Deluxe Memory Man Tone Control Build Report, was ATTN: Mark Hammer

[Primus]

Mark,

I love the sound of a delay pedal that rolls off more highs with each repeat. I read your suggestion for doing this w/ the Deluxe Memory Man:

"If you want to cut the bass increasingly with each repeat (great for dub music and all Augustus Pablo fans), identify the cap in series with the regen path, just before where it gets mixed back in to be shipped to the BBD again, and reduce its value.  It will probably be set for maximum bandwidth (we'll say that's 20hz rolloff just off the top of my head) so you will want to cut it by about 80% for starters to raise the low-end rolloff by 2 octaves or so.  If, hypothetically, it were 0.1uf, resulting in a 20hz rolloff, you would probably want to replace it with .018uf to raise that rolloff up to around 110hz, or .012uf to raise it to 167hz."

I must have this sound : ). Lyle found this schematic for a Deluxe Memory Man: http://www.freeinfosociety.com/electronics/schemview.php?id=381

I was wondering if you might be able to point the cap out to me. I am not great at understanding analog delays.

Also, I did more searching and found a post by Skreddy where he suggested a clever little way to make a pseudo-variable capacitor. He included this diagram:


I would then use the maximum bandwidth cap that's in there in series with a pot (of what resistance?) and put my maximum rollof cap in place of the old max bandwidth cap. What do you think?

John

[Mark Hammer]

Looking at the schematic you link to, cast your glance to the portion of the schematic where you see the two halves of the NE570.  You can see a 10k feedback level pot.  Just ahead of it you see .47uf cap in series and a 47nf cap to ground. The .47uf cap blocks DC and rolls off a smidgen of bass.  The 47nf rolls off a bit of top end.  Following the wiper of that 10k pot to the left, you'll see it goes to a 22nf cap and 100k mixing resistor back to the mixing stage where input signal and recirculated signal are combined to form the "new" signal going to the compander and delay chip.

The 22nf/100k combination provides for mixing regen in equal proportions with the input signal, with a 72hz low-end rolloff imposed.  If that 22nf cap was reduced to 10nf, the rolloff would shift upwards to about 159hz, and if the cap was dropped to 4n7 the rolloff would be shifted upwards to 338hz.  A 6db/octave rolloff starting around there would be just about right for shaving off the right amount of low end to make each repeat a little thinner.  If not 4n7, then maybe 3n9 (408hz).

Doing this in a pop-free way is another matter.  These days, I prefer to use series caps for cap-value switching since you can get different cap values by simply using a toggle to shunt one or more caps and achieve different values, and you avoid the popping.  In this case you want to go from stock (22nf) to mod (4n7).  If you had a cap just a bit bigger than 4n7 in series with 22nf, their combined series capacitance might hit the desired value.  Then all you would need to do is to shunt the smaller-value cap with the toggle to restore stock tone.  If you stick a 5600pf (5n6) cap in series with 22nf, you'll get 4n46, which is close enough enough to 4n7.  If the target of 3n9 is closer to your needs, then a 4n7 in series with 22nf will get you 3n87, which is pretty damn close.

The 47nf cap to ground just ahead of the feedback pot can be altered to get different degrees of brightness/dullness on repeats.  Shifting up to a 100nf cap would get you a duller version each repeat.  Shifting down to a 22nf would keep the signal about as bright on subsequent repeats.

Now, do I know where these caps are?  No.  You'll need to do that detective work yourself.  Is this schematic accurate?  Don't know.  Looks like Stellan Lehrburg's handiwork, and while he's generally trustworthy, this IS a big circuit and you can easily see some errors or the 570 alone (note the dual pin 10 outputs).

[Primus]

Mark,

This has been very helpful. I might end up doing the pop-free switching scheme you mention. However, do you have an comment on trying to use two caps in parallel with a pot to mix in the contribution of each cap?

[Mark Hammer]

Sure.  It works!    For some folks its perfect, and for some folks it's overkill.  If you wanted to dial in low end, I'd say use a 3300pf cap in place of the 22nf on the board, and stick a 250k-500k variable resistor (pot) and 22nf in series and place that in parallel with the 3300pf one

[Primus]

Mark,

Perfect answer. Thanks mark!

[Primus]

Well, it looks like the schematic is different. I'm looking at my pedal and don't even see an NE570. Coming off the wiper (center lug) of the 10K feedback pot, I find a 100 µF cap, then a trace that goes all the way across the board to what looks like a 160 K resistor comming off one of the 4558s. Maybe I can sub the 100 µF part for a 22 µF cap and get the desired effect?

[Mark Hammer]

There were Memory Man's produced early on that had no companding but I don't have a schem for that with me here at work.  I can't remember if E-H ever stuck house numbers on the 570/571 the way they did with CA3094's.

100uf? That doesn't seem even remotely close to what is needed there.  Are you certain that's the actual value?  Is it maybe scratched, hard to read, or is that "u" possibly an "n"?  All that should be needed there is, at most, 10uf or less.

[cd]

There were Memory Man's produced early on that had no companding but I don't have a schem for that with me here at work.

Any chance you can post that up through www.imageshack.us if/when you've got the time?  Or anyone else?

[Primus]

This is a 2001 DMM reissue. I'll take another look at the board and see if I can get some numbers off the chips. Also, I'll triple-check that cap value.

[Primus]

Okay, so here's the skinny... I got the value of the cap wrong. It does indeed say 100UF on it, but on closer inspection so do all the caps. Unfortunate labeling designtation! Still, it is a fairly large poly film cap. The poly films I got to replace it are maybe 1/2 the size. So it goes from the center lug to this to a 160K resistor to U1 (the input dual op amp). This is starting to look a little more consistent w/ the schematic we are looking at. Would it hurt to pull this cap and try one of my replacements? I dont' have a DMM that can measure capacitance. More details:

The PCB reads PICO 0033

There are two chips (1 chip each) that read:

HEF4047BP
D7327PS
Hnn0044 3

and

SA571N
J66372H
Jnr0022

[Primus]

So I decided to experiment and replaced the cap coming off the wiper (whatever it is) with a 3300 pf cap in parallel with the original cap and a 500K pot. I am happy to report that this gives an excellent effect of low-fi sounding repeats. It sounds a lot like a space echo, which I love. Two unexpected things:

1. When the smaller cap is dialed in all the way, the pedal will not overload. It will still feed back for 100+ repeats, but it doesn't do that "oh shit! reach for the level knob!" type of overloading.

2. When the pot is turned all the way up, lots of bass gets through. When it is all the way counter clockwise, lots of bass is cut. This is fine, but it is the opposite of what I was expecting. IE, I was expecting that as I incrased the resistance in the path to the big capacitor by turning the pot clockwise, I would force more signal into the small signal cutting capacitor.

I give this mod an A+ becuse it is easy and has dramatic, great sounding results. Thank you, Mark!

One other question... So it seems that I can control the rolloff frequency w/ this capacitor. How can I control how much is rolled off on each repeat. Say I wanted to rolloff twice as much per repeat as I am now, as it takes some time before a lot of bass gets cut away.

[Mark Hammer]

There are two chips (1 chip each) that read:

HEF4047BP
D7327PS
Hnn0044 3

and

SA571N
J66372H
Jnr0022

The first is a 4047 chip for the clock pulse.  This is what you see in the Small Clone and a number of other BBD-based circuits.
The second chip (SA571) is simply an NE571 with a different prefix.  So, this is a companded delay line.

As for the rolloff, the filtering that you have installed is pretty shallow, and deliberately so.  The idea is that one has to strike a balance between leaving enough frequency content intact that there can be something else to filter out on subsequent repeats, and producing so subtle a change that you have to wait for 6 repeats to notice anything.  Ideally, I suspect the amount of filtering used should produce an obvious tonal change by the third repeat.  I certainly can't quote any research literature that SAYS real echoes produce such a decay, but my hunch tells me that it would sound more natural if the "treble half life" were only on the order of 500msec or so (i.e., after 500msec there would be only half as much of what we think of as treble, as what you started out with).

The way to achieve this would be to simply up the value of the cap to ground a bit more.  The rolloff will still only be 6db/octave, which is good in this application.  On a sidenote, the reason why you can crank the regen and still not get screeching feedback is because the filtering reduces the overall regen signal amplitude, so that it is full tilt...but not really.