multiple BBD's in series

[Serge]

Are there any downsides to putting several BBD's in series?  What are the limits?  I know that how far you can stretch the delay time of a BBD depends on the minimum clock frequency, but how does this work when you put several BBD's in series?  Can I just use one clock for an unlimited amount of BBD's or do I need a separate clock for each BBD?  How does this trick work?

[Mark Hammer]

1)  Every BBD requires a bias voltage on the input.  Unfortunately, for a variety of reasons, one can't simply take the output of one BBD and "re-cycle" the bias voltage it was fed with.  The bias has to be wiped with a DC-blocking cap and resupplied each time for each chip in sequence.  Some companies find it good enough to have one single bias voltage source and distribute that voltage to every BBD chip.  Others find it more appropriate/suitable/optimal/precise to have a separate bias trimpot for each BBD.

2) The same clock CAN drive multiple BBDs, but some caveats apply.  The clock input pins on the BBD have what is described as an input capacitance.  In the case of the Matsushita chips, that capacitance is around 700pf per 1024 stages (350pf for a 512-stage device, 2800 for a 4096-stage device).  In the same way that the capacitance of a very long patch cable can erode the high end of your guitar signal, excessive input capacitance on the clock pins tends to "round off" and "de-square" the clock pulse.  Similarly, while cable capacitance may have no impact on lower frequencies, input capacitance on the BBD pins only starts to show itself as the clock frequency goes higher.  Below a certain point the clock pulse arrives every bit as crisp and square as it started out.  So, when you look at datasheets for the MN30xx and MN32xx series of BBDs, you will see that the clock chip (MN3101/3102) is only spec'd to drive up to 8192 stages (e.g., a pair of MN3005s), and the BBDs themselves are only spec'd to be clocked as high as 100khz.

Why is that "smudging" of the clock pulse important?  Because the BBD is essentially a switched device that hands off brief snapshots in a manner that mimics a continuous flow of signal.  If the internal switching depends on the clock pulse exceeding some critical voltage, then with a rounded clock pulse peak the FETS will only be switched on for a mere fraction of the time they are supposed to.  Imagine the sonic impact of dropping 1 or 2 out of every three samples, or frames of a film, and you get the sense of how drastic an effect it can have on signal quality.

The disadvantage of the high input capacitance, and the weaknesses/limitations of the clock chips, however, CAN be overcome.  If a buffer is inserted between the source of the clock signal, and the clock input pins of the BBD, the same clock can drive many more stages and drive them at faster rates.  The greater current of the buffered clock signal is sufficient to overcome the hindrance imposed by the input capacitance.

In the case of echo-length delays, the inclination is to be able to stack as many BBDs in series, and clock them at a relatively slow speed, so as to achieve very long delay times.  Here, the buffering is not essential for getting over the clock-rate limitations, but rather the hindrance created by the sum of all that capacitance even at low clock frequencies.  You can use one clock but you probably want to have separately buffered lines going to each BBD.  In the case of things like flangers, or high quality choruses where a high-capacity (e.g., 2048 or even 4096 stage) chip is run at very high speed for greater sampling rate and bandwidth, and inaudible clock noise, there may be minimal capacitance to overcome but what capacitance there is seriously challenges the quality of clock signal at the clock rates aimed for.


So, the long and the short of it is that there ARE hurdles to be overcome when multiple BBDs are used in series.  Thankfully, there are many practical and workable solutions.

[Serge]

thanks Mark!  that's quite an answer!  I guess I should have specified in my first post that I was talking about echo-length delays

[Mark Hammer]

No problem, I always figure that there are folks scanning such threads who represent all of the various camps (flange, chorus, echo, etc.), and try to stick in stuff for them because I know it's gonna get asked eventually.

[Sir H C]

A 2.8 nF  load!  That is huge!

[puretube]

here`s a "how" 

[Serge]

you probably mean "how " 

[StephenGiles]

http://img.photobucket.com/albums/v517/uncle_boko/TDA1022_c.jpg