[sdiy] Schematic/PCB layout best practices

[John Luciani]

On Wed, Dec 3, 2008 at 10:40 PM, Aaron Lanterman
<lanterma at ece.gatech.edu> wrote:
> Greeting schematicians and PCBeings,
>
> 1) Is there a Best Practice for numbering components? I'm putting in
> schematics into Eagles in a semirandom fashion, and am hence getting them
> numbered in a semirandom fashion.
>
> I was thinking:
>
> A) It might be best to renumber them based on how they appear on the final
> PCB, maybe number with increasing numbers left to right or something, to
> make things easier to find there, or

This is common practice which helps in troubleshooting. A number of schematic
packages allow you do a refdes renumber based on position.

> 2) What's the Best Practice for choice of trace width for (A) typical
> signals and (B) power and ground for typical synth module operation? I
> realize I've been using Eagle's default, which is 0.016 for default, for
> signals, and 0.024 for ground and power since it was the next step bigger
> from 0.016 from the drop down menu, but then I figured I should have better
> reasons for picking values than they happen to be some Eagle defaults.

Larger traces will let you use lower cost PCB houses or etch your own
boards. For low density boards I use 12mils for signals.
For denser boards I use 9mils. Most of my power is done with planes.
For traces I would use either 25 or 50mils depending on the voltage
drop I can tolerate.

>
> 3) Buchla quite often has separate "Quiet" and "Noisy" grounds. He uses the
> Noisy grounds in digital circuitry, but he also seems to use the noisy
> grounds sometimes in circuitry related to control signals, in places where
> it seems hooking things to a "Quiet" grounds would do no harm. If a circuit
> I'm working on is pretty much all analog, is the a compelling reason to keep
> separate "signal" and "control" grounds, or can I just lump them all
> together? Is it related to Buchla using 1/8" for signal but bananas for
> control?

For large signals at low frequency I would use the "let ground abound"
approach. You may get into more trouble by slicing and dicing your grounds.
My back-off-the-envelope numbers --- copper at 100MHz has a
surface resistivity of ~3 milliohms/square. You will need a fair amount
of current to make a difference on a 10V signal (if you have a mixed signal
board with large switching currents you may need to look at this
closer). A 3cm diameter loop  has an impedance of ~50ohms. Keeping loop
area low would provide a much bigger benefit.

(* jcl *)

-- 
http://www.luciani.org