[sdiy] Why Bands for String Filters and Narrow and Numerous

[rsdio at audiobanshee.com]

On Jul 1, 2017, at 11:25 AM, Michael Zacherl <sdiy-mz01 at blauwurf.info> wrote:
> On 1.Jul 2017, at 11:14 , rsdio wrote:
>>> I say, discrete BBD madness!
>> OMG. I'm trying to imagine a PCB with 1024 capacitors, 1025 FETs, and a pair of non-overlapping square wave oscillators.
> 
> I seriously considered that some time ago and found it won’t work on the comparatively large scale of an SMT-PCB and potential varying tolerances of the components. Besides the sheer number of parts involved. But then again it looked like a perfect job for automated SMT, at the first glance.

Not surprising.

I think that the smallest BBD chips were 512, so perhaps a PCB with 512 would be feasible. The largest BBD chips were 4096, if I recall correctly, and those would obviously require a ridiculous amount of circuit board space.

As you say, the varying tolerances would be a significant issue. The BBD chips not only had the advantage of all transistors being on the same die, but they also adjusted the gain of the FET stages to be slightly greater than 1.0 to compensate for the loss of voltage as it was passed from one stage to the next. Replicating this on a PCB would probably require 0.1% resistors that would also add space. I suppose you could just let the tolerances fall where they may and have only final makeup gain stage that could be adjusted for unity gain across all of the stages, but that would obviously compromise the signal-to-noise ratio since each stage would not be operating at optimum levels.

p.s. Thanks for the MOnSter 6502 link, Ben. I had only seen the Visual 6502 page before, and wasn't aware that someone had built a discrete transistor version. Too bad it can't run at 1 MHz.

Brian