information on digital reverb structures

[froombosch]

Hello All,

This will be a post that will be edited all the time and it is far from ready. I add more information soon.

Topic of this post will be the different structures of digital reverb simulations. If you want me to add something or have comments please reply to this post and I'll add the information. There is a lot of information on this topic. It is interesting to understand how things are built and why some decisions where made with the different types of digital reverbs. I only know my bit, being only a person interested in digital effects, trying to make all the scientific information more approachable.

Harrie

Before digital Reverb

Adding a reverb to a dry sound has been very polular and there are different ways to add reverb. Reverb is used to make a something sound like in a space and to add a nice colour or more power to a sound. The obvious way to make a reverb is to build a room and place speakers and microphones. In a lot of 50-60 studios these rooms are build and with use reverb dampening devices  the colour or the needed reverb time could be reached. Other ways of making a reverb is by using a plate reverb like the EMT-140 (1957)/240 (1970) where a metal plate is exited and the movement (=sound) of the plate recorded. These kind of reverb has a very dense sound and needs a predelay and added early reflections. Tapedelays or the Binson Echo machine was used for predelay or early reflections.

Digital reverb

In the 60-ties Moorer and Schroeder worked on different structures build in digital domain to recreate reverbs. In some papers the classical reverb structures where setup. These structures involve series and parallel all-pass and combfilters. When you are trying to test these structures they mostly are not really usable for audio applications because they are very metallic sounding. Or as quoted in the SP2016 website: We experimented with a couple of different 'patches' creating several variations on the Schroeder theme but with no real improvement. But the basic can be used and where used with some of the early digital reverbs.

The first commercial available digital reverb: the EMT 250
Barry Blesser and Karl-Otto B?der designed the algorithms of this machine. The resources in computertechnology were very basic as there where only
16K memory chips available. The algorythm structure is based on the next series: predelay –> three diffusers build around allpass filters –> reverb tank build around three series of 4 delays (tanktime between 80-120 ms). The output of the third delaytap is feedbacked to the input and in this feedback loop there are high and low filters added. The outputs of the delays are put into a schroeder decorrelator and has four outputs. Dynatron has build a remake and there are several replicas in different machines. The patent gives more information on the delay times. The amplitude is exponetially decaying. I'll add more information later. A rebuild can be made in Vsig. Jusst by using these building blocks it is possible to make it sound like the EMT. Calculating the gains for every delay tap and the filters in the feedback is the biggest work. I am happy with spreadsheets…. 

The second one
Quad Eight CPR-16
There is no information on this reverberator, beside that it sounds pretty crude.

the third:
Ursa Mayor Space station STS-282
Chris Moore explained a lot on this structure on his website. The structure is based around a 24-tap delay. fifteen of these delaystaps are modulated (with 62 microseconds steps) and 8 are not. These 8 are probably summed and form the output. The space station is know for a gritty but usefull reverb. There is a remake build. How deep the mudulation is and the waveform of the modulation is not clear to me. Need to do more research on it.

Later Ursa Mayor was sold to AKG and they build the AKG ADR 68K

Lexicon 224 (1978) The Lexicon 224 was the design of Dr. Griesinger. I have not found a lot of information on this structure. The d'Attoro paper gives more insight on the basic structure of the lexicon structure.  http://homepage.hik.se/staff/tkama/audio_procME/papers/matlab_reverb.pdf

After these first three units others where jumping onto the reverbs:

Dynacord vsr-23, Sony DRE-2000, MXR etcetc

Eventide SP2016 (1982) The Princeton site gives not a lot of information, besides that it is not a variation on the Schoeder theme.

Quantec QRS (march 82) This German compagny does not give a lot of information on the structures.

The fun thing about these early reverberators is that some of these sound so good. The EMT 250 is still in use in a lot of studios and some swear by it. A lot of remakes are build, some in the form of hardware machines, some in the form of software.

[rmaxwell]

Hi Harrie,

Thanks for the well researched thread on the history of digitally synthesized reverb. It is nice to see the information compiled with a timeline and the theory filled in.

I look forward to further posts to fill in the gaps. Nice work!

Ray

[froombosch]

Thank You! As soon as my busy schedule allows it, i'll fill the gaps and add more detailed information.

 Harrie

[froombosch]

Some interesting information on the different reverb structures

http://ccrma.stanford.edu/~jos/Reverb/Reverb.pdf

Harrie

[eventideathon]

Hey Thanks Harrie!

I am looking forward to reading the stanford paper…. Looks very interesting!
I have another DSP research paper on reverb around here somewhere 😛 I will see if I can find it for you 🙂

-Colin

[froombosch]

Thanks Colin!

 There is a lot of information on this topic. The biggest problem is getting information on structures used in the different machines. Every manufactor has spend a lot of time and effort develloping their algos. So they do not give away a lot of information. But it is interesting to know a bit of information on the topic.

My guess with the different reverb blocks inside the Eventide machines is that these are build around FDN structures.

http://ccrma.stanford.edu/~jos/Reverb/FDN_Late_Reverberation.html 

This because the manual says they are build around like 32 delaylines. And there are some tricks inside them that make them sound typical Eventide….. 

 Harrie

[wschira]

Hallo Harrie,

it’s a lot of time since your post, and I’ m not sure you’r already in the community. If so, I wonder what are the results of your effort? I’ d like to hear more about it.

Best Regards

Wolfgang

 

[froombosch]

Yes, I made a resoanably working patch for it and more. Maybe better to send me a PM.

[froombosch]

froombosch wrote:

Yes, I made a resoanably working patch for it and more. Maybe better to send me a PM.

I see it is not possible to send PM’s. My name is Harrie Munnik and I can be found on a lot of places on the internet.

[Author]

Posts