Tring to understand low latency. Is latency fixed?

[sionnach]

I have been wondering a bit more deeply about latency settings recently. I never really thought about it too much, rather i just found a setting that works and is usable but recently i have been thinking.

I know that hardware can play a part and software config is important too but are buffer settings, at least in theory, fixed? In other words, if i set my sample rate, buffer and periods in QjackCtl, will the reported latency be the same on every system? These are just mathematical calculations after all, right? In which case the whole battle for low latency is about finding a setting that gives you the lowest latency with stability?

If i was to use something like AV Linux that is properly configured, will each combination of sample rate, buffer and periods will give me the same latency as on a system that isn't configured at all for low latency audio? It'll just be a case of being able to set a lower buffer before my system can't handle it? Am i correct in my understanding of this?

[Michael Willis]

Am i correct in my understanding of this?

Yes. The idea is that tuning your system performance allows you to run your audio with lower buffer sizes without getting xruns. Let's say you configure two different systems for 48kHz, 1024 samples, and 3 periods. In theory they will have the same latency. Of course, in practice there are some things that confound this, for example different audio interfaces have different amounts of latency overhead in addition to that reported by JACK.

[CrocoDuck]

You are correct. There are many sources of latency in a computer audio system, and buffering is one of them. If the buffer size, the number of buffers and the sample rate are the same, then the latency associated with buffering is the same in every system. The additional latency sources in your system are pretty much dictated by your soundcard, and possibly data transfer over the bus the card is on. Other sources of latency are perhaps totally negligible. Be aware that latency can vary widely between sound-cards or by how the drivers are operating on them. You can measure the full-chain latency of your system with jack_iodelay as described here.

[Openmastering]

I'd like to add, don't fret over latency except if you overdub something or play something where you need live ITB monitoring (hardware monitoring usually doesn't cause latency).
Many think it's a kind of benchmark how good their system is. That's just not true. It's just faster with more risks. Audiophiles for example don't need low latency at all.
So, yes, it's sometimes really important, but most of the time, not.

Knowing your roundtrip latency (as stated before from CrocoDuck) is quite important if you're serious about overdubs.
Example: you set Jackd for a displayed latency from 10ms (it's a standard threshold for latency free monitoring) but your soundcard adds 2,5ms on the way in and 2,5 ms on the way out. You now have a roundtrip latency from 15ms. Your singer now hears itself with a slight delay and it'll ruin your takes.

[sadko4u]

You can measure the full-chain latency of your system with jack_iodelay as described here.

Or use LSP Latency meter developed by CrocoDuck.