TCM Reverse Engineering Project

[ShelGame]

The .bin I read off the '89 TCM is complete. I finished the 1st pass dis-assembly on the plane ride to Paris.

The 8x8-bit table format is this:

X0,Y0 ; point 0
X1,Y1 ; point 1
X2,Y2 ; point 2
Xn,Yn ; point n
0xFF   ; termination byte (end of table)

So, it's similar to the Engine Controllers' 8-bit table in terms of the data it stores, but the format is more like the 16-bit tables; using 0xFF to signify the end of the table.

I'm going to setup my dis-assembler to properly decode these tables.
Morris, can you add a new table format to MP Tune?

[Force Fed Mopar]

Awesomeness! I'm picking up my transmission for the black Daytona this week also, so perfect timing. I'll have to hunt down a '89 TCU now, unless the '91 is flashable somehow already? Both my 604 vehicles right now are '91 models. The van has a update sticker on the TCM from 12/99 though...

[ShelGame]

Morris, here's an actual table output from my dis-assembler. The last byte pair seems to be repeated in every case. I think this must be the same case as the 3D tables - IE, it's used for the last point calculation. I'll have to read the interpolation routine in detail.

EDIT: Actually, that's all wrong. The first byte is the number of data points in the table (just like the point-slope tables). The first byte always seems to return 0x00, and the last 0xff. There's no repeated bytes. Correct table format is now updated below.

Table11:
	.byte	5			; number of points
	.byte	0x48, 0x00		; point 1
	.byte	0x48, 0x2f		; point 2
	.byte	0x4b, 0x34		; point 3
	.byte	0x7d, 0x3e		; point 4
	.byte	0x7d, 0xff		; point 5

[bakes]

Dam im trying to find my old dealer ship cds that had the bins on them let hope they show up in my storage room.

[wowzer]

rob - can you post up the latest disassembly and bin. the last one only had 19k size?

[ShelGame]

Done.

[wowzer]

also, since the first and last bytes have the same x address how does the system know which y value to return? do these tables have slopes?

[mopar-tech]

Dam im trying to find my old dealer ship cds that had the bins on them let hope they show up in my storage room.

You know, I might have some stuff there.

[Aries_Turbo]

So in anticipation of this, what tcu would be good to obtain? 89?

- - - Updated - - -

And also, in guessing the 2.4l versions of this gearbox had a different transmission range sensor. I remember replacing the early version of that on my 90 3.0l 604.

Brian

[ShelGame]

also, since the first and last bytes have the same x address how does the system know which y value to return? do these tables have slopes?

Not sure what you mean? But, no, they don't have the slope encoded. It's interpolated by the lookup routine.

[ShelGame]

So in anticipation of this, what tcu would be good to obtain? 89?

Brian

No Idea. The '89 is noticeably larger than the later model TCM's. I suppose they made a bunch of improvements.

On the one hand, I know my flash module will work in the '89's. On the other hand, the '92+ are flashable from the factory and similar to flashing the Neon FCC. Which I can do, but it's not without it's quirks. For certain types of chips,the flash process isn't so reliable.

I'm going to try and collect as many .bin's as possible. Maybe there's an advantage in one code generation vs. another.

- - - Updated - - -

Morris -

Is it possible to update the MPTECU program to read starting at 0x0100 as an option? Right now, it starts at 0x8000 for SMEC/SBEC, and at 0x2000 for the FCC. Just for the purpose of getting all of the data out, I'd like to do a full-read for the TCM's. Starting at 0x0100 will give everything in the addressable space except the RAM (which would contain the boot program).

[wowzer]

rob - i'll take a look at that mptecu program soon and let you know. i can just put an "option" in there for now.

on the table points --- maybe it's no big deal but if the ecu does a look up for the x val of 0x48 (table 11) how does it know whether to use 0x00 or 0x2f for the y val? also, i ran it through my disassembler also and will trace out that interpolation routine to see what it does.

Guten Tag

[Vigo]

And also, in guessing the 2.4l versions of this gearbox had a different transmission range sensor. I remember replacing the early version of that on my 90 3.0l 604.

IIRC the early 604s used a range sensor AND an old style 3-pin neutral safety switch whereas later ones combined all functions into a single transmission range sensor and got rid of the 2nd switch and that hole in the case. There is definitely a change in the range sensors sometime in the ~93-95 time frame.

[countryathlon]

They made Tons of update units (later flash updates) for these cars. I am interested in getting into this stuff -I don't know a lot, but am eager to learn. I have watched at a distance on the tuning work you all have done on the ECUs, but not having a turbo, I didn't want to mess around too much with one that worked fine. On the TCM though, all 89-91s are interchangeable (for a large number of cars). I believe the change for 92 was they removed the speed sensor on the RH CV shaft because it was technically redundant (with the output speed sensor on the atx). Newer flashible units had a larger memory IIRC a 28F512. I have an old '89 P/N 4557120, a '91 P/N 4672105, and a few of the newest flashible ones (the final update for the 89-91 cars) P/N R4761846AA AKA P/N 4796121. Earlier in the year I swapped from the '91 to the newest, an noticed it didn't coast to a stop as nice (even after lots of 'learning') -the PWM behavior results in a quicker, harsher downshift -so I'm not sure the newest is the 'best'...

[Vigo]

The separate vehicle speed sensor is preferable for a modified car in the sense that you can change the transfer gear and ring/pinion ratios without affecting speedo/odo accuracy. If there is a way to 'hack' the tire size factor and make it infinitely adjustable to compensate for gearing changes, then it wouldn't matter either way.

[Aries_Turbo]

looks like rockauto lists that transmission range switch (the neutral safety switch looking one) up to 95.

so that means that i could possibly use a 95 stratus box on a 2.4 with AWD. that would be cool.

Brian

[ShelGame]

The separate vehicle speed sensor is preferable for a modified car in the sense that you can change the transfer gear and ring/pinion ratios without affecting speedo/odo accuracy. If there is a way to 'hack' the tire size factor and make it infinitely adjustable to compensate for gearing changes, then it wouldn't matter either way.

There is apparently a 'pinion factor' that has to be programmed into the later TCM's when you swap them between cars.

[ShelGame]

What analog sensors does the TCM have on it's own? IE, not data it gets from the CCD bus.

[Vigo]

I've been able to change it through a snap on scan tool, but i was only able to pick between the different factory tire sizes, which is a fairly narrow range of diameters. The gear-driven VSS is preferable there because the same style of gear was used with a huge range of tire diameters so you can take a car that came with a 25.5" tire and put a 23" tire and correct your speedo with ~$15 and a 10mm socket. If the software correction was made more adjustable, that would be better. 604 output speed sensors are both more reliable and cheaper than the older VSS designs.

[Force Fed Mopar]

What analog sensors does the TCM have on it's own? IE, not data it gets from the CCD bus.

I think just the range sensor, and the input and output speed sensors. Might be a trans temp sensor also.