Challenge to VST reverb developers

[greendoor]

Why - when so many reverb plugins are available - do hardware reverbs still sound better? I think i've found out why, which means some plugin developer will get my respect (or money) when they finally nail it ...

The whole thing about convolution reverb has proved several things to me:
1 - the best reverbs are predominantly artificial, from high end hardware boxes
2 - samples are not really flexible for real music applications. (I need to tweak everything to match the groove, especially with reverse or inverse sounds)
3 - algorithm VST plugins aren't at the same level as hardware algorithmn reverbs (even hardware from the 1980's sounds better than current plugins) - BUT
4 - convolution reverb can't capture the full effect of the classic hardware reverbs (for example - convolution can't model any pitch modulation that might occur in a hardware reverb tail - should you want that).

I just found out recently that some of these hardware reverbs require two stereo impulses to capture what they do. This is because a sound directed to the left input creates one full stereo tail, and a sound directed to the right input creates a different, complimentary, full stereo tail. That means 4 output channels are required, even if they are summed to a stereo pair. This means a stereo mix sent to this reverb will give better stereo imaging than most "stereo" reverbs that simply have two channels of wet processing.

I see no reason why a VST plugin reverb cannot nail, if not blow away, the great hardware reverbs from the past. So far this hasn't been done, and i'm wondering if this true stereo processing might an important missing ingredient.

I write some code, but not DSP, so my understanding is very limited. But I believe many reverbs are based on the freeverb model of 8 delay lines. Obviously the timing of each delay is a critical part of the sound and, as with real rooms, it's important to avoid nasty sounding nodes that result from too many similar delays reinforcing certain frequencies. (would the use of prime numbers yield the ultimate reverb tail? - I can only speculate).

I wondering what would happen if a vst reverb plugin was designed using essentially two stereo algoritmns in one reverb. The delays would have to be carefully selected so they wouldn't clash, and in fact would sound especially lush when used together. The tails created by each side would be tailored so that the tail created by the left side input alone created a full stereo tail, but skewed to the left. The right hand tail would likewise be skewed to the right.

The use of modulation in a verb tail is controversial, and frankly I hate the sound. It doesn't occur naturally, but apparantly was used by some hardware makers. As long as it could be turned off, I don't mind it as an option. The new Glaceverb sounds pretty good, providing the modulation is turned off!

Although convolution is popular at the moment, it's a clumsy method and I think a new breed of excellent algorithmic reverbs could, and should, leave them for dead. Maybe the the coming Classik reverb is a sign of things to come, but I don't want to deal with a company that uses challenge/response or PACE or anything that will prevent me using my investment down the track if they go bust.

The big thing missing in VST reverbs is excellent Inverse or Non-Linear tails. That requires full editability of the timing. I want a big slow reverse reverb tail that peaks right before the snare hit - that sort of thing. The relationship between reverb and groove should be explored more fully - so perhaps reverb parameters could be linked to host tempo in the way that delays are now?

Numbers are important - the hardware GUI's, although crude, were often better than software GUI's. Nothing worse than fighting with a mouse to set 100, and all you can get is 99 or 101 ...

Sorry to rant, but I have a dream ...

[Kim Lajoie]

Yeah, I'm hanging out to hear how IK's Classik reverb sounds.

Forever,




Kim.

[mandolarian]

Compared to what hardware? With whose ears? LexiCON? No thanks. Quantec. Maybe. Real rooms still sound better than any hardware. Oh wait, they are the ultimate hardware. Convolution to my ears is the next best thing to being there, or having the mics there. But not much tweakability and where's the fun in that? Modeling and convolution could be very powerful combo once we have a few extra teraflops kicking around.

But verb is such a personal/style/genre/context thing that I don't believe there is one solution. Just lots of options. UNLESS you include Smart Electronix Anechoic Room Simulator. It really is the ultimate expression of technology.

[wsaidah]

greendoor i was looking to mod some tails what i came up with is to use two convolution plugs one for the pre and the other for tail the trick is to put your mod plug in between the convolution plugs and if you dont like your pre verb going throught the mod plug it will take energyXT. one more thing you can capture Modulation from hardware just sample both left an right at the same time.

[greendoor]

Mandolarian - no argument that real rooms are the ultimate (Abbey Road chambers ) but not practical. I'm just stating that convolution reverb seems to be state of the art for VST reverb, and 99% of the time it's being used to recreate hardware digital reverb less than perfectly. This seems unnecessary, but where are the good sounding VST algorithm reverbs that can equal a 1980's Lexicon or TC?

wsaidah - yes, maybe there are workarounds. I don't actually like modulated reverbs - probably shouldn't have mentioned it, but the convolution experts are quick to say that convolution can't do chorusing. I think the impulses of modulation that appear to work aren't actually "chorusing" in the true sense of modulated pitch change. I think they give an appearance of movement a little like using eq to simulate phasing.

Anyway - my point is simply that, despite a plethora of VST reverbs, ancient hardware boxes are still better, and it seems we are stuck with convolution of samples taken from these boxes until somebody in VST land can raise the bar. Please, oh please.

[kritikon]

Some good points. I'm sure it can be done with plugins - it's only coding with both h/w and s/w. Don't know what particular reasons force the sad lack of decent reverb plugins though - monetary gain, or lack of it would be the strongest reason, I imagine. You can charge $2000 and upwards for a reasonable h/w reverb - but look around within the DAW world...people moan and groan if something costs $200 (a tenth of the price ).

Princeton have shown that it most certainly can be done - theirs is a top notch plugin. Limited, but excellent quality nevertheless. And any Lexicon coding can be reproduced - there's no special magic h/w involved....they just have chosen not to for the present, unfortunately .

Don't know the business figures, but it must pay for Lexicon etc to stick with h/w and target the more serious user. Alot of musos serious about their music consider a decent reverb as a good investment...whereas the proportion of users within the s/w world is more heavily tilted towards amateur/hobbyist etc (call it whatever). So the level of experience or discernment is not as high overall. You see evidence of that by the number of postings that claim Ambience etc is every bit as good as a Lexicon ( ). Inexperience is all. One day, those people will hear the difference.


Having said that...some of the plugins now are as good as the previous h/w to my ears. M1 etc can stand up to most of the 90s standard black boxes. And it seems to me there is less fierce competition for budget h/w reverb units nowadays (that could be in part due to the changing nature of music production, but to me it seems also likely it's because the s/w is equal to that level of h/w). I'd have to think really hard between buying a new budget Alesis, Zoom, Digitech etc multiFX for reverb and a good choice of s/w plugins. I don't hear alot of difference. Although for sure, I completely agree that mostly the best of the best s/w reverbs (with only one exception, and convolution) don't hold a candle against any Eventide, Lexicon or even TC upmarket model (although I don't like TC reverbs at all).

But it can be done. Princeton have done it.

And if piracy is a concern for the upmarket manufacturers...I would buy a UAD or other DSP card like a shot if Lexicon were to release a PCM plugin on that platform. Even though the dynamics plugins on that platform tempt me, they don't tempt me enough, but I'd happily pay $500 or more for a PCM plugin on top of the price of the card itself. I suspect others would be of the same mind too. And warez don't seem to be an issue for DSP cards.

Don't know about the quad outputs though - I have an old unit that has quad outs with all kinds of fancy options for routing outputs (front/rear) and cross-modulating them...and I never use the facility. It doesn't sound any better going out quad than it does mixing to stereo....although I can definitely see an application for it if I was doing surround sound stuff. But if stereo is your ultimate destination, the only benefit I see to 4 outputs is a marginal easier access to control. But you can mix the front/rear within the h/w just as easily.


But yeah......I would like to see some improvements on the s/w reverb front. Convolution for me (although I'll buy the Princeton very soon) - but as you rightly say...convolution has too many limitations on it at the moment...not tweakable enough by far.

[greendoor]

Reverb is a very subjective thing, and different genres require different things. I'm looking for rock/pop musical sounds, and I personally love high end TC verbs and can tolerate some Lexicon sounds. I'm guessing Mandolarian is involved in classical or film post production, as he likes Quantec or real rooms.

I trialled the Princeton VST reverb earlier this week, and was disappointed (partly responsible for my ranting here). The Princeton demo VST has some nice warmth, but is hardly smooth. Sounded like some artifacts you would expect from a low-end box. I just feel that VST reverb developers are re-inventing the wheel - it's all been done before, but for some reason we are being withheld the good stuff ...

[kritikon]

As you say Greendoor...a matter of taste. There's a definitive character difference between TC and Lexicon - I prefer Lexicon, and I like the Princeton too....I suspect that may apply in general too? Lexicon can be quite "in-yer-face" reverbs with a hint of artificialness - I think the same applies to Eventide, so probably if you like one, you'll like both. Whereas I find TC reverbs to be "cold" and inobtrusive to the point of being timid. If you're a TC fan, you're probably more pissed off by native reverbs than a Lexicon fan ... I've not heard any native one that even comes close to simulating TC.

Although I suppose you have the benefit of TC DSP cards. Us poor Lexicon people don't even have that comfort Excepting the Lexicon Core32 soundcards (which I nearly bought...pleased I didn't though, as they folded that operation pretty quickly...I had hoped it might take off, and eventually lead to onboard PCM algorithms instead of the MPX ones ...wasn't to be)

[greendoor]

Don't know about the quad outputs though

Sorry - I didn't mean quad outputs at all. I'm talking about how the internal processing is done. As I understand it - there are various reverb options available, and maybe this is where current VST plugins are falling short. Consider these options:

1 - mono to mono (hardware plates, springs, single mic chamber). There is a lot that can be done with a great mono reverb. Most digital reverbs are stuck in stereo mode, and usually sound inferior when collapsed to mono. This is where convolution of a nice mono plate sample, for example, is worthy of experimenting with.

2 - mono to stereo - most cheap hardware boxes summed the left and right dry inputs. So no matter whether you feed the left, right, or both inputs, you get the same stereo output. This skews the stereo field - a hard panned guitar will sound wierd because the stereo reverb tail is obviously comming from centre stage ...

3 - independant 2 channel stereo - this is where most plugins are at the moment. The left and right hand sides are process independatly. That means, a guitar panned hard left will get some reverb tail on the left side only. Nothing comes out the right hand side, because nothing is being fed into the right hand side processor.

4 - "true stereo" - a misnomer, but it seems to be applied to 4 channel stereo to stereo reverbs in the hardware realm. Also multi-channel convolution processors can use seperate stereo impulses (4 channels) to recreate these hardware boxes better. This is what i'm suggesting the VST plugin developers might need to consider. I think it might explain why the high end hardware boxes sound better.

With one of these, a guitar panned hard left would get a stereo reverb tail from the stereo output bus, similar to what two mics in a real room would pickup if the guitar was close to the left wall. A guitar panned hard right would also get a stereo reverb tail, but different. A centred guitar part would get a rich, diffuse combination of all four channels of reverb.

I hope this makes sense - especially to the people who can code VST plugins.

[kritikon]

Thinking about it ...... that says it all really. It boils down to the new PC generation feeling ripped off unless everything's the price of pocket money.

The fact that Lexicon ventured into the soundcard market, but had their fingers burnt is sad, but true. The great unwashed public obviously thought it too expensive. Yet they got pretty good ADCs, dual MPX processors onboard, and onboard Db soft limiting at input (switchable). Yet it folded before it even got going.
And the TC Spark cards...It doesn't sound like they do immense business (although I could be wrong there). So unfortunately, the general message the good companies are getting is that the majority prefer nasty but cheap reverb.

Which means they're going to stick with h/w for the foreseeable future - at least we can hope for DSP cards though. It's sad but true...search around many of the topics at KvR, and the general view is that users are happy with things like Ambience, M1/M2, Warez versions of Waves reverb, etc. Those users are not going to be prepared to shell out for a truly good reverb. It saddens me.

[greendoor]

I thinks that's true of VST in general. TDM is for the Lords, and VST is for the peasants. Occasionaly the Lords throw the peasants a bone or two, but generally the peasants are growing restless and starting to feed themselves But to anyone interested in making their music the best possible, a stunning reverb is the sound of money. I agree that VST reverb is as good, if not better, than cheap hardware. I sold my Lexicon Alex a long time ago! Convolution is giving us a taste of what the Lords have had for years, but until you hear the real thing (well, the real artificial thing that is), you don't know what you are missing. Understandably, the shareholders of Lexicon and TC don't want to give their trade secrets to the peasants at a price they can afford. But ultimately, it's just code and somebody has to crack it eventually. I think we already have the processing power, if somebody can just find the magic combination. Convolution is like using a bulldozer to crack open a nut.

[kritikon]

Ok...got it about the stereo field.

I wonder if there's any actual real stereo stuff going on within most of the plugins - I suspect both outputs simply have exactly the same reverb output, but mixed at different levels. Like a summed stereo.
Certainly you're right...some of the even quite old black boxes had different reverb character coming out of the front/rear processors - so it would be no surprise to find out they actually processed the stereo field separately as well.....in fact I don't even remember seeing front/rear as an available option on any s/w plugins.

I don't know....I think it's probably more likely to be just sloppy coding, taking the easy route. Lexicon have made reverb from ages ago with presumably simple-ish algorithms - probably they just paid more attention to the details. Same with TC and Eventide. When you look inside an Eventide - there are countless numbers of modules you can fiddle about with, but each actual module is reasonably simple. I can't see that those company's algorithms are too radically different from some used within s/w plugins - at least not in their conceptual basis. Just that the developers in s/w don't take the extra steps.

As a mitigating factor though...to be fair, companies like Lexicon have spent large amounts of cash into R&D and they have had the luxury of being able to devote whole aspects of their business towards one particular direction. Same with TC - only reverb, with the odd extra effect thrown in - their finalisers etc don't use the whole spectrum of FX technology. Whereas maybe it is not cost-efficient for developers in s/w to go that route. Even the Princeton plugin - it's only a code that was already used in h/w simply ported to VST coding. Whereas a coder doesn't have access to the 30-odd years experience some of those other companies have amassed.
Just playing the Devil's Advocate.

[kilroy]

The use of modulation in a verb tail is controversial, and frankly I hate the sound. It doesn't occur naturally, but apparantly was used by some hardware makers.

Lexicon started that. Nothing natural about the sound of Lex revs...but they do a far better job of providing some all important localization cues that kind of make you *think* you are hearing a real room, when it gets mixed in. VST plugs, and in particular convolution approaches, fall a wee bit flat in this regard. Samples in general...well, they have no life, you know...they just kind of..."are".

Although convolution is popular at the moment, it's a clumsy method and I think a new breed of excellent algorithmic reverbs could, and should, leave them for dead.

If you can't have a real room, the next best thing is full modelling. Sampling is an awkward stepping stone...static and cadaverous. You have this one moment in time, what are you going to do to make it a dynamic experience? No matter what you attempt, it's a bit like waving the arms round on a dead guy and saying, "Look! He's alive!"

The big thing missing in VST reverbs is excellent Inverse or Non-Linear tails. That requires full editability of the timing. I want a big slow reverse reverb tail that peaks right before the snare hit - that sort of thing. The relationship between reverb and groove should be explored more fully - so perhaps reverb parameters could be linked to host tempo in the way that delays are now?

Well, at least there is still hardware.

I have an old Ensoniq DP-4 that does superb non-lin stuff. This is actually an excellent box all round. If you troll e-bay you see them, from time to time. *Very* programmable...a veritable tweaker's dream. Kilroy recommends it, he does.

Numbers are important - the hardware GUI's, although crude, were often better than software GUI's. Nothing worse than fighting with a mouse to set 100, and all you can get is 99 or 101 ...

Nuther box I have is an old Korg DRV-3000. Ever seen one? High end studio unit. Very, very expensive when it came out...too expensive. But the interesting thing is that it is totally interfaced with using a remote. I can tell you, *that* is the coolest way to tweak your fx...standing across the room, pouring yourself a bit of the grape whilst adjusting the parameters on your vocal plate...it doesn't get any better than that, mates.

[greendoor]

A lot of the R & D was into the hardware - the analog audio components, A/D and D/A convertes, lcd gui's and power supplies and stuff. VST levels that playing field by removing the need for all that. We are just left with the ghost of the machine - the fun stuff. I'm sure that VST developers can deliver the goods, which is why i've posted this to stimulate some ideas. Just look what zioKiller has done with modeling electric pianos - truely inspirational. Scarbee samples are good, but require brute force PC power to run (like convolution of reverb). But with a goal and some determination, a SynthEdit creation has smoked them. I think it's time that VST reverbs smoked the hardware guys.

[birrbits]

I would have to agree that there is little in the way of competing with hardware reverbs on the whole. But I would add that this is not by any means limited to the 'high-end' market. I have both a TC M300 ($200) and Lexicon MPX100 (actually an MP100 on a Core2, but the MPX went for roughly $100). These are not high end but they certainly have no competition in thier price range within the VST world (at least that I have heard) . Add to the fact that my TC has digital I/O (therefore consumes 0 cpu) and even has midi in, the question kind of answers itself. It is its own dongle.