how do you measure latency?

[dobro]

How do you measure latency on your system? Is it simple and straightforward, or do you need a degree in computech?

 

[Slackmaster2K]

Once again, latency is only a concern when mixing, and it's not really a big deal.

If you're concerned about LAG, that is, the time offset between recorded tracks, then the best way to go about measuring it is with a real life test. Create a short track consisting of 8 or so hi-hat ticks. Load it up in your multitracker, and route an output from your soundcard to an input on your soundcard. Hit record and what you should end up with, in a perfect world, is a near duplicate of the original track. However, in the real world, you'll probably see between 0.5 and 1ms of lag from track one to track two. Not a huge deal all things considered. Can you use a similar technique to benchmark your converters? Yes!

If you really are concerned about latency, that is, the time between moving a fader on screen and hearing the results of that move, then the mathematics are simple enough. If your 24bit playback buffer size is 6Kilobytes, or 2000 samples (6,000bytes / 3bytes per sample = 2000 samples), and you're playing back 44,100 samples per second.....well, you don't need a degree in computech to divide.

Slackmaster 2000

 

[dobro]

Hm, okay, that's how I'll do it.

Reason I ask: I'm gonna start overdubbing, and I thought latency/lag was an issue in overdubbing. No?

 

[Slackmaster2K]

No. Well, ocassionally...but it's usually a hardware/software/driver problem.

Almost every decent soundcard on the planet is going to support live input monitoring. That is, when you're recording, the inputs are routed to the appropriate outputs at the source, thereby nearly eliminating lag. The hi hat click test will demonstrate this....in fact I posted some cool graphics of my results a while back. No matter how high you set your buffers, the lag between tracks should be a millisecond or less.

Slackmaster 2000

 

[cordura21]

Latency test, from a Digi engineer

I read this one on Digidesign's forum, and I thought it could be of use. It's pretty much what Slack said:

Good question! It can be tough to measure the total I/O latency with DAW applications because most of them (including Pro Tools) do some auto-time-adjustment to compensate for some of the buffering delays (mostly in the disk buffers).

Here's how we measure Total I/O Latency...

First of all, you need two physical inputs and two physical outputs. Take a small patch cable and connect output 1 to input 1. Take another cable and connect output 2 to input 2.

In the DAW software (like Pro Tools), create 3 audio tracks and one Aux track.

On disk track A, put a small audio file/region with a nice square wave (or impulse). Route the output of this disk track to output 1.

On disk track B, route the input from channel 1. Route the output to some unused bus (you don't need it).

On the Aux track, route the input from channel one, and the output to channel 2.

On disk track C, route the input from channel 2, and route the output to some unused bus (you don't need it).

Record enable tracks B and C.

Record the impulse.

Look at the results on tracks B and C. The distance between the peaks on these tracks is the Total I/O Latency in the system (including converter delay, hardware delay, driver delay, and audio engine delay).

The reason we use 2 separate disk tracks is that any "auto-time-correction" used by the software will be cancelled out when you subtract tracks B & C, leaving you with the total I/O latency.

Regards,
Kenn LeGault
Digidesign