An interesting question was asked recently. Why is latency affected by the sample rate? Isn’t the sample rate time-neutral? Doesn’t it introduce more sound data, and do so over the exact same time period? After all, one second of sound data is one second of sound data, regardless of whether you have 44,100 samples or 192,000 samples.
This number of samples, while unimportant when considering sound in general, is very important when dealing with the computer’s ability to handle the sound data. 44,100 samples takes less processor power than 192,000 samples. However, even when working with the smaller number, the processor is likely going to hiccup once in a while. If there was no room between input and output, then every single little hiccup will become audible in no time, essentially ruining a recording.
The Holding Tank
Enter the buffer. Jack has a buffer that stores sound once it comes in, and then releases that sound when the system is ready to handle it. Because of this buffer, hiccups can be smoothed out before they reach the output, making the sound perfectly clear and clearly perfect. It, in essense, averages out the processor’s speed to all sound data.
This buffer has a specific size, determined by the “Periods/Buffer” setting. The larger the buffer, the longer the time between input and output, but the less chance that processor hangups will cause xruns (under- or over-runs) in the sound, ruining the recording.
Here is where latency comes in.
Size and Space
As was mentioned, one second of audio does not change based on the frequency of sample collection. However, the buffer is not time-based, but memory-based. This means that the buffer has a specific amount of data storage for sound data. One second of 44,100 Hz data takes up much less space in memory than one second of 192,000 Hz data.
Because of this, the higher the sample rate, the less actual “time” worth of sound will fit in the buffer. And, with less time fitting in the buffer, the less latency there will be, although this, once again, increases the risk of xruns in the sound, as the buffer is there for that explicit purpose.
I hope this helps to explain why latency is affected by sample rate, even when the sample rate is definitively one second worth of audio. With this, I hope you have a clearer picture when determining the best settings for ensuring that the system is primed and ready for you to make noise.