Behringer VD400 analog delay - trimpot adjust and it's awesome!

[bolero]

well thanks again, everybody

I just spent more time monkeying with it, the middle trimpot also affects the overall wet volume of the center mix knob: it can get quite distorted!! it's interesting because the trimpot fizzles out completely at each extreme. is that a byproduct of biasing? but that distortion isn't the brittle sound I'm trying to get rid of

it is easy to find a quiet setting for the 3rd trimpot, far right, which controls the whine

the first knob can mitigate it to some extent, but only around 200ms and lower I would guess. Even there I can still hear it but the delays are so short it's not quite as noticeable

maybe I'm expecting too much of this thing: even on shorter delays I can hear that brittle sound. maybe the circuit needs better filtering. I would be interested to know what opamps that fellow used earlier in the thread

[Fancy Lime]

I listened to your sound sample again and I think I understand now what you mean. If that "brittleness" does not go away or get worse when adjusting the BBD bias, it could be aliasing because you are running into the Nyquist limit of the fixed bandwidth. That would also explain why you only hear it on the first few repeats and it is getting less noticable with shorter delay times. Can you trace the low pass filters after the BBD (with help of the DM-2 schematic)? That would help in figuring out what the maximum possible delay time is before Nyquist aliasing starts.

I highly doubt that changing the opamps will do anything at all to the sound or to fix your problem. Modding the post BBD LPFs is much more promising imo.

Cheers,
Andy

[j_flanders]

Can you trace the low pass filters after the BBD (with help of the DM-2 schematic)? That would help in figuring out what the maximum possible delay time is before Nyquist aliasing starts.

I can't find a schematic of the Behringer vd400 but it's supposed to be a Boss DM2 or DM3 clone.
The Behringer VM1 analog delay also copies the filters from the DM3.
There was a thread over at the other forum a couple of years ago where I compared its filters to the ones in a Deluxe Memory Man:

I also compared the anti-aliasing and reconstruction filters:
I did this in the online Falstad web app. Since it only lets me to use opamps and no transistors in the frequency plot mode, I converted the VM1 filters, which use transistors, to opamps.

Anti-aliasing filter (pre delay):
https://i.postimg.cc/CFMt2QZ7/Screen-Shot001.jpg


Reconstruction filter (post delay):
https://i.postimg.cc/1yr2JQz4/Screen-Shot003.jpg


Depending on what frequency you choose as cutoff, based on the graph, you'd arrive at the following max delay time:

3kHz cutoff:
1/(2*3000) * 4096 /2 = 0.341 seconds max delay. (2*3000 = 6khz as slowest clock frequency = longest delay time)

4kHz cutoff:
1/(2*4000) * 4096 /2 = 0.256 seconds max delay. (2*4000 = 8khz as slowest clock frequency = longest delay time)

[Fancy Lime]

I couldn't find the schematic either. What I meant was, can you trace it yourself from the device in your hands? Not trivial and sometimes impossible with SMT, if the resistors and caps are not properly labeled. But let's assume that the VD400 filters are identical with the DM-2/3 and VM1. Anything over 250ms is going to be problematic. I also wonder if that bump at 2kHz may be a problem. It may be possible to (mostly) get rid of that by increasing one of the cap values ever so slightly by adding another cap parallel to it.  Have you tried simulating that? Not sure if it makes a world of difference but it may make it a little better.

Andy

[j_flanders]

I also wonder if that bump at 2kHz may be a problem. It may be possible to (mostly) get rid of that by increasing one of the cap values ever so slightly by adding another cap parallel to it.  Have you tried simulating that? Not sure if it makes a world of difference but it may make it a little better.

Andy

At the time I wondered if those little (resonance?) bumps were due to the fact that I converted the filters from transistors to opamps, where in the transistors I thought the gain would be slightly less than 1 compared to unity gain with opamps.
But when I later on actually audio probed the circuit, those little bumps were showing up in the frequency plots as well.

[bolero]

 well that is very interesting!! that peak in the freq response could well be what I am hearing?

here is a link to the DM2 schem, if it's any use:

https://ia801909.us.archive.org/26/items/boss_dm2-sm/boss_dm2-sm_text.pdf

I will explore the thing today & see if I can find room for a cap: in fact maybe a pot is even a better idea? a tone control for the wet echo only

am not an electronics whiz but I can solder

thx again!

[bolero]

 actually it doesn't look that bad at all, to work on: no SM components, big caps everywhere, even appears to be some unused sections on the pcb?

hopefully these image links work:

[j_flanders]

actually it doesn't look that bad at all, to work on: no SM components, big caps everywhere, even appears to be some unused sections on the pcb?

I don't see any resistors, nor any small value caps.
That means that probably over 50% of the circuit is on the other side of the pcb and most probably all smt.
Comparable to the 'underside' of the Behringer UV300 Vibrato pedal:

[bolero]

 haha I was just going to post that, no resistors I bet there's stuff underneath!

makes sense as that's where all the soldering is

[Mark Hammer]

actually it doesn't look that bad at all, to work on: no SM components, big caps everywhere, even appears to be some unused sections on the pcb?

On the part you're looking at, yes.  It takes more than those parts to produce an audio delay, however.  The other side of the PCB is where you'll find all the SMT parts.

(Flanders beat me to it.  Stupid Flanders!)

[j_flanders]

One thing we haven't really addressed is the very limited headroom of the bbd chips.
When I read your first post I just assumed it was bbd clipping. I didn't reply until I heard your mp3. The clock noise was so obvious that from that point on it was all we discussed.

Especially those Coolaudio V3205 chips are hard to get a good, clean sound out of. Even with a perfect bias.
Compared to other brands (Xvive, Belling, Panasonic) they seem to clip earlier and they also seem to have worse bandwidth, sounding darker/muddier.

Also, bbd clipping in general is very harsh and as you desbribe very apparent on the attack.
To fix your issue(s)
- Limit the maximum delay time for less clock noise
- Aim for perfect bias for less clock noise and less distortion
- Reduce the input signal level for less bbd clipping and the harsh distortion that comes with it.

Forget about the opamps. If there's clipping that sounds like opamp clipping, it's most likely coming from the bbd chip.

[bolero]

thanks!

one thing I might try is putting a pot on the echo, as a tone control. I am not too concerned about muddy delay, as my other delays I roll it off as well

at least the bottom of the board is well labelled:





higher res images here:

http://el34.coolpage.biz/images/vd400/003B1.jpg

http://el34.coolpage.biz/images/vd400/003B2.jpg

[bolero]

I wonder if Behringer put that EQ peak in there deliberately, to counter the dark character of BBG's?

hard to imagine it would be accidental

But didn't see a scale in the graph Flanders posted, so hard to tell how dramatic it is?

In any case, I will post my results here for posterity, if I figure anything out

[j_flanders]

But didn't see a scale in the graph Flanders posted, so hard to tell how dramatic it is?

Yes, I should have added that. It's 20dB I think.

Here are the links to the online simulations, so you can play around with it.
You can change a component's value by either double clicking it and entering a custom value.
Or by hovering your mouse pointer over a component and using the mouse scroll wheel to scroll through standardised values.
At the top right there's a slider for 'Response dB scale', slide it to the left for a better view.

Deluxe Memory Man pre bbd filter: https://tinyurl.com/6kff2sap

Deluxe Memory Man post bbd filter: https://tinyurl.com/tv4wu3k3

Behringer Vintage Time Machine (Boss DM2/3) pre bbd filter: https://tinyurl.com/yz9nv88v

Behringer Vintage Time Machine (Boss DM2/3) post bbd filter: https://tinyurl.com/37e4zhc6

Here's a version where the filter is not converted to using opamps but with the original transistors: https://tinyurl.com/54nt3pk5
But I don't understand how the frequency response plot in that version of Falstad works. 
The other one I do but it won't let me choose transistors under 'active components'

[Fancy Lime]

That peak is 20dB? That's a lot, isn't it?

Anyway, about that being on purpose or not: the thing is that it is almost impossible to design a third order LPF with standard cap and resistor values at a specific frequency without ripples in the passband. The nature of this beast is that it is very sensitive to even minor deviations from the ideal values for a predetermined frequency. If you would need a 268pF cap but only have 220 and 330pF, there will be a bump or a dimple somewhere. So unless you happen to have access to some extremely unusual component values, this sort of ripple is just unavoidable. Unless you combine caps in parallel to get close to the ideal value. There is s reason why third order Sallen-Key filters are relatively rare. The Japanese companies seem to be fond of them, though. I doubt Behringer put any thought in the filter design. The probably just copied the Boss design verbatim.

Andy

[Mark Hammer]

Do note that the flat 16-pin chip on the underside is a 570 compander, deliberately used to manage transients, in addition to keeping noise low.  In short, headroom is pretty much a non-issue.

In the pre-compander days, most analog delays would include an input level control, and an LED to indicate reaching clipping levels.  Once companders entered the picture, those LEDs pretty much disappeared.

[bolero]

Here are the links to the online simulations, so you can play around with it.
You can change a component's value by either double clicking it and entering a custom value.
Or by hovering your mouse pointer over a component and using the mouse scroll wheel to scroll through standardised values.
At the top right there's a slider for 'Response dB scale', slide it to the left for a better view.

Deluxe Memory Man pre bbd filter: https://tinyurl.com/6kff2sap

Deluxe Memory Man post bbd filter: https://tinyurl.com/tv4wu3k3

Behringer Vintage Time Machine (Boss DM2/3) pre bbd filter: https://tinyurl.com/yz9nv88v

Behringer Vintage Time Machine (Boss DM2/3) post bbd filter: https://tinyurl.com/37e4zhc6

Here's a version where the filter is not converted to using opamps but with the original transistors: https://tinyurl.com/54nt3pk5
But I don't understand how the frequency response plot in that version of Falstad works. 
The other one I do but it won't let me choose transistors under 'active components'

very cool, thanks for that!

[j_flanders]

That peak is 20dB? That's a lot, isn't it?

dB/div is 20dB
peak is still around 7dB though

[bolero]

 
yes if that's 7dB, it is pretty obnoxious. I wonder if that's what I am hearing?

when I was adjusting the bias pot I could get severe distortion from the BBD, but it seemed different & distinct from the edgy sound I was looking for, very muted in comparison

thx again, everybody, it is very interesting

[PRR]

.... it is almost impossible to design a third order LPF with standard cap and resistor values at a specific frequency without ripples in the passband.... access to some extremely unusual component values, this sort of ripple is just unavoidable....

It's not value sensitivity. Even at 2nd order, you either have a bump or a very slow roll-off. Even if your parts are ideal. At 3rd order you can get another bend and we start to say "ripple". The marginal case of no-ripple we call Butterworth after the guy who solved the math. Steeper filters "must" have ripple and we thank Chebyshev for much of this math. A specification for a Chebyshev usually includes a limit on ripple. The ripple may be swopped from pass-band to stop-band in ways I do not understand. (Usually less interesting for audio.)

A softer filter is Bessel.

Small ripple may be perfectly inaudible. I used a cookbook recipe for about 0.3dB ripple which is enuff steeper than a Butterworth to be interesting.

All these have sensitivity functions, to tell you how whack they get with part tolerances. When you try to fit 10 poles in a half octave, 1% is not good enough. Some "simple" types are impractical because of high sensitivity to tolerance.