Tremulant types

Is there any trade-off between using Hauptwerk's LFO tremulant model and using samples with real tremulants embedded?

I wonder if one is more efficient than the other in terms of memory usage and/or processing power.

Hello Alan,

I think the tradeoffs in performance are obvious (and we are not speaking about the sound, that has been discussed quite a bit on this forum and elsewhere):

On the memory side the LFO trem model uses only little memory. It needs to load the trem waveforms, which are three channels of wave data for each waveform. Most samplesets use two or three such samples per octave, some (most that I did the ODFs for) use one waveform per sample. Since these are usually with a smaller samplerate and also rather short (one trem period), they don't require a lot of memory. Sampled trems on the other hand will require a full set of tremmed attack/sustain samples, and, for wet sets, also the release may sound different enough to demand extra releases, so there is quite an additional memory requirement.

On the CPU side, the sampled trem won't require any additional overhead, as exactly the same number of samples are played with and without trem. For the LFO model, the trem samples are used to modulate the played samples in real time, and this includes amplitude, pitch and harmonic content modulation. I don't know how big this load is, but it is quite obvious that polyphony will be reduced by some margin (I assume this is more than 10 %).

My suggestion would be though to go with what you feel sounds better! It's the result that counts.

Best regards
Reiner

Hello Alan,

I think the tradeoffs in performance are obvious (and we are not speaking about the sound, that has been discussed quite a bit on this forum and elsewhere):....................

..................My suggestion would be though to go with what you feel sounds better! It's the result that counts.

Best regards
Reiner

Hi Reiner.

I guessed as much.
I'm not thinking of changing trem types, I was just acquiring informed knowledge, in case it is needed in the future.
I too was talking about performance, not sound quality.
Apart from where there are memory restrictions (such as the Hauptwerk free version), I guess memory loading is not really of prime importance.
Polyphony is a major concern and for CPU loading, I guess natural tremulants win the argument.
Obviously polyphony is important, but as you say, it's the final sound that really counts.