[sdiy] Patchell's modules

[harrybissell]

I think we are on the same page, but expressing our thoughts in
different ways.

I agree... it is the inductive reactance that is the problem and this
dominates
the impedance. The straight resistance is rather small... and would
probably
help STOP any ringing.

There will also be some parasitic capacitance in there as well.

So to a step change in current... there will be some sudden DROP in the
voltage on the bus, followed by some overshoot or small ringing caused
by
the parasitic capacitance.

If you add a LOT of capacitance (decoupling)... in most cases you will
completely
swamp the inductive reactance, and there will not be any overshoot or
ringing.

In the short term... the capacitor IS the power supply, and the
inductance between
it and the chip is VERY low.  Later, the cap will recharge, but the step
change is
mostly absorbed between the cap and the chip... and very little of it
propogates
back onto the power supply.  Assuming that the next demand for a current
spike
is a long time away (assuming low audio frequencies...).. then the
inductance
in the main bus will slow the recharge of the local capacitor, but
probably will not
ring.

If the next demand for power comes when the capacitor is still in a
state of recharging
(higher frequency) then there is a good chance of ringing (as you
pointed out).

Staggering the capacitor types and sizes WILL help because each of the
caps has
a different equivalent series resistance, and parasitic inductance. The
Electrolytic
will have a high ESR and high inductance and will tend to resonante at a
lower frequency. The large size makes it capable of storing a large
amount of energy.
The Tantalum has low ESR and fairly low inductance. Ceramics (usually -
assuming
you selected a suitable one) have very low ESR and very low
inductance... otoh they have small capacitance as well.

Done right, these self resonant points of the capacitors will NOT occur
at the same frequency, but will be stagger-tuned across the frequency
that the circuit operates at.

Decoupling (done right) is exactly that... it uses the fact that the
trace inductance between a local capacitor, and the local power drain...
is very low and the trace
inductance between that local (let's call it) node and the next one is
fairly high.

I guess that inductance is your friend as well as capacitance... as long
as you
recognize them and keep them in the right places...

H^) harry

karl dalen wrote:

> --- Harry Bissell Jr <harrybissell at prodigy.net> wrote:
> > --- karl dalen <dalenkarl at yahoo.se> wrote:
> >  Decoupling caps
> > > > are your
> > > > FRIENDS !!!)
> > >
> > > Only "if" power planes are low impedance.
> > > If power planes are high impedance bypass caps are
> > > terrible enemies.
> > > They worsen the cirquit behaviour. So always low
> > > impedance your
> > > powerplanes.
> >
> > nah nah... they are still your friends. Its the power
> > supply impedance that is your enemy !!!  :^P
>
> Well, well, i forgot to include the supplies, sorry, but could
> we say that inductive reactance of the tracing that usually
> dominates the impedance characteristic?
>
> I had no idea of what Ian was designing, however a caos
> gen cant be high power so check inductances, if you do
> a caos of high power check impedances! :-)
>
> > This is true only in very rare cases.
> > In most cases
> > adding the decoupling caps lowers the power supply
> > impedance anyway.
>
> Yes but asume we have a board trace net that is high in
> impedance, or un even impedance,now adding caps moves
> the cirquits resonant points closer to various cirquit
> frequencies wich in bad luck will create spurious resonances.
> Wich would be less of an event if the impedance would be low,
> because impedance varies with frequency!
> Right, or im out and peddaling in the green?
>
> The rule has been for long time that ground planes to be
> of low impedance 0.1ohm, and less.
>
> > The "bite-in-the- at ss" comes if the power supply lines
> > are HIGHLY INDUCTIVE (like big loops of wire enclosing
> > a large area between current feed and return).
>
> Well, agree! but these two are closely connected,
> would it be wrong to say: inductance play larger
> role in most low current,designs but imedance a
> larger role in power applications?
> (asuming same frequency for both). To some extent!
>
> > Adding
> > a cap here might possibly make the circuit RING (a
> > resonant condition).
>
> Yeast, but what does the resonat condition depend on,
> is it easyier to get resonant peaks in high impedance
> nets rater in low, asuming low as 0.1ohm, and high as
> in at least kilo ohm nets, however you could have a net
> of very high inductance and never go into ringing, wich
> i sugest depends on the resonant frequency, wheter you
> hit the resonant peak or not.
> (i cross my fingers im not totally wrong here!!)
>
> > Using staggered sizes of decoupling cap helps here, an
> > electolytic, a tantalum, and a ceramic for low to high
> > frequencies respectively.
>
> > Most of the ringing problems would be at very high
> > frequencies that the average synth-diy'er will never
> > see... but for RF and some high-speed computer aps
> > there WILL be an issue.
>
> And OP ams have Bw of several 100Mhz today, a current
> feedback OP easily does 500Mhz.So the average synth diy can
> see this problem if not properly designed.
>
> Remembering the oscillator soft sync problem where the  slave
> oscillator synced due to master osc integrator triggered the
> slave trough the power lines, trough noise level at power
> supply lines, i wonder was this a inductive problem or
> impedance problem or both?...........or something else!!?
>
> I dont say this PCB thing is easy, it aint, and its complex.
> I have a baad feeling that i have mixed up something here!
>
> Reg
> KD