Tuning Evo789 ACD
Jan 17, 2014, 01:30 PM
#106
Evolved Member
I think there is a calculation made by mitsubishi itself in the PDF about super AYC..
Jan 17, 2014, 02:31 PM
#107
Newbie
Join Date: May 2009
Location: California
Posts: 29
Likes: 0
Received 0 Likes
on
0 Posts
I think manufacturers can become vague at times. I like the raw data!!!
Jan 18, 2014, 09:50 AM
#109
Finally got around to checking my wire I had added to pin 25 of the ACD connector, and after playing around with it a bit I was finally able to read off of the ACD ECU!
romid is 401702. I hope to play around with small modifications to the maps over the coming days and see how the car reacts.
romid is 401702. I hope to play around with small modifications to the maps over the coming days and see how the car reacts.
Last edited by Raptord; Jan 18, 2014 at 10:03 AM.
Jan 18, 2014, 12:41 PM
#110
Nice work guys, I'm really happy to see all that information coming out. This same knowledge that has been kept secret by many ACD tuners since many years. Now we can learn the tuning process by ourselves. With all the snow we have here in Canada, I think we can provide some answers and test the behavior of the system more easily than in florida i.e.
Personally, on snow, what I think the car need the most is a quicker initial lockout of the center diff, and sooner release. It's hitting late, and for too much time. That's why I feel the car is more responsive in gravel mode. I never ride it on snow mode.
Personally, on snow, what I think the car need the most is a quicker initial lockout of the center diff, and sooner release. It's hitting late, and for too much time. That's why I feel the car is more responsive in gravel mode. I never ride it on snow mode.
Last edited by domyz; Jan 18, 2014 at 12:46 PM.
Jan 19, 2014, 01:43 AM
#111
Evolved Member
Thread Starter
Well, I think we now have enough info to go for a more complete operating description.
So here are my current notes:
ACD GENERAL TUNING NOTES
Accel maps are used under acceleration and (presumably) throttle lift. Decel maps are used when the brake is applied.
When the hand brake is applied the ACD is set to fully open ie no lockup.
ACD A MAP THROTTLE LOCKUP ADDER TUNING NOTES
The A-Maps take Throttle and Speed as inputs and are the main ACD maps.
The A-Maps are multiplied with the B-Maps and then the C*G*H result and D*E or F results are added to A*B.
Larger values increase the lock, smaller values decrease lock (increase slip).
When throttle is more than 50% the ACD is progressively locked.
Accel and Decel are generally identical pairs.
Loose surfaces require more ACD lock so the SNOW setting may be typically 3x the TARMAC setting.
Rally driving is enhanced with some added constant base pressure, fill all zeros above 14%TPS with 2-5.
Always leave the parking cells set to zero to prevent binding.
ACD B MAP G-FORCE LOCKUP REDUCTION TUNING NOTES
The B-Maps are G-Force multipliers to the A-Maps, their function is to reduce lockup during high G force corners.
The B-Map value is multiplied with the A-Map, using scaling CF128 which has a range of zero to one.
A B-Map value of 1.00 will have no affect on ACD lockup (ie lockup will follow the A-Map).
A B-Map value of 0.00 will cause the ACD to go fully open (ie the A-Map is cancelled allowing maximum slip).
Decel maps are generally set to have no decrease in ACD lockup below 140 kph, this is to aid stability under braking/decel.
ACD C*G*H STEERING INPUT TUNING NOTES
The CGH group of tables combine four variables to increase ACD lockup, based on:
Wheel Speed Delta (difference), Wheel Speed, Steering Angle and Steering Angle Velocity.
The G-Tables provide the additional LockFactor, the C-Maps provide the Steering Angle corrections and the H-Tables provide the Steering Angle Velocity corrections.
Both stock Evo8 and Ralliart C1 ROMs are setup so additional lockup is only added when front/rear slip exceeds -4.3 kph. The Ralliart K2 rally ROM is more aggressive, see tables G1, G2, G3.
The Steering Angle Wheel Slip Adder (G) is multiplied with the relevant C-Map and Steering Angle Velocity Multiplier (H).
The result is added to A*B.
Steering and Wheel Slip calc: C * G * H
The un-filtered C*G*H result can be monitored at F2AA+F2AB
The filtered C*G*H result can be monitored at F2AC+F2AD
On stock ROMs the G tables are set to zero except for -6.9 kph setting, therefore the C-Maps do not contribute until there is more than -4.3 kph wheel speed difference.
On stock roms the Steering Angle Velocity tables are fully set to 1.00 and thus do not modify the result.
The 0-40 deg/Sec elements can be reduced a bit to reduce lockup with rapid steering wheel movement.
C-Maps - Larger values increase the lock, smaller values decrease lock (increase slip).
Accel and Decel are generally identical pairs.
At low speeds (typically less than 20 kph) the ACD is set loose (not locked).
When the steering angle is less than 40 degrees the ACD is tight (locked).
When the steering angle is more than 40 degrees the ACD is progressively loosened.
Gravel and SNOW settings will have more lock than TARMAC setting as steering angle is increased.
ACD D E F SPEED DELTA LOCKUP ADDER TUNING NOTES
The D and E tables function to increase ACD Lockup (Higher pump pressure) when the ACD Controller registers a positive or negative difference in wheel speed front to rear (SPEED DELTA).
They appear to operate independently to the C*G*H Steering/Wheel Slip adder combination.
Tables D and E operate together, table F operates independently.
The controller decides which will be added to the total lockup calculation based on which is the larger (D*E or F).
Table group D*E operate on positive and negative wheel speed differences.
Table F operates on only positive wheel speed differences.
The three correction shaping tables (D1 D2 D3) are normalized to 255 (using CF255 scaling) and can have a value from zero to one.
The six Accel/Decel Speed Delta Lockup Adder tables (E.1 E1.2 E1.3 and E2.1 E2.2 E2.3) use LockFactor scaling (the same as A-Maps as they are added together).
Tables D and E are multiplied together (D*E) .
This result is then compared with the Speed Delta (+/-) Lockup Adder tables (F1, F2, F3).
The larger of the two results is then added to the total result: A*B + C*G*H + (D*E or F).
As you can see from the corrections made to the tuning notes, out understanding of how the maps work and operate together is getting much better.
So here are my current notes:
ACD GENERAL TUNING NOTES
Accel maps are used under acceleration and (presumably) throttle lift. Decel maps are used when the brake is applied.
When the hand brake is applied the ACD is set to fully open ie no lockup.
ACD A MAP THROTTLE LOCKUP ADDER TUNING NOTES
The A-Maps take Throttle and Speed as inputs and are the main ACD maps.
The A-Maps are multiplied with the B-Maps and then the C*G*H result and D*E or F results are added to A*B.
Larger values increase the lock, smaller values decrease lock (increase slip).
When throttle is more than 50% the ACD is progressively locked.
Accel and Decel are generally identical pairs.
Loose surfaces require more ACD lock so the SNOW setting may be typically 3x the TARMAC setting.
Rally driving is enhanced with some added constant base pressure, fill all zeros above 14%TPS with 2-5.
Always leave the parking cells set to zero to prevent binding.
ACD B MAP G-FORCE LOCKUP REDUCTION TUNING NOTES
The B-Maps are G-Force multipliers to the A-Maps, their function is to reduce lockup during high G force corners.
The B-Map value is multiplied with the A-Map, using scaling CF128 which has a range of zero to one.
A B-Map value of 1.00 will have no affect on ACD lockup (ie lockup will follow the A-Map).
A B-Map value of 0.00 will cause the ACD to go fully open (ie the A-Map is cancelled allowing maximum slip).
Decel maps are generally set to have no decrease in ACD lockup below 140 kph, this is to aid stability under braking/decel.
ACD C*G*H STEERING INPUT TUNING NOTES
The CGH group of tables combine four variables to increase ACD lockup, based on:
Wheel Speed Delta (difference), Wheel Speed, Steering Angle and Steering Angle Velocity.
The G-Tables provide the additional LockFactor, the C-Maps provide the Steering Angle corrections and the H-Tables provide the Steering Angle Velocity corrections.
Both stock Evo8 and Ralliart C1 ROMs are setup so additional lockup is only added when front/rear slip exceeds -4.3 kph. The Ralliart K2 rally ROM is more aggressive, see tables G1, G2, G3.
The Steering Angle Wheel Slip Adder (G) is multiplied with the relevant C-Map and Steering Angle Velocity Multiplier (H).
The result is added to A*B.
Steering and Wheel Slip calc: C * G * H
The un-filtered C*G*H result can be monitored at F2AA+F2AB
The filtered C*G*H result can be monitored at F2AC+F2AD
On stock ROMs the G tables are set to zero except for -6.9 kph setting, therefore the C-Maps do not contribute until there is more than -4.3 kph wheel speed difference.
On stock roms the Steering Angle Velocity tables are fully set to 1.00 and thus do not modify the result.
The 0-40 deg/Sec elements can be reduced a bit to reduce lockup with rapid steering wheel movement.
C-Maps - Larger values increase the lock, smaller values decrease lock (increase slip).
Accel and Decel are generally identical pairs.
At low speeds (typically less than 20 kph) the ACD is set loose (not locked).
When the steering angle is less than 40 degrees the ACD is tight (locked).
When the steering angle is more than 40 degrees the ACD is progressively loosened.
Gravel and SNOW settings will have more lock than TARMAC setting as steering angle is increased.
ACD D E F SPEED DELTA LOCKUP ADDER TUNING NOTES
The D and E tables function to increase ACD Lockup (Higher pump pressure) when the ACD Controller registers a positive or negative difference in wheel speed front to rear (SPEED DELTA).
They appear to operate independently to the C*G*H Steering/Wheel Slip adder combination.
Tables D and E operate together, table F operates independently.
The controller decides which will be added to the total lockup calculation based on which is the larger (D*E or F).
Table group D*E operate on positive and negative wheel speed differences.
Table F operates on only positive wheel speed differences.
The three correction shaping tables (D1 D2 D3) are normalized to 255 (using CF255 scaling) and can have a value from zero to one.
The six Accel/Decel Speed Delta Lockup Adder tables (E.1 E1.2 E1.3 and E2.1 E2.2 E2.3) use LockFactor scaling (the same as A-Maps as they are added together).
Tables D and E are multiplied together (D*E) .
This result is then compared with the Speed Delta (+/-) Lockup Adder tables (F1, F2, F3).
The larger of the two results is then added to the total result: A*B + C*G*H + (D*E or F).
As you can see from the corrections made to the tuning notes, out understanding of how the maps work and operate together is getting much better.
Last edited by merlin.oz; Jan 23, 2014 at 04:19 PM.
Jan 20, 2014, 02:47 PM
#114
Evolved Member
Thread Starter
We have now identified most of the map/table combo attack and decay timers.
They are 2-byte parameters (not tables as originally expected) and seem to work as time constants (like RC time constants for the electronic engineers reading this).
The exact scaling has not been verified but I have something that seems to be close, but the values seem to be very short, which is a bit confusing as people report the timers are typically to long on stock roms.
There does seem to be a minimum A*B value change before the valve current moves off the rest value (approx 60mA) and jumps to approx 260mA. So the initial valve current change is not linear. After the step jump to 260mA it then seems to move linearly.
On the decay side, the valve current seems to hang at 260mA for a while before stepping straight back to 60mA.
So there is a threshold value to be exceeded before change.
They are 2-byte parameters (not tables as originally expected) and seem to work as time constants (like RC time constants for the electronic engineers reading this).
The exact scaling has not been verified but I have something that seems to be close, but the values seem to be very short, which is a bit confusing as people report the timers are typically to long on stock roms.
There does seem to be a minimum A*B value change before the valve current moves off the rest value (approx 60mA) and jumps to approx 260mA. So the initial valve current change is not linear. After the step jump to 260mA it then seems to move linearly.
On the decay side, the valve current seems to hang at 260mA for a while before stepping straight back to 60mA.
So there is a threshold value to be exceeded before change.
Last edited by merlin.oz; Jan 20, 2014 at 02:50 PM.
Jan 21, 2014, 04:55 AM
#117
We have now identified most of the map/table combo attack and decay timers.
They are 2-byte parameters (not tables as originally expected) and seem to work as time constants (like RC time constants for the electronic engineers reading this).
The exact scaling has not been verified but I have something that seems to be close, but the values seem to be very short, which is a bit confusing as people report the timers are typically to long on stock roms.
There does seem to be a minimum A*B value change before the valve current moves off the rest value (approx 60mA) and jumps to approx 260mA. So the initial valve current change is not linear. After the step jump to 260mA it then seems to move linearly.
On the decay side, the valve current seems to hang at 260mA for a while before stepping straight back to 60mA.
So there is a threshold value to be exceeded before change.
They are 2-byte parameters (not tables as originally expected) and seem to work as time constants (like RC time constants for the electronic engineers reading this).
The exact scaling has not been verified but I have something that seems to be close, but the values seem to be very short, which is a bit confusing as people report the timers are typically to long on stock roms.
There does seem to be a minimum A*B value change before the valve current moves off the rest value (approx 60mA) and jumps to approx 260mA. So the initial valve current change is not linear. After the step jump to 260mA it then seems to move linearly.
On the decay side, the valve current seems to hang at 260mA for a while before stepping straight back to 60mA.
So there is a threshold value to be exceeded before change.
As I said some days ago it could be interesting to manipolate that start and stop values, that now I think shoud be at Min 260mA and max to 1000mA. Lower limit is not so important but upper can let us gain some working bars more.
Look at the pict, this is a tipical pressure/current caratteristic:
Jan 21, 2014, 05:09 AM
#118
Test n° 4:
I've tryed some AYC trick:
1- most of the map set to 0%
2- most of the map set to 100%
Log both of that sets.
Surface very low grip: wet on slippery asphalt
I really didn't felt so many difference while drifting, maybe something in enternig in the corner.
A think tha map works in non slip condition, but in few moment in some (unkown) conditions...
Can anyone understend something alse from these logs?
I've tryed some AYC trick:
1- most of the map set to 0%
2- most of the map set to 100%
Log both of that sets.
Surface very low grip: wet on slippery asphalt
I really didn't felt so many difference while drifting, maybe something in enternig in the corner.
A think tha map works in non slip condition, but in few moment in some (unkown) conditions...
Can anyone understend something alse from these logs?
Jan 21, 2014, 12:37 PM
#119
Evolved Member
Thread Starter
Thats an interesting data sheet there Floppyz, both the at rest current and the step response time are relevant to out cause here.
Regarding the 3D AYC map, there are two sets of three tables (JI.1 J1.2 J1.3 Accel) (J2.1 J2.2 J2.3 Decel) that get multiplied with the AYC map.
The Accel group of three are at 122D0, 122DC, 122E8, the Decel group of three are at 128E0, 128EC, 128F8.
At the moment I am tagging these tables with CF128 scaling as a multiplier to the main 3D AYC map.
On stock roms, these multipliers are typically set to zero from 140 kph and above. The Accel and Decel tables are different (all the Mitsu documentation says the AYC behaves differently under deceleration from acceleration).
As this is a multiplication function, I am guessing that when the result is zero (typical result above 140 kph) then no AYC valve action takes place and the drive is evenly split between both rear wheels.
But it may be that the diff is set to run as an open diff. More work required.
If someone out there knows the answer to this I would like to here it.
Regarding the 3D AYC map, there are two sets of three tables (JI.1 J1.2 J1.3 Accel) (J2.1 J2.2 J2.3 Decel) that get multiplied with the AYC map.
The Accel group of three are at 122D0, 122DC, 122E8, the Decel group of three are at 128E0, 128EC, 128F8.
At the moment I am tagging these tables with CF128 scaling as a multiplier to the main 3D AYC map.
On stock roms, these multipliers are typically set to zero from 140 kph and above. The Accel and Decel tables are different (all the Mitsu documentation says the AYC behaves differently under deceleration from acceleration).
As this is a multiplication function, I am guessing that when the result is zero (typical result above 140 kph) then no AYC valve action takes place and the drive is evenly split between both rear wheels.
But it may be that the diff is set to run as an open diff. More work required.
If someone out there knows the answer to this I would like to here it.
Last edited by merlin.oz; Jan 22, 2014 at 04:19 PM.
Jan 23, 2014, 11:07 AM
#120
when no valve has no action the ayc diff is as an free differential, may be a little bit meccanically slip limited.
So that is a confirm that the 100% in the 3d map indicate a power tranfert to the external weel, isn't it?
bye
So that is a confirm that the 100% in the 3d map indicate a power tranfert to the external weel, isn't it?
bye