Homemade Reverb

[WGTP]

I guess the trick with a reverb is to make it sound like the most wonderful fantasy room in the cosmos, complete with hot babes, cool drinks, black full moon nights and the white light of paradise.  :shock:

[Nasse]

Thanx for lifting up this thread. I have been thinkin the possibilities of using more than one those affordable but nice spec chips. I think these have huge diy potential.

[shawn]

Wow! Being that I am a programmer I would LOVE to sink my teeth into this reverb project. How long do you project befor eyou announce it? I'm dying to buy the parts already!

[DaveTV]

Wow. I can't believe it's been a year since this thread was started. I'm currently in the process of putting a reverb pedal together using the Alesis chip set that Peter was talking about. These chips aren't distributed by Alesis anymore, but are instead available through Wavefront Semiconductor.

The reverb circuit I'm building is pretty simple and is based on the designs shown in the data sheets for these chips. I'm pretty satisfied with the  built-in reverb algorithms as they are, so at the moment I don't plan on doing any additional programming, but it's nice to have the option. If it turns out okay, I may post what I've done if people are interested.

[SeanCostello]

Getting back to the original post: Why 6 PT2399 delays? You could probably get away with 4:

3 delays in series, with allpass feedback/feedforward around each delay (gain coefficient fixed at 0.7 - figure out the resistors)
1 delay following the 3 series allpasses
Output of delay goes to outputs, & into lowpass filter + gain control
output of gain control fed back into input

If you set the delay times right, this will give you a much nicer reverb sound than 6 parallel comb filters.

Of course, 5 allpasses followed by a delay, & feedback around that system, will give you an even smoother sound. However, the decay time of 5 allpasses will limit how "small" of a reverb sound you can get.

Sean Costello

[markr04]

If it turns out okay, I may post what I've done if people are interested.

I'm *very* interested.

[Peter Snowberg]

Wow... it sure has been a long time.

I haven't seen Jack around here for some time now and heard that he might be in "retired" status. I hope you're doing well Jack.

I would love to see a PT2399 reverb. I'm guessing from what he said in the first post that his design was based on serial 2399s without companding and without a complex reverb style architecture. I'm guessing it was probably more of a complex echo. A totally cool effect for sure and probably more within the reach of the average DIY builder than a full-tilt-boogie hybrid reverb.

Sean, you bring up a totally cool design idea there. It sounds like 8-10 dual opamps and 4 PT2399s would do the trick. That's a board for a 1590BB for sure. I hope a design like you suggest also sees the light of day. You can never have too many options.

Dave, I would LOVE to see your project. Having built essentially the same setup I was quite pleased with the built-in effects too. It's too bad there's no way to adjust the LFO speed of the flanger or rotor effect, but the reverbs are fun and the variety is good. I also like the chorused reverbs. My big wish for an effect to add to the stock set is an echo/delay that uses all the RAM for one channel to get .68 seconds delay.

As far as my DIY-DSP project goes, I just don't have the time these days. I hope it's getting much closer but it's a ways off and being a 2 layer PC board with surface mounted components, it's going to take at least another month to collect enough spare cash to get a few boards made.

In the near term I hope to share a design for an echo with loooooooooooong sample memory and that's going to take some cash too. I think I need to find some paying gigs. :lol:

[Mark Hammer]

I was tinkering around (conceptually, at least) with how to achieve something that might make a passable reverb from BBDs, and it occurred to me that the key is to incorporate feedforward and feedback simultaneously.

Consider the venerable MN3011 chip, which was originally intended to achieve a more authentic multi-reflection sound from a BBD by using multiple non-harmonically related taps.  So, you feed in the dry signal at point A, and you get outputs at delay times B, C, D, E, F, and G.  If you sum all of those outputs together, since none of them is a multiple or divisor of any of the others, you should get something sort of like the sound where the acoustic signal bounces off many different surfaces and arrives at your ears at different time (rather than the yelling-into-a-10-gallon-drum sound of an analog delay).

Just about every design I've ever seen using one of these, however, simply sums up the outputs at a mixer stage in some fixed proportion (usually a little more amplitude for the initial reflections/delays), with no bandwidth adjustment/change by delay time (as would exist in the real world) and no means for changing the sense of space by varying emphasis of different reflections or regeneration.

The other thing is that the proportional gap between reflections/delays is fixed.

What I proposed was something that used several discrete BBDs in series, with a mixer stage preceding each one and a splitter stage following each one, so that you could, say, feed back some of chip 3 to the input of chip 2 (but not to chip 1, and feed forward some of chip/delay 1 to chip 3.  And so on.  The net effect is to have a very flexible array of delay/arrival times that are sums and multiple of the various delay times each chip provides.  Stick some simple-but-flexible lowpass filtering between the chips and you should be able to mimic a ridiculously large spectrum of spaces and ambiences.

As Brother Ton Barmentloo and others have pointed out, though, there is noise-risk entailed by having multple BBDs on the same board wanting different clock frequencies (though I *think* that risk is decreased if they are all driven by a master clock and each receive simply a divided-down version of the same master HF clock, but I may be wrong).

In any event, Jack (and Hi, by the way), if you can figure out a way to use series and/or parallel implementation of PT2399's, and make use of feedforward/back capacity, that may let you get away with a decent simulation using fewer delay chips.  I only worked out in my head what I'd do with a bunch of MN3007s/3207s because I have some, but 2399's could work also...albeit in a slightly bigger board.

[puretube]

...concerning heterodyning clocks:
it`s not impossible to use diff. clocks in such projects for advanced builders;
-- but one should be aware, that the risk increases, when
[A]-clocked output gets fed back into -clocked input -- ;

the "time-chopping" of a BBD occurs at both in- and output!  :wink:

[Mark Hammer]

But that would be true of both analog and digital shift registers, no?

Is the heterodyning thing MUCH more of a problem when delays are in series as opposed to parallel, or is it fundamentally a problem of mixing such multiple signals ANYWHERE on the same board?

[puretube]

http://www.pat2pdf.org/patents/pat2105318.pdf


haha: btw:

Warning: this topic has not been posted in for at least 120 days.
Unless you're sure you want to reply, please consider starting a new topic.

[amz-fx]

Reports of my retirement are greatly exaggerated...  (to paraphrase Mark Twain)   

Why 6 delay lines, you ask?  A proper reverb emulation not only must have the proper delay times and feedback ratios, but also it must have enough non-harmonically related impulses to build a density of reflections to make the sound convincing and thick. 

You want the sound to be dense with discrete echos at a minimum...  the more reflections (delay lines) the better up to a point.

Sure you can make one with 4 delay chips, but as you cut back on the number of echo paths the realism of the reverb will suffer. 

Series connection of the PT2399 is probably not a good idea because so much quality is lost in a pass through the chip.  Clock interference won't be a problem with proper grounding and low impedance power supply lines.

The MN3011 can be imitated with a PIC and a few support chips...

Other projects have made me push this idea on the back burner. 

regards, Jack

http://www.diystompboxes.com/smfforum/index.php?topic=38248.0

[toneman]

BBDs vs DSPs

and......the chips are *still* undecided..

Does anyone remember the RatShack "Ambience Synthesizer"??

3 SAD1024s

I started building one ages ago.
Still have the 95% completed pcb.
I Have the instructions somewhere(?)
Does anyone else remember this, or have it scanned??
Mark?   PT?  Pete?  Jack?
Another note, the Rockman X100 used  a "tapped delay" BBD for it's reverb.
Definitely gave the x100 the "hall effect"  a-la "Boston".

no boinging----It's ......springless
stayverbbbbbbbedd
tone

[Mark Hammer]

Although Fender spring reverb is not the only benchmark for what reverberation "is", for a great many players it is what reverb *should* sound like.  The classic Fender stand-alone reverb, and a number of amps built since that time which used the Fender unit as their template, has a "dwell" control.  The dwell control sets how hard the springs are driven.  I suppose the name for the control comes out of the fact that if you kick the springs harder, they wiggle longer - i.e., the reverb "dwells" longer.

Beyond being merely a sort of level-matching control to adapt different input signals to the needs of the springs and circuit (it WAS a stand-alone unit, after all), the Dwell control also permits one to extract different "behaviour" out of springs, simply because - as mechanical devices - they behave differently depending on signal level fed to them.  It is that dynamic "character control" that is often missing from a great many completely electronic attempts to mimic reverb. 

Again, I'm not saying that is a NECESSARY feature of a well-designed reverb, because certainly the ambient qualities of a room don't change when you scream louder.  However, as much as the quirks of springs have become embedded in many musicians' conception of what reverb "is" (and that depends on what the meaning of "is" is ), building in dynamic character control would be a desirable feature.

[R.G.]

... but an Accutronics spring unit is only about $18...   They come in the normal long lines and a short 8" line with three springs as well.

It's not like germanium - they still actively make springs. 

[puretube]

of course, one could start to select them...

[Peter Snowberg]

It's nice to see this post again. It reminds me that my "fun" projects are at a two year backlog now.

Let's take a step into the question of what reverb really is since we're deconstructing in order to build the perfect beast.

DELAY: This involves just delaying the signal by X time period.
ECHO: This is the same information being fed back into the delay, with attenuation of the repeated information, and often with HP filtering in the feedback loop.
REVERB: a VERY complex phenomenon where sound takes an essentially infinite number of different paths from the source to the destination, being modulated by every surface encountered along the way.
SPRING REVERB: A device that attempts to mimic the phenomenon of reverb by bouncing sound down a spring. A very poor emulation of real room reverb, but a wonderful thing when considered as its own effect rather than an attempt to copy something else.

We're talking about using multiple echos here along with filtering to emulate reverb, which is another category to add to the list above.

This stuff gets so complex so quickly that programming is the only way to explore a large number of options in less than a lifetime. That's one of the reasons that DSP becomes so nice in a place like this. Now with that jump to DSP you gain so much capability that I think there is a whole range of mid-complexity level algorithms that are skipped over and not as available as sonic tools, either in the digital or the analog world. Jack is talking about such a device..... too simple for complex DSP systems, yet too complex for analog systems that can be sold at a profit.

It would be wonderful to hear the differences in the topologies discussed so far in a side by side comparison.

The first rev of my DSP thingy was a washout because of feature creep and RFI issues with too much stuff in too small a space. Rev 2 simplifies things a little, switches to a different (and more expensive) Atmel controller, and most importantly it reduces all the digital logic to 3.3V. The new layout also allows for copper tape shielding to be easily applied to the digital area.

I would love to see a layout for up to six delay chips and some modular filtering blocks. There is a LOT of territory to explore out there in reverbland. The only worry I would have would be heterodyning from six different clocks.

[puretube]

he, who defeats* two clocks, doesn`t fear four more...


(*= sorry - don`t know the accurate english expression...)

[Joe Kramer]

Hi!

FWIW, I think the obvious way to go for a DIY stompbox reverb is the MN3011.  About the same amount of circuitry and board space as a BBD delay, single-supply powerable, and the chips are available at Small Bear.  I've been tinkering with the guts of a broken Rockman X100 for a while, meaning to breadboard a circuit based on its reverb.  It's really a pretty nice sound--sort of a hybrid of spring verb and analog delay.  The concept is solid, the chip is available, and the glory is up for grabs to the energetic puppy who can tie up  the loose ends and realize a decent circuit. 

Joe

[StephenGiles]

That's good enough Ton, we know what you mean!
Stephen