Best practice biasing for MN-series delay lines

[ElectricDruid]

Hi all,

I'm working on a flanger design, and I'm wondering what the best way to do the biasing of the delay line is. I've seen various approaches on different schematics, and wondered if there's much to choose between them.

Some circuits use a preset as a voltage divider on the input of the delay with the signal coming in via a DC-blocking cap. Some don't bother. Some use a large capacitor (100uF) and a small resistor (68R) to derive the Vgg voltage, which is supposed to be 15/16th of Vdd. Others use a pair of resistors. I've got a 1K and a 15K in my circuit currently and that seems to work. What's the relationship between Vgg and the input biasing if any? What does the Vgg *do* anyway?!

Similar questions arise at the output. Some schematics simply tie the two output phases together. Others have 100Ks to ground, and 3K6s or 5K6s to the output. Some replace these two resistors with a 10K preset to balance the two outputs.

Has anyone done any experiments or know which of all these various options is supposed to be the best?

I'm not expecting it to be silent, but I'd like to get the best noise performance I can.

Any pointers appreciated. Thanks.

Tom

[Mark Hammer]

It can depend on which MN series we're talking about.

• Typically, the bias voltage is derived from the supply voltage by dividing it down, either by a trimpot, or by fixed resistors.  The challenge that creates is what the bias voltage ought to be as the battery ages and the supply voltage changes.  The trimpot is a best guess, and may become invalid over time.  Of course, if the pedal/circuit uses a more stable AC source for its supply voltage, bias becomes a set and forget thing.
• When Panasonic developed the 3rd generation chips (MN3001/3002 being the first, and the MN3007 et al being the 2nd), one of the tasks they appeared to set for themselves was maintaining a valid bias in all operating conditions.  The MN32xx series could work at much lower supply voltages than the MN30xx series.  What this made possible was the use of a stable 5V supply from a 9v battery.  A humble LM7805 3-pin regulator could provide a rock-solid5V source, and the bias could be derived from that, either with fixed resistors or a trimpot.  Since that type of regulator needs at least 2Vdc more at its input than at its output, that meant the 5V supply would be stable until the battery had dropped down to around 7V, by which time there likely wouldn't be enough juice in it to power the circuit anyway.  So, low voltage functionality meant that the bias could be set as if the circuit was running off wall current, with no adjustment needed after it left the factory.
• Balancing the two complementary outputs is partly a matter of whether the result justifies the effort.  One needs a balance between the two complementary outputs so that the "tick" and "tock" of the clock pulse cancels out when the two complementary outputs are combined.  But how much reduction in clock noise is achieved by that means, in comparison to all other potential sources of clock noise and aliasing in the circuit and layout?  Some companies use companding as a shotgun means of dealling with various sources of noise, and others adopt a more audiophile approach, tinkering with the details.  The epitome of this was the first issue Maxon AD-999 that used eight 1024-stage chips (the 4096-stage chips had not yet started being second-sourced via Beiling and Cool Audio), and provided separate biasing and output balance trimpots for all 8 BBDs.  Naturally, all those trimpots, and all that adjustment, raised the price.  Good thing for Maxon they had a little bit of a reputation at that point or they wouldn't have been able to sell the things for the price they were asking.

[/list]

[Fender3D]

Just a few side notes:
MN32xx are N-channel BBDs whereas MN 30xx are P-channel.
This means 32s will have a pull down resistor to GND in a single supply at their output whereas 30s will have a pull up resistor to VCC.
The same for the VGG: 14/15 VCC for 32s and 1/15 with 15V VCC for 30s. This is not so strict though.
I didn't bother too much with different VGG voltage, (hell, if EH tied VGG with GND for MN30xx... ) since datasheets suggest it.

In my designs/adaptions I always provided a filter capacitor for both VGG and VBias.
Maybe it's avoidable... hmm, better safe than sorry...

VBias will also change with temperature... Some designer provided diodes to correct this.
This will be more noticeable with SAD BBDs though.

The output null trimmer will be a must when clocking the chip with frequencies inside the audio spectrum (under 40KHz).

With guitar, I prefer a gate noise reduction ala ADA than a compander, because I feel compander (570 series) altering wet signal, this might be more acceptable/usefull with bass.