Deluxe Memory Man

[lowell]

This is most likely a question for Mark Hammer... Mark if you're out there in cyberspace I need your expertise.

This pedal is Rev C which is actually quite close to Rev D. I believe the difference is that Rev C uses NE571 and Rev D uses NE570 companders.

This pedal has no "effect" output. I've traced the signal all the way to U8A. Actually the signal stops at the junction of R38/39. Should I get signal on my scope here? Sometimes I know that signal won't show on a scope on op-amp inputs. V- is good on U8 pin 4.

[Mark Hammer]

I heard my name called....

If I've understood you correctly, your pedal is highly similar to the attached schematic, but is not exactly the same. The 570 and 571 are like two twins, but one has slightly longer sideburns than the other, so I'm assuming there would be more substantive differences than mere chip substitution to warrant being labelled a different issue.

The one you attached has a quartet of MN3008s for a total of 8192 stages of delay. The next most recent schematic I have, dated 2002, has a pair of MN3005s for the same delay capacity. But with fewer chips, the "U8a" nomenclature from your schematic bears no relation to the other one I have. So I'm not sure if the "U8a" signal you refer to is at the first pole of the 3-pole lowpass filter shown in your schematic, or some other sort of signal.

If your pedal does correspond to the schem posted, do you have signal at the output of U3b? If so, then you clearly have delay signal.

I think we need a few more pieces of information here to move forward.

[lowell]

Your nose itched?

I believe so yes, I have signal at U3B out. I had to play with the bias trimmers to get the thing to wake up. The signal however is a bit distorted in that it looks between a regular sine wave and a triangle wave. Maybe this is inherent in this type of analog BBD delay circuit. I also forgot to mention that the 4 MN3008s are all getting the clock signal on pins 2 and 6 as well, as seen on my scope. It was fun to watch the clock signal change as I adjusted the "delay" knob. Not totally sure the schem is the same as mine I was assuming/hoping that it was. (shame on me) Is U8, in the schematic I attached, a 3-pole low-pass filter?

[MarkusBass]

Sometimes I know that signal won't show on a scope on op-amp inputs.

If you already know it, why don't you check the signal on the outputs ? I mean U8 pins #1 and #7. Is there a signal? I would also check DC voltages on these outputs (without input signal).

Mark

[Mark Hammer]

U8 appears to be that. It's a little wonky, but I see one buffered pole in the first stage and a 2-pole in the second stage.

One of the servicing difficulties of the multi-BBD units is that if one is improperly biased, it holds up the whole show. I gather they oblige one to bias each BBD in sequence, using a scope. That is, scope out the first BBD output and tweak the bias. Then bias the second one, under the assumption that the signal it receives from the preceding BBD is optimized. And so on. I pity those poor buggers who produced the Maxon AD-999 that used eight MN3007s.

[lowell]

Mark I should have been more clear. I also checked the output and there is not signal on U8 pin1 or pin7. I'll check the DC later on.

[lowell]

Mark I realized today when I came in that I had biased all the trim pots with a 800mv signal going into the pedal. Whoops. Anyways I rebiased everything with 400mv going in and the thing works. It's aliiiiivvvee!!!!

[lowell]

I would like to understand this circuit as best I can. From what I gather thus far the opamps are gain stages, the NE570 is a compander (compressor/expander) which compresses the signal feeding the BBD devices and then expanded signal again before the final gain stage, the MN3008 are BBDs. I think I understand that these basically move the signal from capacitor to capacitor all within the chip... in effect storing it for a short burst of time. The amount of time is dependent on the 4047 clock pulse being fed to pins 2 and 6 of the MN3008s. U8 opamp is used as a low pass filter thus removing the clock pulse from the output signal. Do I have this much right?

[Mark Hammer]

You have it 99% right, so good on ya!

The companding is also directed at removing audible clock noise coming from the delay side of the circuit. The lowpass filtering is intended to remove clock, but also "aliasing", which we might think of as "fake signal" created by the stairstep nature of what is essentially a sampled input. The stepping nature of sampling an analog input, even at high rates, creates the appearance of high frequency components that weren't in the original signal. The lowpass filtering attempts to take them out.

The clock signal is also removed by means of the 2k4 resistor pair you see on the output of the first three MN3008s, and the balance trimmer on the last one. They perform a function analogous to what a humbucking pickup does, only muuuuuuuucccchh higher up in the spectrum, by combining opposite phase versions of the clock pulse in equal proportions so that it cancels out. Some companies just use a pair of equal-value resistors. Some use a trimpot to provide precision balancing for maximum clock-whine cancellation (Maxon used eight of those suckers in the AD-999!). EHX uses a combination here.

If you look at a "sample & hold" module on an analog synth, what you'll see is basically an FET and a cap feeding a bifet op-amp. The FET is turned on and off by a clock - just like a BBD - and the cap hangs onto whatever charge it accumulated for that brief moment that the FET was turned on. The Bi-Fet op-amp has a high input impedance, such that with high-impedance paths on either side of it, the charge in the cap has nowhere to go, so it stays put, according to whatever leakage properties the cap has. A BBD is essentially wads of that packed into a chip with two interleaved paths that are combined at the output. I like to call them the tick and tock paths. When one is busy opening up the front door to charge up the caps in the little cells, the other is shutting the front door and opening the back door to et the stored charge move on.

[lowell]

Thanks for confirming Mark. I'm a bit confused about the compander. Does it literally "compress" the signal before the BBDs? IME this raises the noise floor so this doesn't make sense to me. Finally, how does the "expander" part of the IC work? What exactly does it do?

[Mark Hammer]

You can usually find out more about a component by searchng for app-notes and datasheets, using the part number and "pdf" in your search. Here's one for the NE570: http://www.datasheetarchive.com/dl/D...DSA-461687.pdf

The rationale behind dbx, companding, and similar strategies is that the path between the compression and expansion stages is inherently a source of additional noise. When the dynamic range is restored via the complementary expansion, all levels of the input signal are restored to where they oughta be, but any noise added to the output in that "risky signal path" gets downward expanded selectively without being offset by the compression. So it makes whatever was acquired between compress and expand quieter but does not have the same effect on whatever came in through the compression stage.

[lowell]

Wow that's REALLY cool. Thanks!

[Dirk_Hendrik]

. I pity those poor buggers who produced the Maxon AD-999 that used eight MN3007s.

Bump of an old thread.
Bin there Mark. It's a shitty job indeed. And it gets worse. After finally having all BBD's correcty trimmed one gets a guitar and starts playing.... distortion. Back to the generator and o-scope.... perfect at the last BBD but distorted at the compander's output. What the H..ll?

The BBD's in the AD999 are powered with 5 volts. That means that the last BBD's output is, if lucky, around a 2.5 volts DC output level. This DC voltage biases the 2 emitter follower filter stages after the last BBD. That in turn means that after the 2nd filter the DC level is at 1.3 volts because 2 times 0.6 volts was substracted at the base-emmitter junctions. At that point the last emitter follower is at such a poor bias point it's very non-linear and therefore distorts like crazy. Solution is to add decoupling caps between the BBD and fisrt filter as well as between the 2 filter stages and give the filters their own biasing to a proper 4.5 volts.