11
\$\begingroup\$

I recently received a run of 10 PCBs from a manufacturer in China, and am worried they have begun to cut corners and source counterfeit parts. Here is why:

I had them do full turnkey production (PCB Fab, component procurement, assembly). I've used them in the past and they've been pretty good, albeit with an occasional mistake.

I noticed on 4 / 10 boards, the circuit below not behaving as expected:

schematic

simulate this circuit – Schematic created using CircuitLab

On the faulty boards, whereas we expect the gate voltages of Q3 + Q5 to be either ~0 V (if NOR output = 1 (5V)) or ~12 V (if NOR output = 0 (GND)), the gate voltages were anywhere from 3-7 V...

Here is why I'm suspicious of the parts:

  1. We've used this exact circuit in previous iterations of the PCB, with the same manufacturer, and haven't seen this issue. Only changes are minor difference in PCB layout.
  2. After I hand-removed Q1 Q3 and Q5, and replaced them with parts I had from Digikey, the circuit functions as expected. I've done this on 3 boards, and all 3 went from non-working to working.

The relevant NPN + PMOS part numbers are given below, here are datasheet links: DMP3010 MBT2222

Alternatively if something seems fundamentally wrong with the circuit, I'm all ears. But it is a pretty common, simple circuit, and as mentioned one I've used in previous iterations with no problem.

\$\endgroup\$
  • 8
    \$\begingroup\$ You did the wrong thing in replacing all three components. Know you don't know which of the three is the true fault. Assuming even if they used fake FETs and that they are still insulated gate FET devices, it would seem more likely that the Q1 is your problem. Your R1 value is rather high (10k) so I'd imagine the most likely problem is that the Q1 leakage is high. This might be caused by overheating during soldering and perhaps they have a PCB reflow problem and are hand soldering as touchup/repair. \$\endgroup\$ – Jack Creasey Oct 4 '17 at 3:26
  • 7
    \$\begingroup\$ 10 boards may be hand assembled depending on complexity, may just be human error in placement / soldering \$\endgroup\$ – sstobbe Oct 4 '17 at 3:51
  • 1
    \$\begingroup\$ Do you have any kind of test fixture to generate performance graphs of the components? If you can consistently find deviations from properly sourced parts, they are counterfeit. If they just behave like broken components, they are broken components. \$\endgroup\$ – PlasmaHH Oct 4 '17 at 13:52
  • 1
    \$\begingroup\$ @Jim yes visual/continuity/resistance etc. Note your design has a power sequencing requirement of 12V before 5V otherwise the gate pull-up voltage is undefined and may half turn on Q5 causing it to burn up \$\endgroup\$ – sstobbe Oct 4 '17 at 14:12
  • 1
    \$\begingroup\$ @Jim. When you next have a faulty board with this type of fault ... 1) Short Q1 collector to ground ...if the output switches turn on then they are functioning ok. 2) if Q1 collector does not rise to 12 v, remove Q1. If the voltage rises to 12 V Q1 is leaking, if the voltage on R1/Q1 collector track does not rise, then Q3 or Q5 is bad. I'd also suggest you need to change the value of R1 to about 1-3k Ohm in future builds. \$\endgroup\$ – Jack Creasey Oct 4 '17 at 14:20
11
\$\begingroup\$

You have to prove the parts are bad. They may be damaged by ESD.

If you had only pulled the Q5 or Q3 and measured V(Q1-C), that would have isolated the part as the problem. Then verify R1 is 10k and not 10M or something else.

The only weakness on the design is that the circuit turn off is slow and load reactance is unknown.

Normally any FET (such as this one rated for 8mΩ @ VGS = -10V) is driven by a gate driver resistance of about 1000x (8Ω as used in datasheet) but your R1/RdsOn ratio is about 10 million. This makes it slow and prone to oscillations with stray inductive/capacitive feedback to gate voltage depending on layout.

  • also you are putting ~(4V-0.7)/1k=3.3mA into base and and capable of driving >>100mA into gate capacitance Ciss, until Vce is saturated. But pullup to turn off is only 12V/10k = 1.2mA which results in spurious turn off behavior. More design margin would use 1k for R1 at most.

Conclusion:

Test FETs for ESD damage, leakage in gate as per above. Reduce R1.

There is no assumption as to if/when the ESD damage occurred.

Article on Clone parts.

http://www.sae.org/aaqg/audit_information/2010/Atlanta/Impact%20of%20Counterfeit%20Parts%20NASA.pdf

\$\endgroup\$
  • \$\begingroup\$ Thanks Tony. I used a higher R1 to minimize power consumption, and because switching speed isn't a performance requirement for us; this is a reset switch used for a few seconds every few hours. However, I didn't consider the protection aspect. I'll have to take closer look with the scope. \$\endgroup\$ – Jim Oct 4 '17 at 13:55
7
\$\begingroup\$

I suspect they may have substituted a MOSFET that has low voltage gate protection zeners for the DI parts. There is no motivation to sub Q1, they're dirt cheap in China and any similar part would work as well. The MOSFETs, on the other hand, are expensive.

Your circuit switches off slowly so it could put a lot of stress on the MOSFETs if the loads are low impedance, but that would not likely cause the observed effect (though it could conceivably under some conditions).

If you have another misbehaving board try swapping out just Q3 if you want test my theory. You can also contact DI with a high quality photo of the part markings and ask them if they match parts made for sale anywhere in the world. Of course you can compare them visually with the parts you bought through distribution but there are often multiple packaging facilities used and the markings may vary somewhat, especially (but not exclusively) for very different date codes, so a difference is not conclusive. If they look exactly (to a critical eye) the same including method of marking, font, and small features on the transfer molding, it's a pretty good indication the parts came out the same factory.

For such a tiny quantity they probably sent someone down to the market (in Shenzhen at Huaqiang bei lu) and got whatever parts were available from one of the many retail vendors. If you want to be 100% sure, send your own parts to them, especially if the quantities are modest).

\$\endgroup\$
  • \$\begingroup\$ Great advice. A couple follow up questions: 1. what would you consider low impedance in this context? The 5 V Fet is switching anywhere from 25-750 mA, the 12 V FET from 100 mA - 5 A..... 2. I think I'll follow up with DI. 3. Good thinking regarding sending our own parts; we've done this in the past and haven't seen the issue. \$\endgroup\$ – Jim Oct 4 '17 at 13:56
  • \$\begingroup\$ Given that I haven't gotten to the root of the problem yet, should I consider this problem unanswered? Both you and @Tony gave great, actionable advice but I don't know for sure at this point. \$\endgroup\$ – Jim Oct 4 '17 at 14:01
  • \$\begingroup\$ If the power dissipation in the MOSFET is excessive during the switching. From what you say about I doubt it's a problem. You can select whichever answer is more helpful or none at all, or defer and do it later. It's up to you. \$\endgroup\$ – Spehro Pefhany Oct 4 '17 at 16:00
0
\$\begingroup\$

I bet you do have a parts issue, However the issue is likely due to your MBT2222. A widely used generic N2222a transistor is likely being swaped and that has been fabricated by hundreds of manufacturers to who knows what specifications .

\$\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.