In-Circuit Testing of a Diode 1

[rjgoebel]

Hello,

I am trying to test a diode on a PCB for work in forensics, concerning a product failure that caused some batteries to explode. I have posted on this topic before but I have more specific questions this time. Does anyone here have any experience with in-circuit testing? I have stumbled on a large in-circuit component tester that looks to have an intended purpose of being used in manufacturing - ie: testing products with PCBs as they come off an assembly line.

I think that something like this will work very well for testing a diode in a circuit that I am not allowed to mess up. The issue is, I cannot constuct any destructive testing of this PCB because of legal issues and spoilation of evidence. If I wanted to test a diode that is in a PCB, I would need to de-solder the diode and remove it then use a DMM to conduct the test. Of course I cannot do this because it might be construed that the component sustained failure while being removed from the circuit.

So my question is, first of all, has anyone worked directly with one of these machines? The item that I stumbled on was a low-cost (for high volume manufacturing corps.) in-circuit component tester doing a google search on the internet, made by CHECKSUM.

Here is a URL of the product I am talking about.

http://www.checksum.com/analyst_ft.html

Does anyone know if I can pay to use one of these products, or anyone that is an expert on them? Thanks!

 

[itsmoked]

Sorry I have not worked with a machine like that..
That said, you shouldn't need a machine like that. Take a standard Digital Multimeter set it to diode test and use it.

With the probes hooked up one way the diode will read some really high reading. Hooked up the other way you will usually read the diodes correct normal forward voltage drop.
If this is your result, look else where for the problem cuz the diode isn't it. And yes this is all done in circuit!

 

[sdmays]

To add to itsmoked's post, verifying the diode in the forward biased direction (should read some small voltage 0.7V to 2 or 3 volts) will be accurate as long as there are no components connected in parrallel to the diode (atypical case). Verifying the diode in the reversed biased direction (typical DMM should read "OL") should also be accurate if the above condition is true.

 

[BrianR]

The DMM is appropriate to test the diode, but ensure that it has a diode test mode as this ensures there is enough voltage across the terminals to forward bias the diode. The resistance mode on a DMM will not work reliably for testing diodes.

Most DMMs apply a small current to the diode at several volts and read the voltage. A forward biased rectifier or power diode will read a maximum of 0.7V but could reasonably read as low as 0.5V depending on type. WIth the leads connected the reverse direction the diode should be an open circuit although but this will be affected by parallel circuitry but you should read a higher reading than before.

You don't need that machine.

 

[sdmays]

I would assume the actual current the DMM supplies varies from DMM to DMM, etc...In my experience, the diode voltage is usually slightly lower than what you might expect. A DMM I use at work supplies about 16uA if I remember correctly (measured it with a second DMM).

 

[itsmoked]

Yeah I agree. As an aside, most DMM try to use a low voltage on the resistance measuring because they don't want to *activate* any semiconductors which tends to leave the resistive component isolated, allowing in circuit measurement.

 

[BrianR]

16uA is a bit low for a useful diode test when there might be leakage present I think but it would explain a 0.5V reading. 1mA is a better value and many meters use that.

 

[VEBill]

I've worked with the old Factron ICT systems. They can deal with parallel circuits provided that there is another node in the unwanted parallel path.

In other words, if you have two components in parallel then you are obviously forced to measure the parallel combination. But, if the unwanted parallel path is made of two or more components in series, then the system can apply a 'guard' (basically a ground) at that node that basically takes that path out of the measurement circuit.

Here is a general link to the 'guarding' concept:

http://www.google.com/search?q=in+circuit+test+guarding

For legal cases, isn't it more normal practice to invite the opposing team to send over their own subject matter expert to witness, and even participate in, the examination? That way, the two experts from the opposing teams might actually come to agreement and thereby take the case directly to settlement.

 

[VEBill]

Another thought in case you need it...

The old Radio Shack used to sell a special clip-on heatsink that you clipped onto the component lead (right next to the component body) so that it would keep the lead temperature down while you spent 20 minutes ;-) struggling to desolder the device from the PCB.

 

[VEBill]

To properly test a diode, it might be better to use a Curve Tracer. These gadgets will show you the complete V/I curve (including both polarities). Such information would be much more complete.

Link:

http://www.google.com/search?q=Curve+Tracer

(Re. the first hit - I don't think a homemade curve tracer would be appropriate for a legal case...)

If I was overly concerned about a diode (legal), I'd also rerun the Curve Tracer test at various ambient temperatures.

 

[OperaHouse]

I agree that you need a curve tracer. I have seen a lot of strange transistor problems that show up on a tracer. I forget, but there was one brand of in circuit tester that had a scope to view waveforms. You need to electrically disconnect the diode from the circuit for the best results. Since this is in a power circuit, it should be easier to "remove" other components by cutting traces to them. I agree that any mechanical or thermal stress can alter the diode. Lightly grinding away a trace with a Dremmel tool may be an option.

Home Made may not be appealing legaly, but in circuit testing to simulate normal conditions may be the only way to recreate the "failure." Proving something didn't do it is difficult.

 

[Comcokid]

There are production-type in-circuit testers made by many companies. Checksum is one of many.

There is also a bench-type in-circuit tester more intended for troubleshooting, repair, and low-volume products. These are based on a sine-wave stimulus and response measurement displayed in a X-Y pattern and interpreted by the operator. Comparisons can be made with a known-good board and circuits containing L, C, R, and semiconducting junctions can be evaluated. These originated years ago and in the military were known as an "octopus". Modern more capable units are made by companies such as Huntron Inc with the trade-name of Tracker. This may what OperaHouse is think of.

 

[rjgoebel]

Hello Everyone,

Thanks for the responses. We have a few people that think I can get by with a simple DMM and a few more that are suggesting a curve-tracer and a few more that are saying the in-circuit tester will do what I want.

I think the DMM is pretty much out because of the complexity of the circuit. There are parallel parts connected to this diode and I cannot realistically use a diode test function on the DMM because of the traces in parallel, and I cannot really remove the diode from the PCB without much trouble. It is possible to do, but if the diode does not turn out to have failed, what then? There has to be another way to proceed in testing the product and determining the primary failure mode, but if an important diode in the power circuitry part of the mixed signal PCB is gone, that might make it difficult.

VE1BLL, you hit it right on the head with the examination procedures. For all of those who don't know the forensics engineering and legal procedures that we have to go through, basically every single party that is put on notice for potentially causing a problem (their insurance carriers) hire experts to attend an evidence inspection/evaluation. This is nessacary if ANY destructive testing is done, because if you do destructive testing by yourself then spoilation of evidence is a big issue in court and you might as well have lost your case for your client.

And you can't hold a destructive testing or evidence inspection unless you have a clear idea of what procedures you might be taking to test something. Often times you have 5 to 10 engineering failure analysis consultants at one place and you don't want to waste their time!

Jim Goebel,
Electrical Engineer
Mid-West Forensics, Inc.

 

[analogkid2digitalman]

An alternative to removing the diode, if possible (non SMT and enough real estate/access) and legally permissible, is to cut one lead in half so as to break all parallel connections. Now use the DMM across the diode. When done testing, solder the cut lead back together.

Wheels within wheels / In a spiral array
A pattern so grand / And complex
Time after time / We lose sight of the way
Our causes can't see / Their effects.

 

[OperaHouse]

I would suggest that you build a duplicate circuit and simulate different diode failures. This will give you experience in what the signals would look like. Also this would give you information on whether a diode failure could cause the same damage. If a diode failure would cause this problem, there should have been a redundant diode or some other protective device in the circuit. This wouldn't be a cell phone would it?

 

[itsmoked]

I also have a bit of a problem with there being a diode with lots of parallel path stuff when we are talking power supply stuff. I can hardly believe it... What would the purpose be of parallel stuff around a diode? The current can go both ways around the diode? Then why have a diode?

In signal suff this would be more likely but then battery busting energy probably wouldn't be present.

 

[ScottyUK]

Hi Itsmoked,

Here's a couple of suggestions:

Flywheel diodes are in antiparallel with the switches they protect;

Diodes are frequently found in FET gate drives to allow different switch on and switch off times.


There are many more possibilities. Without knowing what the circuit does it's lergely speculation.



----------------------------------

If we learn from our mistakes,
I'm getting a great education!

 

[itsmoked]

Hehe ScottyUK, I knew I was asking for it...

I can actually think of a few more cases too.. It is hard with out a schematic.


rjgoebel; You should post a schematic. We would have a blast speculating.. You might find our suggestions very helpful in your analysis. It could provide you with a lot of insight.