I'm sure i remember cliff or one of the moderators here mentioning in the past that making the cab samples any larger than 1024 would make very little difference at all to the sound?
That would have been me. Let me elaborate on what I said:
1. Assuming for the moment that you have a reflection-free IR of a
guitar cab that is 42ms long (2048 points) with no leading silence, the audible difference between using the entire IR or just the first half (1024 points) will be nil.
2. Nobody presently has a collection of such IRs. In order to acquire some, I will have to test in a larger space than I normally use. Nearfield IRs intrinsically attenuate reflections from nearby boundaries, but anything after the first 15-20ms in such an IR will consist entirely of those reflections and will therefore contribute nothing to the accuracy of the
cab sound.
3. The portion of the "room" you get in a 42-ms IR is inadequate to accurately characterize a shower stall, let alone a desirable acoustic space. In many cases, the
earliest reflections in a hall occur later than 42 ms. Ergo, even with a 2048-point IR length, you won't be able to capture a convincing "room sound." You could possibly get some early reflections, but you'd have to supplement those with a reverb block.
4.
If you play bass and have really convinced yourself that you've got to have the characteristic sound of a bass cab, the additional 1024 points will enable you to capture the highpass (IOW the low-frequency rolloff) behavior of cabs with a cutoff frequency of 23 Hz or higher. Given that most recorded bass parts - and many live ones as well - are DI'ed from the instrument itself, this capability would at best be of limited use. Additionally, if your monitor cuts off higher than an octave
below the cutoff frequency of the cab you're simulating, its cutoff behavior will defeat the additional accuracy you got from the longer IR.
