# Audio stability of a computer system

Well, suppose we have a computer of any kind, a sound card and an operating system installed and ready to go. Nice, we think we are ready to make music then! However, this may not always be the case. In fact, the system (in the following I will refer to system as the union of a computer, with a sound card, and an operating system configured for audio) may not be able to perform stably enough from the audio point of view. In simple words stability is the capability of the system to run over a prolonged period of time without encountering issues. Imagine we are trying to record a song. To be able to do that not only the system will have to respect our requirements on latency (if any) and sampling variables, but it will have to do it for a long time span, hopefully way longer than the time over which we need good audio performances, so that sample errors, losses, pops, clicks, xruns and crashes don’t ruin everything, wasting our time. We may intuitively define the audio stability (just stability in the following) of a system as the capability to perform audio tasks without encountering issues over time. For sure, we prefer a system as stable as it can get.

But how to define audio stability?

This is a hard question. A way to technically quantify the audio stability of a system for sure exists… but it would be restricted to that system only first. Second, it would be probably based on over-complicated technical data that are quite hard to understand and use, since a computer and the operating system running on it can sum up to a very complicated system actually. Then the complete benchmarks of audio performances of computers would prove themselves to be probably useless, as hard to interpret information are obtained… and they are not even general. Instead to use this approach, I prefer to give this definition of audio stability, based on my experience. In the way it is formulated it can be applied only to systems running Linux, but those are the ones under my scope anyway…

A system is said stable from the audio point of view if on $N$ repeated continuous recordings over a time interval $\triangle t$, taken from all the available inputs of the sound device(s), providing simultaneous monitoring to all outputs, in at least practically lowlatency operation, the frequency of xruns

$f_{xruns}\doteq\frac{X}{N\triangle t}$

with $X$ the total number of xruns counted over all the records, is less then 3 mHz.

where practically lowlatency operation has been defined in another post. The recording can be made with any software and depends on the particular system as expected. I have chosen these conditions as, when $\triangle t$ equals 5 minutes, these are the conditions that one needs to make the simplest possible simultaneous multi-track record of an average song (which is 5 minutes long) relying on the system for the monitoring while having to fear less than a xrun each take of the song. This condition could then supply a “fairly stable” system. At this point it is better to note again that latency is not everything. Lowlatency, for example, is not required if only mastering or post-production are of interest. Also, it is useless if the monitoring for the musicians is not supplied by the system but it is by other means. So, according to the goal, many systems that are not audio stable according to this definition may be indeed very stable for their purpose. However, I think that it is fair to pretend that a system can convert itself to a lowlatency one if needed, so it is fairly multi-purpose. Also, if a system is audio stable when operating in lowlatency, then it will be stable also in not lowlatency conditions (hopefully…). Last notes: I have chosen the xruns as a parameter as the more general digital errors are inherent of any digital system and are way less disruptive of the overall quality and workflow with respect the xruns, that really makes impossible to work and listen to the results. Also, a system could be very stable if recording with just command line utilities, without launching a GUI, but proven unstable when recording in a full DE with a DAW. Hence, not even this definition gives a general information. However, it is quite easy to manage, makes measurements easy and maybe it can work as a good rule of thumb. As always, even if a lot of technicalities can be pulled in, as music is art, everyone has a different goal and every definition is a matter of common sense as well. At the same time, it is quite hard to find a balance between definition of quantities easy to manage enough to be operationally useful and representative of complex systems in the same time. A big lesson given to me by acoustics…