guidelines for laying out high gain circuits?

[gaussmarkov]

please share any general principles for laying out high gain circuits, like the bsiab2, dr boogey, thor, and tornado.

one particular question is whether keeping +9v and ground far apart is a good idea.  some advise the opposite.  also, how should the signal path be treated? is it beneficial to keep it short and direct across the board?  how would you trade this off against the size of the layout?  how about with feedback loops?

principles already identified are to use lots of ground fill (regions of copper that are grounded) and keep the input separated from the output.

cheers, gm

[John Lyons]

Hey Gauss.. I posted about this a while back.
http://www.diystompboxes.com/smfforum/index.php?topic=52406.0

Basically it's a keeping the stages left to right (or linear across the board) and not crossing signal paths and keeping the input as far away from the the output....

About ground and +9v traces.  The Dr boogey as laid out by Buck sears runs 9v super close to ground and my build is pretty quiet. I'm not sure this is the rule but it that one is a very high gain build and it works quietly if you layout the wires and pots neatly without crossing them...

John

[gaussmarkov]

funny.  i thought you might've.  that's a great topic.  i'll have to put it in the wiki, if it's not already there.  thanks john!

anyone else about the power rails?  close or far? 

[gaussmarkov]

in the thread that john cites, R.G. says (among many things)

6. be very, very careful with high impedance points; know what the input impedance of each active device is on your circuit board and lay out accordingly; a high impedance can be fed into by a vanishingly low capacitance, especially if you have a lot of gain; there are reasons why RF circuits are all done at impedances which are in the 50-377 ohm range.

what should you do to be "careful" and "lay out accordingly?"  apologies in advance if i am being dense. 

[R.G.]

Power and ground as close together as you can get them, preferably "stitched" together with 0.01uF to 0.1uF monolithic caps at every IC. Signal path short and direct. Minimum trace length is always correct.

what should you do to be "careful" and "lay out accordingly?"
1. know what points are high impedance points (i.e. FET inputs to opamps, FET gates, darlington inputs, bootstrapped bipolar inputs...)
figure out what the input impedance of each place a signal possibly could come in through capacitive coupling. If you don't know this, within at least an order of magnitude, you are not being careful.

2. Lay out accordingly by making it hard for capacitive coupling to happen. Capacitors are made higher capacitance by making the plates bigger and closer together. So to separate things, you make the voltage carrying places smaller and further apart.  If you can't move them apart, put a grounded piece of metal between them to shield them. I use guard traces all the time.

3. Know and use grounds properly. Ground is not ground, as I'm always preaching. Ground may be (a) safety ground (b) voltage reference ground or (c) sewer ground. You must be able to point to any ground trace and say which of these it is. Mixtures are bad.

However, on tiny circuits, this may all be wasted. The smaller and more importantly lower gain the circuit is, the less sensitive it is to cross coupling. One transistor circuits can get away with murder. Super high gain FET distortion pedals with many stages ARE murder.

[gaussmarkov]

thanks R.G.  that would seem to wrap this up.  layouts just got harder ... and more interesting. 

cheers, gm