KVR Audio

Krakatau · Post by **Krakatau** » Mon Dec 05, 2005 9:30 am

I didn't find a similar post on the database actually, so apologise if the subject would have already debated

My question is :

Can some of you explain us ,shortly and explicitly, what are the quality and weakness of both kind of reverberation units and in what situation, one kind would, at least, be more recommanded than the other...

Considering that in both case, you would use anyway something of high quality in its particular field ( algo or IR )

Music Engineer · Post by **Music Engineer** » Mon Dec 05, 2005 6:12 pm

the convolution approach is superior when it comes to the most authentic simulation of real rooms (or other linear time invariant systems). the algorithmic approach on the other hand gives more flexiblitiy to the user because here the user can adjust the parameters (like pre-delay, damping, etc.) all seperatly - these are hammered in stone in IR-reverb. it is somehow analogous to the difference between samplers and synthesizers - a sampler can reproduce real instruments EXACTLY, but lacks the option of sound-transformations whereas in synthesizers, everything can be tweaked, but it will hardly be possible to get a convincing piano out of a VCO-VCF-VCA structure

soniccouture · Post by **soniccouture** » Mon Dec 05, 2005 6:23 pm

yes, i agree. also, Traditional reverbs ( lexicon, TC etc), give a 'fake' sound, but it's also a very attractive sound, and is well suited to pop vocals etc, where they give a polished artifical sheen that is often very desirable.

hifiboom · Post by **hifiboom** » Mon Dec 05, 2005 6:33 pm

the best algorithmic reverbs easily wins against every convolution reverb.

convolution quality depend on the recorded impulses and seems to sound dead without liveness...

many impulses sound dull, because of the mic characteristics....

Music Engineer · Post by **Music Engineer** » Mon Dec 05, 2005 6:39 pm

hifiboom wrote:the best algorithmic reverbs easily wins against every convolution reverb.

convolution quality depend on the recorded impulses and seems to sound dead without liveness...

...that's of course a matter of taste. and we are considering only HQ-impulse responses here. it is also straightforward to take the impulse responses of the best algorithmic reverbs and to use them in a convolution reverb.

chagzuki · Post by **chagzuki** » Mon Dec 05, 2005 6:47 pm

I agree: convolution sounds smooth and artifact-free, but somewhat dead.

hifiboom · Post by **hifiboom** » Mon Dec 05, 2005 6:51 pm

braindoc wrote:
hifiboom wrote:the best algorithmic reverbs easily wins against every convolution reverb.

convolution quality depend on the recorded impulses and seems to sound dead without liveness...
...that's of course a matter of taste. and we are considering only HQ-impulse responses here. it is also straightforward to take the impulse responses of the best algorithmic reverbs and to use them in a convolution reverb.

did you ever compare a recorded impulse of a real Lexicon with the hardware effect unit. There are differences, you can easily hear...

Music Engineer · Post by **Music Engineer** » Mon Dec 05, 2005 6:58 pm

chagzuki wrote:I agree: convolution sounds smooth and artifact-free, but somewhat dead.

mmmh...actually there is no such thing like "the sound of convolution". it all depends on the particular impulse response - unless, of course, you do a time varying modulation of some of the algorithmic reverb-parameters, which would be impossible with convolution. ...but O.K. ...i don't want to start a new war here

Music Engineer · Post by **Music Engineer** » Mon Dec 05, 2005 7:00 pm

hifiboom wrote:
did you ever compare a recorded impulse of a real Lexicon with the hardware effect unit. There are differences, you can easily hear...

i have to admit, i didn't

eduardo_b · Post by **eduardo_b** » Mon Dec 05, 2005 7:19 pm

Krakatau wrote:Can some of you explain us ,shortly and explicitly, what are the quality and weakness of both kind of reverberation units and in what situation, one kind would, at least, be more recommanded than the other...

I think there's no best choice unless one is trying to replicate a live instrument in a real space, in which case you have (in theory) a standard by which to judge the virtual version. The reality of what reverb sounds best is completely subjective in many situations. If you're using a synth running through one or more effects to create a unique sound, how authentic a reverb might sound is essentially meaningless because the sound itself has no real world counterpart. Better to try different reverbs with different settings until you find something that works with the sound and the mix.

kritikon · Post by **kritikon** » Mon Dec 05, 2005 8:39 pm

Can some of you explain us ,shortly and explicitly, what are the quality and weakness of both kind of reverberation units and in what situation, one kind would, at least, be more recommanded than the other...

The short and sweet answer for me is
1.Sound quality
2.CPU useage.

The slightly longer answer is...

1. I don't much like the quality of most native reverb plugins for most types of reverb (but not all). Apart from the Princeton ones (PACE

) and the UAD EMT plate (DSP card...

...but I'm going to get one anyway). The sound quality issue would come into it especially if you're doing accoustic music or orchestral or anything where you want a more lush or realistic reverb or pretty well anything with long reverb tails.
For more dense mixes on electronic styles, algorithmic reverb plugins are....acceptable sometimes.
Also for mixes where you use alot of reverb on many layers - to me most of them soon add up up to a nasty metallic or grainy mush - although that's only my personal taste.

But...as mentioned already - they can be much more flexible. Things like predelay and RT60 can be adjusted on convolution reverbs - some even have Eqs that can be used to a degree for damping etc, but you have no control over density, early reflection envelopes etc (mind you...not that many algorithm reverbs go into early reflection deeply enough either). Even a simple thing like changing patches - which is something I'd do for dub breakdowns etc...you simply can't do that running live with convolution, but some algorithm reverbs can switch seamlessly.
Other things like low and high damping - not usually well implemented on convolution. If you want to fit a reverb to the tempo of your music...it's just more difficult with convolution (although not impossible).

2. CPU...not really too much of an issue with shorter reverbs on most up-to-date hosts. But I'd struggle with long reverbs - anything over 3 or 4 seconds starts to chew up CPU. I tend to render to audio with long convolution reverbs, and for obvious reasons, that's not always ideal. There's also the side issue of latency...your host will compensate for VSTi and audio reverb, but if you have a track running and then want to record a midi part, or simply noodle out a new idea on a keyboard - 1/2 a second latency makes it impossible with the convolution running. So you have to switch it off - it doesn't make things impossible, simply annoying if you've got a sudden inspiration you want to get down quickly.

If you use reverb over a master recording...I haven't used an algorithm (s/w) reverb yet that is good enough. And for mastering, CPU is a non-issue, so to me it makes complete sense to go for convolution, and the highest quality available.

To some extent it also depends on how you work - for whatever reasons, you may want to work at 16bit. Convolution impulses can be a bit dodgy at 16bit (although with a decent host, you can work your FX at 24 or 32 bit, even on a 16bit song). And if you work at high samplerates - to use convolution effectively you'd need a pretty decent PC, or resign yourself to alot more rendering to audio. So if you're a 96KHz slut, you may have to use algorithm reverb through necessity, or your PC will blow up. Mine would

I.R. vs algorythmic reverbs - what to choose and when ?