Page 1 of 1

Sampled Tremulant Improvement?

Posted: Sun Nov 12, 2023 10:00 am
by rayjcar
One of the criticisms of sampled tremulants versus synthesized is that each sampled note has a different tremulant phase relationship. The point in the tremulant cycle where the sample starts will be different for each note. As a consequence, the amplitude and pitch shift will be different. When many notes are played in the same rank, there will be a mushing effect not present on real organs. With real theatre organs, all the pipes in a given rank (and on a given pipe chest in fact) will have their amplitude and pitch shifts generally in sync, realizing that the speaking speed of a given pipe will be somewhat different.

There are some software utilities which are meant to allow phase alignment of two different audio signals. My question for Martin or others is whether using such a utility would resolve this issue, or would it create other problems related to attack and release? In the case of tremmed samples, my idea is to match the phase of the tremulant envelope, which has a rate of maybe only 2 or 3 Hz, as opposed to matching the zero crossings of the actual audio frequencies.

Ray

Re: Sampled Tremulant Improvement?

Posted: Sun Nov 12, 2023 1:04 pm
by mdyde
Hello Ray,

The problem is that if the phases of the tremulants recorded in the audio samples were to be aligned then one would either have to use an artificial amplitude envelope for the pipe attacks and releases, instead of real sampled pipe attack/releases (which I think would involve a significantly greater loss of realism than non-synchronised tremulant phases), or one would have to delay the speech and release of each pipe until its sampled tremulant phase matched those of other pipes sounding at the time (which could introduced a large and variable latency/jitter, e.g. up to 1/2 second for a 2 Hz tremulant). Personally I don't think either would be acceptable to most people.

Re: Sampled Tremulant Improvement?

Posted: Fri Nov 17, 2023 7:23 pm
by dkoschinski
rayjcar wrote:One of the criticisms of sampled tremulants versus synthesized is that each sampled note has a different tremulant phase relationship. The point in the tremulant cycle where the sample starts will be different for each note. As a consequence, the amplitude and pitch shift will be different. When many notes are played in the same rank, there will be a mushing effect not present on real organs. With real theatre organs, all the pipes in a given rank (and on a given pipe chest in fact) will have their amplitude and pitch shifts generally in sync, realizing that the speaking speed of a given pipe will be somewhat different.
In my opinion there is no way to record (sampled) tremulants in sync because of the time differences per tone between the swings. The swings (when playing multiple notes) therefore will never run simultaneously and without certain phase differences !
The best solution to simulate a1 tremolo is to live-record a specific tone with the real tremulant and simulate this effect with a specific LFO-filter including the necessary ADSR beating follow-up. Then overlay it in a certain mixing ratio over the original sample.
I have tested this method sufficiently and it comes closest to realism.