Calculating latency

Question

I want to know what sort of latency my system has. Can you tell me if I'm calculating this correctly?

At a buffer size of 128 and a sample rate of 44100, the input latency on my system is 150 samples and the output latency is 228 samples, which makes a total of 378 samples. I'm calculating that 228 out of 44100 is 0.0086 seconds, or 8.6 ms. Is that right?

When I up the buffer size to 512, the input and output are 534 and 612, making a total of 1146 samples of latency. 1146 divided by 44100 is .026 seconds, or 26 ms. Is that right?

SteveG_AudioMasters_ · Accepted Answer

Paul_Ferguson wroteSo, it's worse than the software indicates...It must be possible to measure the real value in real time by cabling an in to an out and sending a measurable signal through the system, yes? Yes, but you'd need an entirely independent system to record and measure it on!Some things to do with latency you can measure directly quite easily. For instance, I can measure the delay between two mic arrays (one close, one distant) picking up the same signal by using a clapper board, just like they use in film. You put the clapper board where the sound source is, 'clap' it once, and when looking at the resulting multi-track recording with the mics on different channels, it's dead easy to slide everything back into sync again. You can do this perfectly accurately as there are no dependent variables - everything stays the same except the acoustic conditions. This also gives you a pretty accurate estimate of the reverberation time of the building...But when you want to measure the actual delay through a complete system, you need an independent recorder - just a portable device will do. You can do it very similarly - start with a clap at a sensible distance from a microphone, and have the portable recorder's mic positioned close to the headphones, which you put close to the clapper. When you set it recording, you will have, on your portable recorder, the original clap and a reproduction of it from the headphones. Dump the file into Audition, and it's then dead easy to measure the delay directly from the waveform.It's also worth noting that if you include effects in your monitoring chain, this will also affect the system latency, as you have to include their processing time too.

SteveG_AudioMasters_ · Answer

You quoted 228, but actually calculated the value for 378, which is correct - the round-trip value. it's also worth noticing that if you are a foot from a microphone, you need to add another millisecond, as this is how long it takes the sound to get from your mouth (which is pretty near to where your ears are) to the microphone, and is technically part of the latency calculation as well. But unfortunately that's not the end of it; the figures you're quoting don't include the A-D and D-A latency introduced in your sound device, and there's absolutely nothing you can do about this. A typical figure you have to add for this is typically 5 milliseconds. So with your original figure, you've now got the true latency up to around 13-14 milliseconds, and this is the sort of value where you will notice it quite significantly if you're using headphone monitoring on the round trip, instead of directly from your sound device.

Sign up

To post, reply, or follow discussions, please sign in with your Adobe ID.

Sign in to Adobe Community

To post, reply, or follow discussions, please sign in with your Adobe ID.

Scanning file for viruses.

This file cannot be downloaded