Talk:Jitter

Rewriting of Jitter Types
I changed a fair amount of the guts of the article. In some cases, as in the definitions of "period" vs "cycle-to-cycle" jitter, I changed it back to what it was in Nov. 2004. If someone thinks that definition is wrong, please comment here so we can discuss before changing it. I'm pretty sure the article is more right than it used to be.

other changes.


 * 1) rejected the false paradigm shift.  Jitter has always and will always be a concern for all links.  Added some additional interconnects
 * 2) the qualitative or quantitative thing seemed a bit strange the way it was.  It's like saying you can measure something as speed or as km/h.
 * 3) jitter does not have to be rapid or repeated.. and in fact, the article later says so.  Hence, that had to go.
 * 4) the standard deviation of peak-to-peak was confusing.
 * 5) phase departure and phase perturbation..  are these industry-standard terms?  I've never heard them as such.
 * 6) usually random but if cyclc it may be expressed in hertz.  Well.. anything cyclic can be expressed in hertz (actually.. no.. it's cycling can be expressed in hertz.. the thing itself can strictly not be).  Also, there's no reason to say that phase jitter is usually random, as far as I know.
 * 7) peak to peak period jitter is the worse case of cycle to cycle jitter.  This is not true.  Read that JEDEC jitter definition.  It defines period vs cycle-to-cycle jitter.

I also changed the name of this section from jitter types to phase jitter metrics because I think the term jitter types is more suited to describing actually different types of jitter (eg. deterministic jitter, sinusoidal jitter, random jitter) as opposed to just how it is measured.

Jurgen Hissen 07:16, 22 August 2006 (UTC)


 * I think this could do with some additional terms being added. E.g. Time Interval Error (TIE) seems to be a common measurement in test equipment. Also, JEDEC talks about phase jitter. Long term jitter, accumulated jitter, etc. — Preceding unsigned comment added by 77.99.44.113 (talk) 16:30, 23 April 2018 (UTC)

"jitter"/"chitter"/"chatter"
Jitter is a word in general usage for the tendency of something to jump abruptly backwards and forwards. It supposedly comes from the Scots word, "chitter", meaning to shiver, which might in turn be related to the onomatopoeic word "chatter", as in, "Her teeth chattered due to the cold".

See also "jitterbug", a dance with fast jerky movements, and "jittery", to be prone to sudden movements when startled, due to anxiety

"chatter" can refer to a constant stream of Morse Code signalling (signal chatter), or a continuous stream of fast staccato speech (-> chatterbox) … that then gets shortened to chat (as in, "We had a nice chat"). This has been adopted by the internet community (eg chatrooms)

"jitter interpolation"
Anyhow, back to jitter and telecoms …

As well as referring to unwanted clock timing errors, jitter is also a deliberate technique used for obtaining sub-bit resolutions in digital audio and image processing, using spare bandwidth to trade off unneeded "timing" resolution into additional "value" resolution.

"jittered" interpolation on output signals (obsolete term?)
If we have an audio DAC running at twice the rate of your sample frequency, we can reproduce output values with "half-bit" resolution by deliberately "jittering" the lowest bit between 0 and 1, and then low-pass filtering out the jitter signal to leave an smoothed "averaged" intermediate value. If the DAC runs at eight times your sample frequency, we can adjust the proportion of "ones"s to "zero"s, to achieve one eighth of a bit accuracy, and so on. This output can be in simple PWM form (see: "one-bit convertors"), or, to filter out these high-frequency artefacts more efficiently, the output stream of ones and zeroes may be deliberately organised to reduce the occurrence of consecutive ones or zeros, to shift the frequency of this jitter noise as far above the base frequency as possible.

Jittered interpolation on sampling
For sampling at sub-bit resolutions, you can mix the signal with low-level high-frequency noise (out of the audio range), oversample, and then the statistical proportion of ones and zeroes in your lowest bit as it gets toggled at ultrasonic rates by the noise, tells you roughly where the underlying signal level was within that lowest bit range. A majority of ones tells you that it was a high sub-bit value, a majority of zeroes tells you that it was a low sub-bit value. So you just average a succession of jittered sample values together to give you the final higher-resolution sample value. The noise source is required to "tickle" the ADC into jittering, so that it doesn’t just slowly step between discrete values as the input signal value rises and falls.

This technique was probably only worthwhile when component counts were a significant factor. The drastic drop in prices with chip-scale integration probably makes this technique redundant, a DAC/ADC feedback loop with an oversampled output signal and a "successive approximation" approach is probably simpler and more reliable to implement, since the hardware has now gotten so cheap.

jitter and information theory/QM
As a general statistical method for shunting information between domains (e.g. the frequency domain and the accuracy domain), jitter principles show up in information theory and quantum mechanics. The technique is useful in a variety of situations where we want to take measurements below the level of an apparently unavoidable quantisation threshold.

jitter interpolation and image sampling
An analogous effect can be used in digital image processing to obtain images with sub-pixel resolution. Instead of taking a single image with a standard exposure (which would be limited to the resolution of the sensor, and might also suffer from blurring if the camera is unsteady), jitter techniques can make positive use of camera shake to take a succession of less blurred shorter-exposure images, which can then be enlarged, and then aligned and rotated with sub-pixel accuracy to obtain the best fit, with the adjusted images then being superimposed to produce the final composite image. While the noise levels in each individual picture will be higher, this background noise will tend to cancel in the final image. The degree of extra image resolution that can be achieved this way will depend on the number of images taken, and the accuracy of the sub-pixel realignment process. One might expect the best results if the angle of the image varies slightly between images as well as the offset of the image. Rotating and offsetting multiple images requires a certain amount of processing power, but this is not an insuperable limitation when cleaning up and enhancing image data from, e.g., the Hubble space telescope.

A more "consumerised" application of jitter interpolation might be to fit the CCD image sensor of a consumer digital camera with a piezo transducer to sweep the CCD sensor sideways across the image field while a photograph is being taken in order to produce an orderly series of offset images that can be processed and assembled by the camera's embedded software to achieve a "virtual resolution" for the camera that is higher than the resolution of is sensor array. I don’t honestly know if jitter interpolation is already being used in mass-market electronic cameras yet, but if it's not, it's probably just a matter of time before it shows up. ErkDemon 05:38, 23 March 2007 (UTC)


 * Aside from the first part, regarding etymology, all of these sub-sections refer to what is normally called "dither". Perhaps it would be helpful to offer a link to that in the article. --212.44.20.129 (talk) 15:08, 4 January 2010 (UTC)
 * There's a link to Dither in the See also section. ~Kvng (talk) 22:28, 16 April 2022 (UTC)

Merge phase noise here
pros? cons? that page needs rescue; for one thing, the distinction between time- and frequency-domain in the lead of phase noise is disputable. Fgnievinski (talk) 01:21, 3 July 2014 (UTC)


 * Support - Phase noise actually seems to be in pretty good shape. Jitter needs more attention IMO. There is nothing incorrect with the Phase noise lead. We have time and frequency domain representations of the same phenomenon; they should be in the same article. Articles should be demarcated by concepts not terminology. ~KvnG 13:58, 7 July 2014 (UTC)


 * It is incorrect when it states that "phase noise is the frequency domain representation ... caused by time domain jitter". The time-/frequency-domain division seems artificial. Fgnievinski (talk) 00:43, 8 July 2014 (UTC)


 * The time and frequency domain division is always artificial - different ways of looking at things. That doesn't mean that the lead is wrong. ~KvnG 23:07, 8 July 2014 (UTC)


 * Oppose merge. The topics are related, but they are often treated separately. Phase noise is more specific to stable linear oscillators. It is a well developed subject (and yes, there are ways of measuring phase noise using time domain measurements). Jitter is a more general topic that can encompass other issues such as a transmission path. Jitter is also more concerned about digital transitions and the clock edge. Glrx (talk) 19:12, 8 July 2014 (UTC)


 * It would be like merging digital and film photography just before less people are using film these days. There are technical teams that use phase noise exclusively without mention of jitter. User: ChrisD Jan 2015. — Preceding unsigned comment added by 72.48.98.212 (talk) 16:21, 22 January 2015 (UTC)


 * Oppose merge It's not our job to teach the engineering community the "right" way of looking at the phenomenon. We should represent how reliable sources treat it and for better or for worse most source focus on one viewpoint or the other, rarely both. I'm in favor of a prominent hat note or a "See also" to cross reference the two articles though. SageGreenRider (talk) 15:47, 30 April 2015 (UTC)


 * Yes, it is also my experience that these are treated separately; Your point is valid. It was useful and improved my engineering when I recognised that they were two sides of the same coin. ~Kvng (talk) 14:33, 3 May 2015 (UTC)
 * I have added a Phase noise link in See also. Phase noise already links back here though somewhat obliquely. ~Kvng (talk) 22:32, 16 April 2022 (UTC)

"deterministic jitter is bounded and random jitter is unbounded"
The article makes the claim "...deterministic jitter is bounded and random jitter is unbounded..." which is a) unsourced and b) not quite accurate (I believe). I think a more accurate statement is "...deterministic jitter is always bounded but random jitter can be unbounded in some circumstances..." An example of bounded random jitter is crosstalk (aka bounded uncorrelated jitter). The aggressor is random from the point of view of the victim if the data pattern on the aggressor is never correlated with the victim's pattern. I don't have a citation though. Thoughts? SageGreenRider (talk) 14:38, 30 April 2015 (UTC)


 * I support this improvement. Do you have a reference to back it up? ~Kvng (talk) 14:30, 3 May 2015 (UTC)


 * I have made these proposed changes and changed the Dubious tag to Citation needed. ~Kvng (talk) 18:13, 7 November 2017 (UTC)


 * I have found and added some refs that clear this up. The refs see random jitter as unbounded. ~Kvng (talk) 13:40, 17 July 2018 (UTC)

External links modified
Hello fellow Wikipedians,

I have just modified one external link on Jitter. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:
 * Added archive https://web.archive.org/web/20120207103425/http://documents.exfo.com/appnotes/anote119-ang.pdf to http://documents.exfo.com/appnotes/anote119-ang.pdf

When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.

Cheers.— InternetArchiveBot  (Report bug) 01:19, 23 April 2017 (UTC)