You are here

A/D and D/A converter latency issues

For everything after the recording stage: hardware/software and how you use it.

A/D and D/A converter latency issues

Postby Bob Moose » Tue Mar 26, 2013 2:11 pm

Hello

Generally speaking, I could not find much information about A/D and D/A latency.

I think starting from dummy practical examples will help.

Suppose we use a PCIe to multiple ADAT I/O interface set to 32 samples of latency (0.7 ms at 44.1 kHz). It is connected to:
* ADAT/analogue converter A that has an I/O latency of 15/12 samples
* ADAT/analogue converter B that has an I/O latency of 48/48 samples

We ignore the following latencies:
* ADAT and PCIe latencies, because they are very small as far as I remember
* program latency (operating system, audio interface driver, DAW, plugins, etc), because this is something else
* constant latencies related to speaker distance, cable length, etc


Problem 1 - live and tracking

We play a live instrument, like an electric bass, into a realtime software FX, then plugged to a bass amplifier.
If we use converter A, the latency is 12+32+32+15 = 91 samples (2.1 ms)
If we use converter B, the latency is 48+32+32+48 = 160 samples (3.6 ms)

A difference of 69 samples corresponds 1.6 ms, which looks short enough to be ignored perhaps. But when using converter B the conversion itself adds more latency than the proper computer/audio interface, and this is something salesmen would rarely tell you about. Even "zero-latency monitoring" (ie when the audio signal does not go through the OS and programs) would actually be 96 samples (2.2 ms) in this case.

When playing in live with a computer, the converters themselves should be fast enough. But there is a trade-off between speed and quality, as Dan Lavry explained several times (which is not saying that a faster converter always sound worse). How to find the right compromise? Are there converters that are really too slow for live? Very few manufacturers publish A/D and D/A latency figures.


Problem 2 - recording

When recording 2 tracks called A and B, using converter A and converter B, then track B should be 33 samples late from track A.
Should we add a 33 samples delay on track A?


Problem 3 - playback

We import a 2-track recording and play L through converter A, R through converter B. R should be 36 samples late.
Should we add a 36 samples delay on L?


Problem 4 - multipass setup

Consider a setup in which the signal flow goes several times through the A/D and D/A from converters A and B. Should the A/D and D/A latencies be compensated like in problems 2 and 3? Even when using a modern DAW, this one should only know about the computer/audio interface latency, not the A/D or D/A latency.


All the best
Bob Moose
Frequent Poster
Posts: 695
Joined: Thu Jan 17, 2008 12:00 am

Re: A/D and D/A converter latency issues

Postby Jack Ruston » Tue Mar 26, 2013 2:20 pm

Not sure I quite understand what you're getting at...

In terms of latency we're always concerned with the total value...AD, software and buffers, DA. It matters not how we arrive at those values, but rather whether we can deal with that total, and how.

It makes sense to measure the roundtrip latency for yourself.

NEVER split stereo files or multi mic'd sources across two converters if you can possible avoid it. Not only do you absolutely have to compensate for the difference, but you may also find that there are subtle differences in the frequency or phase response of each which affect things in a negative way.

J
Jack Ruston
Frequent Poster
Posts: 3512
Joined: Wed Dec 21, 2005 12:00 am

Re: A/D and D/A converter latency issues

Postby Sam Inglis » Tue Mar 26, 2013 3:59 pm

Most DAWs will automatically compensate for latency caused by buffering, but not for latency caused by A-D/D-A conversion. However, many of them let you enter a manual offset in samples if needs be. I think Pro Tools is the exception here.
Sam Inglis
Frequent Poster
Posts: 1978
Joined: Fri Dec 15, 2000 12:00 am

 


Re: A/D and D/A converter latency issues

Postby Hugh Robjohns » Tue Mar 26, 2013 4:15 pm

Bob Moose wrote:Generally speaking, I could not find much information about A/D and D/A latency.


Latency in A-D and D-A converters comes principally from the digital filtering used for the anti-alias and reconstruction filters (respectively), and is typically of the order of about 0.75ms per filter stage at 44.1k, and progressively shorter at higher sample rates. IN general, the converters add a very insignificant chunk of time to the overall latency figure for a computer-based interface.

Even "zero-latency monitoring" (ie when the audio signal does not go through the OS and programs) would actually be 96 samples (2.2 ms) in this case.


Zero-latency monitoring is normally an entirely analogue domain monitoring path from analogue input direct to analogue monitoring output -- avoiding not only avoiding the OS and programmes, but the converters as well. There is no way you can achieve zero-latency if you go through converters!

H
User avatar
Hugh Robjohns
Moderator
Posts: 17475
Joined: Thu Jul 24, 2003 11:00 pm
Location: Worcestershire, UK

Technical Editor, Sound On Sound


Re: A/D and D/A converter latency issues

Postby The Korff » Tue Mar 26, 2013 4:17 pm

I think HD, HDX and Native (with a capital 'N'!) do take converter latency into account, because they necessarily use Avid converters (or third-party ones that have been made compatible with the Digilink interface — and I'm pretty sure Apogee designed their converters to have the same latency as the official Avid boxes for this reason).

But as long as you track things through the same set of converters (eg. all drum mics through the same interface, etc), it's not something you should be losing much sleep over.

Cheers!

Chris
The Korff
Frequent Poster
Posts: 1776
Joined: Fri Jan 08, 2010 11:28 am
Location: The Wrong Precinct

I AM KORFF


Re: A/D and D/A converter latency issues

Postby Goddard » Tue Mar 26, 2013 6:51 pm

Hugh Robjohns wrote:IN general, the converters add a very significant chunk of time to the overall latency figure for a computer-based interface.


Hugh, I believe you meant to say that the converters add a very insignificant chunk of time to the overall latency figure, yes?


Hugh Robjohns wrote:Zero-latency monitoring is normally an entirely analogue domain monitoring path from analogue input direct to analogue monitoring output -- avoiding not only avoiding the OS and programmes, but the converters as well. There is no way you can achieve zero-latency if you go through converters!


Oh dear.

Yes, absolutely. But there's an old "near zero latency monitoring" vs. "zero latency monitoring" marketing rhubarb at play here. On many interfaces (especially with onboard DSP/FPGA routing/mixing) the monitoring path includes the converters (and probably should more properly be termed as "near zero" or "low latency").

BTW, blame RME for this marketing-speak wordplay with their "ZLM":

http://www.rme-audio.de/english/techinfo/lola_lomo.htm

"Monitoring in hardware on the other hand works virtually without delay..."

"...ADM also allows for monitoring the input signal via the hardware in real-time..."

Hmm, dunno about "real-time" or how much a "virtual" zero is. ;)
User avatar
Goddard
Frequent Poster
Posts: 999
Joined: Tue Apr 03, 2012 11:00 pm

 


Re: A/D and D/A converter latency issues

Postby dmills » Tue Mar 26, 2013 7:12 pm

As long as everything used in a multi miced setup has the same latency the actual fixed latencies are not a big deal down here, however 3ms corresponds near enough to 3 feet in air, which is more then enough to stuff up stereo or REALLY stuff up M/S. One possible exception might be headphone mixes for vocalists where the latency can cause severe comb filtering with the bone conduction path, turning up or down by 3db or so will usually solve that.

Interestingly some of the Wolfson micro parts (among others) have a low latency mode where the filter impulse response is shortened to improve the latency at the expense of either passband ripple or aliasing, not that sure it is a net win for most things.

As to not having these latencies in the analogue domain, if you were filtering sharply in the analogue system and wanted linear phase you would have exactly the same latency... Of course nobody does high order linear phase filters in the analogue domain (because they are a bear), so we tend not to see the tradeoff that is implicit in the maths.

Regards, Dan.
dmills
Frequent Poster
Posts: 1552
Joined: Thu Aug 24, 2006 11:00 pm
Location: High Wycombe, UK

Audiophiles use phono leads because they are unbalanced people!


Re: A/D and D/A converter latency issues

Postby Jack Ruston » Tue Mar 26, 2013 7:16 pm

We're always going to have to pay attention to the phase of stuff where latency is concerned (just to make sure it's right hopefully, rather than actively DO anything), but I think we're coming now to a time when you can do cue mixes ITB without worrying about the latency. HDX is 0.5ms in and out at 96k. That's really good. For most of us we're quite happy with much longer latencies than that, but some musicians are very sensitive to it, especially singers for whom there is no natural delay between their instrument and their perception of its sound.

Obviously there's no way around the situation where you wish to overdub onto a mix with loads of instruments and plug ins with latency, but you can always print a couple of stems and turn the tracks off if that's what you need to do. It's no great hardship.

J
Jack Ruston
Frequent Poster
Posts: 3512
Joined: Wed Dec 21, 2005 12:00 am

Re: A/D and D/A converter latency issues

Postby Hugh Robjohns » Tue Mar 26, 2013 8:09 pm

Goddard wrote:Hugh, I believe you meant to say that the converters add a very insignificant chunk of time to the overall latency figure, yes?

Whoops! Thanks. Corrected.

H
User avatar
Hugh Robjohns
Moderator
Posts: 17475
Joined: Thu Jul 24, 2003 11:00 pm
Location: Worcestershire, UK

Technical Editor, Sound On Sound


Re: A/D and D/A converter latency issues

Postby Goddard » Wed Mar 27, 2013 10:38 am

Hugh Robjohns wrote:
Goddard wrote:Hugh, I believe you meant to say that the converters add a very insignificant chunk of time to the overall latency figure, yes?

Whoops! Thanks. Corrected.

H

Right-o!

FWIW, otherwise known as the converter "Group Delay" time, typically specified in the datasheet in terms of [number of samples/sampling frequency] and expressed in seconds.
User avatar
Goddard
Frequent Poster
Posts: 999
Joined: Tue Apr 03, 2012 11:00 pm

 


Re: A/D and D/A converter latency issues

Postby Mixedup » Wed Mar 27, 2013 10:52 am

Sam Inglis wrote:Most DAWs will automatically compensate for latency caused by buffering, but not for latency caused by A-D/D-A conversion. However, many of them let you enter a manual offset in samples if needs be. I think Pro Tools is the exception here.

Worth pointing out that most DAWs now feature a semi-automatic feature, whereby you can 'ping' to detect the latency of DA/AD and any signal path in between for setting up 'external plug-ins'. Not that this feature actually *works* on all DAWs with all hardware setups (as I know to my cost!), but it does on several.
User avatar
Mixedup
Jedi Poster
Posts: 4017
Joined: Tue Sep 02, 2003 11:00 pm
Location: Laputa

Re: A/D and D/A converter latency issues

Postby Bob Moose » Fri Apr 05, 2013 7:02 pm

Thanks everybody

I asked about this because sometimes I use 2 different microphone preamplifiers with built-in A/D conversion (it's only because I don't have enough microphone inputs). In this case I compensate for the delays manually.
Same for D/A.
I did not know some DAWs could be configured for compensating the converter latencies, I will check mine (Reaper).

About 'zero latency monitoring', yes it's not possible with a digital device. It is simple when recording and monitoring from an analogue mixer, but when using rack preamplifiers or even the line inputs of the A/D converters, what is the best solution? Inserting an analogue mixer in the recording path would require a very clean one (or at least you really like the colour), and for mobile recording an additional mixer may be a bit cumbersome.
I guess some people connect jack sockets in pairs on a patch panel for duplicating the signals (one branch connected to the recorder, the other one to the monitoring path). And maybe there are dedicated active boxes that do the same.

Regards
Bob Moose
Frequent Poster
Posts: 695
Joined: Thu Jan 17, 2008 12:00 am

Re: A/D and D/A converter latency issues

Postby Martin Walker » Mon Apr 08, 2013 8:05 pm

Bob Moose wrote:I did not know some DAWs could be configured for compensating the converter latencies, I will check mine (Reaper).

Hi Bob!

Yes Reaper has some really handy features built-in to take care of stuff like this - for instance. I often find myself using the Auto Detect Delay Compensation button in its ReaInsert effect, so I can insert external hardware effects and get them (pretty well) lined up so they act like plug-ins 8-)


Martin
User avatar
Martin Walker
Moderator
Posts: 10100
Joined: Wed Jan 13, 2010 8:44 am
Location: Cornwall, UK

Steampunk & Synth News | Mad Scientist Mode



Who is online

Users browsing this forum: No registered users and 3 guests