Joined: Fri Aug 30, 2002 1:09 am Posts: 8546 Location: Southern California
|
Dan Moos wrote: I you look at my most recent schematic, I thought I fixed the switch placement problem. Yes, my apologies. I guess I forgot to click on the diagram to enlarge it, and just assumed the top was Vcc (or VDD) so you were switching the pull-up resistors rather than the IRQ\ lines.
Quote: I was wondering more what the theory was behind putting power and ground close to all the other pins on a connector. I realize its an AC performance thing, just not exactly the specific reason. I imagine its to do with capacitance? Is it similar to why one might have every other wire on a ribbon cable be ground?
I know from audio work, that ground and Vcc are the same for AC purposes, so I figure that is a clue.
Care to elaborate, or maybe point me to the relevant part of my "High Speed Digital Design" book? For this subject, I have toyed with creating some animated .gif images to illustrate what happens; because although the concepts are not difficult to understand (leaving out the heavy math), they can be difficult to communicate.
It has to do with inductance, quite different from what goes on in audio circuits which are comparatively high-impedance and low-frequency. Imagine a pulse going out a signal line, and its ground return current going the opposite direction two inches away because all the ground pins are clustered together at one end of the connector instead of being distributed. Suppose the load is three inches away. Now you have kind of like a single-turn coil, two inches wide and three inches long. That inductance is unwanted, and makes it take longer to charge up the input capacitances of the loads, and, along with the capacitance, forms resonant circuits that can ring badly, such that clean square waves become sloppy and their ringing following each intended rising and falling edge may cause unwanted multiple triggering or false inputs, or, less likely, even damage. If you put the ground return current really close to the signal line, it's like reducing the width of that coil so it is very small, and the total inductance drops way down.
But if the ground return connection is far away, an unwanted opposite reaction voltage will be induced in nearby signal lines instead.
I'll avoid getting into transmission lines just yet. It is interesting however that if you have, let's say, a place where the ground return current can return to the transmitting IC 2" away and one 0.2" away, the return current will take the closer one, because of mutual inductance. Unlike the other inductance, mutual inductance here is your friend. Consider this illustration:
Attachment:
tranLine2.gif [ 12.44 KiB | Viewed 5399 times ]
Current in one conductor builds up a magnetic field around it, and a change in that field is opposed, and there's a pushback, which is exerted on the adjacent conductor as well. If that conductor can allow an equal change in the opposite direction, the pushback collapses, and it's easier to get good signal integrity. In fact, if you put the pair through a ferrite bead or core of some kind to concentrate the flux lines, that core can handle practically a limitless amount of current without saturating, assuming there's no common-mode current. In mechanical terms, I like to think of it this way:
Attachment:
tranLine1.gif [ 22.22 KiB | Viewed 5399 times ]
The large weight slides freely on a greased rod; but it has a lot of inertia. If the two racks accelerate equally in opposite directions, you don't have to accelerate or decelerate the weight. But if the two racks accelerate or decelerate in the same direction, that change is opposed, because the large weight wants to maintain whatever state of motion it is in.
Similarly, if you have a PC board with a true ground plane, a signal trace's ground-return current will not take the shortest path back, but instead will flow directly under the trace, taking on the shape of the trace. For this to happen though, there cannot be any interruptions in the ground plane—which is why ground pours do not qualify, since they are not continuous. In audio circuits, the coupling between adjacent wires is by electric field, caused by voltages. In RF and high-speed digital circuits, they're by magnetic field, caused by currents. There is a way to use pours to supplement real planes; but if they're not done correctly, they can actually make things worse, according to experts in the field like Rick Hartley, Eric Bogatin, and Suzie Web whose lectures you can see on Altium's YouTube channel.
Edit: Here's a related post: viewtopic.php?p=32133#p32133 (It's pretty short.)
_________________ 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?
|
|