Home-made Modular Synth teething trouble II

[<<marjan>>]

Phil Callaghan (Core Design Ltd.) wrote:
> 
> Hello guys,
> 
> I mailed this list a week or two ago about problems with my home-made
> analogue synth concerning the unwanted interaction of modules.
> Most of the advice I have received suggests that putting capacitors over
> the supply rails is the way to go. However, my power supply is regulated
> so I cannot see why this should be.
> Can anyone explain briefly why I need these capacitors, what they are
> doing and why I may need more than one capacitor over one supply rail.


> Are these capacitors acting as filters, and therefore different
> capacitor sizes filter different (unwanted) supply frequencies?
nope...

There are lot disturbances that may be in any circuit.
PS main problems come from as it has some impedance in it's rails
to circuit board (also aplies for traces on board) so when circuit 
takes high current you get voltage drop on rails which affects
prformance
of components (snr,noise margin,...) also if there's bad ground
connection
you can get some potential on connection to ground with increasing
current
(so you get 1V instead 0V on reference point) - PS must be low out 
impedance and good gnd.

For capacitors : large electrolytic caps have significant  impedance so
you connect some ceramic 100nF in parallel which has minor inductivity
(it's dielectric is feroelectric,meaning higher dielectric const. and
high specific capacitance) for supply with some parasitic impedance
Why : in oscillator circuits and esp. modern components there are very
high current pulses inside it for making better speed performances 
(eg. for fast charging of parasitic cap on cmos gate or for Miller
capacitace iside compesated opamp ,fast comparators ,pulses with
fast rising edges) so instant current spike can't be obtained from 
big filter cap with high indictance (and inductance disables sudden
current changes) without 2-3V drop ,so little ceramic cap enables 
wanted electricity Q for that spike with just some 0.3-0.5 V drop as
it has small inductance .Thats why you also need those caps across fast
switching components onboard itself - closer to it - better as there's
also unwanted impedance in copper traces too.It's good to solder them
directly to IC on -/+ pins,on every ttl,cmos and comparator chip.
Well it's maybe too much but it can't hurt.
eg. using 10-100nF cap and 0.5V drop you have 5-50 nC charge.
One logic gate has 50pF input,and for 0->5V rise it needs some 150pC .


There is also crosstalk between parallel traces due parasitic
capacitance,
noise,RF from around can also disturb PS,or inducting from high
amplitude (even LF)
going trough unshielded wires and near traces...
Also some pulses from AC network which must be filtered before
transformer
as it can get from primar to sec via capacitance between,blah,blah...




marjan