Full time lock up mode

[Bernard Fife]

Still no luck duplicating what you are seeing. I have tested using your MSQ, and the TCC works as expected in both on/off mode and PWM mode (no taper in on/off, of course). I have checked both the TS indicator (and TCC PWM% gauge) and used a scope on the TCC output and all looks good.

It's hard for me to fix anything unless I can duplicate it here, so I am kind of stuck.

I don't know what else to check at this point. Perhaps you can post another MSQ and datalog made with that exact MSQ and I will look some more.

Lance.

[Bernard Fife]

gurov,

I apologize, I figured out how to recreate the problem (not surprisingly, the clue was in your original post). To disable the full time TCC mode, set the 'Lock At or Above RPM' very high (8000), and the 'Unlock At RPM' high as well (7000 or so).

But leave the 'At or Above Gear' at 3. This will work and re-enable the taper on the PWM. (It appears that the gear being out of range is the problem - sorry for the bad advice on this earlier!)

There is clearly a problem in the logic, and I will sort this for the next release (soon - a day or two - there are a few other things to fix-up as well).

Lance.

[Bernard Fife]

gurov,

The fix turned out to be simple, so I have made 2.00Dy code for immediate release (same INI as 2.00Dx). It is here:

http://www.msgpio.com/manuals/mshift/fi ... Dy.abs.s19

or here:

http://www.msgpio.com/manuals/mshift/V2code.html

This properly recognizes the gear, so you can set it out of range. The problem turned out to be a mis-placed bracket in a long conditional statement.

Sorry about the hassle, I hope this works as you hoped.

Lance.

[gurov]

cool, glad you found it. i will be loading this firmare on the controller within the next few days. i was driving around today very slowly, which caused the trans to start heating up. i saw temps over 200 degrees (205), at that point i lowered the TCC engagement to 2nd gear and 20 mph, which let me know that i need to make it taper from below 60% to locked, as there's a noticeable bump in the ride when it starts to engage, and then scale that out to a higher % in the higher kpa, if i need to make it slip less under boost.

and yes, i know 190 kpa is high for TCC. i'm running straight up TCC engagement into boost (it maxes out at 4.5 psi now on the wastegate spring) i've verified it to hold up to 7-ish psi. one thing i was pondering, the spot where a converter locks, does that depend simply on the engine RPM, or does there have to be a difference between engine rpm and input shaft ? my reasoning for riding the TCC is i can tell when that is slipping from the graphs, and back off where that does happen.
i am also still skeptical of how much power the torque converter itself can take.

25 mph for shift hysteresis is me trying to make it not hunt for gears. as i feel it still does quite a bit. i will see if i can get a more concrete log of that. for now, i only use auto mode rarely, and if i'm in stop and go traffic. if i'm on the highway, i most of the time just shift by hand, which makes me wish i had paddles on the steering wheel, but that's a $400-600 ish expense to get the right steering wheel that can have paddles added to it.

it also doesn't help that i'm still having a hard time wrapping my head around the table for gear shifts. i've been meaning to write some software to analyze the logs where auto_mode is manual, and basically plot out gear vs kpa vs mph, to see if there's a common pattern, basically make an auto-tune like offline dealie for the trans.

[Bernard Fife]

gurov,

Let us know how it goes.

To stop the trans from hunting in auto mode, it is better to space out the bins than to increase the hysteresis. When I first started this code, I thought of using fewer rows and columns, as this reduces hunting quite a bit. However, you already have many duplicate columns, so I don't know what to say. If you can get a log that shows the hunting, that would really help analyze the situation. However using a large hysteresis will cause the shifting to do some strange things (like if you have the gear table set up to shift at redline at full boost, but also have a hysteresis of 25 mph, the shift may not occur until you exceed the redline by quite a bit - definitely not recommended!).

It is possible to 'auto analyze' the shift table. The biggest problem is that the bins are critical, unlike a VE table where you have interpolation and a relatively smooth set of VE values. The gears are obviously discontinuous, and *must* jump at certain speeds/loads, so the analysis must calculate bins to be useful. There is a calculator in the tuning document to help set the MPH for given rpm (and tire size, final ratios, etc.). This can be used to determine the mph bins given your redline. It is here: http://www.msgpio.com/manuals/mshift/V2tune.html#ag

Lance.

[gurov]

gurov,

Let us know how it goes.

To stop the trans from hunting in auto mode, it is better to space out the bins than to increase the hysteresis. When I first started this code, I thought of using fewer rows and columns, as this reduces hunting quite a bit. However, you already have many duplicate columns, so I don't know what to say. If you can get a log that shows the hunting, that would really help analyze the situation. However using a large hysteresis will cause the shifting to do some strange things (like if you have the gear table set up to shift at redline at full boost, but also have a hysteresis of 25 mph, the shift may not occur until you exceed the redline by quite a bit - definitely not recommended!).

It is possible to 'auto analyze' the shift table. The biggest problem is that the bins are critical, unlike a VE table where you have interpolation and a relatively smooth set of VE values. The gears are obviously discontinuous, and *must* jump at certain speeds/loads, so the analysis must calculate bins to be useful. There is a calculator in the tuning document to help set the MPH for given rpm (and tire size, final ratios, etc.). This can be used to determine the mph bins given your redline. It is here: http://www.msgpio.com/manuals/mshift/V2tune.html#ag

Lance.

right, well i would have it show me stuff graphically, and then decide on what bins should be.

and yes, i fully understand that the bins are very important, and the edges are very very jagged, as you can't exactly interpolate between 3rd and 4th gears.

i will try the calculator.

[gurov]

VSS noise was causing TCC to unlock and taper back up to proper level. so i had to very quickly change the scaling to not scale. i've got some logs of this happening as i'm driving constant speed i can post later.

[Bernard Fife]

gurov,

Yes, the VSS noise will do that (among other bad things), and this will be worsened by having a very short TCC hysteresis time.

If this was mine, I would set the VSS masking to 70 to 100% (and never more than 100% - higher values appear to cause VSS n oise on my bench), set the TCC hysteresis time to at least 2.5 seconds (preferably higher), and the load smoothing factor to about 2000.

I will think about how to add a user-set smoothing factor to the VSS signal.

Lance.

[Bernard Fife]

gurov,

I have added a vss smoothing factor to the 2.00E code (here: http://www.msgpio.com/manuals/mshift/V2code.html). It is with the other VSS parameters (http://www.msgpio.com/manuals/mshift/V2tune.html#gv).

The factor works like this:

new speed =( current speed * (VSS smmoothFactor-1) + new speed)/VSS smoothFactor

The default VSS smoothFactor is 16, so if the current speed is 20 and the VSS indicates it has risen instataneously to 25 (usually because of noise), the calculation is:

new speed =( 20 * (16-1) + 25)/16 = (300+25)/16 = 20.3 mph

It recalculates this 100 times per second based on the then current vss periods, so the speed can still rise or fall quite rapidly.

The smoothing factor can be from 4 to 255 in the ini. The value in the old code was effectively 1 (which is no smoothing at all).

Lance.

[Bernard Fife]

gurov,

I managed to f-up the changes for 2.00E, and speeds above 102 mph cause the speedometer to act very strangely with the default values (it rises and wraps around continuously). I will post fixed 2.00Ex code shortly.

Lance.