Greetings all.

I'm not an audio engineer but I do understand how software VSTi algorithm reverbs work and that most consumer hardware reverbs use a DSP chip running some algorithms.
But I do not have a clue about how convolution / Impulse Response reverbs work.

There's 100'a of VST reverbs which have a great fan base eg. Liquidsonics-Reverberate, Overloud-Breverb, Redline, Pianoverb....
http://www.kvraudio.com/q.php?search=1&q=reverb&ty[]=e

And there's Impulse Response (IR) hosts eg. the flood of Henry Olonga releases lately etc.....

Can anyone give their view on the pro's / con's of software (algorithm ) reverbs vs Impulse response / Convolution types ?
eg. in terms of CPU load, file / disk storage... ?

Also any opinions on favorites and why ?

Thanks.

 

Question is more on sound side ;)

Once upon a time...10 years ago a lot of ppl thought EVERY reverb/echo/ambience in general could be reproduced ACCURATELY (this is the word) with static impulses.

We know history now, and Lexicon sold their reverb native plugins like hot cakes, whereas thousand ppl tried to reproduce their DSP based hardware reverbs with IR/convolution, without any REAL success.
Because DSP hardware makers introduce more non-linearity in their algos than most ppl think...

So, to me, static linear impulses (not Nebula, witch is non-linear) are good for special FX and acoustic reproduction, not capturing algo based reverbs.
And algos are good for almost everything, but impulses are more realistic on the "acoustic" side of course.

CPU charge can be very low on algo and IR. RAM usage is negligible in both (an impulse is only a few MB).

Nebula, witch is the cream of "captured" reverbs can kill your RAM and CPU at the same time But this is the best mean to capture hardware springs, plates and even some DSP algos reverbs...and guitar cabs (lot of non-linearity here).

Real life example : for orchestral work, i will tend to use Altiverb (IR) for each instrument/sections and use a final touch of algo reverb ... or a Nebula reverb to "glue" all tracks in the same space. Because non-static algos reverb (some will call it "modulated") will make the sound more "alive". And sometimes, we don't care about the famous "ACCURACY"

For general IR work, Altiverb because some rooms are nice and features too, or Reverberate, compatible with a lot of formats.
For general algo work, all ValhallaDSP reverbs. Those ones are very well implemented, easy to use, and cheap.

Why ppl use plates on vocals and springs on guitar ? Not because it sounds always better...but because ppl were used to do this in the hardware days

There are no rules, really

 

It depends on what you're doing, and what your personal taste is. For example, in my scoring stuff, I use a lot of orchestral samples that capture whatever concert hall they were recorded in. So I wind up using Altiverb with the early reflections turned off (and a second instance with the early reflections on for everything else). For regular music -- rock, pop, electronic, etc. -- I tend to use Lexicon PCM, for a "larger than life" sound. Some people, however, prefer a realistic, "no reverb", this-is-exactly-how-it-sounded-in-the-studio sound for their bands, and they'll end up using an IR, and some people want to bathe their orchestras in lots of thick reverb and use algorithms.

Other than the "realism" aspect of using IRs (or getting what is theoretically the sound of specific sampled analog and digital reverb units, plates and echo chambers), I tend to prefer algorithms. Think of it as the difference between using a sampled distorted guitar and recording a real (or sampled DI) guitar and running it through an amp sim. The latter will probably sound better because the amp sim is generating the distortion, compression and everything else in real time, whereas the sampled distorted guitar, which might sound plenty real at first listen, is static and will only ever sound the way it did at the moment it was sampled.

 

Thanks for your perspectives Iggy & Pipotron.

For my taste, I like things such as organic / ethnic percussion to be quite raw, whereas keys, sometime vocals, and synth (strings, pads, leads, arpeggios) to be immersed in reverb of the spacey type - from a short tail to a vast open void effect.

I don't think I like reverbs which impart a metallic character. Is that what you call 'plate'?
I prefer sizes from halls to infinite spaces.

So I'm not really seeking to recreate a copy of a specific location eg. the xxx Opera House.

 

Congrats to Pipotron for his explanation.

IRs captures the frequency span response of a room/hall/etc... very accurately, but only at one volume level. But like Pipo said, non-linear. Is like "sampling" a reverb. This is done applying the convolution (math operation) of the IR to the signal.

That means that the response doesn't change with the input level. But is CPU inexpensive and there are IR based plugins which works with zero latency. Also, you can combine several at once. The metalheads and guitarists in general use a lot IRs for emulating the "speaker part" of the guitar Amplifiers.

Nebula also performs a "sampling" of reverbs and other effects but is on the next level. Non-linear. Captures the frequency response at every volume level. It works with a very sophisticated type of convolution.

If Nebula programs (and other Acustica Audio Acqua plugins with its engine embedded) were less resource intensive would be a no brainer (if they're sampled properly). But they're. Latency can't be avoided. Many people use IR reverbs and Nebula ones in the mixdown.

I haven't much experience with algorithm based reverbs. I know they can do a good work, but in this case the quality depends entirely of the algorithm, so if you use this take your time and make a good research (like for example this thread ;-) )

Cheers all

 

Totally agree about personal taste and workflow arguments here

Dalmation : for your raw reverb, convolution will do wonder, but so will do Lexicon native or Valhallah Room

For out of this world, "spacey" ones, some years ago i would have said convolution (there are some "already stacked" effects in one single impulse, to save CPU) or algos.
But now algos reverbs are so good, i would say algos all the way

valhalla shimmer

About "metallic" tone, yes it can be related to real mechanical plates/springs or algo simulations.
But it can be bad algos programming, too

This is the most usual reproach done to bad algo reverbs, sounding too "metallic".

On this page, there are several real springs "sampled" in Nebula :
Springs

On this page, same with plates Nebula (warmer and less metallic than springs) :
Plates

If you hate those examples...it is sure you don't like mechanical springs/plates
I like plates, but springs not so much...

 

The sad thing is many people reproach algorithmic reverbs without even testing it.

 

Same applies to a lot of things
There are audio myths all over the place.

A day, i was totally bored having too much reverb plugins. I took some dry vocal/ acoustic guitar tracks, and all my algo reverbs full wet, one after the other.
I trashed every bad reverb bluring my signal, bad stereo image...

Only a few algo survived soundwise : Lexicon (nice stereo enhancing), Valhallah (lush sound), 2C Aether and Breeze (flexible and creative)

Now i'm trying Native Instruments ones, Softube based RC24, RC48 (very good for now)