02-26-2021 04:23 PM
Dear forum members,
Please, help me.
I'm building my prototype board with Spartan6 SLX75 devices. To cut the expenses, I've bought aforementioned ICs at Chinese broker via AliExpress (a very nice person). When I've mounted both chips initially I thought that they were damaged during the reflow. Also initially I couldn't find anything abnormal apart from Impact refusing to program the IC saying:
Identifying chain contents...INFO:iMPACT:1588 - '0':The part does not appear to be Xilinx Part.
But after all the basic tests (diode connectivity, current consumption, all the config pull-ups, power supplies, ...) the device seems functional but not accessible (doesn't even attempt to pull the data from the SPI config flash - no clock). It's not the signal integrity issue, since I've changed the clock speed, termination etc on JTAG interface...
So, I switched it to the slave config (pulling both M0 and M1 high) and read the device ID:
on the both SLX75 chips show the following ID:
0x0406E093 instead of 0x0400E093. By looking at the UG380 page 87, table 5-13, I can't identify the chip with the same signature.
By using the decoding method by breaking the code on V-field (revision), F-field (family code), A-field(subfamily), B-field(device identifier), C-field(company code) we have the following:
So, the device subfamily says 0x3 instead of 0x0.
Hence there are 2 questions:
1) Does anyone came across such a device number? Do you know what was the device purpose (I hope it's not the unobtanium mil spec or aerospace devices). Obviously the code is unknown to IMPACT tool.
2) Is there a possibility to generate the configuration bitstream with this code and/or somehow adapt it to the ISE 14.7 toolchain in order to be able to configure the device?
Thank you and all the Best
02-27-2021 04:44 AM
I have bad news..... The chip is the counterfeit. It was removed from the board, re-balled and etched and remarked. After rubbing the top package with the Acetone I can see the original writing under the new one:
Chip as sold to me:
After rubbing with Acetone. As you can see the original "TAWAN" is visible above the new fake "TAIWAN"
The "SPARTAN" is written over "SPARTAN"
Last thing is to read the actual grade of the device, since it was removed from some equipment (likely military) and re-balled
02-27-2021 08:44 AM - edited 02-27-2021 08:45 AM
Well done for getting back,
I am sorry you have found the hard way that there are very very many counterfeit parts out there,
Unfortunately the machines used to label parts are readily available,
The only answer is to purchase from the autherised distributors,
You were in many ways lucky,
it failed at the first step,
Many times parts only fail when hot / cold or at extremes of the voltage supplies.
which is much harder to debug, costs a lot more to sort, and looses you customers really quickly.
02-27-2021 11:47 AM
You know, this is a very first experience with fake chips in my case. They are not only reconditioned, but polished, remarked, reballed and repackaged. I've read about fake chips in a press, but never experienced myself. And it looks like the fake production now goes at the industrial scale. IMHO the seller didn't know himself that the chips are fake. He was certain and sure that they are not (the seller is a big business).
AliExpress was my reliable source up to now. But I've "burned my fingers" badly now. In total I've spent around 10 hours to figure out what's going on. I made lots of designs with smaller Spartan6, but this design has the first FPGA of this kind. So, I've suspected the absence of the chip pre-baking, not controlling well the profile. Then not accepting the failure, I took the ohmmeter and checked every ESD protection diode to make sure that everything is soldered correctly, manipulated in all the config modes, recorded the power rails during the startup condition etc. The only thing I haven't done - XRAY. And finally so simple answer - polished and laser marked recycled chip.
Also the counterfeit makers badly screwed up with their "fake quality":
- size of the BGA balls was slightly bigger than nominal - 0.65mm-0.7mm instead of 0.55mm-0.6mm (I was a bit on alert when I saw it under microscope the first time)
- balls weren't just dull matte but with crystals and not very evenly shaped.
- Since this is the overmolded top side die BGA package, the PCB part of it is fragile, but I had an impression that the PCB layers were dilated in corners
- at the end the total failure of selecting the wrong chip as the donor of JTAG identity.
If the above factors were not visible and obvious, I would never question the chip quality.
And that's great that the fake was detected at the very first step as you have mentioned. I can't imagine what nightmare I could get into if I would have some chip performance problems later on.
Now finding the right and low cost samples is a quest on its own. It's not a problem for big companies, but almost impossible for the amateurs and individuals since the price of a single sample is $$$$ exorbitant.
Looks like I've to start the "sample" program to find a good supplier and a good lot of chips to keep my prototype project in budget.
For me this is the new reality and I don't have a good recipes ready.
02-27-2021 11:57 AM
I'm afraid its only going to get worse
the next few years are going to be interesting,
I'm receiving quotes for some chips ( not Xilinx ) with 80 weeks lead times,
The industry went through the same in the 80's, and survived,
chips on allocation, placing orders for 10K of something and only being allocated 5K,
but we were a lot less dependent on chips then,
In the past, I remember receiving some chips that just did not work,
they turned out not to have any silicon inside them,