ad3208 report and qs

[swt]

Hi!. Ive built the ad3208 in perfboard!. It works!. Just want to know what are the balance trimmers for and how to set them. Thanks a lot!

[Mark Hammer]

Every BBD has two parallel sets of storage cells.  Each of those storage cells acts like a sample & hold on an analog synth, where it takes a snapshot of an analog voltage and prevents it from leaking out too quickly.

The snapshots taken for each sample get moved to the next cell in line on the command of the clock signal fed to the chip.  Each set of storage cells alternates between grabbing a sample and moving the samples it has along in line.  One of the advantages of this system is that sample rate (hence signal fidelity) is maximized by having the sample-taking duties effectively divided between these two sets of cells.  It's a bit like the way my wife and I alternate getting the groceries out of the back of the car: while I'm bringing a bag in she's going back for another, getting the same job done at double the speed.

All of this takes place by feeding complementary (opposite polarity copies) clock pulses to the two clock input pins on the BBD.  Via what I assume is some sort of a flip-flop arrangement, pulse N tells it to take a sample, pulse N+1 tells it to move the sample over a notch/cell, pulse N+2 tells it to take a sample, etc.  Because the pulse trains coming from the clock driver chip (typically, though not always, an MN3101 or MN3102) are antiphase, one half of the cells are always being "told" to do whatever the other half are not being told to do.

Although most BBDs (SAD1024 and MN3010 being notable exceptions) have a single audio input pin, they have two audio output pins, one for each set of storage cells.  This provides for a kind of "audio interlacing".  Back when video monitors could not scan all that high, higher-res images were created by alternately feeding line N, N+2, N+4, N+6, etc, then line N+1, N+3, N+5, etc., to the monitor, so that the entire screen was made up of two different scan halves.  Fortunately, the image alternation was done so fast that all you really noticed was a small (though irritating) bit of flicker.  In an analogous way, the two halves of the *audio* signal are stitched together at the outputs of the two BBD output pins.  As you can guess by now, the balance trimmers are for matching those two signal halves to form a nicely balanced complete audio image, where each set of samples contributes equally.  

The big question is how can you *tell* if you have actually done so.

Fortunately, you have clock bleedthrough to help you out here.  BBDs are notoriously noisy.  Because that noise is clock related, and especially because the two channels in the BBD have opposite phase clock noise, when you combine them in perfect balance, the clock whine is cancelled, exactly the same way hum received by each of the two coils in a humbucker pickup cancels out if the two coils are well-matched and combined out of phase.

Ideally, you want to use a scope to do the balance adjustment, but if you lack a scope, you can still do the adjustment by ear.  Tune the delay for the longest possible delay time, so that the clock bleedthrough will be at a low enough frequency to hear (and be annoying).  Now adjust the balance trimmer to minimize that whine.
 
Since the AD3208 is forced to use a pair of BBDs rather than a single MN3005/3205, this makes the task a little harder: you need to listen for reduction of clock noise in one chip, all the while putting up with clock noise from the other unadjusted chip.  (Incidentally, this is probably one of the main reasons why high capacity 4096-stage BBDs were such a godsend in the old days.  Imagine trying to adjust a quad of MN3007's or SAD1024's.  The labour costs would likely add more to overall production costs than the parts.)  Doing this with a scope and a probe on the trimpot wiper would oibviously be the most straightforward method, but if you use linear trimpots and set them for close to the midpoint, that should make it easy enough to hear noise level differences.

The bias trimpots are, of course, set so that you can actually hear reasonable-quality delay signal coming through.  If the bias trimpots are set right, but the balance ones set wrong, you *will* hear delay, just not the highest quality or most noise-free delay.

Hope this makes sense and helps.

Mark

[swt]

so my next question will be: i'm getting a good delay sound, no distortion, but the delay loses lows repeat after repeat, like going higher in pitch. Is this caused by the balances?. Or maybe just a matter of design?. Thanks mark for the already GREAT explanation!!

[Mark Hammer]

To prevent stray DC voltages from interfering with functioning, most devices that use a feedback/regeneration loop will stick a capacitor in the loop, so that the stage which mixes the input and feedback signal only sees AC input, not AC on top of DC.

In the case of the AD3208, that cap is a 1uf cap in series with the 100k resistor on pin 12 of the compander chip.  There are other caps in the signal path which may also limit low end, but the 1uf cap identified would be the one that is responsible for each repeat having less bass than the one before.  I've actually installed selectable caps in the feedback path as an intentional effect in past.  Some dub artists have used echoes that produce increasingly shrill repeats on each iteration.

I am not suggesting you have sloppy skills, but you may want to check the value of that cap on your board.  Possible it is a 0.1uf or 1nf by mistake?  If so, that could produce the attenuation of low end on each successive repeat since the 100k resistor and series cap function like a highpass filter.  A 0.1uf (100nf) would not do all that much damage since it produces a rolloff starting around 16hz, but a 1nf cap installed by accident would produce a low-end rolloff starting just under 1.6khz!

Alternatively - and here I confess to ignorance - is the cap the right value but the wrong orientation?

[Mark Hammer]

Just to add, since the rolloff point is a function of the resistor AND cap, accidentally using a 10k, rather than 100k resistor would also change the rolloff point (raise it higher), in additional to yielding more feedback.

[swt]

thanks again Mark!! I'll give it a try and let you know...

[swt]

Hi Mark!. I've checked all the values and connections, and they look ok. I've wrote this to Scott:I have two analog delays, an ibanez ad99 and a deluxe memory man. Both of those delays behave similar, and in the eh, when the feedback is high, it does the regeneration thing, but it keeps the pitch. In my Ad3208, if i set for longer repeats, it sounds as a guitar feedbacking, going higher in pitch like a distortion feedbacker, and if it does the regeneration thing, it goes really high!. I've changed the ne571, for a ne570, but it does the same. Any ideas?.
What do you think?. I can definitely send you an email with a small mp3 to show you what it's doing. I really want this pedal to work because i think it's a good design, and want it for gigs, so as not to risk the vintage ones. Any help will really be apreciated. Thanks a lot!!!!!!!

[quarara]

hi,
I have the same exact problem that swt had 5 years ago with his ad3208. during the self-oscillation the pedal doesn't retain the pitch of the note/chord it's played, but it starts to feedback reaching high pitched sounds, not like other delay where the feedback it's rounder and darker. I think it's a design issue since several builders had the same problem. what should I do/check?

[cathexis]

Upload a sound clip, if possible,to let us hear what you mean. Mine sounds like this:

http://www.aronnelson.com/gallery/main.php/v/cathexis/Temporary+files/AD3208+fb.mp3.html

I really like this successive loss of what feels like both lows and highs on the returns, gives this delay it's character, and sets you up for some serious dub possibilities.

LARS

[quarara]

Upload a sound clip, if possible,to let us hear what you mean. Mine sounds like this:

http://www.aronnelson.com/gallery/main.php/v/cathexis/Temporary+files/AD3208+fb.mp3.html

I really like this successive loss of what feels like both lows and highs on the returns, gives this delay it's character, and sets you up for some serious dub possibilities.

LARS

first of all thanks for your demo, it's really useful. as soon as I can (I cannot record anything at home) I'll upload a soundclip of what I mean. however, imagine that the pitch of the forth or fifth starts getting higher: that's what happens to my feedback. your feedback it's really "blurry" (and definitely cool, I have to say), mine it's more... hectic, I'd say.
I forgot to write that I used 2 mn3205 instead of two mn3208 for this delay. another modification I did was to swap the 100k resistor coming from the pin 12 of the compander to the 1uF cap before the repeat pot with a 22k for more repeats. Lately this afternoon, after having written my previous message here, since I was unsatisfied with the noises described above, I put a low pass filter (a 10k resistor, a 3,9n cap and another 10k resistor in series) instead of this 22k resistor. The situation slightly improved, but not as much as expected.

[quarara]

I think I've solved my problem.
When I did yesterday the low pass filter I was very tired and somewhat frustrated by the ugly day I had yesterday. Today I've checked the low-pass filter again and -ashamed to admit it- I've noticed I didn't connect the cap to the ground. I did the wiring of those components right and now the repeats seem more harmonic, organic and in self oscillation the highs I had yesterday disappeared leaving my poor eardrums -and my amp- in peace.
If someone has my same problem, s/he can find the answer in this message in which the omniscient Mark "Her Majesty" Hammer explains how this low pass filter influences the sound of this unit.

Thanks again to cathexis for his help!