PCB worries/ decoupling

Tim Ressel Tim_R1 at verifone.com
Tue Mar 14 01:51:01 CET 2000


We here in the Compliance lab deal with these questions all the time. All the
experts have the same answer: "It Depends".

Here is our current wisdom; I hope it helps:

Bypass caps: 0.1uF is the "pulled from air" default guess from the TTL days. 100
nF is fine for any CMOS circuits you are building. The tantalums tend to have
low internal resistance (ESR), and therefore work well as bulk decoupling caps.
In a mixed analog/digital design, the most important layout issue is keeping the
analog and digital ground lines separate. This means run a different ground line
for analog and digital circuits, and keep them apart electrically for the
longest distance possible. Tie them together at the incomming power connector,
or even better, have a separate set of power wires for the analog and tie them
together at the power source. Chips that have analog and digital together will
often have separate analog and digital power pins. Pay attention to these, they
will make your job easier.

Wow. Brain dump. All questions welcome.

Tim Ressel--Compliance Engineer
Hewlett-Packard
Verifone Division
916-630-2541  
tim_r1 at verifone.com                     



-----Original Message-----
From: Roel Das [mailto:Roel.Das at student.groept.be]
Sent: Monday, March 13, 2000 9:35 AM
To: Synth DIY
Subject: PCB worries/ decoupling


I'm currently working on the pcb of my midiCV project. I need ALL the tips
you can give me. It's a mixed digital/analog design, and I haven't got any
experience on pcb design. First time I ever do this, actually.
I use the eagle lite software for it. It's got lotsa libraries, and is quite
easy to work with. The size is somewhat small (160x100mm) but I can split up
the schemo easily.

And a question on decoupling capacitors. Which values should I use? Will 100
nF do fine? And which kind of capacitors? The DAC716 datasheets suggests 1uF
Tantalum capacitors at the power pins. Does the tantalum part matter? Why?

Thanks
Roel


----- Oorspronkelijk bericht -----
Van: "Harry Bissell" <harrybissell at prodigy.net>
Aan: "Per Mattsson" <XinclXsynth-diy at mkv.mh.se>;
<synth-diy at node12b53.a2000.nl>
Verzonden: zaterdag 11 maart 2000 18:08
Onderwerp: Re:unwanted sync


> Mosk "unwanted sync" events involve iceberg, submarine, etc
>
> see "titanic", "lusitania", etc...
>
> But seriously folks....
> The things to watch for are mostly power supply routes. Use the routes to
force
> current
> to flow where you want it. The best layout would have the oscillators
> physically separated
> at three corners of the board, with common circuits in the fourth...
(makes
> sense yes??)
> The power should start from a single point (the main filter caps right
where
> power enters the board) and run separately to each oscillator. Use
decoupling
> caps for each active (IC) device very near the chip (keeps the capacitance
near
> the current drain, with very little inductance). Inductance will stop the
caps
> from delivering current in time to meet demand, causing voltage drop...
> unstable operation....
>
> IMHO the lack of attention to power supply concerns (esp. placement of
> decoupling) is the greatest cause of a good (schematic) design to perform.
>
> The NEXT BIGGEST ERROR is to share a single IC (maybe a quad op-amp)
between
> separate VCO's. This guarantees that the power supply distribution is WRON
G!
> and also
> puts sensitive circuits within microns of each other, instead of mils
(.001").
> A sure prescription for interaction !!!
>
> Good circuit boards are art. A proper layout looks good, feels good.
Keeping
> symmetry
> between all three VCO's will assure similar performance. Think about
thermal
> issues too, the VCO nearest the top of the enclosure is probably the
warmest
> one...
>
> OTOH... some circuits work because of (intended or unintended)
interaction. The
> MXR
> Bluebox is a prime example.... the 'sneak' path through the power supplies
> actually makes the unit perform 10X better. The designer was either very
lucky
> or very skilled.
>
> IMHO I perfer to keep circuit functions separate... so that changing a
gain
> resistor doesn't
> screw up three other parameters.
>
> Are you going to build and sell these... or are they a kit ??? SMT can be
hard
> with a
> "100 watt American Beauty" (soldering iron).
>
> H^)  Harry
>
> Per Mattsson wrote:
>
> > Hi guys!
> >
> > I'm drawing loose sketches on a three ASM VCO board. To reduce size I
hope
> > to make use of as many SMDs as possible.
> >
> > Are there any specific "demands" on that (or any) VCO board design? A
part
> > from the lin-log components beeing close/glued together. Will I get any
> > unwanted soft sync effects from traces going parallell to long? Or are
the
> > traces on a PCB the size of a ASM VCO simply to short to have any
effects
> > of that kind at all? Or does unwanted sync things has more to do with
> > porely regultated/filtered power than "badly" placed components .
> >
> > This will be a real compact design since I aim to build a four voice,
three
> > oscillator module.  I hope to make room for "global" CV ins as well
voice
> > specific CVs . I have yet no idea on how to patch this unit together
with
> > other polyphonic modules. Maybe matrix switch chips or five pole DIN
> > suggested on the list before.
> >
> > If I were to give board a name it samply has to be  'a ASM VCO SMD PCB',
> > which I find rather fun...
> >
> > ---Per Mattsso
>
>



More information about the Synth-diy mailing list