Talk:Latency (audio)

From Wikipedia, the free encyclopedia
Jump to: navigation, search
WikiProject Professional sound production (Rated C-class, Top-importance)
WikiProject icon This article is within the scope of WikiProject Professional sound production, a collaborative effort to improve the coverage of sound recording and reproduction on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
C-Class article C  This article has been rated as C-Class on the project's quality scale.
 Top  This article has been rated as Top-importance on the project's importance scale.


"One example of latency is a musical keyboard connected to a computer. When the user hits a key, an audio signal, which is analog, is transferred along the connecting wire in the form of electrical current." Most "musical keyboards" that get connected to a computer are MIDI keyboards. The keys trigger MIDI events, not audio signals. (talk)

I read somewhere that 13ms is an in audable latency amount to trained ears. Wonder if someone knows of any articles relating to this. —Preceding unsigned comment added by (talk) 19:32, August 24, 2007 (UTC)

I did some measurements of professional live production systems in the BBC some years ago (Ref: page 14 of I found that within a production centre about 1.5 milliseconds each way was common and in wide area around 9 to 10 milliseconds each way was acceptable between different production centres provided that audio jitter (or latency variation) performance is kept below 5 microseconds for production audio paths and within 125 microseconds more generally. Keeping the jitter low is vital in order to keep the audio data receive buffering low and thus minimise end to end latency. All this work was done to provide the specifications for the development of linear wide area low latency audio standards and this led to the development of AES47. This standard has now been installed across the BBC to provide high quality audio contributions between all major broadcast centres across the UK and does in practice provide latencies of around 9 milliseconds. —Preceding unsigned comment added by Chrisc (talkcontribs) 21:49, 24 March 2008 (UTC)

The article appears to be biased against Windows. —Preceding unsigned comment added by (talk) 08:48, 6 February 2008 (UTC)

This is not biased against Windows. It merely points out that windows does have more issues with latency, and describes the usual solution. Latency is the main problem for Windows audio. For Linux, it's software and plugins... Ab8uu (talk) 15:01, 26 February 2008 (UTC)

I edited the page to remove a confusing ambiguity. "keyboard connected to a computer" brings to mind the thing I used to write this comment. Looking at the link, it is "musical keyboard", so I changed the link to reflect that. Ab8uu (talk) 15:01, 26 February 2008 (UTC)

Recent edits regarding live event audio[edit]

The new section "Audio latency in live performance venues" contains some problematic assumptions. I don't think the speed of sound in air has much to do with the topic of latency. There's no great problem associated with concert sound in an arena and digital latency, unless the venue is very small and the latency is very large, as it might be in a speaker management system that uses digital FIR filtering for each of the pass bands including the lowest frequencies where the latency is largest. At the most extreme, a latency of 30ms would be fairly easy to accommodate on an expansive arena stage; one would simply time the main speakers back to the kick drum or bass guitar ten (or so) meters away. The speed of sound in air would be equalled by such a long latency. The more common situation of about two to ten milliseconds of IIR filter latency is very easily dealt with at concert venues of all sizes. Perhaps this section would read better if greatly reduced in size, eliminating every mention of speed of sound, humidity in air, absorption of high frequencies in air and venue reverberation. Note that audio cables of all sorts have insignificant latency. Binksternet (talk) 22:18, 24 March 2008 (UTC)

I do agree that the speed of sound in air has a marginal connection with audio latency in systems generally speaking. I was hoping to try and link the design needs of digital audio systems such as those used in live venues to tangible benchmarks and all the multi-speaker sound reinforcement designs I have been involved with over the years do indeed make this correlation. (comb filter effect) In rigs where additional PA stacks are placed half-way down the auditorium in order to keep the front stack SPL down to safe limits. (where they could not be “sky hooked”) These additional stacks would be timed at around 3 mS per metre from the stage and would have appropriate latency inserted into the amps feeding these. I think this is a good example of designed in latency and its link to the speed of sound in air. Currently OFCOM in the UK is trying to force the entertainment industry here to move to using digital radio microphone and in-ear monitoring and failing due to the latency and cost of such a move is a good example of why latency management is vital in PA structures. I was just trying to be complete in mentioning absorption, reverberation and so on and agree all that could come out if you feel it should. Yes of course audio cables have negligible latency however a number of modern digital replacements do have a small impact. Maybe this should be reworded to convey this in a better way? As always with digital structures, it not that any one section is necessarily an issue, but the accumulation end to end that produces the problems. This is a great subject and I would like to help get it as useful as possible. If you were the person who started this up, well done! —Preceding unsigned comment added by Chrisc (talkcontribs) 23:24, 24 March 2008 (UTC)

Speaker systems that are delayed on purpose don't fall under the topic of latency, in my opinion. Latency is the unwanted by-product of digital sound processing. Anything done on purpose is just sound system design. The part of latency that intersects with sound system design is the part that interferes with hoped-for goals or that forces the designer away from one sound system architecture and toward another.
Even digital audio cables don't have significant latency, though the sending and receiving circuits on either side do. I've measured ADAT latency at 96kHz sampling rate to be about 0.5 ms on one particular product.
Digital radio latency does deserve a mention here, as does normal wired/wireless digital processing latency that makes the option of in-ear monitors less attractive to some users. Whatever you have on OFCOM's lack of success in persuading the entertainment industry would be helpful here.
One thing that's not mentioned here (so far) is the latency that is associated with each processing block within a DSP. When I assign a new EQ filter, compressor or limiter element to a generic DSP unit (such as BSS Soundweb, A&H iDR, Peavey MediaMatrix etc.), I get a slightly longer total latency. This topic definitely deserves more effort from all of us. I'll be able to dig in within a few days but right now, no. Feel free to sculpt it as you see fit. Binksternet (talk) 23:54, 24 March 2008 (UTC)
I see where you are coming from. However, while sound system design latency is expected, it provides the constraints by which equipment has to work and system designers have to be very aware of these. Sound system design is also an important example of where audio latency is a key factor. You mention a very important point with the issue of the changing latency with the insertion of additional block of processing. I am sure there is a good bit more to be explored here. I will do some digging around on the digital radio issues and see what I can come up with over the next week or two. —Preceding unsigned comment added by Chrisc (talkcontribs) 00:32, 25 March 2008 (UTC)
I've been thinking about this section quite a bit and it just seems to be overwhelmingly dire in tone. Digital audio latency is not that big a deal in live performance spaces. It's noted, then incorporated into system design. After that it isn't a concern.
Though I have very few references, I'm rewriting the performance section (which at this time has NO references.) This version is rushed and could use more additional information. I'm going to leave all the straight delay parts out--any discussion of delay adjustments made on purpose to live performance loudspeakers should go over to Delay (audio effect)#Straight delay, or to its own article. Binksternet (talk) 20:43, 30 March 2008 (UTC)


The word "ideally" has been brought in via recent edit: "...local circuits should ideally have a latency of 1 millisecond or better." Ideal latency is zero, nonexistant. How else can the recent sentence be stated? Binksternet (talk) 18:38, 30 March 2008 (UTC) OK, but a zero or non-existent latency is not actually achievable in practical digital audio or system designs. Therefore a practical "ideal" statement has to set a range for audio latency such as "1 millisecond or better" which also covers the situation where latencies tend towards zero. While zero audio latency may be the perfect ideal condition, it is not possible to achieve practically so would not be a helpful statement in my view. Theoretically, the only place audio could has zero latency is when measured at the precise point of emission. (i.e. not traveling any distance over any medium.) —Preceding unsigned comment added by Chrisc (talkcontribs) 02:00, 27 January 2009 (UTC)

What are the reference points for latency measurements?[edit]

When we talk about latency, we need to condition that by describing the component or system for which the latency is being specified or measured. The lead in this article probably needs to state as much. My recent edits to the lead attempted to enumerate all the potential sources of latency in a system. I included speed of sound as one potential source. User:Binksternet reverted that edit and has stated that this is not a latency contributer. Speed of sound comes into play if you consider propagation through air to be part of your system. With no explicit definition of the system, we have to assume that in some cases it is included and in some cases it isn't. In any case, I would argue that it is always a potential source of latency that should be given consideration. --Kvng (talk) 22:39, 27 June 2010 (UTC)

Francis Rumsey writes that "latency is the delay incurred in executing audio operations between input and output of a system." I imagine if the system includes air, then the speed of sound is involved. I can see returning speed of sound to the lead. I deleted it earlier because my professional encounters with it are all electronic ones, not natural ones. Binksternet (talk) 05:19, 28 June 2010 (UTC)
You do need to consider air as part of the audio system and a contributor to latency in cases such as sound reinforcement. --Kvng (talk) 15:30, 28 June 2010 (UTC)
Certainly sound reinforcement must take speed of sound into consideration, and since that is my career, I am well aware of the physics. It's just that the word latency is not used by practitioners in my field to describe the delay in air, so I have resisted the word's extension that direction. The word latency is instead saved for electronic processes, mainly digital signal processing and format conversion; processes that apply to all listeners. The speed of sound applies differently to each listener depending on position. Again, I grant that the speed of sound is an undeniable part of a system that has air as a sound-carrying medium, so the loosest definition of latency will include the speed of sound. Binksternet (talk) 15:57, 28 June 2010 (UTC)

1 ms[edit]

I believe it is beneficial to have sub 1 ms latency when monitoring through headphones. The Introduction to Livewire ref does indicate that <3 ms is inaudible. I personally find comb filtering can be an issue starting at 1 ms. 1 ms is, however, difficult to achieve with digital audio systems as AD and DA conversions alone can consume this amount of time. If you can't get below 1 ms it is arguably better to go out to 10 ms than it is to operate in the 1-10 ms area where comb filtering is most annoying. No ref for this yet. --Kvng (talk) 17:49, 21 June 2011 (UTC)

From empirical research into this I concur, it's very frustrating to have to deal with 5-10 ms latencies when monitoring audio. Even a 1 or 2 ms latency is slightly audible if I pay attention (which I do so rarely... ;-) Chris W. (talk) 22:59, 6 February 2013 (UTC)
Uhm, you realise that 1ms is the difference in timing you will experience if you move your head about one foot in either direction from the source, right? Low latency *is* important, but mostly because it cascades into higher latency quickly once you've got more than one hop. -- — Preceding unsigned comment added by (talkcontribs)
1 ms is no problem on its own but when mixed with live sound a comb filter is created. A 1 ms comb filter creates an obvious notch at 500 Hz which some headphone wearing DJs describe as feeling like their brains are being sucked out. ~KvnG 14:34, 20 November 2014 (UTC)
While not quite as crazy as what some people claim they can hear when using oxygen-free USB cables, that still has a very tenuous grip on reality. In order for that to happen, you would need to have _both_ point sources for that 500Hz signal *and* be a very exact distance from *all* of them (including all of their reflections). Since neither musical instruments, nor loudspeakers are point sources (and in most/all cases will radiate sound from a source larger than the wavelength of a 500Hz tone), this seems no more probable than the normal effects that can be seen in any room even without a source delayed by 1ms. Even if you happened to be sitting exactly at the locus of such a local minima for some frequency from one source, your ears are far enough apart that they won't both be in it, and moving your head just a few cm would take you out of it. If you're listening to "live sound" through headphones. then with a 1ms "delay", the sound would actually get to you faster in the headphones than it will from any performer or speaker that is more than one foot away from you. If this effect were as you say, then the performers themselves would also be at risk of having their "brains sucked out" while they played as they moved around :) Sorry, but if you really can hear something, you're going to need a better explanation for it than this. There is nothing magical about 1ms wire latency, phase cancellation depends on the *relative* timing between multiple sources, not the absolute time from any one of them. -- — Preceding unsigned comment added by (talkcontribs)
Point sources and exact level matching are required for a perfect comb filter. We don't need a perfect comb filter to mess up the perceived frequency response in an uncomfortable way. Any motion would not take you out of the comb filter, it just changes the frequency response slightly.
I agree that this effect depends on relative timing. We're talking about a DJ who is, say, 3 inches from the microphone and headphones which are less than an inch from the ears. As the DJ speaks, sound conducts through the bone from the DJs mouth to the ears, that path is perhaps 0.25 ms. The path from mouth to microphone through electronics to headphones to ears is 1.25 ms if we assume 1 ms for the electronics. The difference is 1 ms producing the comb filter with first dip centered at 500 Hz. ~KvnG 14:27, 26 November 2014 (UTC)
Telos' Livewire protocol (based on multicast RTP/IP) also provides "less than 3 ms"[1][dead link] RTT IP transmission of PCM audio over a LAN and they tout it as the ideal choice for various scenarios (radio, mixing, surround work) to that end. Leo Laporte uses Telos stuff in his studio and seems to be happy with it, and he has a fair deal of background in broadcast and radio. Details:
Also this document may be of some relevance (it's interesting reading irrespective): [2][dead link] Chris W. (talk) 22:59, 6 February 2013 (UTC)
Christopherwoods both the links you provide above are dead (404). ~KvnG 14:37, 20 November 2014 (UTC)

────────────────────────────────────────────────────────────────────────────────────────────────────This discussion goes against the guideline at WP:NOTAFORUM. Let's stick to discussing how to improve the article. Binksternet (talk) 22:05, 23 November 2014 (UTC)

Sorry, I didn't mean for that explanation to become a side-show. The point in question "local circuits should ideally have a latency of 1 millisecond or better" seems like a relatively uncontroversial statement, though "should ideally have the lowest latency possible" is possibly more strictly accurate. It's never wrong to try to reduce latency, though sometimes there are technical constraints that are in conflict with being able to do so. The idea that there is some black hole between 1ms and 10ms is incorrect though, and should not be included in the article.
I don't find the original statement "dubious", but finding a reliable source to cite for it would indeed be an improvement. -- — Preceding unsigned comment added by (talkcontribs)
I removed the statement under discussion, about "local circuits" which were in any case not defined. Binksternet (talk) 00:40, 24 November 2014 (UTC)
I think that's a good solution. The previous sentence already covers this, and is less confusing without this uncited qualifier. Thanks. -- — Preceding unsigned comment added by (talk) 20:26, 10 December 2014 (UTC)

External links modified[edit]

Hello fellow Wikipedians,

I have just added archive links to one external link on Latency (audio). Please take a moment to review my edit. If necessary, add {{cbignore}} after the link to keep me from modifying it. Alternatively, you can add {{nobots|deny=InternetArchiveBot}} to keep me off the page altogether. I made the following changes:

When you have finished reviewing my changes, please set the checked parameter below to true to let others know.

You may set the |checked=, on this template, to true or failed to let other editors know you reviewed the change. If you find any errors, please use the tools below to fix them or call an editor by setting |needhelp= to your help request.

  • If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
  • If you found an error with any archives or the URLs themselves, you can fix them with this tool.

If you are unable to use these tools, you may set |needhelp=<your help request> on this template to request help from an experienced user. Please include details about your problem, to help other editors.

Cheers.—cyberbot IITalk to my owner:Online 23:01, 18 January 2016 (UTC)

External links modified[edit]

Hello fellow Wikipedians,

I have just modified one external link on Latency (audio). 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:

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

You may set the |checked=, on this template, to true or failed to let other editors know you reviewed the change. If you find any errors, please use the tools below to fix them or call an editor by setting |needhelp= to your help request.

  • If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
  • If you found an error with any archives or the URLs themselves, you can fix them with this tool.

If you are unable to use these tools, you may set |needhelp=<your help request> on this template to request help from an experienced user. Please include details about your problem, to help other editors.

Cheers.—InternetArchiveBot (Report bug) 21:26, 17 December 2017 (UTC)