Author Topic: Interesting fault using PCA82C251 CAN- Transceiver, defective from the supplier?  (Read 1333 times)

0 Members and 1 Guest are viewing this topic.

Offline Electr0nicusTopic starter

  • Regular Contributor
  • *
  • Posts: 131
  • Country: at
Hi

I have made a PCB which uses a AT90CAN64 microcontroller in conjunction with a PCA82C251 https://www.nxp.com/docs/en/data-sheet/PCA82C251.pdf CAN- Transceiver. I have finished populating the board by hand (because it's a prototype) last week and wanted to test it this monday.
The board is powered by a Linear Technology/Analog Devices LTM8052 µModule regulator, configured to 5V/5A output.
Because it's a first test, I tuned to current limit of my lab power supply to 600mA and the voltage to 6V. Why 600mA, because the LTM8052 needs some current to get it's things startet, and this is a value which I found is a good starting point, from previous projects, where I also used the LTM8052. The current draw should immediately fall back to <100mA in less then 3 seconds after startup.
In this case, the current draw persisted at 600mA and the voltage was pulled down to odd 2,5- ish volts, but not fluctuating. The output of the LTM8052 was only 0,12V. So I tought, well maybe there is a solderbridge beneath the LGA package of the LTM8052, although I do my soldering IPC complient. So I desoldered the LTC8052 and as the chip was desoldered, I decided to measure the output directly with an  ohm- meter. I was suprised to measure only 0,17\$\Omega\$ on the output.
So following to this, I removed the ceramic output capacitors, same thing.
Then I thouroghly inspected the PCB again for solder bridges, as I did the week before when I finished soldering it, but maybe I've missed something. But I couldn't find anything.
So, as I have a thermal imaging camera handy at work, and because the nature of the short was very low impedance and the same in both directions, I decided to pump current into the fault and see on the therma lcamera what is getting warm. First I upped the current limit to 4A, and hardly could see anything getting warm on the PCB. This can be due that it is a 4 Layer board with 70µ copper on each layer, and the VCC and GND layer are literally a floodfill across the whole PCB.
Then I increased the current to the maximum my PSU can deliver, which is 10A. When I activated the output the voltage displayed on the PSU was around 0,3-0,4V and current indeed 10A. With the thermal camera, I could see the whole PCB slowly heating up, but also a tiny spot in the vicinity of the PCA82C251 CAN-Transceiver getting warm, but everything <40°C because of the low resistance of the short.
OK I thought, let's desolder the CAN- Transceiver and see what we are up to.
After desoldering the CAN- Transceiver and switching on the PSU, i was really suprised that the current draw was a around 30mA @ 5V and the short was completly gone.
So I've found the culprit. I could not explain why the chip was shorted, and how it was possible that it was shorted that completly. I had unintentionally destroyed many chips in my career, and the short they had always was much higher impedance and also extremly temperature dependend. I also checked that I've used the correct pinout in the schematic and PCB- footprint.
Then I measured the desoldered Chip and found the following:
  • VCC-GND: 0,05\$\Omega\$ = SHORT
  • VCC/GND-CANH: 0,06\$\Omega\$ = SHORT
  • VCC/GND-CANL: 0,06\$\Omega\$ = SHORT
  • CANH-CANL: 0,05\$\Omega\$ = SHORT
So this shows, the chip is a complete short. The other four pins showed values in the tens of k\$\Omega\$s largely fluctuating.
Then I took a new chip directly from the reel and measured it directly with the multimeter, without running it in circuit. I was gobsmacked that i measured identical values to the 2nd digit after the comma.
  • VCC-GND: 0,05\$\Omega\$ = SHORT
  • VCC/GND-CANH: 0,06\$\Omega\$ = SHORT
  • VCC/GND-CANL: 0,06\$\Omega\$ = SHORT
  • CANH-CANL: 0,05\$\Omega\$ = SHORT

So can it be, that I got defective chips right from the factory/distributor(Farnell). Is this even possible? I don't know the chip industry that much, but isn't there a 100% test after they get shipped to the customer? This is the first time this occured to me.

I added the schematic and layout screenshots FYI.
« Last Edit: April 15, 2019, 08:24:12 pm by Electr0nicus »
 

Online Ian.M

  • Super Contributor
  • ***
  • Posts: 13216
Its not impossible for counterfeits to get into the supply chain, even at reputable distributors.   Scenario:  Shady customer buys a reel of expensive genuine ICs, makes sob story that they misordered,  and need a RMA to return them for credit (less supplier's standard restocking inspection fee).  Returns fakes, re-reeled onto original genuine reel in expertly resealed original packaging.
 

Offline HwAoRrDk

  • Super Contributor
  • ***
  • Posts: 1607
  • Country: gb
Sounds like you got completely fake chips, that have no silicon, just a leadframe with every pin shorted to every other.

By the way, your schematic and layout as-shown don't seem to match: on layout you have RXCAN connected to TXD, and TXCAN to RXD - the wrong way round. Hopefully this was fixed before you had the PCBs made. I assume so, as the schematic is correct. :)
 

Offline Electr0nicusTopic starter

  • Regular Contributor
  • *
  • Posts: 131
  • Country: at
Sounds like you got completely fake chips, that have no silicon, just a leadframe with every pin shorted to every other.

By the way, your schematic and layout as-shown don't seem to match: on layout you have RXCAN connected to TXD, and TXCAN to RXD - the wrong way round. Hopefully this was fixed before you had the PCBs made. I assume so, as the schematic is correct. :)
Actually they were back to front. Also on the PCB. But because that is only a one maybe two-off prototype it is no big deal. I was currently in the process fixing the issue, when I decided to write this post, therefore the schematic is correct, but the PCB isn't.  :palm: I have testpads nearby, which i can use. So I'll have to cut the traces and connect a bodge wire from the testpads to the CAN Transceivers pins.

Yeah the thing with the chip is interesting. As there is actually a dead short between the pins 2,3,6,7. The other pins show some resistance in the k\$\Omega\$s- range when measured to GND. So maybe they are connected somewhere.

Maybe I will try to depot the chips with 65% nitric acid, which I also have access to. Would be interresting to see what is going on inside. If I have the time I will give it a try i think.

I've ordered a pin compatible replacement Transceiver (MCP2551), and not a big wonder it works just fine, no short whatsoever  :-DD

This guy on YT had the same problem in some sort, but worse for him, with much more severe consequences. Interesting to watch! :-+
« Last Edit: April 16, 2019, 09:54:52 pm by Electr0nicus »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf