09-04-2019 05:56 AM
We bought 10 FPGA boards (with xc7a100) from an FPGA board manufacturer in Germany.
In this small batch we had 4 pairs with the same FPGA DNA ID!
We know that up to 32 devices worldwide can have the same FPGA DNA ID, but for this small batch it is strange to have such a high number of identical IDs.
Is it true that there are a maximum of 32 devices worldwide with the same FPGA DNA ID? Does Xilinx mix FPGAs (with the same ID) before sending them to customers?
Of course there are other solutions for higher security (e.g. EFUSE_USR ...), but we wanted to keep the configuration effort for our product as low as possible! Xilinx produces a lot of these FPGAs and only 32 of them have the same ID. Therefore the probability to get the same ID multiple times should be very low! (We got 4 pairs out of 10 FPGAs)
Are these ID clusters are normal?
09-04-2019 07:15 AM
I suspect these are early silicon, or engineering samples. Look at the date code and markings. As they are probably not purchased through an authorized Xilinx distributor, you will be unable to get help from Xilinx. As the DNA is on efuse, that are blown in test at TSMC, it is also possible that someone programmed un-blown efuses after receipt.
The value programmed is supposed to be wafer lot number (n of 25), XY position of die on wafer, as well as other information. It is not unique, but 10 of the same code, in one small lot? Coincidence is unlikely at best, or as I suggested, non-production engineering sample silicon.
09-06-2019 12:20 AM
the fpga's of our modules are from Avnet Silica an authorized Xilinx distributor.
You expect 10 time the same code, but Florian has wrote 4 pairs with the same FPGA DNA ID. So not all 10 are the same. It's only two the same, but it's occurred four times on his order.
I've told Florian, he can check LOT code of the FPGA package (https://www.xilinx.com/support/documentation/customer_notices/xtp424.pdf), but unfortunally he has orderd with heatsink, so he must removea a head sink to see the LOT of the package.
We have also contact Avnet and ask if they have some more information about DNA.
I havn't any experiance with DNA number fo the FPGA, I only read this one:
From UG470 page 122:
And I found this posts:
But there's somehow no definite answer.
It seen you has more expericance. Do you know what Xilinx means with "up to 32 devices within the family can contain the same DNA value". What's the case that they are same? If it comes from the same production batch? Various or same production sites? Or is it really a coincidence?
09-09-2019 01:26 AM
I've got response from Avnet Silica.
Please check FUSE_DNA (64Bit) and not DNA_PORT (57Bit).
Fuse_DNA is a 64 bit unique identifier in 7 Series and Zynq devices.
The DNA_PORT is a 57 bit value that can be shared with up to 32 devices.
The DNA_PORT is a sub-set of the FUSE_DNA value.
DNA_PORT [0:56] aligns to FUSE_DNA [63:7].
This means that bits 63,62,61...8,7 of FUSE_DNA aligns to bits 0,1, 2...55, 56 of the DNA_PORT.
09-09-2019 06:45 AM
As the DNA has wafer in lot (1-25) and XY die posisition, and week of year date code, as well as a final test code, it isn't unlikely to have identical DNA on high volume parts. A7 is a high volume family (least expensive), so this could be the new normal. All I know is that before I left, it was a rare event that had not happened yet (but as no one does anything to prevent it, it was possible). The extra USER efuses are there so one could add a sequence number, which would when combined with the DNA make the device unique (at least in your production). The DNA is there so Xilinx can track the device back to thge die probe, final test, in the event the device ever comes back for failure analysis. It is unlikely that Xilinx would ever get a device back that they could not uniquely identify.