1
\$\begingroup\$

I have been in a small factory today, they produced 4 kind of PCBs that all are simple analog boards, each month they produce 1000 of them. The problem is that till now, after mounting them almost 10% of them were damaged. So they prefer not to test each of them, as the cost of time did not worth for them. But now the percentage of damaged has increased and they are looking for a solution. I suggested to add some test point to the PCB and I will design a Tester. But the problem is something else, they say that they don't need a module tester. They need some thing that can tell them if they have to buy a series of component or not , for example they order 10,000 of capacitor , they need a tester to test specific number (I also have to suggest the number ) the capacitors and tell whether or not buy these batch or not , so basically I need to design a tester that can test these things : Resistors , Capacitors , Different kind of Diodes , Relays and LEDs?

So the questions are mainly these 2 thing :

  1. For each device what are the best way that I can test them , even a document would be useful but , tips here is really appreciated because I can ask questions about.

  2. Do we have to test all them before mounting or not?!! how can I get the efficient number for testing to get the best result? (if I have to change the question topic just let me know please)

\$\endgroup\$
  • 7
    \$\begingroup\$ They ought to buy their components from reputable suppliers and improve their manufacturing techniques. That should virtually eliminate failures. \$\endgroup\$ – Leon Heller Jun 14 '15 at 18:05
  • 1
    \$\begingroup\$ The cost of the testing equipment and time to do it presumably will have to be absolutely minimized, otherwise it would have been more economical to buy better quality components for which large numbers are not defective. If you are looking at having to 100% test resistors, for example, you should seriously consider whether the money is better spent buying from a more reputable supplier. The most important thing in practice will be to identify the major causes of failure and eliminate those, rather than testing everything exhaustively. But it depends on the acceptable failure rate, cost, etc. \$\endgroup\$ – Oleksandr R. Jun 14 '15 at 18:05
  • 1
    \$\begingroup\$ There's a whole statistical science to sampling. For leaded components you can use various spring clips, etc. For surface mount, tweezer-type solutions are common. But also consider that any components you take out of the tape to test aren't going back in for the pick-and-place machine to actually use, so you could also just (automatically?) solder them to test coupon boards. \$\endgroup\$ – Chris Stratton Jun 14 '15 at 18:05
  • 1
    \$\begingroup\$ What you schetch (assuming out-of-the-box suggestions like "buying quality components" are off-limits) is a statistical problem. But you will need some figures, like and estimate of the distribution of failure rates for all components, and cost of a failed board versus cost of testing more components, to do a meaningfull calculation. \$\endgroup\$ – Wouter van Ooijen Jun 14 '15 at 19:14
  • 1
    \$\begingroup\$ There is no "different complicated thing". It is pretty simple - purchase crap unknown materials or trusted materials from reputable known suppliers. \$\endgroup\$ – Michael Karas Jun 14 '15 at 23:11
1
\$\begingroup\$

Unfortunately, the situation you describe does occur for various business reasons. In your question you don’t indicate if the parts are SMT or through hole. In today’s day and age it is hard to imagine widespread use of through hole parts. So I am assuming the manufacturing process is SMT.

As widely discussed by others, it is standard for all manufacturers of electronic components to ship parts that are tested and are in specification. I must say that I have encountered electronic component manufacturers shipping low cost untested parts. Some ship a percentage of parts above the order to compensate for the defective parts in the shipment. In almost all situations it is best to purchase reliable electronic components. In almost all cases the cost outweighs the benefits.

In the event one encounters such an undesirable situation, a combination of components testing and the use of statistical tools can benefit greatly. Let’s assume that you have to verify and validate a 10,000 piece batch of 1k resistors that has tolerance of 10%.


Method

  1. Select random sample of 30 resistors
  2. Measure and tabulate the data. Below is random set up values to help illustrate the concept. \begin{array}{| c | c | c | c | c | c | c | c | c | c |} \hline 809 & 1117 & 1059 & 1162 & 1045 & 939 & 1043 & 997 & 879 & 1144\\ \hline 1060 & 899 & 959 & 968 & 1004 & 1127 & 886 & 817 & 1190 & 1095\\ \hline 948 & 1144 & 1167 & 933 & 865 & 1136 & 947 & 955 & 902 & 1038\\ \hline \end{array}
  3. Calculate the average, and standard deviation.
    Average:= \$1007.8 \Omega\$
    Standard Deviation:\$=110.44 \Omega\$
  4. Calculate the, 1-\$\sigma\$, 2-\$\sigma\$ and 3-\$\sigma\$ values \begin{array}{| c | c | c | c |} \hline & \text{Low limit }(\Omega) & \text{High limit }(\Omega) & \text{Percent} \\ \hline 1-\sigma & 936.7 & 1049.1 & 68.27\\ \hline 2-\sigma & 880.5 & 1105.3 & 95.45\\ \hline 3-\sigma & 824.3 & 1161.6 & 99.73\\ \hline \end{array}
  5. So what it means for the data set is that \$68.27\%\$ of the population will fall between \$936.7 \Omega\$ and \$1049.1\Omega\$. Similarly \$95.45\%\$ of the population will fall between \$880.5 \Omega\$ and \$1105.3 \Omega\$.

Standard Deviation

  1. Based on this type of simple analysis you can accept or reject the parts. This type of decision making is cost effective and will greatly benefit your client.

A similar question and response supported by mathematical equations can be found at "How can the vendor be approved/disapproved based on validation test data?".


References:

\$\endgroup\$
  • 1
    \$\begingroup\$ great , but the only question that remain is that which parameter should i check ?!! for example about resistors u just said about the value , but we know that a bad capacitors usually have a leakage that is the main problem not the calculated capacitance of it in the first time , mainly i have to test diodes , capacitors , resistors but till now the main problem was capacitors \$\endgroup\$ – Majid khalili Jun 21 '15 at 10:33
  • \$\begingroup\$ Here youtube video of how test capacitors? youtube.com/watch?v=TFa6JfVu3B4 You can employ the same strategy discussed the the post for analysis. How about measure 30 capacitors of one kind. Post the data. I suggest that you post new question with real measured values, and we can do the analysis via SE. The engineering SE question an example your situation \$\endgroup\$ – Mahendra Gunawardena Jun 21 '15 at 10:49
3
\$\begingroup\$

It sounds like you have two distinct problems:

1) Incoming inspection

2) Production testing

Incoming inspection is easy but tedious: you simply test some percentage of the components as they arrive in your facility. If you are using SMD devices and the components are inexpensive, you simply discard the tested samples afterwards (or put them on eBay as bagged, loose components for prototyping use).

If more than a certain percentage of that batch fails testing, you reject the entire batch.

For example, if you are testing SMT capacitors that arrive on spools, you test parts from every single spool that is part of that batch.

This is painful but if you can't trust your supplier and cost drives your purchasing, it's about the only method that you have.

Production testing is somewhat similar in that you test some percentage of your final production.

Some companies test 100% of their production - the extra cost is offset by the high reliability that those products require. Others test only a small percentage of the production run but if there are too many failures, every unit from that production run is tested.

Some companies don't do any testing at all. These are generally much larger companies that are producing products by the millions. Instead, they rely upon having quality components and a variety of statistical process measurements.

\$\endgroup\$
  • \$\begingroup\$ +1 Sometimes distributors get the parts mixed up- it can be funny at times (a switch marked as a terminal block) but it is not amusing at all with small SMT parts that have no markings. If you're buying less than a reel you're more vulnerable than if you're buying whole reels. \$\endgroup\$ – Spehro Pefhany Jun 16 '15 at 20:26
1
\$\begingroup\$

Hmm, as far as I know there are no PCB manufacturers that test the components they use before mounting. Why not ? Because that would be very expensive. They rely on the manufacturer to test the components. How long do you think a manufacturer of untested (and therefore sometimes faulty) components will stay in business ? Not long !

Imagine SMD components in a reel, taking them out, testing, putting them in a reel again ? Because how else will you feed those components to a pick and place machine again ?

So yes, complete motherboards with hundreds of components are soldered with manufacturer tested components ! Of course the PCB is tested before (sometimes) and after assembly.

So if you go the "test all components" way then I think you are doing something wrong.

What if the components are 100% OK but they break during manufacturing of the PCB ??

You should pinpoint the problem and use common sense to solve that.

\$\endgroup\$
  • \$\begingroup\$ I'd bet the military does 100% part verification, but they're not exactly the epitome of financial minimalism. \$\endgroup\$ – Connor Wolf Jun 15 '15 at 4:02
  • \$\begingroup\$ Sure they even have their own "military spec." that says enough. Indeed I was only talking about consumer/low cost products. \$\endgroup\$ – Bimpelrekkie Jun 15 '15 at 7:40
  • \$\begingroup\$ it was interesting , i i didn't know about it , i always thought if u want to send something to the market u should have test it and make sure that everything is fine , here in this company the just rely on the be fore mounting test , and they believe that the cost of testing doesn't worth the time and they just exchange any damaged product returning from market , i really didn't know its an option \$\endgroup\$ – Majid khalili Jun 16 '15 at 5:15
  • \$\begingroup\$ No testing of end products ??? Geez, obviously they don't care about their reputation among their customers. Also this way they have almost no way to improve their manufaturing process. From the field returns you cannot tell if it was a manufacturing mistake or that something went wrong at the customer. What a strange attitude this company has. I would run away screaming !!! \$\endgroup\$ – Bimpelrekkie Jun 16 '15 at 6:22
1
\$\begingroup\$

This is not an uncommon problem, actually, but one which many engineers have not encountered.

Some Chinese manufacturers offer 'B' grade parts at a large discount. Many parts are out of spec and some don't work at all. This is apparently an attractive bargain to some South Asian customers who have much lower labor costs than in China so they can be sorted out. Some US IC manufacturers offer similar untested dies, sold in China. Even complete modules with low quality are offered- the user is expected to weed out the bad units and rework many of the ones that are not working. The difference between a high quality 50 cent product and a low quality 20 cent product can be significant, and chances are that not every customer cares if the product stops working at high temperature or low supply voltage etc., or if one LED out of a 9-LED flashlight quits, or if an indicator LED is very dim but still visible.

They really need to do an analysis of whether 100% incoming inspection of some parts, or thorough testing of boards or buying better quality parts makes sense. A quantity of 12,000 per year is pretty small by most standards. Without using parts with a high AQL the quality of the product will always be suspect- it's axiomatic that you cannot test quality in. If the parts came from a batch with 70% fallouts at the manufacturer and 80% of the remaining parts barely meet the specs that you can test for, eventually you're going to see some really big failure rates. The high failure rate at the manufacturer can mean that some part of the process is not in control so you may get intermittent failures and/or early failures.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.