I found useful info rearding discussed problem here: http://www.keil.com/forum/20221/

To prove this, on a dissasembled ECU, I routed the data lines between C167-proc and AMD-flash. Here is the result:

C167CR-LM	AMD29F200BB
AD0	DQ4
AD1	DQ12
AD2	DQ5
AD3	DQ13
AD4	DQ6
AD5	DQ14
AD6	DQ7
AD7	DQ15
AD8	DQ11
AD9	DQ3
AD10	DQ10
AD11	DQ2
AD12	DQ9
AD13	DQ1
AD14	DQ8
AD15	DQ0

So, the problem with write/erase fails is that the data lines are crossed (for routing simplification): when the C167 "wants" to write the "sequences" (see the AMD datasheet), the memory doesen't see the right values.

Actually, this problem is well known in Siemens automotive ECU units.

For example, in some cases you must write "44h" on the proc side, if you want the memory to see part of the unlock sequence "A0h".

So, I have to rewrite (and recompile) the minimon driver in order to correct the memory patterns.

Progs as WinOls and ECM2001 have integated option to code/decode dumps BIN<->SIE, due to swapped data lines.

For comparision, find two attached files: one is in-circuit read with Minimon, the other is read by eeprom programmer (flash chip desoldered).

Could someone help me to write a new driver for AMD29F200BB and AMD29F400BB, which will work in the case with swapped data lines?

Thanks.