On Sun, December 26, 2010 5:05 pm, William \"Chops\" Westfield wrote: > >>> A ground plane > > Is the loop in the GND/Power signals in this design really so much of > a problem? With many modern chips having power on each side, I've > done (or rather, thought about doing) similar things in order to keep > a PCB single-sided. Umm... if you look at most modern chips, the power pins are usually located really close to each other (old PICs, TTL, and analog parts are the exception) but all the newer parts have the power pins located close together- and on the quad parts (leads on 4 sides) there is usually power on each side. A magnetic antenna (a loop) is not as sensitive (frequency selective) to size as a traditional dipole- it is more sensitive to cross sectional area- why a ground plane really helps. The number one thing to remember is that every current has a return current. If you can keep the path of the current going in very close to the path of the current going out, you're going to radiate far less, and be much less susceptible to interference. You'll also find that your rise-times are much faster. I did a circuit for my 2010 Master's class that had on the order of 800ps differential rise time out of a PIC32. It can be done- and at 800ps rise time, there *is* energy in the GHz range. (This is with the first generation of the PIC32, which did not have as much slew rate limiting) > 1) The chips involved are small, so it's not a BIG loop. > 2) As an antenna, you'd be looking at relevant frequencies of ~1GHz or > more; well above the operating frequency of the circuits. > 3) Bypassing at each end (or at multiple points?) ought to mostly > negate inductance issues of the "long" traces. > 4) I like having a loop of GND around the edge of my board even if > it's not specifically required by power wiring. It feels like cheap > protection against ESD (contact with 'stuff' quickly shorts to GND > rather than entering a chip signal pin.) Does this become a bad > idea? (This includes a loop of GND on the PCB side opposite the > Ground plane, when doing a double-sided PCB.) #4 can potentially mess with your rise-times, as you would have multiple paths for the return current- and with all these paths, you very likely will be inducing current in nearby traces- at high frequencies it doesn't take much capacitance between traces to induce signals on nearby traces that have nothing to do with the trace you are sending the signal down. Signals will find a way to couple, and a few pF at high frequencies is a relatively low impedance. As Olin pointed out (and I pointed out as well, but I believe he inadvertently left that part out :) it will probably be good enough for a board you're learning with. But it is good to know *why* your rise-times are not what you expect, and why the signals on unrelated lines are not as clean as they should be. You may not be able to see them on a low bandwidth scope, but don't let that fool you- they're there. > Trying to keep the PCB single-sided, is star-shaped gnd/power made of > jumper wires any better? In the interests of short straight jumpers, > I've shaped GND/PWR into loop-like structures around the chips and > components on a board, rather than splayed or grid-like structure. No ? I've been under the impression that star grounds are more important with high currents, to make sure that you have consistent references. It is important to understand where your high currents are- current across an impedance means voltage- which can cause your ground reference to change- which can lead to very weird outputs, particularly with respect to analog parts. Matt Bennett Just outside of Austin, TX 30.510843,-97.919286 The views I express are my own, not that of my employer, a large multinational corporation that you are familiar with. --=20 http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist .