Harness checker

[Bryang]

Hi all
Nice to see many of the same names in the forum from when I was last "playing" with PICAXE's a couple of years ago

At the moment we get wiring harnesses made, which we subsequently fit over a period of a few days. Unfortunately when we switch on we often find there are errors at the plug/socket pin-outs despite the suppliers supposed rigorous QA process!, which often results in hours tracking down the error(s), spliting the harness coverings, and re-pinning and/or re-routing. ...and no, changing harness makers isn’t an option! :-(

I have been asked to make a harness tester, the suggestion being supplying power to a couple of rotary switches, the second being a gang rotary switch, off the first, in effect making a matrix eg: 12 x (8 x 12) = 96 pinouts, and a matrix of 96 LED’s, to GND, one from every ganged switch pin, thus by sequentially rotating the switches any connected lines will illuminate the appropriate LED. Simple, but clumsy, and also reliant on people following a check sheet.

It’s a couple of years since I played with PICAXE & the like, but I figured I could use 2 or 3 pins on a PICAXE to trigger a few decade counters matrixed to replace the ganged rotary switches to sequentially feed power to a pin/wire one at a time , but how can I get the PICAXE(s) to determine which of the other 95 pins/wires are subsequently now active?
With a bit of homework I think I can work out how to store the results on EEPROM & send to a screen/PC based on my PICAXE projects a couple of years ago, but for the life of me, I cannot get my head around reading what pins are subsequently connected to the selected supply line! (there is probably a simple solution, but obviously I’m even simpler!! I have been looking through the forum for something similar, but no joy so far!)

Any ideas?

 

[techElder]

Arrange your INPUT to place a known pattern on 8 known lines. Look for that pattern on 8 known lines. If that pattern doesn't show up, record it for later troubleshooting.

Repeat for 8 more until you get to 96.

Look back at the recorded data to determine which lines were coming out at the wrong place (if required.)

Do this at your own receiving and ship the wrong ones back for rework before you get them into the production area.

 

[Bryang]

Thanks for your reply
Ok, so you’re saying use 8 inputs (eg: using a 28X), power up 1 wire, look at the pattern of the 8 inputs, then cycle the inputs to the next 8 wires, and so on, then when you’ve gone full cycle, power up the next wire, and run thru the input cycles again. etc. But how do I cycle the inputs to the next 8 wires?

The only method/idea I’ve come up with so far (haven’t tried it out yet!) is to use a matrix of decade counters to supply power sequentially, and duplicate the matrix for “reading” the pins, but instead of powering the wire(s), the 2nd matrix would supply the base of a transistor, and if there is power in the wire it can cascade to supply a signal back to the PICAXE, but as each decade counter pin would require a transistor, that would equate to quite a few transistors! Pity there isn’t a decade counter (that I know of) that works in reverse, sequencing pins to accept a signal rather than providing a signal.

Cheers
Bryan

 

[techElder]

I'm not saying "power up one wire", I'm saying "power up" 8 wires.

You already know where those 8 wires are supposed to go.

I don't know what this harness looks like physically, but when I had cables to verify I could build "a connector" for each end. Perhaps you can't get to each end? (My crystal ball is smoky.)

 

[Haku]

Could you setup a Picaxe with a bunch of shift registers to get 95 output lines plus ground, and on the other end a string of 95 LEDs?

Then get the Picaxe to cycle through the lines one by one so you can easily see if one or more of the lines isn't working or is out of sequence.

 

[Bryang]

OK, more background….The harnesses are for trucks that we fit hydraulic rams, etc to. So on a typical example a “main” harness might be plugged into say 14 electric solenoids for the hydraulic valve bank, and various proxy & pressure (etc) sensors, emergency stop buttons, etc. So a loom might have 30+ plugs/sockets, from single-wire pins to 47-pin plugs, & as many as 140 wires, with a few of them typically spliced within the harness. The “main” harness is usually about 10-13 mtrs. The harnesses are similar from truck to truck so usually based on the same generic layout, but there is often something different from customer to customer’s requirements.
We then also have “sub” harnesses for tail lights, flashing beacons, etc, etc.

I can easily get to each end before it’s mounted in a truck, & that’s where I wanted to tackle issues – before they are fitted to the truck!

My thoughts were to build a “universal” sequential line checker box & use a selection of jumper cables between the respective harness plugs/sockets & the test box.

Haku, yes, I could do it with a bank of LED’s but I wanted to get it away from a light board & get it to a printable connection readout.

(I get trouble with smoky crystal balls too…. I think it’s from the smoke escaping from the wires too often! ;-)

 

[graynomad]

but as each decade counter pin would require a transistor

You would'nt need transistors, the output will drive a Picaxe pin.

Anyway there is a kind of reverse decade counter, a multiplexer. Admitadly you have to provide an address but the result is the same. 6 4067 chips will do the trick.

I would do all of the above, make a rig with plugs to match your harness. Feed a walking 1 into shift regs on one end then read the wire at the other end of the 1 and also it's neighbours and any other wire that could realisticaly be shorted to it.

It might also make sense to have special patterns to test for common faults etc.

If you use 8 inputs it will run faster but you would have to be careful with your pattern, for example 01010101 won't detect wires 0 and 2 shorted. May as well just use a walking 1 I think.

Either way you'll have to demux the 96 to something the Picaxe can handle.

A compromise is to use 6 Picaxe inputs and drive the four 4067 address lines so you can look at 6 wires at a time. That's 10 pins ro read + 3 to write to the shift regs.

 

[graynomad]

Cross posted there.

If you have several harneses then same as above, but your test rig will have a stack of different connectors by the sound of it. That makes no difference except that you may need more than 96 IOs.

The beauty of using a Picaxe for this is that you can have multiple pattern/result sets (for the different cables) selectable by switches on the rig or from a PC. Add an LCD and/or serial link to a PC and you'll have a nice test rig. You could even have it learn from a known good cable.

 

[Haku]

Instead of a bunch of LEDs to show the status of the lines, could you use a bunch of parallel-to-serial ICs to then read the incoming signals?

 

[graynomad]

could you use a bunch of parallel-to-serial ICs

Yep, good idea and cuts down the pin usage a lot. Say using 74165, 74166 or similar. Then the entire ptogram could be as simple as

shiftout x times
shiftin x times
compare "in" bytes with stored pattern

 

[manuka]

Thought: Don't forget traditional wiring check techniques - capacitive especially. It may also be worth looking at the likes of a Fluke network UTP cable checker. Although these have only 8 wires to check, their versatility & ability to identify bizarre wiring blunders is quite mind expanding, & may help focus your PICAXE approach.

 

[Bryang]

Thanks guys, I was floundering there a bit before giving in & trying the forum.
Sounds like some great leads for me now to work with Thanks again

 

[hippy]

My thoughts were to build a “universal” sequential line checker box & use a selection of jumper cables between the respective harness plugs/sockets & the test box.

That's how it was done at one company I worked for when we built a similar system. The same tester could be used on any cables, just connect the cable tester to the right panel with appropriate plugs and sockets attached. We started with one cable tester and separate panels, eventually built more cable testers so they became standalone units for particular cable assemblies.

The way we did it was to determine how many pins there were in total for both ends, built a multiplex which could read all those pins, then, one at a time, set a pin high and low and read all the pins - really a glorified keypad matrix. Every input had a resistor which were collectively pulled high or low, inverse to the test signal.

Note that this sends signals both ways through the cable; important if there are diodes in the cable or loop-backs at either end.

If you have 'X' pins at one end of the cable, 'Y' pins at the other you get an (X+Y) * (X+Y) * 2 bit-sized array back. Then it was simply a matter of did the array match what the array should have been. The front panel had signals to tell the tester which cable assembly was being tested, which array to match with.

We also inverted the array so we could tell if a cable with same ends but not bi-directional was correctly wired but simply inserted the wrong way round. With some clever coding the arrays ( obtained versus expected ) could be analysed and give a good report on what the fault were; "pin 1 and 2 on plug A swapped", "pin 3, 4 and 5 on plug B shorted" etc.

The array can be large, but if you only want a pass / fail indicator you can do it a byte at a time.

A good way to start would be with a PICAXE with 8 I/O which can all be bi-directional; that will allow a 4-wire to 4-wire harness tester to be built needing a 512-bit array. It's then just a matter of scaling it up.

 

[Bryang]

Thanks Hippy, great points
I hadn't thought as far as uni-directional wires, & yes we do use a couple of diodes built into connectors, ditto resistors for the PLC - I'll look to factor that it to.

Thanks again everyone for all ideas
cheers
Bryan

 

[John West]

I built a rig similar to the one described by hippy. Used a Commodore 64 for the controller. It worked very well. And now it can be done for the cost of a PICAXE, some mux's and connectors. Along with a $10 LCD or an output to a printer, or both. Dirt cheap! I love modern technology.

 

[geoff07]

Why have the bother of multiplexing? As long as all you are looking for is Go/Nogo for crossed/missing connections then why not have a bunch of picaxes all looking at a subset of the wires? Then send a loom back if any fault is found and let the supplier do the work. Code replication is free and picaxe chips are next to free compared to your time or to designing complex pcbs..

If you fit loopback plugs on some of the ends you might be able to test many wires in one go and not need the large no of channels anyway.

Plus you could measure resistance if you did set up long sequences of looped wires, or diode properties.

 

[premelec]

this is a fun topic on a prosaic problem... ! put a bunch of precision resisistors in series through harness links and put a known current through the lot - look for correct voltages - will find shorts - opens and high resistance links... whatever works...