[sdiy] ladder filter questions?

[Tim Stinchcombe]

Hi Dan,

This... 

> the odd thing was that it was working when i had nothing more 
> than a pot to +v and -v hooked up with a 1k resistor to the 
> emitters (for cutoff) but when i switched to a real cutoff 
> circuit the filter stopped working and I had to change all my 
> resistor values.

...is most likely linked to this:

> I am also just sort of randomly deciding where to put 
> coupling caps (do they have to be on the inputs? do they have 
> to be on the outputs into the differential opamp, etc)

...which with this:
 
> i fear the thing i dont understand well is how the  gain 
> structure/feedback works. it seems if i put too much feedback 
> into the - side of the ladder, i get no or little sound, and 
> if not enough, i get thin resonance.

first leads me to say that designing a filter from the ground up, even if
taking an existing circuit and playing with resistor values etc., can be a
tricky prospect - how some people manage to do it mostly by trial and error
eludes me, and suggests they have a more intuitive grasp of what is going on
than I can ever hope for. Me, I stick to the maths, as (for me) it yields
results more surely than any other method, though simulation helps here too!

So here is a quick whip round the function of key (coupling) caps in a
typical diode ladder filter - I'll work from this, as it is easily
referencable:

http://www.timstinchcombe.co.uk/synth/diode_18_24/tb303_vcf.gif

First up the frequency control. This stems from Q9/10/11: to work correctly
the emitters at the bottom of the ladder, Q12, must be high enough so as not
cause any problems with Q11 collector. In turn, this is (hand-waving)
'somewhere above Q11 emitter' = Q9 emitter, a diode drop above approx 2.6V,
the 5.3V ref halved by R64/65. This means Q12 bases must be a diode drop
higher still than the emitters: R68/70 set these at the 5.3V ref voltage,
ample to eliminate any problems back down to Q11. This leads to several more
points: first we have to ensure that any DC in the input signal applied
through R62 doesn't mess with the 5.3V level at Q12 base - this is why C17
is needed; the input is attenuated by R70/R62, and we need to prevent the AC
of the input signal upsetting the 5.3V ref set at Q12 bases - this is the
function of C16, which lets the AC through, and so gives the 'AC ground'
point for the ladder. The same applies for the feedback side, where cap C23
is again 'decoupling' the DC level of the feedback signal (whatever it is)
from the 5.3V set for Q12 bases. (The attenuation here is by R97/R98, and
you'll notice it is much less than the input side.) In a transistor ladder
you'll see R68/C16's function taken by a large cap somewhere up the whole
resistor chain biasing the ladder; and the 5.3V set at the ladder inputs is
also a bit 'extreme' - for a simpler expo converter, the output collector
may only need to be a volt or two above ground so as to avoid trouble.

The voltages gently increase up the ladder, where we have similar problems
with the DC levels in the differential 'take off' amp formed around Q21. It
has DC biasing levels set by R116/112/109, again with C29 preventing the
ladder signal from upsetting these. Caps C25 & C27 allow for the DC
difference between the diff amp and the signals picked off the ladder. The
diff amp output at R110 is referenced to the 12V supply, but is buffered by
emitter-follower Q19/R98 and then C14 takes the 'unwanted' DC out so that
the signal is now swinging about ground (this being the actual filter
output). This is then fed through the resonance pot (clipped from that
schemo), back through *another* decoupling cap, C15, because there are more
level-shifts around the buffer Q18/R94 - I leave these as the dreaded
'exercise for the reader', in the hope that you are now starting to get the
hang of it!

The TB303 circuit is perhaps more complicated than others, due to no op amp
involvement, which in general terms lead to quicker understanding of what is
going on (certainly I find it much easier to work out the gain from any
given op amp set-up than I can for a transistor-based diff amp such as
here). And it is also a bad example from the gain setting point of view, as
there are many places around the feedback loop which are having an impact,
and so which could be a reason for why the resonance isn't what anyone who
builds it would like it to be - there is no substitute for understanding won
at the expense of a lot of hard grafting!

Tim
__________________________________________________________
Tim Stinchcombe 

Cheltenham, Glos, UK
email: tim102 at tstinchcombe.freeserve.co.uk
www.timstinchcombe.co.uk