Convolution reverb Vs Surround ranks

[magnaton]

Hello:

I apologize if this topic has been discussed before. A couple years ago a HW user explained to me that with his limited RAM, he resorted to using HW's built in reverb for the rear surround channels. He trimmed off the reverb of the middle (i.e the Nave) recorded ranks, add just a touch of reverb to match the original, then additionally route them to the rear channels with a appropriate convolution choice, that closel matched the original with regards to decay times. I heard it and it sounded fine. By doing this method, he saved on RAM usage (he had no choice) and reduced organ load time.

I understand the benefits of Close & Near recording perspectives for play back as the producers provide faders in their "Settings" tab allowing you to tweak percentages between the two, usually by division which is nice! So the question I have is there anything to gain by loading in the actual "Rear" or "Surround" ranks vs the 'poor-man's-surround' design?

[mdyde]

Hello Danny,

I would say that the two main compromises potentially involved in applying reverb to release-truncated samples are:

1. Truncating release samples can't remove the real reverb from the attack/sustain portion of the samples, so for those portions you would effectively be applying (convolution) reverb on top of the reverb that is already in the samples. With samples that weren't recorded especially wet, that may not be too noticeable, but nonetheless it does sacrifice some realism.

2. Unless you apply a different reverb for every individual pipe, you sacrifice some spatial impression (and thus perceived clarity), since any given impulse response effectively places everything fed through it at a single point in the virtual acoustic space. If you can use multiple impulse responses that were recorded from different positions in the same real acoustic space (e.g. applying different ones to different ranks, or to different portions of ranks, by using groups) then that helps mitigate that.

[mnailor]

Yes, there's a reason, but I don't want to discourage you from enjoying a sampleset that wouldn't fit in your memory without using release truncation and convolution reverb, which also reduces polyphony and so can run with a less powerful CPU. (Releases count as voices for polyphony and continue to demand CPU time until the sampled reverb is finished.)

The best way to add convolution is to start with Close/Direct/Dry samples and truncate releases, then route them to a bus or bus group with reverb added. In HW9, the reverb can be added or not as a per organ setting on the bus. The advantage to using not-very-wet ranks is that the added reverb isn't amplifying the reverb bloom the room adds to the sustain sample while the note is held, which you can't truncate away. That can give a delayed boom effect that sounds unnatural. So I wouldn't use middle or far ranks as a source for adding reverb -- the sustain samples are usually too wet.

Why isn't this quite as good as sampled releases? With convolution reverb, you add one impulse reverb (IR) to each bus, going to a stereo pair of speakers. That pair of speakers has the exact same impulse reverb applied to every source sound, based on the IR's sample of a single noise from a single source position in the building. In my case, 12 stereo pairs allows for up to 12 different IRs, and some of SP's IR packs include that many variants for one building, with different mic distances, source positions, and widths. That's only 12 slightly different reverbs for thousands of pipes, and many people will only apply one or two IRs.

Real sampled releases vary the source position by pipe, so there are subtle differences in the reverbs from pipe to pipe. The accumulated releases give a more lively impression of the room's effect, where the limited few IRs is a more static sounding environment. Whether that bothers you is a personal thing. It bothers me, but I have a big computer and can always load all the sampled releases.

[Martin and I were typing at the same time. He's right, as always. ]

[magnaton]

Thank you Mark and Martin for the quick reply.

My take away from both replies is; convolution reverb is impressive, however its still not as accurate as what was actually recorded. Having all the ranks mix down to a single type of reverb in the surround speakers (HW5 -HW8) the rank response is treated equally. How the different ranks and pipes actually interact with the the acoustics of the room is what is captured in the Surround ranks.

Mark:
The setup I described belonged to someone else not me. Like you, I have sufficient CPU horsepower and 22 channels of audio to load any sample set. I have a successful part time VPO business so I have to keep up with the latest offerings.

Best,

Danny B.
VOCEInstruments.com

[mdyde]

Thanks, Danny.

Having all the ranks mix down to a single type of reverb in the surround speakers (HW5 -HW8) the rank response is treated equally.

Just to clarify, there's no difference between v8 and v9 in that regard (except that it's much easier to apply different reverbs to different organs in v9). (However, as previously discussed, in v5+ you could potentially spread the rear perspective's pipes amongst multiple buses/groups, for applying different reverbs to different ranks or groups of pipes.)