Evo 8 post reflash - RAM not cleared
Mar 23, 2009, 12:00 PM
#1
Evolved Member
Thread Starter
Evo 8 post reflash - RAM not cleared
I thought we might run into this problem given that I've noticed (as have others) that Evo 8 fuel trims are not reset by a flash.
My test for a ROM to RAM copy of the alt maps uses Tephra's 0xDEAD test, which on the 9 works reliably for ECU middle plug disconnect or post reflash.
The first alpha tester of my Evo 8 MR JDM live mapping has indeed found that the RAM is full of junk, but the 0xDEAD test doesn't work, so nothing triggers my copy from ROM to RAM, so the alt maps are also junk, he has to pull the middle ECU plug to force a RAM clear so 0xDEAD is not read.
I could just put an Evo 8 post reflash button in my app that changes from 0xDEAD to something else to force everything Tephra related to clear.
Other ideas?
My test for a ROM to RAM copy of the alt maps uses Tephra's 0xDEAD test, which on the 9 works reliably for ECU middle plug disconnect or post reflash.
The first alpha tester of my Evo 8 MR JDM live mapping has indeed found that the RAM is full of junk, but the 0xDEAD test doesn't work, so nothing triggers my copy from ROM to RAM, so the alt maps are also junk, he has to pull the middle ECU plug to force a RAM clear so 0xDEAD is not read.
I could just put an Evo 8 post reflash button in my app that changes from 0xDEAD to something else to force everything Tephra related to clear.
Other ideas?
Mar 23, 2009, 07:47 PM
#2
Evolved Member
iTrader: (22)
Join Date: Nov 2003
Location: Northern KY near Cincy
Posts: 2,408
Likes: 0
Received 6 Likes
on
6 Posts
I just wish someone would post up a clean evo 8 rom. There has been so many patches, findings, and updated that my 94170015 is all screwy.
As for the fuel trims not resetting, most of us just disconnect the battery and start over. Been doing it this way for a long time. Every time a new part is added that could effect the tune. Disconnect and reset.
I wouldn't mind having a "reset button" put into EcuFlash. What would be even better is the capability to set on the fly as some other ECUs out there can do.
As for the fuel trims not resetting, most of us just disconnect the battery and start over. Been doing it this way for a long time. Every time a new part is added that could effect the tune. Disconnect and reset.
I wouldn't mind having a "reset button" put into EcuFlash. What would be even better is the capability to set on the fly as some other ECUs out there can do.
Mar 23, 2009, 08:41 PM
#4
Hi John,
So the actual flashing process first loads the incoming ROM segments into RAM before burning it to flash. Does that sound right to you?
Hence why after you do a flash the RAM is sometimes screwed.
I suspect that the block of memory you are using for altmaps gets screwed over but my DEAD check is fine.
You can probably modify the dead check to look at a word near the start/middle of your altmaps.
This might explain why valet mode sometimes goes funny (after a flash) on 96530006...
Cheers
D.
So the actual flashing process first loads the incoming ROM segments into RAM before burning it to flash. Does that sound right to you?
Hence why after you do a flash the RAM is sometimes screwed.
I suspect that the block of memory you are using for altmaps gets screwed over but my DEAD check is fine.
You can probably modify the dead check to look at a word near the start/middle of your altmaps.
This might explain why valet mode sometimes goes funny (after a flash) on 96530006...
Cheers
D.
Mar 24, 2009, 03:11 AM
#7
Evolved Member
Thread Starter
The 9s I have done always do reset the trims, although the RAM still has junk from flash blocks in it.
Worse still the first 8 tester has found that the RAM maps seem to be corrupted overnight.
Worse still the first 8 tester has found that the RAM maps seem to be corrupted overnight.
Trending Topics
NP
Mar 24, 2009, 01:08 PM
#12
All the stock ROMs can be found on the ECUFlash Wiki.
http://evoecu.logic.net/wiki/Downloads
The Tephra patch threads also have stock ROMs with the patches implemented.
This isn't too good for us Evo 8 guys as the live tuning will be a great addition. I hope you can figure out a good solution to this problem. This explains why some flashes run great right away and others take a day or two of driving on to get driveability back.
http://evoecu.logic.net/wiki/Downloads
The Tephra patch threads also have stock ROMs with the patches implemented.
This isn't too good for us Evo 8 guys as the live tuning will be a great addition. I hope you can figure out a good solution to this problem. This explains why some flashes run great right away and others take a day or two of driving on to get driveability back.
Last edited by 03whitegsr; Mar 24, 2009 at 01:10 PM.
Mar 24, 2009, 01:17 PM
#13
Evolved Member
Thread Starter
I had chance for a brief test on an 8MR tonight. Everything appears to work for a while until the RAM is getting corrupted by something. Need to work out what that is!
The 8 MR is an hour away, this sort of thing needs extended testing. Will have a think.
The 8 MR is an hour away, this sort of thing needs extended testing. Will have a think.
Mar 24, 2009, 01:55 PM
#14
Evolved Member
Thread Starter
It looks like my FFFF8480 area variables are probably being overwritten, we thought this was universally free on all ECUs, but it seems not, and tephra evicted me from the FFFF8400 area 
When the variables that control my comms are overwritten, the DMA memory transfers end up screwed and there are offset portions of maps that get written in weird places.
Whilst I cannot see other routines that address my variables directly, I suspect that variables are passed from other routines and blocks are being written, which is why this area appears mysteriously vacant.
When the variables that control my comms are overwritten, the DMA memory transfers end up screwed and there are offset portions of maps that get written in weird places.
Whilst I cannot see other routines that address my variables directly, I suspect that variables are passed from other routines and blocks are being written, which is why this area appears mysteriously vacant.