06-28-2018 07:59 PM
I've seen other posts about the resistance from VCCINT to GND being low (0.3 ohms) and I understand why it is so low. Does Xilinx have a recommended manufacturing test to differentiate between a normal board and a defective board that has a short circuit? Our CM cannot detect the difference and the only test is to power it up and see if the regulator goes over current. This seems dangerous and I'm looking for suggestions for another way to test for shorts in a production environment.
06-28-2018 11:09 PM
We can't detect a difference either on the boards that have large FPGAs, so our CM does the same thing as your CM.
The resistances are so low that you'll need to use 4-terminal measurements at multiple test points across the board if you want to find a short.
06-29-2018 07:13 AM
A properly designed power supply will not damage anything,
Verifying power is good is as simple as checking that the voltage and current is what is expected when the board is powered up.
Contact Avnet for all the power supply designs for your application: designs are free (and verified), including bills of material, layout.
07-01-2018 06:47 PM - edited 07-01-2018 07:46 PM
> A properly designed power supply will not damage anything,
Before making broad statements like that, you might like to consider that for a large FPGA it's difficult to tell the difference between an unpowered FPGA core and a short circuit, and also that the power supply might need to be able to supply a current of many tens of amperes.
For example, the biggest FPGA core supply I've ever designed had a short circuit current limit of about 100A.
That said, some of the newer multiphase controllers can have a very rapid response to overcurrent events and other faults.
[Warning: anecdote] I had one production fault that involved a short across a switching inductor in a multiphase supply. The pullup FET in the poor DrMOS driver IC was trying to connect +12V directly to the FPGA core. It only did this very briefly (one or two switching cycles, IIRC) before being shut down, and everything survived and worked once the short was repaired.
That one had single turn inductors, and no (static) measurements I took could tell the difference between a properly working circuit and one with a short. We had to use X-Ray to find the fault.
07-02-2018 06:51 AM
Good points. We are entering a new era of 100+ watt core power levels in the larger, newer technology parts. I agree that this is new territory for FPGA devices, and feel that at these power levels, the difference between a 'short' and a 'working device' is difficult to detect.
I encourage engineers to take advantage of known working and tested power supply designs available through distribution. Xilinx has always worked closely with their power partners to provide a reliable solution.