BBD delay as simple as possible

[Scruffie]

One of my first (if not possibly my first?) PCB layouts was the EHX double tracking to fit a 1590B, I think someone did eventually verify it so it should be doable.

If you're going with 2 dual opamps you may as well do input pre-emphasis > sallen key > BBD > BBD > sallen key > de-emphasis, look at the small clone for an example... that'd be a more standard way to do things.

Edit:You'll have trouble with bias though, the BBD output is ~+/-1V (can't remember which, look at the datasheet) so the second BBD will need a trimmer or to tap off the V.Ref... 2 BBDs improperly biased isn't going to be great for headroom.

[Fender3D]

Buffer after 1st BBD is not needed, better fit a null trimmer on last BBD's output pins.
I'd rather put a buffer/mixer on output, instead of just 2 resistors.

You may set up a differential amplifier @input, +in as high impedance guitar input and -in as return for repeats...

[Scruffie]

Buffer after 1st BBD is not needed, better fit a null trimmer on last BBD's output pins.
I'd rather put a buffer/mixer on output, instead of just 2 resistors.

You may set up a differential amplifier @input, +in as high impedance guitar input and -in as return for repeats...

An inverting buffer between BBDs does offer some benefits... but probably not as much as pre/de-emphasis for noise.

[allesz]

Hey, the filter cap substitution was good: now I have 22n and 4,7n; it works very good, but always olny untill you get to longer time delay, at that point the faint wining is unstoppable. I say unstoppable because it bleeds through even if the echo line is disconnected; I don't knsow if it's a problem that comes from the breadboard or the layout... but it's ok to me: I never wished for something sounding this good (in a nerdy diyer sort of way, of course).

The idea of two 3208 came because, visually speaking, I would like to see five 8 pin bugs in line; of course the circuit will get a little bit more complicated... and I really feel I never did something so big and difficult.

If you look at the GGG ad3208 schematic, you see he did not use buffers, only a null trimmer, a cap, and a bias network. Madbean, on the dirtbag, use an inverting buffer, a bis network, and the null trimmer is only on the last bbd... now I don't find it, but I swear I saw a schematic where the two 3208 (or 3008, can't recall) only had a plain non op amp inverting buffer between them, no bias or capacitors.... but since i can't find it again I admit that my memory may be faulty.

About the null trimmer, I tried a 20k trim, a 50k one, then a couple of 4,7K 1% resistors... I felt no difference at all, so I went with the easier way; but I admit that not having a scope, or other instruments (apart a guitar, of course  ), I can't appreciate the effect of that pot.   

[PRR]

> null trimmer..., I can't appreciate the effect of that pot.

Think how the original "bucket brigade" worked. A row of men passing buckets of water from a lake to a fire. The man at the end threw a bucket of water on the fire, then waited for the next bucket. Water went on the fire only part of the time.

The very simplest BBD would work like that. On odd beats it passed signal, on even beats it moved another bit along. You had output only half the time.

That's not good, and in this case easily fixed. The last stage is duplicated with an extra beat delay on one output.

If CMOS were "perfect", the two outputs would be identical. But CMOS devices vary a lot. One output might be only 9/10 of the other. This is better than nothing but still leaves a glitch. In effect it is clock leakage.

The earliest commercial (and preproduction) BBDs let you mix the two outputs and trim for least clock leakage. I know that most production BBDs have very nearly equal outputs, and the trimmer will be very near midpoint. I think a few later designs omit a trimmer here (perhaps the CMOS process improved so much the two outputs are equal-enough in most cases).

[allesz]

A small update: I am still breadboarding around the damned bbd. It's fun in the end.

I learned something about Sallen Key filters (always a nice thing) and other things like, for example, that breadboard layout is very important on big projects 

Also I discovered that a big filter on the power rail (470uF, just like in marteen's schematic) works wonder limiting the clock whining bleed through at longer delay settings.

But keeping the project simple makes impossible to reach longer delay settings (and we are talking about 300 msec, when I say long  ).

I am refining the project right now, but I will post something just a little more complicated than marteen's schem that will have a repeat control and it will be quite silent up to around 150/200 msec. Probably nobody will whant to box a thing like that, but it's a funny project to try on the breadboard.

I also made some experiments with a Vb control but the simple two 10K resistot to ground and a smoothing cap really seem to work the same way.
 

[samhay]

Sounds like you have an analogue slapback delay, which plenty of people may find quite appealing.

[allesz]

I agree, but maybe digital delays got to a point of perfection, even emulating analog echoes  , that there is no point in the exercise.

And a well used pt2399 can give a wonderfull echo with not too much work.

Anyway here is the last "complicated" circuit:


It is verified on the breadboard.
Only watch for * and ** on the schematic: the pot in combination with the cap set the max time delay and, as draw, times are longer and the device quite noisy. Right now my favourite values are 50K pot (just stick a 100K resistor in parallel) and 330pF.
You can keep those values and, simply, don't use the noisy part of the pt's travel.

Actually I am testing an even simpler design that is more noisy; not bbd whining this time, but a little bit of hiss.
I decided I can live with some hiss.

[allesz]

Hera I post my two latest (and probably last) breadboard adventures reports on bbds.

The first is quite hissy, the second is a bit better in this regard  thanks to some advice from mictester on the other forum.





I added a bias control, quite useful, but I admit that the coolaudio 3205 and the bl3208 I have biased right even with the standard couple of 10K resistor bias network.
Then I added a 500K trimpot to "tune" the time pot and avoid whining at max delay time: set the 100k pot to max, then turn down the trimmer untill it stops whining: you should get around 170/180 msec of decent delay.

[samhay]

You have your 'repeats' control doing double duty as wet mix as well as controlling the number of repeats. I guess there is a sense of economy there.
The input impedance is quite low. I assume this is not designed to sit at the start of your signal chain?

[allesz]

  Well, the first thing to think about is: do I really need to put this "thing" in my beloved signal chain?
But I admit I did not try to put a buffer, or buffered pedal, before.

You are right samhay: if you turn the repeats to zero you have no repeats at all; this has another side effect, the level of the echo increases as you increase the number of repeats... to be perfect it should work the other way around  .

The echopatic sounds way better, but the max delay time is almost the same.

[allesz]

 

IMPORTANT UPDATE: I draw the tiime network with a mistake: the 330pF (or 470pF) time cap has to be between pins 5 and 7 of the 3102 chip, not in series with the 22K resistor.

Sorry.

[allesz]

https://soundcloud.com/allesz/echopathetic

[allesz]

Here is the latest super simple schematic, you can liste to it in the recording I posted yesterday, and a pefr layout.

[thomasha]

Hi,
I guess I could add some discussion here...

I tried the last schematic, and it indeed is a nice short delay. Problem is the high pitch noise at longer times (I tried it with a 1M potentiometer ...)

So after some tests I came up with a small twin T design, which cuts a lot of signal, and with it the noise.



I used this tool here to calculate the center frequency
http://sim.okawa-denshi.jp/en/TwinTCRtool.php

to match the frequency I was reading for max. time at 1M: 3kHz! If my math is right, that should be like.... 682ms (1/3k/2*4096)

Of course at that frequency the signal has just a slight resemblance of what it was before. I think what I like is how it fades away anyway, just like a beat that gets quieter.

I built the twin T with a potentiometer so that I could adjust the center frequency to barely match the high pitch.
What also helped was adding a balance at the output of the BBD (not in the schematic). Mine is slightly off center, where the high pitch is also slightly reduced.

The drawback of the twin T approach is that the echo got really quieter. Still something to improve.

My next step is add another TL072 to use as sallen-key filter. A previous test showed a good reduction of the noise until 6kHz without the twin T! That is time enough for a single V3205 IMHO. Let's see if it fits in a 1590a... 

Now something that I was questioning myself: is it really possible to get rid of the clock noise, or is working with higher clock frequencies the only solution, so that I would have to use more BBDs in series?

The compander won't fix the clock noise right? Or does the DM2 and similar designs use lower clock frequencies in the 5k range?

[Mark Hammer]

Where the heck is the bias for the BBD?

[allesz]

Hallo tomasha, I was lurking unlogged (as usual) and noticed your reply.

Are you still on the breadboard?

Since I did the last one, long time ago (I boxed a couple), I always wondered if elevating the power rail of the clock chip would allow more noise cancellation.
Never had the chance to try it out... .

[allesz]

I really don't understand why the BBD bias correctly. But this is the part I stole from another forum member (link on the first page).
If I remember right not all double ICs work.

[Scruffie]

Even a single pole filter on the input of the BBD will help greatly with the quality of the resulting output signal, just tack a cap off the 10k on the input.

The BBD output isn't low impedance, I bet it's not enjoying all that load on it, hence your loss of signal.

You can pull an absolute maximum of about 500mS (realistically 400mS and 300mS if you want decent quality) usable out of a 4096 stage BBD (I've done a second but it was not worth it) and it's absolutely possible to get rid of clock signal... just not with a design like that. If you aimed for 400mS absolute maximum you could get 2.5kHz bandwidth which isn't to be sniffed at.

Companders do help a hair with clock noise through S/N improvements but more with overall hiss. The DM-2 pulls 300mS, it's somewhere in the 7kHz clock frequency range and IIIRC the filters are about 3kHz so just a hair under nyquist.

I would look at 3rd order MFF filters rather than sallen-key if low parts is your aim and for the higher Q factor they can offer.

[Scruffie]

Where the heck is the bias for the BBD?

From the op amp output, same as the zombie chorus, crappy way of doing it but apparently the 3205 biases fairly close to mid supply if this works!

Hallo tomasha, I was lurking unlogged (as usual) and noticed your reply.

Are you still on the breadboard?

Since I did the last one, long time ago (I boxed a couple), I always wondered if elevating the power rail of the clock chip would allow more noise cancellation.
Never had the chance to try it out... .

The best way to do it IMO is to run the clock direct from after the diode with a ceramic or tantalum bypass cap directly on its power rail and then r/c filter for the audio.