Hello again, Jukin. You've apparently got a noise problem. Not unusual in the real world. The thing is, sometimes problems like this are almost impossible to deductively troubleshoot without some major test equipment. Here's a few suggestions you can run through on inductive logic troubleshooting without taking out or renting the big guns.<p>* First, take a break and clear your mind. To start fresh, you need to look at the problem with fresh eyes. Go outside, have a Coke, take a walk. When you come back in, start a lab notebook, and jot down everything you do, as well as any ideas you may get along the way. Go back and read it occasionally, especially if you get stuck again.
* Take a good look at your circuit again, with the HC595 data sheet by your side. Are any of the outputs driving excessive/capacitive loads? Look carefully. If this is a "one-off" or prototype, be sure to check the schematic and the wiring/board layout again. Good technical rule of thumb -- on prototypes, the most common problem is the short between the ears.
* Look at the power supply again. Put a good scope right on the power supply pins of the '595 (isolate the scope if necessary to avoid measuring the ground loops caused by the scope ground and your circuit, if it's ground-referenced). What do you see? Many factors can contribute to poor regulation at the IC, including trace/wire inductance. Get one of those IC sockets with a bypass cap built in, and install that in the circuit. If anyone put an electrolytic in for bypass anywhere near the '595, remove it immediately. Looking at the power supply first is good practice, and it saves time more often than you think.
*Now that the PS is clear, take a look at your software. Try single-stepping through that section of code if you can, and see if the right things actually do happen in the right order. If the noise upset happens while you're single-stepping, you then have information as to the noise source. Again, good practice. You need to be absolutely sure it's not a software problem (for instance, two things happening at the same time that, because of slight differences in setup/hold times between the ICs, causes one to work and the other not. Single-stepping with a logic probe and a piece of paper will show this kind of problem. Try to make the problem happen.
* Next thing is the real-world loads of the I/O pins. I've used the 595 as essentially a serial in-parallel out output port, and have driven all kinds of loads with this. You mentioned relays. If one of the outputs ends up driving a transistor that drives a relay, make sure there's a diode across the relay coil. (Don't ever use the '595 to drive any inductive or capacitive load directly.) Look at loading, and for non-resistive loads try to limit the max current to something the power supply can handle. Take a look at the power supply wiring, too. Things may be happening on those lines when things are turned on or off which may kick back to affect the IC power supply. If you can, you might want to try separate wiring from the power supply to the loads. Use SSRs if you can -- a relay contact opening an inductive load (like your astable relay oscillator) sets up lots of broadcast energy with its arcing that can be picked up by your chip traces, which won't happen with a transistor- or triac-based SSR.
* If you've gotten this far, it's possible that electrical energy is being coupled into either the input lines (e.g. extra clocks), or output lines. The 8255/-C55 was/is famous for this kind of problem, because of the way its output FETs are laid out on the chip. Any chip with internal latching logic is possibly susceptible to this kind of problem. Hopefully by this time, you've identified a source of electrical noise that might be causing this problem, and how it is being coupled onto the circuit board lines that go to your I/O pins. If possible, move the source of noise as far as possible away from the traces/wires. (The solution...to pollution...is dilution! If the noise source is far enough away, it won't matter.) Moving the wires works, too. Another thing you might try is using small RC snubbers (from the IC pin thru a 33 ohm resistor to a 100pF ceramic disk to GND and your output. you can do this with the data in pin too, but don't do it with the clock pin -- you'll mess up rise/fall time requirements. This strategy works best with a ground plane. It doesn't take a whole lot of energy to make a flip-flop flop, and sometimes, by just cutting the edge of the spike, you can limit it enough so that the internal logic won't upset.<p>You really haven't provided enough information to be more specific, but you should know that you're not the first one to get stuck like this. The lessons learned from "voyages of discovery" like this are a start in learning good construction technique, and help when you're doing it next time. Here are two of the best resources for your problems:<p>
Don Lancaster's Degubbing Techniques<p>Analog Devices' "An IC Amplifier Users Guide to Decoupling, Grounding, and Making Things Go Right for a Change" by Paul Brokaw
Analog Devices AN-202.pdf<p>Feel free to email me if you'd like (no attachments, and put N&V in the subject heading, please).<p>Good Luck (luck is the residue of intelligence and hard work -- you'll be OK).
Chris<p>[ October 02, 2003: Message edited by: Chris Foley ]</p>
