Tricks in finding fault on a corrupted data bus?

Let's say you know that a certain point on a data bus has a defective device that is always trying to drive the bus low; Was wondering if anyone has any good tricks in finding the defective device, other than cutting/lifting pins?

Reply to
JW
Loading thread data ...

Drive the bus high, e.g. with a 1k resistor to Vdd, and use millivoltmeter. You can see the voltage drops in the traces--where the voltage drop is changing, current is flowing.

Cheers

Phil Hobbs

Reply to
Phil Hobbs

Hi Phil, That sound like a good possibility, thanks. I have a Philips PM2534 DMM that has a 300mV range with 100nV resolution. The only thing might be that if the bus is active (changing states) the readings might be jumping around too much. Often I see a tied node on a TTL bus that has transitions from .5 to 2 volts or so. Any ideas for a situation like that? Maybe try to crash the processor? :)

Reply to
JW

Yep. Hit the halt line. Alternatively, disable the clock. As a last resort, you could put an RC filter on the meter.

Cheers

Phil Hobbs

Reply to
Phil Hobbs

e

Does the bus point show low ohms to ground in cold state?(powered down). If so, leave a sensitive ohmmeter connected to it, hover over each device feeding the bus with a hot air blower. When you see a sudden pronounced change in ohms to ground you're hovering over the culprit. Jango

Reply to
jango2

I always check for that - if I see a low resistance to GND when power is not applied, that's an easy one. I have a Toneohm with a 200m Ohm range. Also useful for finding etch/solder shorts. Thanks, though.

Reply to
JW

Hang a CRO off each line in turn, & look for one that's stuck low. Once you've found it, pulling chips out of their sockets is probably the easiest way to find the culprit.

--
    W
  . | ,. w ,   "Some people are alive only because
 Click to see the full signature
Reply to
Bob Larter

If it's stuck low, then the signal won't be changing, will it?

--
    W
  . | ,. w ,   "Some people are alive only because
 Click to see the full signature
Reply to
Bob Larter

Yes, it can. It's not stuck/shorted to GND, but fighting with another latch, driver, or buffer.

Reply to
JW

Hi Bob,

I don't think you understand quite what I'm asking. For sake of argument, let's say I have five ALS245 transceivers all located on a 8 bit bus. One of them should be driving the bus at any given moment, but another one has a defective output that is corrupting data bit 1. I'm looking for any tricks that would help to narrow the fault to one device. So far, Phil Hobbs has had the best idea.

Sockets? Heh. If they were in sockets it would be a piece of cake. Have you taken a look at a modern circuit board with SMD parts lately?

Thanks.

Reply to
JW

anyone

Once found a shorted cap by powering the Vcc with a voltage regulated, current limited source and traced to the hot spot. Once determined that there was heavy traces, fed a couple of amps and watched the bad cap smoke. Saved a $5k board that had been passed around everyone else. There were about 50 or more bypass caps throughout the board and a real pain to have to replace them all to find the culprit.

Reply to
JB

JW wrote in news: snipped-for-privacy@4ax.com:

An ESR tester can be used to trace down low ohm shorts.

I used to use an 'RCA curve tracer' to troubleshoot bus problems. Sick gates usually lack a 'sharp knee' on the curve.

binary search is helpful, if you have to cut a trace, try to do it 'in the middle' (half the circuits on one side of the cut, half on the other). That way you cut the number of possible bad circuits in half, each time you cut a trace.

When you narrow it down to a few ICs, you can cut (or unsolder) pins instead of cutting traces.

--
bz    	73 de N5BZ k

please pardon my infinite ignorance, the set-of-things-I-do-not-know is an 
 Click to see the full signature
Reply to
bz

I've been lurking for a while and this is my first post here. I've been in the electronics industry for 40 years and have picked up a few tricks along the way.

Many faulty parts will heat up faster than their counterparts and can be detected with a liquid crystal sheet. Lay the sheet on a cold unpowered board and apply the power. Whatever heats up quickest is suspect. I've used this technique to find faulty ICs, Caps, and even high current traces.

Here is one source

formatting link

Art

Reply to
Artemus

In message , JW writes

Almost the weirdest fault I ever saw was like that, one PC printed fine to an old Epson dot matrix, another printed garbage.

The printer had a 74LS244 or 245 (can't remember which, was a *long* time ago) on the parallel input, measured with a meter on ohms the offending pin was 2-3 ohms to ground. The old Zenith 8088 drove it fine, the newer 386 machine failed miserably.

One way to trace is to get some thermochromic labels but if it's just one input then it may not work and can be fairly random depending on bus load etc.

Another way that's worked well for me in the past is with a component tester, the sort that injects a signal into the device and displays the 'curve' of the device under test, Huntron Tracker or Hameg 'scope. You quickly get used to the shape of the trace produced by various different chip input/output pins and it's extremely easy to spot a dodgy one.

--
Clint Sharp
Reply to
Clint Sharp

Okay, that's something to work with. Here's how I'd tackle the problem: Hook your CRO channel 1 to one of the 74x245 enable lines, & use that as the trigger, while using the 2nd channel to look at the faulty data line. Probe each of the 74x245 enable lines with channel 1 until you see a stuck line on channel 2.

I guess I'm showing my age here. ;^)

--
    W
  . | ,. w ,   "Some people are alive only because
 Click to see the full signature
Reply to
Bob Larter

detected

apply

Ooh, nifty. I'd like one of those sheets just for the entertainment value.

--
    W
  . | ,. w ,   "Some people are alive only because
 Click to see the full signature
Reply to
Bob Larter

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.