02-01-2018 04:55 PM
Hi Guys. So this is a real head scratcher. Or i may be making a silly mistake.
I have a custom board running a Spartan3AN, and it's running some IP fine, that was first confirmed using a commercial Spartan3AN demo board. The FPGA usage is at 80% at most based on programmer output reports.
The FPGA also 'flashed' fine, and now boots from internal flash on power up - there is no external memory.
However, if i make a non consequential change to the IP code, it complies through to download, completes download to FPGA, but then i get a 'failed to program' error message and FPGA stops working, or it reverts to the code in flash.
THEN, if i un-do that inconcequential change to IP, and once again download to FPGA, the FPGA programs confirms and runs fine once more.
I can change the code for the imbedded processor, but not the ASP block, even if i make the block much smaller. All changes are confirmed to work fine on the reference Spartan3AN board.
So i'm baffled but working it out, and hoping i can get a second opinion or two on what it could be. Is this sort of failure mode indicative of anything?
This behavior persists after PC and dev board power re-cycling.
Specifically, if I change the ASP function from that in the flash, and that I first programmed the FPGA with, it doesn't program, until I change the ASP back to the way it was.
I do get a complete download, from compile of embedded through downloading to FPGA, without errors, but the FPGA doesn't work - 'done' signal never comes on, and I usually but not always get a pop up that reads 'Failed to program device:_____'. At the same time, the same ASP changes can be changed as expected on the reference Spartan3AN board without issue.
If nobody has any ideas i'm thinking i'll try erasing the boot flash, but i'm hesitant to do that when I don't know what the issue is yet, and the board is running an early version of the project at this time. Got any ideas what it is? Boot flash packed too full or something? Hardware non-recoverable failure?
Meanwhile I will keep working on it..
Thanks and Cheers,
02-01-2018 09:26 PM
02-01-2018 10:57 PM
Thanks for your help, kkn.
Here is the device batch and date code:
Purchased from Abacus.
I also have 40 XC3S1400AN in Xilinx vacuum seal, purchased from Digi-Key, which may be from a more recent batch if it makes any difference, but they are yet to be populated on their PBCs.
I'll attach the process log file when I run the next test in the morning. Thanks.
02-02-2018 02:43 AM
thanks for the details, impact log file will provide additional info during programming failure
initially i though the following AR might help, but your device is 1400AN
can you also check the timing report for the failing and passing design and see if there is any critical warning or timing issues.
02-02-2018 04:57 PM
Here is the timing analysis attached. The report is identical for demo board and for custom board. I can also check the other reports, synthesis etc. but they also look similar. and no errors in the bit gen file.
There is only one timing error, and it's based on not being able to read the input pin frequency rather than actual fail. This is an external input signal that appears to be working fine on both boards.
02-05-2018 12:10 PM
So, according to the update, only Flash is affected, and rarely on the 1400 model.
Is it possible that the state of flash memory in the FPGA could prevent subsequent programming of the device? I'd like to rule this out, if someone knows it can be ruled out. It doesn't seem like the malfunction would be related to what happens to be in flash.
Does the Xilinx software programming tools care what's in flash when updating FPGA?
The answers may help as I sort this out.
02-06-2018 12:57 PM
Also, i'm wondering if the 'change advisory' in the link may be involved somehow, but i'm not sure it can be, as the device did flash properly once, boots, and still programs but for a change in the ASP section of C code. https://www.xilinx.com/support/answers/59572.html
|Potential impact for a small number of applications||
For most applications, the new in-system flash is backward-compatible with old in-system flash and its original programming solutions. However, see the Extended Device Information Change Details section for differences in the number of extended device information bytes.
Also, I am using ISE WebPACK 12.1 , invoked through Altium Designer 10. This is the same platform the project was started on and has not changed during development. I think you may be requesting the 'Impact' log file, and that may be a more recent development I don't have, but perhaps I could attach the reports from WebPack 12.1 that are generated.
WebPack 12.1 predates these in-system flash changes.
Are these in-system flash changes reflected in newer versions of WebPack, and/or Impact?
02-06-2018 09:05 PM
looking at the device top marking you sent the design advisory may not be applicable.
Regarding the programming of the internal flash, if you have not write protectected the flash then with impact 14.7 you should be able to program.
can you also check if the ISF is in binary mode or power of 2 addressing.
All the details you can check in ug333
02-06-2018 11:42 PM
Thanks, kkn. This sounds hopeful, but I am not familiar with 'Impact' - the download page for Impact 14.7 indicates that it is for Spartan-6 only - so i'm a bit confused - I am accustomed to using WebPack 12.1 and am not familiar to how it differs from 'Impact' software - can you clarify? Thanks.
02-07-2018 01:08 AM
The download link which you are referring to is for spartan-6 only for windows 10.
Here is the link for lab programmer(impact) for windows7/XP
02-07-2018 11:30 AM
Great. Thanks, Krishna. I am at the page in your link. It is asking for ID to enter, but is not accepting my phone number - I have tried entering normal and with only numbers. It always reports - 'not a valid phone number' - however the number is valid. I tried my head office number and my lab number. I have tried contacting support by phone, but cannot raise any humans to help me out at the support number - 1 800-255-7778 . Any idea how to fix this issue? Looking forward to 14.7. Thanks again.
02-13-2018 02:08 PM
Success. So I have the ISE WebPack 14.7 downloaded now and ready to install. I got a real nice confirmation from Xilinx too.
Time to begin installing. I'll report if I have activation issues. Thanks for your help.