AD-3208 Analog Delay Finished - How Exactly Does One Bias (Calibrate) It?

[Rocket Roll]

I've completed my AD-3208 (the complete project's at http://www.generalguitargadgets.com/index.php?option=com_content&task=view&id=48&Itemid=26), with two CoolAudio V3205 chips, and it looks like this: 




(It's an BBDD enclosure from Banzai, and it was just perfect, size-wise - no need to mount pots underneath the PCB, yey!)

Now for the sound. So far it sounds like this: http://www.box.net/shared/vu6jp9dwk8

I've never had an analog delay, so the biasing part escapes me: how does one exactly "balance" those four trimmers in order to get the cleanest sound possible? Build information says:

"Looking at the PCB and viewing left to right, the trim pots are:
BBD2 Balance--BBD2 Bias---BBD1 Balance---BBD1 Bias"

I'm guessing the "Bias" pots are the one that should be set where signal's the strongest/cleanest, with the help of an oscilloscope, but what's with "Balance" trimpots? Do you use them only when biasing the BBD's, and you afterwards rotate them to 50% and leave them there?

Also, I'm wondering if this is the proper calibration procedure:
1) Start at the first BBD, use a 1 Vrms sine wave test signal, and adjust the first BBD's bias (and balance?) measuring from the pin 7;
2) Go to the second BBD, use a 1Vrms sine wave test signal, and adjust the second BBD's bias (and balance?) measuring from the pin 7.

Any help would be greatly appreciated, as I've never worked with BBDs before!

[Rocket Roll]

Turns out that pin 7 is in fact the input pin, and that pins 3&4 are the output pins. So I went to see a friend who has an oscilloscope and a signal generator, but we didn't dare use 1Vrms signal (we feared it could damage BBDs), and we turned the trimpots until the signal from the first BBD (measured at pin 7 of the second BBD) and the signal from the second BBD (measured on the NE571) didn't got as "sinusoidal" as it was possible.

Not the proper procedure?

[george]

I just did mine by ear - tweaked the biases until I got an echo then tweaked both to be in the middle of the "good" range ...

Don't remember what I did with the balance ones ... I think it didn't make much difference as long as they were set roughly in the middle ...

[Mark Hammer]

Normally, biasing a BBD can be done by ear.  AS one approaches the ideal bias voltage setting, you pass through a zone of no delay signal to a distorted low level delay signal, to a progressively louder and cleaner delay signal and then through all those stages again (in reverse) as you overshoot the bias setting.  All of this, of course, occurrs within robably something like 30 degrees rotation of the timpot so a good ear and a steady trimpot hand are helpful.  Still, it CAN be done by ear by most folks here.  A scope, while certainly helpful, is not essential.

When dealing with multiple cascaded BBDs, the problem one encounters in attempting to bias by ear, is that every other BBD had to be set just right in order to be able to tell if the remaining BBD is bias properly.  Clearly that is next to impossible.  The next best thing, then, is to mimic the circumstance of having only one BBD to bias.  Here's how to think about, and do, that.

Looking at the schematic of the AD-3208 ( http://www.generalguitargadgets.com/diagrams/ad3208_schematic.pdf ), you can see that there are two cascaded BL3208 chips (which can also be MN3208 or MN3205 or Cool Audio 3205s or whatever else is pin compatible).  On the output of the first/left one, you'll see a 10k trimpot, used for balancing the two complementary outputs.  The final mixed/balanced output at the wiper of that trimpot goes through a 1uf cap to pin 7 of the second BBD.  The 1uf cap is there to block the bias voltage from the first one, rendering it an entirely AC/audio signal with no DC bias.  Think of it like a "reset" button which permits the bias voltage for the second BBD to be independently set without any interference from the first chip.  You will note that the output of the second BBD also has a 10k balance/mix trimpot, but its output goes directly to the first 10k resistor of the 3-pole lowpass filter formed around the transistor. 

So, if you wanted to pretend iot was a 1-chip delay line, you would omit the 1uf cap between BBDs, and (temporarily) run the wiper of the first balance trimpot directly to that 10k resistor.  Since you don't want to cut any traces for this procedure, simply leave the 2nd balance trimpot uninstalled for the moment.  With the 1uf cap and the trimpot left off the board, all of the other components related to the 2nd BBD are effectively isolated from the remainder of the circuit.  Run a wire from the pad where the 1uf cap ties to the first trimpot over to where the wiper of the 2nd trimpot woulf normally be, and presto, you have a functioning single-BBD delay line.

You can now set it up by ear, as per usual.  Once that is done, unsolder the temporary connecting wire, install the remaining balance trimpot and the 1uf cap, and adjust the 2nd BBD by ear.

Ta-da!!

[WLTerry]

Good explanation Mark... thank you very much. I built this delay and it do exactly what Rocket Roll's delay does too. But I have a question: Should I isolate the first biased BBD in order to bias the second BBD or the second BBD should be biased along with the fist one?

Thanx
Will

[Mark Hammer]

The problem is like deciding whether you want to buy one television by looking at its image on another television.  If the image of the television you are looking at is not clear, how can you tell if the advertised one is any good?

Do the first BBD on its own, then add the second one and bias that one.  The isolation I described "tricks" the AD-3208 board into thinking there is only one BBD.  When that one is adjusted  just right, it will then be easy to tell if the second BBD is adjusted correct.

If you have an audio probe, and you know how to use it, then you can just go ahead and build the whole circuit board and do the adjustment by testing wih your audio probe at different points.  For people who would rather do the adjustments by plugging their guitar into the pedal and the pedal into the amplifier, it is simpler just to use the isolation trick I described..

[Rocket Roll]

Thanks a lot, Mark!

Just to mention, I've calibrated the clock today using another "wrong method". 

What I did is that I've used a "realtime" VST EQ-analyzer (one similar to Waves PAZ Analyzer, ie. this - http://www.waves.com/content.aspx?id=233) and I had no problem detecting the 10kHz clock noise with it. So, I've tweaked the trimmer until the clock was at approx. 10kHz and now my AD-3205 sounds like this:

http://www.box.net/shared/m85g2z00sc

I've left it on the verge of self-oscillation (the sound dies, eventually, by itself), so that one could check the feedback/repeats as well as the delay time. Right on the 400ms mark, if I'm not mistaken - right?

The problem is, mine ears are still young and I can hear the darned 10kHz clock!  No cure there, right?

[Mark Hammer]

The problem is, mine ears are still young and I can hear the darned 10kHz clock!  No cure there, right?

Sure there's a cure.  Just crank your iPOD up full, all day, all night, for a couple weeks.  That oughta do her.  You'll be oblivious to content over 8khz in no time at all.  Problem solved!

Using an RTA or software equivalent is actually a pretty clever idea.  Obviously it'll work better if the clock frequency is known, and the RTA frequency bands line up with the clock frequency, but still, pretty clever.

[Rocket Roll]

Thanks again! Here's how the clock noise looks in the software. Here's the fundamental 10kHz clock "tone"...



...and then there's a harmonic, twice as loud...



...and here's how they look together:



I suppose I won't be hearing as much of them once I get a decent tube amp, as I plan on playing it pretty loud!