Quartus appears to sometimes, incorrectly report Blank Check Failures

[Randy1]

Our production programming station has reported intermittent blank check failures even though they have performed the erase operation prior to the blank check. The JIC file was read back from the suspect device, EPCS4, after the blank check error was reported. For the suspect case, the entire 512Kbytes of memory shown in the JIC file were blank. It appears that the Quartus Programming software itself has a bug that sometimes incorrectly reports a Blank check failure. The Quartus version is "19.3.0 Build 222 09/23/2019 SC Pro Edition". The FPGA in question is a Cyclone 3 device. It would be good to know if there is an errata for the programming tool to explain the apparent incorrect behavior.

[JohnT_Intel]

Hi,

May I know which blaster are you using to programmed the device, error message shown in Quartus Programmer and could you share with me your full JTAG chain for your board?

[Randy1]

Hello - I am using the USB Blaster in my Engineering setup which I believe to be same as the one that Manufacturing test is using, but I will confirm this. Attached is the JTAG cdf file. The error they described was just Blank Check failure without too much more information, but I will follow up with a screen shot.

[JohnT_Intel]

Hi,

Thanks for the cdf file. I will wait for your screenshot for further investigation

[Randy1]

Can you tell me if there have been other reports of the Blank Check failing when the memory device was confirmed to have been blank?

[JohnT_Intel]

Hi,

I have checked the databased and unfortunately there is no such reported case.

[Randy1]

Update - attached are the files from our production test setup. The text file is a log of the errors observed and the graphic file a screenshot of JTAG chain for the same production setup.

[JohnT_Intel]

Hi,

May I know if you are facing any issue when directly programming the new jic file without the blank check?

Have you try to performed erased 1st and then only performed blank check? In 2 different flow

[Randy1]

Hello, It turns out that the production test folks have tried the two things you mention. 

1) If they Program and Verify only (with no blank check), they have not observed any failures.

2) If they Erase and then Blank Check over and over, they observe intermittent failures. 

In at least one of the cases where they did and Erase and observed a Blank Check failure, they read back the device and were able to show that it was actually blank.

[JohnT_Intel]

Hi,

It looks like there is some issue with the blank check. I will feedback this to the engineering team and see if we could fixed it in the future Quartus released.

For workaround, I would recommend you to not use the blank check features.

[Randy1]

Hi John,

We have come to the same conclusion that the Blank check is not reliable and is not necessary and can be removed from the production test process. However, it would be helpful at some point to have Intel confirm the issue and perhaps let us know if a new version of the tool becomes available with that fixed.

One other note is that we have transitioned from the EPC4ASI8N to the EPCQ4ASI8N version of the memory device since the original non 'Q' version of the part is no longer made.

[JohnT_Intel]

Hi,

I understand your concern. I am not able to commit on when the fixed will be available but I will feedback this issue to engineering team.