do the smae thing, and cause the PS runs INSIDE the chip, along the various PS traces and
ground traces to vary and fluctuate to the digital dance. Even with a PERFECT PS AT THE DAC PS
TERMINALS, the DAC will still mess up, and vary the timing based on the digital data streams varying signal.

None of the above LIM effects, documented in this AES paper,
"Time Distortions Within Digital Audio Equipment Due to
Integrated Circuit Logic Induced Modulation Products"
AES Preprint Number: 3105 Convention: 91 1991-10
Authors: Edmund Meitner & Robert Gendron
is even begining to scratch the whole beast, as the master clock oscillator is usually crystal driven, and not only
is this crystal circuit usually affected by the PS variations to, it is also PIEZOELECTRIC. That means that if the
crystal is vibrated, it will output a voltage in time with the vibrations. Guess what a good speaker system is
doing to the CD player or DAC in the same room.

How, might you say, can a transport affect the sound? After all, it is in a separate box, and it's PS demands are
being supplied with a separate PS unit.
This is where the extreme sensitivity of the digital audio system to jitter problems come into play. As I outline
above, ANYTHING that causes PS variations WILL cause jitter. The SP/DIF format, like the AES/EBU format,
causes jitter in by the very act of transmitting the signal along cables and in the transmit and recieve actions.
See: "Is the AES/EBU/SPDIF Digital Audio Interface Flawed?"
Preprint Number: 3360, Authors: Chris Dunn & Malcolm O. J. Hawksford.

As the signal is received in the DAC, the receive circuit draws power from the PS according to how many
gates are firing when, and this firing pattern is related to the signal transmitted. IF THE SIGNAL
TRANSMITTED BY THE DIGITAL INTERCONNECT HAS JITTER ON IT (FROM THE TRANSPORT OR SEND
RECEIVE CIRCUITS) THEN THE FIRING OF THE RECEIVE CIRCUIT WILL LOAD THE PS IN A CERTAIN
PATTERN THAT IS THEN SUPERIMPOSED ON THE DAC.

By the same token, a CD with jitter added during the mastering process (or the burn of a CD-R) will cause the
servo mechanism to fire in the jittered pattern, and the laser recieve optics, etc. EVERY STAGE THAT
HANDLES THE SIGNAL EVEN IN DIGITAL FORM loads the PS in a particular pattern. With a clean signal,
that pattern is clean, with a jittered signal, that pattern is jittered, and adds to the overall jitter of the reproduced
signal. Hence, the vast majority of jitter is signal correlated, even if indirectly via the digital data stream's
imperfect timing and variations.

Last, but not least, the analog output stage draws current in time with the music, and loads the PS via the +/- 12
(or 15) volt regulators. The +5 volt regulator only has so much line rejection, and so the analog signal gets into
the digital PS as well, once again causing jitter.

The requirements for error free 16 bit data transmission, which all fo the 'perfect sound forever' arguments are
all based on, require jitter levels below what is curently available from even the best gear. Jitter down in the
10-20 pS region is required to avoid jitter from degrading the systems theoretical perfromance limits, and for
20 bit or 24 bit digital at higher bit rates, this must go even lower.
See: http://www.digido.com/jitteressay.html
for more details.

DVD players with their higher data rates require even less jitter than CD players, yet the current crop does not
even match the better CD players, much less the SOTA in CD players at 44.1 kHz..

Jon Risch



Copyright Notice
All website content Copyright Jon M. Risch, 1997 - 2002

All applicable copyright laws apply, all rights reserved, except transmission by USENET and like facilities granted.  Any use or inclusion in print or other media are specifically prohibited.  The informational content is not warrantied in any way or form, and any use of said content are at the reader's own risk, the author shall not be held responsible in any way for any damages or injuries arising from the content of this web site.  Common safety practices are encouraged at all times.  Do not fold, spindle or mutilate.



 ----> E-MAIL ME ! <-----


 Back to home page
Hosted by www.Geocities.ws

1