Flash Extractor
© Soft-Center
About     Shop     Downloads     Manual     Library     Forum     Services     Contacts
multiple ECC combine element   Search  Register  Log in
Reply to topic Goto page 1, 2, 3, 4  Next
Author Message
HDDRCVR



Joined: 29 Sep 2015
Posts: 446


PostPosted: Sun Mar 28, 2021 19:06    Post subject: multiple ECC combine element
Reply with quote

Sergey, would it be possible to make an element that will use multiple dumps for ECC combine for the final corrections?
For example, if we get a stubborn dump, we could read it into a separate folder and read it x amount of times. Then using the layout, we could ECC correct them individually within the same task. Then the multiple ECC combine an element could take corrected values to produce a final united dump.

Currently, I did it by hand by reading the dump 8 times, correcting them all within same task, then splitting to 2x4 dumps and ECC combine, then rename files into 2x2 and combine again, then finally rename 2x1 and making final ECC combine.

multiple combine elements could be very helpful for something like this.
Сергей



Joined: 26 Aug 2005
Posts: 18864

Flash-Extractor developer

PostPosted: Thu Apr 01, 2021 11:52    Post subject:
Reply with quote

//correcting them all within same task

Not need make ECC Fix before ECC Combine, if Fix_Input = ON

//then splitting to 2x4 dumps and ECC combine
// then rename files into 2x2 and combine again

I do not understand this place
Why need split and rename?
Bolo



Joined: 05 Dec 2017
Posts: 468
Location: 506F6C616E64

PostPosted: Wed Apr 21, 2021 12:35    Post subject:
Reply with quote

I think OP means to add possibility to add more inputs at once.... as on atatched photos
how it will works ?

Take folder 1 and combine into MAIN
Then take folder 2 and combine into MAIN
Then folder ...... and combine info MAIN

By this computer and FE can work whole night with combine elements using diffrents input without operator

It's good idea Wink
multi_combine_input.jpg
HDDRCVR



Joined: 29 Sep 2015
Posts: 446


PostPosted: Tue May 11, 2021 8:53    Post subject:
Reply with quote

Bolo wrote:
I think OP means to add possibility to add more inputs at once.... as on atatched photos
how it will works ?

Take folder 1 and combine into MAIN
Then take folder 2 and combine into MAIN
Then folder ...... and combine info MAIN

By this computer and FE can work whole night with combine elements using diffrents input without operator

It's good idea Wink


thats exactly what I mean. Sorry I lost track of this post Smile thanks Bolo

yes, this feature would be very good for those cases that bring very little improvement to each pass. I am 30 passes deep into one CE on current case, and still have 10000 reds in map. If we could Test retry in differnt spots, save data by reading 20-30 times, then let the process run all night we come back to a much cleaner view.
Сергей



Joined: 26 Aug 2005
Posts: 18864

Flash-Extractor developer

PostPosted: Tue May 11, 2021 12:50    Post subject:
Reply with quote

Is is ok if multiple reads will be in same folder? Example:

d:\my case\1
d:\my case\2
d:\my case\3
d:\my case\4
..
jeremyb



Joined: 09 Dec 2008
Posts: 1698
Location: RecoverMyFlashDrive.com Bridgeport, CT, USA

PostPosted: Tue May 11, 2021 13:32    Post subject:
Reply with quote

Why all the hassle with ECC combine? Just re-read straight from the NAND chip.

On a low level there is no reason the NAND Reader can't do page level correction within FE, you have all the functionality in FE except a "for loop" with an "if" statement.

In my NR implementation I do page level addressing all the time, unfortunately my mind isn't smart enough to implement ECC correction so it's a non-starter but functionally all the technology is built into NR and FE.
Сергей



Joined: 26 Aug 2005
Posts: 18864

Flash-Extractor developer

PostPosted: Tue May 11, 2021 18:18    Post subject:
Reply with quote

jeremyb wrote:
Why all the hassle with ECC combine?

Because we can check that memory chip is connected to reader only during "Read ID"

If chip lost connection with reader during reading and start send 00 or ff
All red sectors will be replaced with them

Also if memory give ID, this not mean that it will send normal data

Thats why i always tell:
1. Read new dump
2. Check its ECC map
3. Make Combine

But nobody do )

Okey i will try return ECC Reread
arvika



Joined: 18 Nov 2009
Posts: 2038
Location: Poland

PostPosted: Tue May 11, 2021 18:38    Post subject:
Reply with quote

I agree with Sergey. If something goes wrong you can damage dump (which you for example correct within 2 weeks Smile ).
But current solution is not perfect, because it is time consuming.

I do not know if it is possible, but maybe you can also add function:
reread page/block many times (like 3,5,7,9) and average the result on bit level and save it to dump. Of course reading will be long, but it could give really good result.
jeremyb



Joined: 09 Dec 2008
Posts: 1698
Location: RecoverMyFlashDrive.com Bridgeport, CT, USA

PostPosted: Tue May 11, 2021 21:33    Post subject:
Reply with quote

Сергей wrote:

Because we can check that memory chip is connected to reader only during "Read ID"

If chip lost connection with reader during reading and start send 00 or ff
All red sectors will be replaced with them


Why not create an if statement that only modifies on "Green" ECC

eg:
...Grey...
...Red...
...Green...

Don't write on Grey or Red only if the returned page has 100% ECC Green.
Сергей



Joined: 26 Aug 2005
Posts: 18864

Flash-Extractor developer

PostPosted: Tue May 11, 2021 22:13    Post subject:
Reply with quote

"Sub folders" option added to ECC Combine in FE_1006

If selected, will take dumps from subfolders of selected folder

if not selected, from selected folder

You can add and delete sub folders while it process other folders
jimb



Joined: 18 Apr 2012
Posts: 54
Location: Japan

PostPosted: Wed May 12, 2021 7:02    Post subject:
Reply with quote

I can agree with Sergey, too. Check ECC Map first then do ECC Combine is the safe process, which we often follow. But, I have rarely seen a 00/FF damaged dump.

However, for some jobs the reread process is very time consuming, and the result is sometimes not repeatable.

Case 1: SanDisk v1 reread: red pair of sectors show green in Test Retry: Tune. Use that retry solution in NANDReader and the sectors are still red. Tune again, Read, still red.

Case 2: Micron v2: Test Retry: lots of red and green dancing on the screen, even in Constant Read. Optimal solution does not necessarily fix the red sectors in the dump. Maybe you can catch a retry solution where the red sectors show green. If you are lucky the sector will be green when read by NANDReader.

[How about putting back the "For Combine" option in Test Retry?]

For these cases ECC WriteBack during Test Retry would be a big help. I agree with jeremyb suggestion, write only green ECC sectors.

If ECC Reread, how would one set the Retry and voltage values?

Thanks,
jimb.
Bolo



Joined: 05 Dec 2017
Posts: 468
Location: 506F6C616E64

PostPosted: Thu May 13, 2021 11:55    Post subject:
Reply with quote

Сергей wrote:
Because we can check that memory chip is connected to reader only during "Read ID"

If chip lost connection with reader during reading and start send 00 or ff
All red sectors will be replaced with them

Also if memory give ID, this not mean that it will send normal data



So why not use Status Read 0x70 command before/after read data? The Status Read can also be used during a normal read operation (not only Block Erase and Cache Rrogram on Read Cache Read) to find out the Ready/Busy Status. When we get HIGH (1) on the DATA 6 line we are ready to read and and know that chip is connected. so afer 0x70 we get 0x20 (0010 0000) or more likly 0x60 (0110 0000) because Data Cache Ready/Busy is send on DATA 5 line and have HIGH (1) status if read from chip is possible. If we get 0x00 after 0x70 command we know is something wrong with connected memory.

In case you want to explain more reach me by Skype, since returning back ReRead option is good - it's always an option and novadays it's becoming must for bigger capacity flash.

P.S
You implemeneted Map sometime ago... there is only one small step to use it for Reread now as Jerremy writes
Сергей



Joined: 26 Aug 2005
Posts: 18864

Flash-Extractor developer

PostPosted: Sun May 16, 2021 2:02    Post subject:
Reply with quote

ECC Reread added to FE_1007

Please, test it, if you have good cases
Need check not that ECC map become green
Need check that files become good

It is very easy to make bug in so complex software
Сергей



Joined: 26 Aug 2005
Posts: 18864

Flash-Extractor developer

PostPosted: Sun May 16, 2021 21:45    Post subject:
Reply with quote

In FE_1008 added "Max Diff" parameter

If difference in bits is bigger then Max Diff (by default = 10%) then data from block will not be imported
Bolo



Joined: 05 Dec 2017
Posts: 468
Location: 506F6C616E64

PostPosted: Mon May 17, 2021 11:54    Post subject:
Reply with quote

Works OK as for test now.....

Next upgrade schould be use "Calc stattistic" to reread ONLY sectors marked on Red in map file and combine into main dump
Display posts from previous:   
Reply to topic All times are GMT + 4 Hours
Goto page 1, 2, 3, 4  Next
Page 1 of 4

 

Last added
PS2251-50-F   98 d7 94 32   1x1   v2 IS903   89 84 64 3c   2x4   v4 FC1179   89 c4 08 32   1x1 Monolith Micro SD_37   45 4c 98 a3   1x1 FC1179   2c 44 44 4b   1x1 SM2246XT H AA   45 3a 94 93   4x4   v2 IS917   2c a4 e5 3c   2x2 DM8261   98 d5 98 b2   1x1   Ref Monolith Micro SD_36   ec de 98 ce   1x4 Monolith USB_90   45 3e 98 b3   1x1
News
16.08.2021 Sector Number Slow
15.07.2021 New drivers
21.05.2021 BGA-152 18x12 (smaller)
09.05.2021 Make FAT32 Root
20.01.2021 VT can work with NTFS
Other products
© Soft-Center ltd.