How do I design a foolproof wiring harness tester?

I need to build a tester to verify correct placement of 25 wires that go from one connector to another (different pin numbers) The wires are all the same color, so crossed wires are a common occurence during building of them. Would like possibly an LED display showing that all wires are correctly connected from pin A to Pin B on opposite connectors. Any ideas are appreciated!!

you could use summit similar to a network cable tester. it send a signal thru cable 1, lighting (if correct) led1, then thru cable 2 lighting (if correct) led2 and so on. if any are mis-wiredthen there will be a jump in the output

interesting ...

have to have a little think.

A microprocessor could remember which lines were in error and print out a report. For a quick and dirty solution, this may be an approach:
After the test tech has written down the error number, the counter should be restarted. I have not figured out how to do that, but it should be simple.

Here is a conceptual method for doing what you want. I have shown five cables, but, of course, you can expand to 25 as you wish. In the diagram, all switches are synchronized, and all resistors are precision 1% or better.
Note that a window comparator is proposed in order to sweep all resistor tolerances. Note that the lowest resistor is each leg of the window comparator is 105%, and 95% R respectively. You choose what R is going to be.

Now, as Russ has proposed, if you could use a uC (easily an Atmel 2313!!!), and a DAC (12 bit DACs are very commonly available, and not expensive), you can easily use the micro to get rid of the two banks of switches and the window comparator. In this case, choose the resistor values to give you voltages that are, say, 10 counts or more apart on the LHS of the cable. The use the micro to "shade" the DAC output +/- 3 counts from nominal. Use the internal comparator to detect in you get a match within this "shaded" region. With this you can determine correct one to one connection, or cross connection, or even open (no connection).

Hope that helps,

Jem

Forgot to mention that the cables are the horizontal segments connected to R1, R2, R3 etc., and the switches are just concepts. You could use analog MUXes, and perform automatic scans.

Jem

where i work we use a computer to do the testing of the harnesses (altho i wudnt like 2 get a buzzer nd check over 800 cables accress over 100 connectors....

You can also cascade 3 CD4017 decade counters , use a 555 for a clock, I have an old diagrame here somewhere.........................

Hi retaftech,

Ive had a good think,
and yes you could make some sort of assembly along the
lines of Andrew's post:

you could use summit similar to a network cable tester.
it send a signal thru cable 1, lighting (if correct)
led1, then thru cable 2 lighting (if correct)
led2 and so on. if any are mis-wired then
there will be a jump in the output

Although i would put a half second pause for each LED,
and make each LED depend on the previous being correct.
That way it would stop at the first incorrect wiring.

But i don't think that making up circuits with LEDs is
going to be a good answer.

Andrew also mentions the use of a computer for harness
testing, and i really think this is the way to go.

I think that a computer program could be written which
would give a picture on the screen showing which bits
need correcting. (or whatever you decide).
Maybe checking for shorts too.

I think that a unit could be made which could plug into
the serial port or the parallel port, and this unit
could have two fittings for your harness to plug into.

I think that would be your best answer to this,

Regards, John

Hi retaftech,

Another minor point that just occurred to me,
if the requirements or details of the harness
alter for some reason,
then altering a PC program might be easier
than re-wiring LEDs.

(and putting back the original program would
be easier than putting back the LED wiring)

Regards, John

A problem with checking the continuity of each cable segment in turn is that it is unable to detect shorted pairs. If that cannot occur, then it is indeed simpler.

Jem

this is what we use...

**broken link removed**