Gear Shift Control questions/observations

[DandismRacing]

Yesterday I made my first attempt at getting the Link gear shift controls to work with my automated dog box (H-Pattern dogbox w/ MME pneumatic rams/gear controller).

One thing I noticed was that the Blip Mode configuration/behavior doesn't appear to match the docs.  In all settings (Normal, RPM Target, Predictive) the blip duration table is there although the docs indicate it shouldn't be for 'RPM Target'.  Also, when selecting Predictive there is no 'Blip Response Time' table even though my Start Shift Mode is set to Digital Input.  In the end I can work with Normal mode but it'd be nice to have predictive mode rather that lots of trial/error to get the blip duration correct.

The other thing I noticed and I'm not sure if I have something wrong or if it's just how my engine responds to ignition cut but it was almost like there was a rev hang after the initial dog unload.  I have the RPM Match Limit Table set too 100% cut for the '500' cell but the decay rate of the engine RPM seemed rather slow, like 20ms per 100rpm. Given that my gears have a pretty far spread (~2k rpm), it makes for a long wait to even come close to attempting to rev match.  The gear controller has pretty temperamental logging so I'm not sure if I have data to prove it fully got out of gear but I would expect a speed drop (and feel it) if it didn't, so I don't believe that's what's going on.  

Finally, the gear RPM target seem pretty far off, even taking into account the RPM Target Adjust tables. I manually calculated the gear ratio (using engine rpm/speed kph) so I may try having the ECU calibrate it but I'm not even getting close to the RPM target so it's not a concern at this point.

Here's what the RPM is doing on an upshift.  The initial drop is during the dog unload stage, the gear controller pulls it out of gear just as this happens, then I hold the transmission between gears to allow the engine RPM to drop.  But as you can see the RPM drop rate changes.  The gap between vertical grid lines is 200ms and horizontal is 1000RPM. The final dip was the drive train forcing the engine to slow as it went into the next gear. 

Give it was really wet you could actually see it was causing wheel spin in the wheel speed data as it was going into gear.

The following pulsation seems to be related to the E-throttle not handling the gear controller closing and opening the throttle smoothly.

Unfortunately I didn't log the ignition cut percentages to verify it was holding a 100% cut.  I need to figure out how to cut the log done so I can add just the a few of the shifts.

Any advice/insights would be helpful.  I'll do some more testing/tuning later this week.

[DandismRacing]

I should probably note that the engine is a stock Toyota 2GR-FE with nothing special but a lightweight clutch/flywheel and an intake/exhaust to fit into an MR2. 

[Adamw]

Can you attach the log and tune.  Are you not using paddle shift mode?

[DandismRacing]

I'll try to get you the tune and a truncated log...

Paddle shift mode is not appropriate as this runs 2 independent pneumatic rams with 3 solenoids each.  It's an H-pattern so it's running both the shift and crossgate cables.

[Vaughan]

20 minutes ago, DandismRacing said:

Paddle shift mode is not appropriate as this runs 2 independent pneumatic rams with 3 solenoids each.  It's an H-pattern so it's running both the shift and crossgate cables

Are you using the setup from an MRS?

[Adamw]

3 hours ago, DandismRacing said:

One thing I noticed was that the Blip Mode configuration/behavior doesn't appear to match the docs.  In all settings (Normal, RPM Target, Predictive) the blip duration table is there although the docs indicate it shouldn't be for 'RPM Target'.  Also, when selecting Predictive there is no 'Blip Response Time' table even though my Start Shift Mode is set to Digital Input.  In the end I can work with Normal mode but it'd be nice to have predictive mode rather that lots of trial/error to get the blip duration correct.

The duration table is always present.  For RPM target and Predictive it is only used as a maximum duration, so if for example the target is never reached due to an incorrect gear ratio or mis-shift or similar, you dont have the blip staying on forever.  I would suggest the RPM target mode for most applications, you can then specify more throttle opening than you actually need and the ecu will use a limiter to hold it at the target rpm (the throttle closes as soon as it reaches target also).  This mode adapts very well to how the car is being driven.  

 

3 hours ago, DandismRacing said:

The other thing I noticed and I'm not sure if I have something wrong or if it's just how my engine responds to ignition cut but it was almost like there was a rev hang after the initial dog unload.  I have the RPM Match Limit Table set too 100% cut for the '500' cell but the decay rate of the engine RPM seemed rather slow, like 20ms per 100rpm. Given that my gears have a pretty far spread (~2k rpm), it makes for a long wait to even come close to attempting to rev match.  The gear controller has pretty temperamental logging so I'm not sure if I have data to prove it fully got out of gear but I would expect a speed drop (and feel it) if it didn't, so I don't believe that's what's going on.  

Would need to see a log to give any insight here.  If it is large you can share it with google drive or wetransfer or similar.  Is it synchro or dogs?  I suggest doing both a PC log and ECU log at the same time when testing so you have relevant fast data in the ecu log for looking at the shift in detail and then a full PC log of all channels in case you missed anything you needed.

 

3 hours ago, DandismRacing said:

Finally, the gear RPM target seem pretty far off, even taking into account the RPM Target Adjust tables. I manually calculated the gear ratio (using engine rpm/speed kph) so I may try having the ECU calibrate it but I'm not even getting close to the RPM target so it's not a concern at this point.

You can also log gear ratio (gear ratio calc runtime), drive it gently through all gears then enter the average ratio value logged for each gear.  

This is what you would expect to see for RPM target Vs RPM:

 

 

 

[DandismRacing]

33 minutes ago, Vaughan said:

Are you using the setup from an MRS?

No, this is my own setup on an SW20 endurance car, I think I have the last MME H-Pattern kit they sold.

[DandismRacing]

Here's a link to the log file https://www.dropbox.com/s/l9xphpkadea320i/ECU Log 2023-04-16 3%3B45%3B55 pm.llgx?dl=0, attached is a tune with some stuff cleared to keep my tuner happy.  I've only really configured Drive upshift and Overrun downshifts since my gear controller more or less locks everything else out anyway.  

shareable_4-17-23.pclx

[DandismRacing]

I played with the PID values for the e-throttle and got that smoothed out.  Now to figure out the revs not dropping... 

[koracing]

The log only shows APS, TPS, Gear, and RPM.  No indications of how much ignition cut or fuel cut is being applied let alone RPM target for downshift.  I've found using math to try and calculate gear ratios isn't as reliable and monitoring the ratio runtime in each gear via datalog or live on dyno.

Also LOL @ the fuel and timing tables *Secret Squirrel*.

[DandismRacing]

The log should have RPM target, wheel speed, Gear shift status, ignition trim and some other stuff in there.  I didn't have it logging Ignition Cut %, which I should have.

This is what's logged...

An Temp1 and An Temp2 are the inputs for upshift/downshift from the Gear Controller.   Since the Gear Controller only blips/signals downshift prior to shifting I had to use a Timer and math block (Math Block 2) to effectively extend the blip while the shift is taking place. 

[koracing]

Oops, yeah there's a bit more in there, my bad.  I'm not sure if you need to give it more time or not on the downshift, but it exits the rpm target window on downshift before the throttle even opens to the commanded value.  Perhaps further PID adjustment could improve the response time, but I know there's going to be a limit there.

[DandismRacing]

Yeah, the downshift needs a little more time.  I feel pretty confident I can get that to be in a reasonable statem, it didn't feel to bad in the car as it was.  The upshifts are the frustrating part.  I can give it more time but it'll be slow, and I'll want to switch to fuel cut (I was running ignition cut), as I don't want to dump fuel for that long.

[Adamw]

23 hours ago, Adamw said:

Is it synchro or dogs? 

If its a dog box it looks like there is not enough preload to the shift fork at the beginning of the torque reversal/dog unload so it is staying engaged.  Needs more cut delay after the actuator solenoid is activated.  Typically about 40ms delay is needed to reach full pressure in the actuator with proper paddle shift hardware, with your more industrial-themed hardware it could be significantly longer.  I dont have a lot of experience with synchro shifts but I suspect they would need longer delay and slower movement so that the torque reduction only happens just as the cone brake is about to release.  Also, why does the throttle close during an upshift?

For the downshift there is sometimes an issue with your downshift switch deactivating shortly after the shift request which cuts the blip short (pic below), and for the shifts where this doesnt happen it looks like the blip duration is way too short, the throttle closes before the RPM as barely even risen.  More like 200ms is more typical.

[DandismRacing]

1 hour ago, Adamw said:

If its a dog box it looks like there is not enough preload to the shift fork at the beginning of the torque reversal/dog unload so it is staying engaged.  Needs more cut delay after the actuator solenoid is activated.  Typically about 40ms delay is needed to reach full pressure in the actuator with proper paddle shift hardware, with your more industrial-themed hardware it could be significantly longer.  I dont have a lot of experience with synchro shifts but I suspect they would need longer delay and slower movement so that the torque reduction only happens just as the cone brake is about to release.  Also, why does the throttle close during an upshift?

For the downshift there is sometimes an issue with your downshift switch deactivating shortly after the shift request which cuts the blip short (pic below), and for the shifts where this doesn't happen it looks like the blip duration is way too short, the throttle closes before the RPM as barely even risen.  More like 200ms is more typical.

 

This is a dog box..   I'll see if I can pull data off my gear controller, it's logging doesn't always play nice.  But the air rams have hall effect position sensors in them that will tell me if I'm getting out of gear or not.  Given I don't see a reduction in speed and from past experience trying to get this to work with a stock ECU, I doubt that it's not making it to neutral/out of gear.  Hopefully I'll get log data that will shed some light as that was the only thing I could come up with also.

As for the throttle closing, the gear controller is inline with the APS signal and has the ability to set the APS signals to 0% during the upshift.  That was an attempt to see if I could get a different reaction from the engine by closing the throttle during the ignition cut, earlier in the day I was leaving it open and saw the same RPM behavior from the engine.

On the downshifts I didn't see that sudden drop, I didn't look close enough...  I'm surprised it's doing that though, the math block uses the blip/downshift signal from the Gear controller plus a timer that starts with the blip/downshift signal and some simple boolean logic.  I originally was just using the timer but I noticed that the Gear Shift status didn't show the downshift start until the timer reached about 13ms which is well after the 5ms delay the gear shift controls are set to.  I agree that the blip was too short, 100ms, I think your earlier advice about using RPM Target mode, a larger TPS percentage, holding it longer with a long Limiter Hold Duration is what I should be doing there.

[Adamw]

18 minutes ago, DandismRacing said:

I doubt that it's not making it to neutral/out of gear.

Yes, assuming the ign cut is actually happening, this is your problem, not enough preload on the shift forks at the time of cut.  If you cut too early before there is preload, the torque reversal means the dogs flanks are already locked in the reverse direction before there is force applied.  You need to have full pressure on the dog ring at the time of the cut, the only time the dogs can release cleanly is at the instant the torque transitions from forward to reverse.

The throttle closing is probably not helping either with more engine braking effect it increases reverse torque so they are locked together even harder than if the throttle was WOT. 

 

33 minutes ago, DandismRacing said:

I'm surprised it's doing that though, the math block uses the blip/downshift signal from the Gear controller plus a timer that starts with the blip/downshift signal and some simple boolean logic.

Im certainly no coder or mathematician, but my first thought was that possibly the "&" in the middle of your equation would have precedence over the relational <>.  So it is acting more like b|(a>(0&a)<0.14) whereas I think you want b|((a>0)&(a<0.14))

[DandismRacing]

1 hour ago, Adamw said:

Yes, assuming the ign cut is actually happening, this is your problem, not enough preload on the shift forks at the time of cut.  If you cut too early before there is preload, the torque reversal means the dogs flanks are already locked in the reverse direction before there is force applied.  You need to have full pressure on the dog ring at the time of the cut, the only time the dogs can release cleanly is at the instant the torque transitions from forward to reverse.

The throttle closing is probably not helping either with more engine braking effect it increases reverse torque so they are locked together even harder than if the throttle was WOT. 

I get that but I'd assume the rear wheel speed data would back it up.  It should have a reduction in speed that is proportional to the reduction of RPM.  I only lose 1 kph over the time between gears.

 

[DandismRacing]

Here's the position data from the gear controller for a 3->4 upshift like above.  The Green line is the interpreted position of the ram, 1 -> 3rd gear shifter forward, 2 -> neutral shifter in the middle, 3 -> 4th gear shifter back, (in between positions it drops to 0), and Red is the raw position from the hall effect sensor.  There is a slight delay before it hits the neutral position because I'm activating the solenoid that stops the movement a little too early.  But there's no hesitation getting out of gear indicated by the data.

 

[DandismRacing]

@Adamw @Vaughan Since it doesn't look like I have the appropriate data to really determine what's happening do either of you have an suggestions on what I should log and what changes you might suggest attempting w/ respect to upshifts?  I'll be testing again this Friday and will hopefully have things in a reasonable state to race this weekend. Dealing with a PC inside the car is a little bit of a challenge because I'll be on a live track w/ full race gear but adding more channels to the ECU logging is pretty simple and I have button on my steering wheel to enable/disable logging so it doesn't fill things up.

In the back of my mind I just question if there's something in the firmware that's fighting with the gear shift control cut, without having the ignition cut data it's tough to tell.  My other thought is to turn off the shaft speed matching to see if that changes the behavior.

[Adamw]

%cut is the main thing missing from the logging set, without that you can only guess at the cause of your slow RPM drop. The dog unload should have given you 100%cut for at least 60ms regardless of how close or far rpm target was.  The only thing that would remove the dog unload early is if your upshift digital input ended early but that looks like it is adequate.  Attached is your logging set with a few small changes.  

As for shift strategy, the main thing I would change is increase upshift RPM target adjust to 105% (help file suggestion of 95% is wrong for upshift), for downshift I think your timer will need to be quite a bit longer - more like 250ms, change blip end to RPM target, blip duration to 200ms, limiter hold to 200ms. 

DI shift diag.llsx