ERb301 Verb Lite (version 100 78% cpu)

[adrianf]

erb301 verb lite 002.unit
erb301 verb lite 004.unit
erb301 verb lite 005.unit
erb301 verb lite 100.unit (127.7 KB)

Hello. I have tried to recreate the Erbe Verb from this: https://quod.lib.umich.edu/cgi/p/pod/dod-idx/building-the-erbe-verb-extending-the-feedback-delay-network.pdf?c=icmc;idno=bbp2372.2015.054.

I could not follow along completely without exceeding CPU limits, so I made the following compromises:

• Allpasses series not included
• random modulation when depth to the right is not included
  – i use s&h noise automatically to adjust time of allpass

I don’t actually have an Erbe Verb (I wish I did). So I can’t tell if this even comes close. Can someone compare?

This unit requires the all pass filter unit.

N.B. This unit includes a Feedback Delay Network that you can use in your own experiments. I paid attention to phase and level so that it is stable. Just copy it and remove the embedded delays from the macro unit.

Done in 100:

• using 1 time variant allpass instead of series
• connected Feedback Delay Network correctly this time with 4 doppler delays

Done in 005:

• changed all pass chain to parallel bands
• fixed predelay to 100% wet

Done in 004:

• rerouted LPF and Allpass filters before FBN
  – Added back series of 6 x ms prime all pass filters
• fix pre-delay
• use clocked Doppler instead of just clocked
• fix depth to be variable from sine mod left to “shimmer?” right
  – Changed right most depth to be filtered resonant noise
• implement reverse
• implement Env follower modulation?
  – Env modulation is accessible at the top level controls. Can be then reassigned to whatever top level parameter.

[Starthief]

I’ve got an Erbe-Verb, I’ll give it a try this evening if I get a chance

[adrianf]

Thanks! I tuned the top level controls to what sounded good to my ears. Things like infinite sustain work but use different values than the white paper.

Things that might be different: Depth and delay times were a guess for me. I also put inline limiters as a safety (to replace saturator in paper).

[Starthief]

Comparing this unit with Erbe-Verb, feeding it short pings from QPAS: I can tell there’s some similarity in the topology, but the sound is really pretty different except in relatively narrow ranges.

Erbe-Verb’s decay time goes far longer and more feedback-ier, and that also makes it softer and lusher and more reverb-like. Erbe-Verb never really sounds like a delay in a feedback loop, even at maximum size – there are a few echoes but they’re kind of scattered in a complex way rather than repeating regularly, and still somwhat diffuse even with Absorb at minimum. It can sound quite metallic at minimum sizes. But still, any “sustain” it has is very diffuse and reverby, if that makes sense.

Your unit tends to sound like a delay with a bit some neat scattering to it, especially as you lower the clock rate/multiplier to slower rates. It gets more metallic as you speed it up, but the diffusion just isn’t there to make it sound reverb-like.

Also the envelope follower on the Erbe-Verb isn’t patched into anything, it just goes to the “CV Out” jack; I often forget the feature is there. It’s not really an integral part of the sound as much as the modulation is, especially the “ergotic” random modulation with shimmer. (That paper doesn’t mention the shimmer for some reason; I don’t know if it’s a side effect of how the modulation is implemented or a little extra special sauce.)

Audio example:

I recorded this on Erbe-Verb with quasi-Karplus comb filter pings from Rainmaker. Absorb is at minimum, decay at about 2 o’clock (any higher and it will start to sustain infinitely… actually even at this setting there’s a little bit of sustain “leaking”), tilt is neutral, depth is zero.

I’m starting at maximum Size and working my way down to minimum. (I’m not touching the Decay knob, which seems to be feedback.) The odd pitch shifts on the tails of the first two pings were because I didn’t wait long enough for the decay to fade out before turning the knob

Even the longest size is a scattered, complex pattern rather than a steady echo, and the complexity and “reverbness” increases as feedback further scatters the already-scattered signal.

I think it’d be an extreme challenge to really duplicate the Erbe-Verb, but maybe more reverb-like response is possible within the limitations of the ER-301…

(BTW, this is the closest I’ve ever listened to the Erbe-Verb, especially with short pings that reveal a bit more about its character… so thanks for indirectly encouraging me to get to know it better )

[adrianf]

Thanks for your in-depth trial. Based on your sound example I agree on the diffusion, I don’t think i’ve nailed it down exactly. There’s definitely some adjustments to the all pass section to make the reverb sound like reverb instead of a series of delays. Also, the size of the delay, based on your description, is maybe too long compared to the source.

I’m not sure what a 4 delay version of this would sound like but i’m sure a more lush.

I’m glad i swapped out the clock delay with the doppler because it seems like the doppler shift effect is definitely a character of the module.

The envelope follower I attached to depth for demonstration purposes.

Edit1:

Reading this: https://valhalladsp.com/2011/01/21/reverbs-diffusion-allpass-delays-and-metallic-artifacts/, it sounds like a series of all passes was the correct way of applying it. I’ll make some adjustments and see what we get.

[Zeta]

I would suggest watching (and buying) this: https://youtu.be/o9eMmpuhz1M and all the following ones!
I learned so much about how reverb works!

[adrianf]

Wow thanks. The video you linked was very informative and I like how many demos of configurations he provided. The sound he gets with a couple of combs and all passes in series already sounded good.

[a773]

Great video, only watched the beginning and already got an idea for (and implemented) a reverb on the 301!

[Zeta]

And there are multiple videos!

[a773]

Yeah, I noticed, I’m gonna go through them all!

Thanks again for drawing my attention to these😏

[ringhof]

all pass filter unit

Could someone direct me to this unit? As discussed in here: Should custom units use additional libraries and bespoke units like Accents? At the moment i am not be able to stay up to date on all the units, so i am really would need proper links to all necessary files/units for custom units - especially when they sound so promising like this one

[mopoco]

you’ll find the latest version of @Joe’s Accents here, as far as i know those are the only
middle layer requirements for user contributed presets right now.

SuperNiCd/Accents

[ringhof]

you’ll find the latest version of @Joe’s Accents here, as far as i know those are the only
middle layer requirements for user contributed presets right now.

@mopoco thank you, and good to know!

[mopoco]

as for the allpass filter i just realized:

Middle Layer SDK (aka patching with Lua)

So feedback connections should be good to go in v0.4.02. I’ve taken your Schroeder Allpass unit and used it for it testing (also edited it slightly for correctness and took some liberties). I’ve kept it in the Experimental category of the builtins (where I had it for easy testing) if you want to take a look.

this means that Brian had already incorporated Joe’s allpass filter into the er301 firmware.
but i am not sure at which version. 0.4.11 definitely has it on board.
which firmware are you riding @ringhof?

[ringhof]

I finished my work on my latest album, as well as a live performance
for the release. So for now i’m in an updating process of my whole
system which includes new units.
[edit]
So i would update to 0.4.11 - if this is a good decision.

[adrianf]

Hello,

Still trying to make progress on getting density up while managing CPU.

I made a boo boo connecting up the Feedback Delay Network. It works now. I have carefully adjusted levels and fixed unintentional delays caused by feedback patching to stabilize the network. This among other things does take up 78% of CPU.

Also, because the LPF is resonant, I had to switch to EQ3 and use the lower eq band as a pseudo shelf.

If anyone has optimization ideas or a better more efficient way to create a matrix mixer or feedback mixer, I’m all ears (@odevices : perhaps a more optimal N-size matrix mixer with feedback unit could be in a future update?)

I have had to remove all but 1 of the allpasses but have used s&h noise to act as kind of a diffuser - it’s an interesting effect.

I’ll post the new unit in the OP.

[anon25407821]

thankyou for sharing your work. I cant wait to check this out. …keepin’ the dream alive…