Do higher sample rates generally mean lower latency?

[mjfe2]

Seems counter-intuitive, but whenever I set the buffer size & sample rate for my Mackie 400f I've noticed the higher I put the sample rate, the smaller the latency becomes. Is this a general rule for latency?

[Stan]

64-bit floating point processing on that beast - it could be it is optimised to function at higher sample rates -
i'll wait for the experts

[GaryM]

Nothing to do with optimisation - it's simple maths. At 44.1 kHz each sample lasts for 1/44100th of a second, so a buffer of 256 samples would hold 256 x 1/44100 = 5.8 ms. If you change the sample rate to 88.2 kHz then the same buffer would hold 256 * 1/88200 = 2.9 ms.

[mjfe2]

Ah yes, thanks!

[Hugh Robjohns]

Yes, and the inherent delay through the A-D and D-A conversion filters is also halved if you double the sample rate.

Hugh

[Martin Walker]

Agreed.

Just be aware that since you’re processing twice as much audio data with 88.2kHz than 44.1kHz sample rate, your hard drive requirements will double, as will the processor overheads of any plug-ins/softsynths you’re running.

No such thing as a free lunch I’m afraid

Martin

[Mixedup]

Funnily, this is one of those obvious things I'd never really stopped to think about... I'd never really noticed a huge difference recording at different sample rates (as opposed to bit-depth) in terms of sound quality, but hadn't considered the impact on latency. I suppose it offers a justification for tracking (or playing V-Drums etc) at higher sample rates when in search of lower latency, where previously I'd seen none — even if you later convert down to lower sample rates for a mix.

[Martin Walker]

Indeed it is Matt - I know some musicians who insist on working at 96kHz simply to get headphone vocal monitoring latency as low as possible (presumably they have no ‘zero’ latency monitoring features built into their interfaces, or insist on monitoring with plug-in effects).

Martin

[Mixedup]

Yes... I've never found zero-latency monitoring on any interface much good unless working with outboard gear - compression, comfort reverbs etc. Even when overdubbing a guitar part onto an existing mix, you end up having to fiddle with the DAW output levels to avoid drowning out the unprocessed DI part you're playing in.

It's also particularly useful to get latency as low as possible when playing BFD etc via V-Drums.

[dmills]

If only things were reliably that simple!

AD and DA decimation/interpolation latencies MAY be reduced, but it depends on the filtering.

On the PC side, the key issue is the time between the interrupt and the user space code running. Doubling the sample rate with the SAME size buffers will double the interrupt rate and thus double the overhead while halving the latency, however the same effect could be achieved by simply running with a smaller buffer at the old rate.
If the machine can cope with sample rate 2F at buffer size B then it can probably cope better with sample rate F at buffer size B/2....

CPU cost for convolution effects also rises as sample rate squared given a constant temporal length of impulse, so convolution reverbs and models can become very expensive.

What the best approach is depends strongly on exact hardware and software details and is usually best determined by experiment on a case by case basis.

Regards, Dan.

[Exalted Wombat]

Seems counter-intuitive, but whenever I set the buffer size & sample rate for my Mackie 400f I've noticed the higher I put the sample rate, the smaller the latency becomes. Is this a general rule for latency?

Other things being equal, yes. But, of course,other things never ARE equal

[TEMPUSFUGITSTUDIO]

Hope I'm not getting off track to far; if the CPU can handle it, what are the best settings for the overall sound? Higher sample rates sound better I assume? What do most people sit at for the best sound overall? Buffer sizes probably don't actually affect the sound at all. Am I right?

[Exalted Wombat]

Hope I'm not getting off track to far; if the CPU can handle it, what are the best settings for the overall sound? Higher sample rates sound better I assume? What do most people sit at for the best sound overall? Buffer sizes probably don't actually affect the sound at all. Am I right?

Higher sample rates up to 2X the highest frequency your equipment can produce, record and reproduce sound better. That basically means - don't go lower than 44.1KHz. Beyond that.. well, let's just say it's questionable If someone swears his 96KHz recording sounds better than a CD (which is 44.1 KHz) make sure he can HEAR the difference, every time, and with his eyes closed. I think it's generally accepted that anything even higher than that is just plain silly, or is only justified in very special cases.

You have a LOT of other things to worry about in getting audio quality before even thinking of whether a rate over 44.1KHz might sound better.

Buffer size only affects the sound if it's wrong enough to cause glitches.

[Martin Walker]

Quite right - buffer size don't change the sound at all.

However, there are definitely situations where increasing from 44.1/48kHz to 96kHz can provide audible improvements - depending on what audio interface you’re running, what plug-ins you’re running...

However, as EW mentions, this is what I like to call a ‘suck it and see scenario’ i.e. try it, and if you can hear the improvments and are happy with doubling the CPU overheads then go for it. If you can’t hear the difference stick with 44.1kHz

Martin