What does jitter sound like?

[darrenwm]

There is a lot of chat about the audibility of jitter.

 

Please may somebody describe what it sounds like?

 

[John]

Like the sound of one hand clapping.

 

[drdigital]

Jitter is present on all digital transmission systems, in a digital audio system it is only of importance in 2 places, the sample clock on an ADC or a DAC. However digital audio data is transmitted it doesn't matter how much jitter is present on the data stream provided it doesn't cause a data error.

 

The downstream device would be expected to generate a sample clock of adequate stability whatever the incoming data jitter.

 

Jitter on the DAC sample clock is measurable by measuring noise sidebands on the audio output.

 

Interestingly, purveyors of so called jitter busters never seem to publish such measurements, you can draw your own conclusions from that!!!

 

[darrenwm]

Dr. Digital,

 

Would you expect higher jitter to affect the emotion, musicality, and overall foot-tapping nature of the music?

 

Is it the difference between looking through a net curtain, and opening the window?

 

Will it make my blacks blacker, and sound-stage more three dimensional?

 

Would jitter be more audible in the evenings when the electricity is cleaner and the musicality is at its peak, or would the clean electricity actually reduce the jitter than offset this additional insight?

 

Regards,

 

[Al]

follow this one...

 

[brimac]

I think modern day jitter artifacts sounds like glasses on versus glasses off. If you hear a difference, then jitter will be just as audible. But if you're listening for differences between having your glasses on or not you're probably not listening to the music anyway so it's sorta pointless.

 

[franepici]

"Oh, Ive gotta follow this one..."

 

lol

 

[tfarney]

I hear jitter, people....

 

 

 

Tim

 

[Ashley James]

I've lost the will to live!!!

 

Ash

 

[Roseval]

You might try this one: http://www.audioasylum.com/cgi/t.mpl?f=pcaudio&m=26508

I think the post by John Swenson sums it up nicely

 

[lapaix]

One problem with any discussion of digital audio is that the technical description of error requires metaphors that differ from those that describe, however imperfectly, the properties of an analog system. In digital audio there are at least two major variables. One is bandwidth (16 bit/44 kHz, 24 bit/192 kHz, etc). Increasing the information density increases the fidelity of a recording to the original sound. There is an easily perceived video analogy: a bluray dvd delivers a better picture than a standard dvd for the same reason. Second, the quality of a digital audio system also crucially depends upon the fidelity with which the information is retrieved from the source, a CD or a hard drive. This is where jitter comes into play; basically, it is timing error. How does one hear that? We do not hear "jiiter," per se, but if a series of measurements shows that our perception of the sound of a digital audio system varies in a way that is consistent with the amount of this measured property, we are entitled to say, other things equal, that jitter affects sound quality. Here again a video analogy comes to the rescue. Anyone who has watched an old 8 mm home movie made with a cheap camera has probably noticed that the picture flickers. That "flicker" is the film analogy to jitter. An image that flickers causes video fatigue just as listening to a jittery CD player causes audio fatigue, and perhaps one could say that this fatigue is itself evidence for the audibility of jitter. The fact that we are limited to technical descriptions of a phenomenon does not mean that the phenomenon is not real, only that we need a different vocabulary to describe it.

 

 

 

[drdigital]

Jitter on a DAC sample clock produces noise sidebands either side of the wanted signal, the level of these sidebands can be calculated and unless the DAC system is really bad the noise sidebands are well down at or below the noise floor.

 

Hifi gurus often spout theories and scientific fact without any quantitative data to demonstrate the magnitude of the effect especially if they have a magic product to sell.

 

There are thousands of words by such people on this forum but without solid supporting data they are just words circling a drain!

 

Ignore them and get on with enjoying your music, leave them to their obsessions about Jitter, the quality of the 0s and 1s, the crystal structure of their cables and any other 'problem' they can invent a magic cure for!

 

 

 

[Al]

Tim!

 

[LAMF]

For some people, jitter is a voodoo science of digital audio. Bits are bits right? Perfect sound forever, remember…? It turns out it’s not that simple.

 

How would you pronounce “cho pho use?" Think about it for a while. Sound it out.

 

What if I wrote it as “chop house?”

 

All I’ve done is change the timing in which the data has come to us. When it comes to audio, a DAC expects data to come let's say, every 1/44100 seconds. If the data comes in a fraction of a second too early, the sound card buffer can overfill. If the data comes in a fraction of a second too late, the DAC has to “guess” at what kind of data it needs to send out. This corrupts the audio resulting in worse fidelity.

 

[Roseval]

If the data comes in a fraction of a second too early, the sound card buffer can overfill. If the data comes in a fraction of a second too late, the DAC has to “guess” at what kind of data it needs to send out.

 

I wonder if this is correct.

A traditional DAC slaves its clock speed to the incoming data stream.

Therefore there will be no over- or underrun.

If it happens the DAC is not able to correct, it simply loses its lock.

 

If both source and DAC have word clock, you can slave the source to the DAC.

In recording studios they might use a master clock slaving all devices.

 

Today’s sound card do buffer the input.

Overrun is a bit unlikely or a matter of a bad driver (most communication inside a computer is bi-directional).

Underrun clould happen but I don’t think a sound card will conjecture up a value (it is not CD error correction) but simply play a loud pop/crackle or any thing else clearly audible.

 

Streaming audio players are a different kind, they are build with interrupts (internet technology) in mind. They have a large buffer and read this with their own clock. An excellent method to free the DAC from input jitter.

 

 

[LAMF]

For the ones who do not use asio drivers on windows domain, here's a DPC latency checker that analyses the capabilities of a computer system to handle real-time data streams properly:

 

http://www.thesycon.de/eng/latency_check.shtml

 

It may help to find the cause for interruptions in real-time audio streams, also known as drop-outs.

 

 

 

[audioengr]

Jitter can sound different depending on the clock that sources the stream to the D/A, but in general in my system it sounds like a "veiling" or for a visual analogy, a film on the window that I'm trying to look out. The glasses on, glasses off analog is a very good one too.

 

Jitter is modulation of the D/A conversion, so it is a complicated artifact that should be characterized not only in the time, but frequency domain. There are literally no manufacturers that publish the frequency domain data. I believe this jitter is also affected by the digital filtering inside many D/A chips, so what happens is the jitter is actually more damaging using one D/A chip than another.

 

To give you some examples:

 

AES papers that I have read (with dubious synthesized jitter stimulus I might add) conclude that jitter below about 20nsec is not audible to humans. Well, the specs on my standard oscillator is 2psec of RMS jitter and the Superclock4 is also 2psec of RMS jitter. The difference in sound between these two however is immediatly apparent in most systems. The Superclock4 is more musical (vivid) and the standard clock is very detailed, but a bit dry sounding. The difference in jitter amplitude between what the AES says I should hear and these oscillators, which I can plainly hear is a factor of 10,000. I think there is some kind of disconnect here.

 

I believe it is the spectra or frequency content differences that make one more musical than the other. I dont have $50K to spend on test equipment to attempt to quantify and correlate this. I only know what sounds good. Based on the differences between these two clocks and what the AES believes to be audible, I think measurements are black hole anyway.

 

It's like amplifier measurements. Lots of amps out there that measure perfectly, but sound lousy. Lots of amps that have high distortion and other poor measurements, but sound like real music.

 

The problem of inadequate measurement techniques is pervasive in audio equipment IMO. Dont even mention cables.....

 

Steve N.

Empirical Audio

 

[daglesj]

.....ang judging by these replies, the answer is, it doesnt really exist as an audible quantity.

 

I would say there are other more apparent audio problems that cause audio listening fatigue than little old jitter.

 

Loud mixes are one..........

 

As someone else states on this forum, if it sounds good it is good! If it doesnt then jitter would be well down my list of culprits in computer based audio.

 

[tfarney]

"Jitter can sound different depending on the clock that sources the stream to the D/A, but in general in my system it sounds like a "veiling" or for a visual analogy, a film on the window that I'm trying to look out. The glasses on, glasses off analog is a very good one too."

 

An excellent pair of time-tested analogies. But if they're valid here, surely these effects would be easily measurable as an attenuation of frequencies by a clearly audible amount?

 

Tim

 

 

 

[audioengr]

"But if they're valid here, surely these effects would be easily measurable as an attenuation of frequencies by a clearly audible amount? "

 

You would think so, but every time I go to measure these differences, there is no difference. Particularly with a steady-state signal, such as a sine or square-wave. With music data, what do you measure???

 

I ran into this same enigma when measuring cables that I designed. Two cable sounded radically different, and yet all of the classical measurements showed they were identical. to test this, I suggest that someone reading this builds a cable or takes two identical interconnect cables (digital or analog) and performs the following experiment:

 

Immerse one of the cables or cable pairs in liquid nitrogen. This shocks all of the materials in it.

 

Then listen to both cables and hear the difference.

 

Then go to the bench and measure L, R, and C and do some frequency sweeps to compare them. Use a spectrum analyzer if you have one.

 

The result will be that they are identical and yet they sound different. I've done this experiment.

 

Then, find a Time-domain-Reflectometer and do a TDT (Time-domain-Transmission) measurement on the cable with a high-speed step. You will find only tiny reflections at ultra-high frequency as a result of the broken crystal lattice of the conductor metal.

 

So, how does this affect an audio frequency analog cable? how can you possibly hear this?

 

Only God knows the answer right now.

 

Steve N.

Empirical Audio

 

[tfarney]

Perhaps it's more a matter of use of language than one of divine measurement. When I think of "veiled," or "glasses on, glasses off," I think of great chunks of frequency response missing. You must be speaking of something much subtler than I have in mind.

 

Tim

 

[audioengr]

It's not frequency response.

 

The problem with 99% of consumer (and most pro gear), is transient response, not frequency response. This is the response to an impulse function.

 

Most gear compresses the impulse function. This is never in the measurements.

 

Transient response is anything but subtle. Its how the kettle drums sound when they are beaten with gusto on the start of Also Sprach Zarathustra. What happens with jitter is the transient response is "smeared" and defocused. It's mostly high frequencies affected.

 

Steve N.

Empirical Audio

 

[drdigital]

Steve,

 

Quote:

The problem with 99% of consumer (and most pro gear), is transient response, not frequency response. This is the response to an impulse function.

 

Most gear compresses the impulse function. This is never in the measurements.End Quote.

 

I presume you have measured this and therefore have the figures to back this up? If so could we see them for ourselves please.

 

Also a transient response of your USB DAC with valve in it would be intertesting, come to think of it distortion figures would be useful as well.

 

 

I hope you can oblige

 

[Ashley James]

He said and I quote: "A man who knows how much he knows, knows how little he knows"

 

The problem with this and all audio discussions and also the reason why scientists and engineers are so intolerant of Audio is that knowledge is limited and misinformation rife. If you consider the USB specification which was developed by Compaq, HP, Lucent, Intel, Microsoft, NEC and Philips, plus plus a few more A4 sheets full, that comprise a Who's Who of the Industry, all who are technical leaders in their particular fields and likely to have done a proper job, you may question why anyone of lesser provenance might believe he knows better.

 

The best people to judge the merit of these discussions are those with the greatest knowledge and experience. No one with anything less is able see the disagreements as more than probably irreconcilable differences of opinion. Hence the suggestion that we be more laid back about it. The reality is that the ones who could provide clarification wouldn't be or haven't been recognised and as a result they left audio frustrated. It's been happening for many years and must be a significant factor in the decline of the Industry. The best engineers design the best products.

 

The fact is that contributors to this particular thread have ascending levels of knowledge, don't recognise those with more and so plough on delving deeper and deeper into minutae that means nothing poor old audiophiles who just want their computers to be the main source of their Media and a good sound. Worse than that, it may persuade some of them to buy something expensive for one part of a system that has more serious problems elsewhere.

 

It's not a good situation and it's not helpful, but worse still it is likely discourage the wider and considerably larger audience that is currently switching to home media computers and limit the scope of this Forum to the tiny and marginalised minority of audiophiles. Shenzi had some ideas that could help, but I do think we need to consider these issues more carefully and not try to dismiss them as differences of opinion, because many are actually contradictions of provable fact.

 

Ash

 

 

 

 

 

[shenzi]

Supernatural effects makes a change from quantum effects, I suppose.

 

The trouble with a lot of listening tests is that it is differences which are claimed and without references to the original recording, how do you tell which is the improvement? A device may sound hard and can be modified to sound warmer, more dimensional, etc but what if none of those things were on the original.

 

Back to the jitter bugging.