|
What is the analog? Data converters & amplifiers, or switching power supply?
Either way, they should have their own ground plane, and you join any two ground planes only in one place, and have power-supply bypass capacitors (.01uF) right there at that spot. Bring all the I/O through as close as possible to where the ground planes are joined. Don't make traces go across the chainsaw line, ie, the cut that separates the ground plane areas. Putting a trace across that chainsaw line, away from where the planes are joined, makes an antenna. It will indeed transmit and receive, even though it's right against ground planes, because its return-current path is not immediately beneath it, so there's a loop, possibly with other components or traces inside the loop area. This is indeed how some cell-phone and Bluetooth etc. antennas are made. I should probably scan and email you some articles on it.
In digital work, you want, as much as possible, to have the return current for any given trace to run immediately under it, no farther away than the thickness of the board itself. (Note that that won't usually be the shortest return path, but it will be similar to a twisted pair of wires.) Obviously you can't do this where the signal enters and exits the IC unless every other pin were either power or ground; but the unavoidable loop should be as small as possible.
With SMT, it works best to have the non-component side as the ground plane. If you only have two copper layers, you might have to put short traces in the ground plane as jumpers since you're short on layers; but if you have to do it, make those jumpers absolutely as short as you can.
If you have an audio-only circuit (as opposed to mixed signal), with its higher impedances and low frequencies, the enemy is electric-field coupling, not magnetic, so a break in a ground plane there is not such a big deal. In digital however, with very high dv/dt and di/dt rates, as tempting as it may be to fill all the blank areas with ground fills, those don't really help any. They're not the right approach.
Switching power supplies are yet another animal. The dv/dt may not be as high as the digital stuff, but the di/dt is astronomical. The AC loops there must be kept absolutely as small as possible. Also, if noise getting into ananlog lines elsewhere in the boards could be a problem, double-filter the power supply's input and output with ferrite beads. Wound inductors alone are not enough, because they have capacitance between the windings and they have a resonant frequency beyond which they are capacitive, and they don't present a high impedance at all for the highest harmonics.
I have done only a little in board lay-out with A/D and D/A converters and with Bluetooth (2.4GHz), but a lot with switching supplies and digital on the same boards with audio for our products, and have been quite successful. A switching regulator we have used a lot is the MAX732 which uses a constant frequency and PWM, and, out of necessity, I've gotten the noise down to a fraction of how much is specified. Especially when the load drops down to a low current, the board layout becomes extra critical to keeping the controller behaving well such that the pulse widths aren't varying all over the place between cycles, or even skipping pulses, both of which causes low-frequency noise that wreaks havoc in audio circuits.
_________________ http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?
|
Login
Register
FAQ
Search
