[Bolo]

Hello,

I got idea how to make FE better for ECC correction

There are many siutations that there are some parts of corrupted data on choosen offsets where all other dump are green. Problem are specially on sector assembler but not only and when we work on big card size - 128Gb per bank. For this let's take SanDisk 128Gb card where there are 1 bank and along all green (fixed) sectors we got also corrupted pages.



Right now we need to read by each of pass device - in our case it was required to read it as Slow which takes 10h per pass and then use Combine, then choose another one RR, read, combine without any guarantee that at the end we will get good results.

=============
----------
Test Retry
----------
Sandisk / 0 Green 0 Red 68 Time 312
Sandisk / 1 Green 8 Red 60 Time 265
Sandisk / 2 Green 4 Red 64 Time 250
Sandisk / 3 Green 12 Red 56 Time 250
Sandisk / 4 Green 4 Red 64 Time 250
Sandisk / 5 Green 12 Red 56 Time 282
Sandisk / 6 Green 16 Red 52 Time 250
Sandisk / 7 Green 20 Red 48 Time 250
Sandisk / 8 Green 24 Red 44 Time 250
Sandisk / 9 Green 20 Red 48 Time 234
Sandisk / 10 Green 24 Red 44 Time 265
Sandisk / 11 Green 20 Red 48 Time 250
Sandisk / 12 Green 16 Red 52 Time 250
Sandisk / 13 Green 28 Red 40 Time 250
===============


Of couse over 98% of map got corrected sectors. On each of pass we get some correction but it takes time.

My idea is to use backwrite for ECC together with RR for fixing only this part which we see on screen. This is how system works on normal ECC (without RR) - we change direction to writ, press ECC ON and cells are corrected.



I'm trying to archive the same on RR but got error <cannot write to virtual media> which is ofcourse corect but are you able to add such option that will together with COMBINE allows to fix ONLINE sectors that shows on screen without need to read whole 128Gb dump ?

How it will looks:

1) Customer perform full ECC Fix and Rereads to get most of cells fixed
2) When he will find some small red parts he can change direction to Write back ECC and then use Test Read Retry or Test Voltage together with Combine (or other elements as Slow) to try to fix those sectors. In fact system check if they can be fixed anyway so right now need only to write correct one to dump.

P.S
If you think that process is unsafe you can add communicate - ALWAYS MAKE BACKUP FIRST. but duplicate of dump will take 2 minutes and then we can try and see if this helps instead of reread dump several time for 60-70h to correct only 200-300 pages.

[Bolo]

Any chance to look for such features ? Protection can be not fix if Sector change to gray (0xFF, 0x00)

[Сергей]

Thinking

[Bolo]

Any update here ?

[Bolo]

Got a chance to add this feature into next relese of FE ?

[Bolo]

Up

[jimb]

I would like this feature too.

I currently am trying to fix a Samsung 1x4 (ec d7 98 ca), which needs different RR and voltage values for each CE and page within CE. Each RR/Fix ECC/ECC Combine cycle is taking two hours.

It would be quite speedy to turn on writeback, then Try Retry to fix the area currently being displayed. Of course, we don't want to writeback if the retry result for a sector/page is grey (00/FF)

[jimb]

With RR Constant Read, we can confirm the red/green/grey sector result of the RR values.

Please allow ECC writeback on RR Constant Read.
It shouldn't be too dangerous.

Thanks,
jimb.