Many circuit harness tester design / build

shortbus

Joined Sep 30, 2009
7,600
It would need to be sequential then. That is how the expensive testers work.
That's how the ones where I worked were. Your build board brought back memories they had many of them like that on an oval conveyor, with each operator adding a few wires and taping or other duties. These were for GM cars and trucks. They also had some that were stationary for low volume specialty harnesses, known as "taxi harnesses" though they didn't go in taxi cabs. This was at GM Delphi Packard Electric, they made everything from the wire to the terminals, clips, connectors and even the tape.
 

Thread Starter

tomcat_meow

Joined Jan 5, 2012
12
Yah we were approached by GM a number of years ago just before they moved most of their harness building to Mexico. I along with the other owner didn't want to get involved with car manufacturers because they want cost reduction year over year for 5 years and then they want you to quote and do that all over again... they nickel and dime you down. They change buyers like a guy who has the $hits changes underwear and there is no loyalty at all. I know one of my competitors were doing work for them with a workforce of 250 people and when they moved to Mexico he was down to about 22 people and he had to pay severence up the whazzoo . I think that more or less bankrupted him. Too many eggs in one basket.
 

MisterBill2

Joined Jan 23, 2018
5,763
I was the one responsible for the testing of installed harnesses at a co8uple of different GM plants quite a few years ago. Post #16 makes a lot of good points. A simple continuity checker will not detect short circuits at all. A minimum valid check would "common" all the circuits in the harness and then apply a voltage to the line being checked and verify that enough current could flow. That is the simple way to check for shorts and errors and poor connections. It may also be the fastest. The royal pain in harness testing has always been the connection to every connection of every branch. Also all of the connections add up to a big expense. most of the time. For reasonable voltages and currents one SPDT reed relay per end is needed, or you can use FETs to pull up and resistors to pull down but suddenly you also need A/D conversion and the test code gets complicated.
Lots of questions now: Are these machine harnesses, or for vehicles, or possibly avionics? How many branches and how many circuits at each branch? AND are there splices of conductors in the harnesses? Are the wires simi9lar gauges, or multiple sizes? And ar the circuits in shielded wire or not? AND what sort of fault reporting is needed? Pass/fail, or detailed reports?
 

Thread Starter

tomcat_meow

Joined Jan 5, 2012
12
The harness is mostly 22 gage UL 1007 wire and no shielded wiring ,with several sub assemblies coming together to form the entire product. It is for vending machines. There are about 15 power wires but low power so 18 gage is used for most of that typically 2 or 3 amps powering a few different servos.
There are no resistors or semiconductors involved just pure wiring. I have to make two of these harness testers and more in the future. Connections for the largest require just under 256 so that is good for now but we have another project that will require over 400 connections. Our customer does not require any reporting . As for fault reporting we presently use the cirris brand name harness tester systems (10K per) and they provide a fault code which tells you at the tester which connection is wrong. When I build the test board I create a spreadsheet to document what connection on the Cirris tester goes where on the test board so it becomes an easy look up for troubleshooting if there is an open circuit. The Cirris tester just beeps every two seconds telling the operator all is good as they add more connections while building. If something is wrong and gets plugged in by the operator the tester beeps rapidly, so the tester also becomes a real time building aid. There are several sub assemblies and a number of individual wires so finding something wired wrong is fairly easy because it is whatever was plugged in last. I've been chatting with land boards who have a number of cpu integration boards that will allow expansion of the I/O on say a raspberry pi (which I'm leaning towards using as the cpu) . They have an I2C bus multiplexer and then I/O boards that will create 128 connections per board. The I2C multiplexer will allow up to 8 I/O boards to be utilized so that creates a very large number of connections for me. The rest of the task is coding it up.
 

MisterBill2

Joined Jan 23, 2018
5,763
OK, and thanks for the explanation. Now I am guessing that most of the wires go to a large connector, or a set of connectors, located towards one end, and the other ends go to smaller connectors with fewer connections. It occurred to me that another way to check quickly would be to have the smaller connections provide different resistors to each pin, all fed from a single voltage source. Then the active test end would just need a load resistor and an analog input, and the far end would not need switching, other than selecting which connector to supply power to for the check. That scheme would cut the required number of test channels almost in half, although it would require an analog input and more complex code. That is similar to how the body harness testers worked for the one auto company.
 

Thread Starter

tomcat_meow

Joined Jan 5, 2012
12
No actually there are many in line connectors going to two circuit boards that I am simulating . The picture doesn't clearly show them but there are at least 9 connectors (green) with either 8 or 10 circuits that populate and run to mostly two or 3 circuit weather proof connectors and some other wires are just stripped and we tin them in a solder bath because they (the customer) screws those ends down to other connectors for payment process and power supplies etc. We just use a spring as a quick release mechanism for those wires. harnessboard.jpg
 

MisterBill2

Joined Jan 23, 2018
5,763
No actually there are many in line connectors going to two circuit boards that I am simulating . The picture doesn't clearly show them but there are at least 9 connectors (green) with either 8 or 10 circuits that populate and run to mostly two or 3 circuit weather proof connectors and some other wires are just stripped and we tin them in a solder bath because they (the customer) screws those ends down to other connectors for payment process and power supplies etc. We just use a spring as a quick release mechanism for those wires. View attachment 179349
OK, I see, and it is not like an automotive harness at all. But still, with most of the wires going to one area the concept could still work. It looks like most of the wires all originate in one area, and so that is where the scanning would connect, while the resistors would be at the ends of the branches. And since the checking is real-time, as it is being built, the speed is not so very critical, within limits.
 

ronv

Joined Nov 12, 2008
3,770
I am involved in a wire harness manufacturing business that builds very large wire harnesses with many circuits that are usually branching out over a large area like a 4ft. by 8ft. or even larger build board. Usually we just buy a circuit tester which is very expensive and overkill for what we need which is just to see that a wire circuit is properly housed in a connector and placed and run through the harness to it's other end where it should go. We only need to verify the circuit. The bought harness testers are slick but at 10 k it becomes a financial burden especially when we need many of them for all sorts of harnesses as we expand the business. So I thought it would be worth the look to see if there is advantage to building a DIY system . The bought system can learn a known good harness and store it in memory (called up when you need it again) once it is put on the harness test/build board. Presently I build an interface from the board to the tester which involves running wires from mating connectors to the tester and placing the mating connectors on the front side of the board so the build person can plug in the various parts of the harness and build a complete harness. The bought system also monitors the build in real time so that if a mistake is made it beeps to alert the build person. The bought tester has a standard 256 connections and then you have to buy add on sections each time for another 256 at 1500 for each 256 test points. I need at least 512 and sometimes more. So what I'm looking for suggestions from some knowledgeable folk on which way to approach design of this tester. I'd like to use an arduino if possible to housekeep counting and logic. Not sure but some sort of multiplexing system or if there is some sort of cascadeable addressable logic switch out there that I don't know about or some other LSI that would help. I've attached a simple drawing(paint) to explain the basic idea. View attachment 179182
It seems like you could just build a big scanner and use your micro in a step and scan arrangement.
They make chips like this that seem like they would work. The biggest problem is it's a huge rats nest.
https://www.sparkfun.com/products/9056
Maybe flash memory to store the programs.
 
Top