Hello vpo-organist,
The main benefit of the 96 kHz option (in Hauptwerk v5+) is that it runs the audio engine at 96 kHz, thus avoiding aliasing distortion when pipes' pitches rise by more than a certain amount (because it massively increases the frequency headroom before that could occur, and because Hauptwerk doesn't use per-pipe anti-aliasing filters, which would increase CPU demands by about as much as simply running the audio engine at 96 kHz, with the 96 kHz engine also avoiding any additional distortion that such anti-aliasing filters would themselves entail).
It would technically be possible for Hauptwerk then to down-sample the output of the audio engine as a whole back to 48 kHz for output to the audio interface, but that wouldn't save any CPU overheads (and in fact would very slightly increase them) and any real-time sample rate reduction causes a small (but potentially audible) quality loss, because the frequency and phase response of real-time anti-aliasing filters can't be perfect (especially if they need to avoid adding noticeable latency).
Hence if you the 96 kHz option is selected for the audio engine then Hauptwerk outputs the audio at that same rate, thus avoiding losing quality by real-time down-sampling (anti-aliasing filters). For the same reason, it records audio files at the same rate. If you want to, you could then down-sample the resulting files to 48 kHz within your audio editor, which can be done in much higher quality than would be possible with any real-time filter (and without any latency needing to be added).
This is the excerpt from the v5.0.0 changes section of the Hauptwerk release notice that covers the sample rate setting:
The new 96 kHz option is more CPU-intensive, sacrificing achievable polyphony (typically about a further 50%) for higher audio quality. It gives better signal-to-noise ratio (less distortion, and better clarity), especially for larger upward pitch shifts and deep tremulants (which might otherwise introduce 'aliasing' distortion). (It also has other smaller sonic benefits, even if not raising the pitch.) It is recommended, especially if you raise the base pitch of any organs, provided that you have ample spare computer processing power. The same sample rate will be used for audio output and audio recording, to avoid any further slight distortions that would occur by down-sampling.
To hear the difference that the setting makes to aliasing distortion, try raising an organ's base pitch to 600 Hz and then play a scale on a pure high-pitched rank (such as a flute rank); you will probably hear strange dissonant 'aliasing' tones on some (seemingly arbitrary) high notes without the 96 kHz option. As well as the degree to which the pitch is being raised, how beneficial it is also depends on where the sample set producer has applied an anti-aliasing filter within the samples. Hence, without the 96 kHz option, for some sample sets aliasing might occur even with smaller pitch increases.
(Real-time per-pipe anti-aliasing filters aren't used in Hauptwerk, since the CPU overheads of doing so would be about as much as simply running the audio engine at 96 kHz, whilst 96 kHz also gives other sonic benefits.)
Re. other audio quality options:
vpo-organist wrote:Currently I have set the RME audio interface to 96 kHz, General Settings 96 kHz and the new options in Organ Settings to high quality. Is this all necessary, or is there a possibility to use the good quality and use a 48 kHz output (which are sufficient for our ears)?
- I would always recommend setting '
Organ settings | Organ preferences | Audio engine: Audio engine processing quality' = 'Higher', since that will give a significant and noticeable benefit (significantly lower distortion) in all cases.
- If you can then spare the computing power (polyphony), I'd recommend also selecting 96 kHz. That too will have a quality benefit (some further reduction in distortion), but probably not quite as noticeably (unless you raise the pitch of any pipes, e.g. by using deep tremulants or by increasing the base pitch, in which using the 96 kHz option becomes more important).
- Then set '
Organ settings | Organ preferences | Audio engine: Tremulant/wind supply/swell box/relay model quality' = 'Medium' to start with, which should always be safe, as I mentioned previously:
mdyde wrote:If you then also want to experiment with setting 'Tremulant/wind supply/swell box/relay model quality'='Higher' for some sample sets, then do it with caution, bearing in mind the points in the release notice about the effects on the wind model and also the need for very high per-core CPU performance, especially if trying it with large organs and/or those with intensive wind models.
Here are some excepts from the v7.0.0 changes section in the release notice that covers setting those options:
There is also a new “Tremulant/wind supply/swell box/relay model quality” organ preference, with the following three choices:
o “Lower (not recommended; least CPU-intensive; legacy v2-v4 quality)”.
o “Medium (good balance of quality vs. CPU overheads)”.
o “Higher (caution: needs high per-CPU-core performance for large organs)”.
It controls the time resolution at which Hauptwerk's tremulant, wind supply, swell box, and relay models are processed, and thus their quality, accuracy, smoothness, realism, and response. Higher settings are more CPU-intensive, but give noticeably better results and overall clarity, provided that your computer has sufficient processing power to handle it. Per-CPU-core performance (especially CPU base clock speed) is particularly important for the models, even if your CPU has lots of cores. For the 'higher' option to work well with very large organs your CPU will need to have very high per-CPU-core performance, as well as plenty of CPU cores. Hence the ‘medium’ mode is the default.
The 'lower' mode processes the models at a similar resolution to v2-v4, the 'medium' mode comparably to v5-v6, and the 'higher' mode now processes them at an even higher resolution, again giving noticeably better results.
...
Because processing the wind supply model at different resolutions may slightly affect the behavior of the wind supply, in the Hauptwerk Advanced Edition a new “Wind model organ pipe flow adjustment” preference has also been added on the “Organ settings | Organ preferences | Wind supply model” screen tab, allowing you to reduce the amount of air that each organ pipe draws from the wind system when it sounds, in case the wind system is too unstable or 'wobbly'.
...
In summary, we recommend selecting the new 'higher' mode for the “Audio engine processing quality” organ preference for all organs, and the 96 kHz sample rate option on the “General settings | Audio device ...” screen, if you have sufficient CPU horsepower, since they will yield significantly and noticeably better results.
For the “Tremulant/wind supply/swell box/relay model quality” organ preference we recommend selecting the ‘higher’ mode (only) if:
o The organ is a theatre organ, or:
o The organ is reasonably small (without an excessively demanding wind supply model), or:
o Your CPU has very high per-core performance (including a very high CPU base clock speed).
In other cases, we recommend selecting the ‘medium’ option for the “Tremulant/wind supply/swell box/relay model quality”, and if in doubt, select the ‘medium’ option.
Best regards, Martin.
Hauptwerk software designer/developer, Milan Digital Audio.