MTurboReverb preset making action, let's make the ultimate reverb! ;)

[dmbaer]

Not exactly, it extracts the initial early reflections from it.

So the sample screen of the ER page ... the label says "ValRoom". Being rather liberal with the source of early reflection IRs, are we?

[MeldaProduction]

Not exactly, it extracts the initial early reflections from it.

So the sample screen of the ER page ... the label says "ValRoom". Being rather liberal with the source of early reflection IRs, are we?

Pst

[elassi]

soundcloud...

Honestly, this reminds me to the things I (no worries, it's only me!) don't like when it comes to granular and re-synthesis: that certain timbre... hmm, personal taste.

[plexuss]

Hehe yeah, but I'm looking forward to what YOU can do with it, ideally something that one would actually like to listen to

I will guarantee you will want to listen to it at least 3 times over the next week.

[caulixtla]

I know there’s a lot of people interested in joining the beta program, so I have sent yet another email. My take is this: I am wondering how close we can get to the reverb in Jon Dattorro’s seminal paper on reverbs, “Effect Design Part one,” which was the first time the legendary Lexicon reverb topology was published. I know the algorithm there has been implemented multiple times in free VSTs already (the best implementation is probably the one in Freeverb3), but I would like to see how close I can get in MTurboReverb.

The design in the paper starts with a predealy delay line, followed by a low-pass filter (delay line one sample long), followed by four all pass filters to decorrelate the input. This is followed by the signal entering the reverb tank at two different places; the reverb tank has four allpass filters (two with modulating delay coefficients), four delay lines, and two low pass filters (again, one sample long delay lines). The signal goes through the tank in a feedback loop, where the reverb decay attenuates the signal at four different points in the tank.

The output is created by combining the input at various point in the tank in a fairly complicated way, and the closest way I see of getting that behavior in MTurboReverb is with the SerialTap module, using a modulated delay (since we can’t modulate the allpass filters the way the Dattorro-diagramed reverb does, I would instead use a V/Vibrato module for the delay lines in the tank)

I don’t see a way to have the input in MTurboReverb enter the tank at two points, so we’ll just enter the tank at one point. We will somewhat make up for this by having more of the delay nodes modulating.

It might look something like this:

fl;4a;st[a;v;fl;a;v;a;v;fl;a;v]

Anyway, email sent.

[VariKusBrainZ]

Im enjoying the 'extreme shit' reverbs, great for sound design

[elassi]

Im enjoying the 'extreme shit' reverbs, great for sound design

Yeah, like feeding meticulously recorded 96kHz samples into a VST mimicking 8bit samplers.

No offense, Vari, just kidding.

[plexuss]

The only effects are "10c" and worstreverbever and a bit of delay.

https://soundcloud.com/musicofplexus/vojtechs-nightmare

[Roman Empire]

Can you unlock Phil Spector with this???
No VST reverbs I´m aware of can do this convincingly.

Best Regards

Roman Empire

[Chandlerhimself]

Caulixtla - I believe it that algorhythm is already included as Plate 1 or 2. It isn't exactly the same, but similar. Vojtech said he didn't like it, but I think it sounds alright if you filter it correctly and add modulation. I think it is the basis of the lexicon plate.

Use CC to simulate the figure 8 loop and use the CROSS command to split the frequencies and then lower the level of the top band using L to simulate a low pass filter.

I tried making a few days ago without the CROSS command and I was able to get pretty close to the plate in Valhalla vintage verb. As soon as I finish this first preset I'm working on(I'm tired to hearing static noise every 30 sec) I'm going to try to simulate something like the lexicon random hall.

[Chandlerhimself]

Ok so:

#1 - FDNs are very problematic. Basically the main problem is that every delay line is fed back to every other delay one, which increases the chance of metallic ringing. I did lots of tests with FDNs and eventually realized, that they are just not worth the trouble. They are great for theoretical research, as they can generalize various topologies in a single matrix, but that's pretty much all. I'd suggest using the R and RP modules, which basically implement a cyclic multi-FDN approach, which is kind of a hybrid of classic topologies and FDNs with the aim to minimize chance of ringing and improve CPU efficiency at the same time. It's kind of a new thing, read a paper about it from 2015 or 2016 (not sure which ).
Btw. FDNs can generalize other topologies, which works both ways - if you really like some specific FDN, it's quite possible there would be a topology that would sound similar. But still, I'd avoid them.

I'm surprised you said this. I'm working on a preset right now using FDNs. Of course they can sound metallic and ring, but with correct tuning I think they sound pretty good.

I started with 1 FDN4 and I tweaked the input and delay settings until I heard as little ringing as possible. Then I put 4-20( using complexity) in parallel. IMO it sounds pretty smooth, it has a natural decay, scales well with different lengths and has a nice evolving tail at long settings even without modulation. Of course I added a few more tweaks, but that is the basic idea. It is basically the Schronger idea on steroids. I'm not hearing any metal, but I'll test it on a few more sources to make sure nothing weird is going on. I'll send the preset over today or tomorrow.

[levendis]

Your instructions urge us to maintain the same set of Multiparameters on the Active Preset's GUI, for the sake of preset compatibility. The only reasons that i can think of for why this is so are:

1. Consistent visual scheme for Active Presets/Factory Presets.
2. Smooth operation of the ABCD Morpher.

The first is reasonable, the second, however, doesn't seem to apply. One reason is that the population of "Room Types" on each of the current Active Presets differs in number. Another reason is that the morpher can progress via intermediate Active Presets. eg. (With the configuration you've supplied) If "Plate" is stored in preset selector slot A, and "Room" is stored in slot B, then a morph from A to B will instantiate "Cathedral" and "Hall".

Are you able to offer some leeway on this stipulation, so that the focus of the GUI be pertinent to the preset that we build?
An example of a drastic edit, which i have in mind, is to include an XY pad, linked to the EQ.

My apologies if this query comes across brusquely. If so, it's not my intent. My apologies too, if i have misrepresented the intentions behind the panel configuration.

[bmanic]

#1 - FDNs are very problematic. Basically the main problem is that every delay line is fed back to every other delay one, which increases the chance of metallic ringing. I did lots of tests with FDNs and eventually realized, that they are just not worth the trouble. They are great for theoretical research, as they can generalize various topologies in a single matrix, but that's pretty much all. I'd suggest using the R and RP modules, which basically implement a cyclic multi-FDN approach, which is kind of a hybrid of classic topologies and FDNs with the aim to minimize chance of ringing and improve CPU efficiency at the same time. It's kind of a new thing, read a paper about it from 2015 or 2016 (not sure which ).
Btw. FDNs can generalize other topologies, which works both ways - if you really like some specific FDN, it's quite possible there would be a topology that would sound similar. But still, I'd avoid them.

This is more of an interface problem and lack of "skill". Why do I say that? Because Ariesverb is a purely FDN reverb and it's pretty easy to get a very good reverb out of that one. Even with semi-random tweaking, you don't constantly run into bad reverbs unless you do stupid things and don't keep an eye on the FDN overview page.

So yeah, I'm pretty sure good FDN things can be achieved if the architecture and user interface around it is good.

Also, the free U-he Protoverb is a complex FDN verb and it's very easy to get extremely good reverbs from it.

[MeldaProduction]

The only effects are "10c" and worstreverbever and a bit of delay.

https://soundcloud.com/musicofplexus/vojtechs-nightmare

Haha, nice! It's like a music for a very depressing movie . Like Requiem for a dream, I remember when watching it it was really good, but at the end I just wanted it to stop . Good job! And you masked the horrible panning I made for you with love well .

caulixtla: I noticed your email, let's solve it privately, I think most people would be confused from it anyway . Personally I'm not a fan of these rather ancient constructs, they usually really like ringing . At that times there wasn't enough CPU power for things like R or CC, so they needed to settle with things like that, but I personally think some things should be left in past. Though, you might somewhat improve the concept and maybe something utterly beautiful will come from it!

Chandler: Very good!! Looking forward!
As for FNDs, I wasn't talking about the FDN4 and R modules here. R is specifically made to lower the metallic ringing. But classic FDNs can be any size, not just 4x4, and the results are imho not good. People like them, because they are easy to program, but then you have the big problem - designing the matrix itself, and that's when it falls apart and when it doesn't, one realizes, that with dedicated modules he would be far better off.

levendis: It's rather thinking ahead. The plugin may be instantied inside some other plugin such as MXXX or MTurboMix. And you may build active presets in them too! Now don't get your head spinning - basically the active presets in MXXX or MTurboMix may want to access the active presets inside MTurboReverb inside it including preset selection. And so if every preset would have different parameters, it would just switch the presets, but wouldn't be able to access any of the parameters, because you wouldn't know which is which. But if they stay the same, you would be able to use them without problems.
That said, it's not such a big deal, but at least the main section should stay intact. If you need adding new parameters, use some of the unused numbers. If you need to remove some, do that, no problem, they won't do anything then.

bmanic: Again FNDs, as above . FNDs are not bad themselves, but from my experience whenever they sound good, they only become a generalization of simpler blocks, hence wasting CPU power while being harder to design. But I like the smaller block as they offer a simple and efficient way to build up density. An allpass diffuser inside feedback chain works even better though imho.

[caulixtla]

Well, fellow preset designers, my entry in the competition to make a MTurboReverb preset is one geared towards percussive sounds, but can also thicken up pads and other sounds. I’m not using any early reflections; here is the late reverb algorithm:

a(0.7;11);a(0.7;23);a(0.7;9);a(0.7;19);p[#an[fl;a;fh;a;v;a]]

For reverb designers who aren’t familiar with MTurboReverb’s peculiar syntax, it’s as follows (note that the decay times appear to be affected by the “size” parameter in the reverb designer screen):

• allpass filter, 0.7 feedback/diffusion, 11ms delay time
• Followed by, in series, allpass filter, 0.7 feedback/diffusion, 23ms decay time
• Followed by, in series, allpass filter, 0.7 feedback/diffusion, 9ms decay time
• Followed by, in series, allpass filter, 0.7 feedback/diffusion, 19ms decay time
• Now we split the signal in to N parallel sub-tanks, where N is the “complexity” parameter (# in MTurboReverb’s syntax).
• Each sub-tank is a normalized allpass feedback tank. Each tank has the following in series: a low shelf filter, followed by an allpass with random parameters, followed by a high shelf filter, another random allpass, a random modulated delay line, and a final random allpass.

By having in the front only four allpass filters, and hand tuning the diffusion and length of these first four critical allpass filters (I simply made the decays relatively prime to each other and gave them about as high of a diffusion as I could before the allpasses started making an unpleasant ring), we can get a nice diffuse field quickly (especially when size and decay are small) which, while a bit colored, builds up quickly enough to be really nice for drums (as well as piano sounds or synth pads).

We brute force the decay portion of the reverb; because of the nature of the “an” (all-pass normalized feedback loop) module, the maximum decay is about 4 or 5 seconds. While I’m a big fan of huge reverberant washes, those are less challenging to make (when in doubt, just add some more allpass filters), and I want something a little more tricky for my preset.

Right now, the preset has only one knob: Decay (which affects both size and decay time). I will add some more knobs to the preset before sending it off to Melda for consideration.