-
Change Request
-
Resolution: Unresolved
-
Minor
-
None
-
None
-
None
-
None
I would like to raise a really small problem of frontend messaging, namely the sounds. This problem can be divided into:
- the quality: sampling rate of .wav files is 11 kHz (except “disaster” level – 22 kHz) and quantization is 8 bits; those are poor numbers for sounds and makes the lack of quality audible for human ear – it should be at least 40 kHz (see below for Nyquist theorem) & 16 bits for quantization
- I personally feel the lack of “commonality” in the sounds: it seems like they are taken from different sources; sounds have different loudness levels
- although the effect is growing together with the severity level (“disaster” is more scary than “information”), it doesn’t feel fully consistent – I’m not sure which is scary more – “average” or “high”; “information” level sound is a bit too scary for “information”
According to the Nyquist theorem (https://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem) you can hear up to twice as small frequency as your sampling rate is. Keeping in mind that humans can hear between 20 Hz – 20 kHz, the sampling rate for digital sounds should be no less than 40 kHz to hear the full range. The established standard of 44.1 kHz will be used & 24 bit quantization
In music production, there is this process called “mastering”; it is applied to every music album. You, as a listener, should never have to tune volume up or down while listening to one album, unless you like some specific song more than others That’s what mastering process is responsible for – it makes all songs sound equally to one another, in terms of loudness and in scope of single album. Exactly same approach should be – but is not – applied here in Zabbix: different sounds are provided in different loudness.
Being all that said, my proposal is based on:
- not going too far from the original sounds, but improve them to give more quality
- grow properly in effect together with severity level
- provide sounds in higher than original quality in terms of sampling / quantization
- provide sounds properly mastered
- provide all sounds to be looping-friendly (referring to “Play sound <> Once”)
- introduce the “hidden” and “increasing-as-severity-increases” low-end (bass) component; this should at the very end (“disaster”) not exceed -5 dB peak (common good practice for bass lines); why not use full spectrum if you can?
So I recomposed the sounds by following the mentioned guidelines. The attached sounds might not necessarily be taken as an object of change, more as an example of how it could be.