Information about gear shifting

[federico_raceup]

Hi everybody,

I’m Federico Scacco, drivetrain member of the Race UP team from Padua University, Italy.

During last year competitions our gear shifting system was based on a servomotor, controlled in position. The servomotor was chosen after a static measurement of the torque necessary to disengage and to engage the gear.
The point is that we don’t have the possibility to use a load cell to acquire dynamic data so we don’t know if we have enough torque to complete the gears’ shifting in every condition in the track.
Data acquisition are clear: sometimes we have some problems with the engagement but we are not able to appreciate the cause of the problem. It may be the cut-off logic or perhaps the lack of torque delivered by the servo.

It would be very helpful for us if somebody of you tells us the force measured with a load cell. I know this data are influenced by the operating conditions of the car and also by the kind of engine but we would only use this value to compare our prediction with an effective data.

Thank you for the attention,

Federico Scacco

[apalrd]

I don't have an exact torque or force number, but we found that static measurement was at least an order of magnitude off at high engine load. Torque managed upshifts under high load resulted in less demand on the shifter system, reducing it to somewhere in the ballpark of 4-8x the static measurement. Currently we design around 10x the static value as the nominal load.


This torque on the shifter required to disengage a given gear is highly dependent on the torque transmitted through the gear/dog, as the shifter is fighting the friction of the gear and dog under load. Reducing engine torque will help this significantly.

[Jay Lawrence]

To do this in a simple manner: just get a fishing scale from a fishing shop/ebay (it will just be a spring in a tube with load indicated on the side) and use it to actuate your shifter. Make sure you actuate it in the same direction that your shifting arm does (should be perpendicular for best results. If you are actuating the shifter barrel directly (in rotation) then you might have to do some maths) and you will get a reasonable value. You should then add safety factors as mentioned by Andrew above. Alternatively, simplificate!

[MCoach]

I have a question. Does your servomotor happen to be brushless? I ask because you say that it is position based. If so, are you using an industrial or custom controller?

[federico_raceup]

Thank you for the answers.
@ Apalrd: I understood you measured the static force and you multiplied by 10 to obtain the maximum force to apply,is it correct? May you have some numerical data, perhaps?

@Jay Lawrence: your idea is very easy to try and cheap at the same time. I think we should follow your way to have an order of magnitude of the force. In this way we can measure the force we are transferring, but we don't have an idea of the force necessary to successfully complete the gearshifting. This is the reason why I asked here if someone has already estimated the force, or acquired with a load cell. I make you another question: what kind of actuator do you adopt? Is it pneumatic or electro actuated?

@ MCoach: Our servomotor has got a DC motor and it has got its own controller for the position control.

[MCoach]

Since you're doing your own controller, try adding a current shunt to your system to measure the output to the motor. By looking at the current feedback it'll tell you how much torque your motor needs to overcome as it tries to move from one position to the next. The shunt itself will cost about 10 cents and gives the same data that your looking for as with a strain guage.

You can look for where to end the shift when you see the tellltale torque peak where the torque is high and then drops suddenly as the engine falls into gear and peaks again as it keeps pushing and is in gear. Then release the shifter and continue on with life.

[apalrd]

The data I have:

-We measured the torque of the shifter on the engine. This is how we got the base number. We did this several times with the engine on the bench, in different gears and such.

-I have data captures of a uncalibrated strain gauge which we used for spark retard during shifts. Based on the initial tests above, we sized the strain gauge amp, but with a voltage change of roughly 50 bits (10 bit ADC) with the engine off, we ended up with over 300 bits (maxing out the ADC) under load. We then lowered the gain on the amp and ended up with somewhere in the realm of 10x the voltage when loaded vs unloaded.

-We designed an auto-shifter using the static data from the first step, designed at 5x torque of the static measurement nominal, and it was unable to upshift in some conditions. The maths say this would need at least 8x static torque to happen, depending on voltage.

[federico_raceup]

Thank you a lot.
As far as concern the current measurement nowadays we are working about it and some interesting results are coming out. We also realized that gear shifting problems are also due to to the servo velocity.
428
The picture shows that in correspondence of the vertical yellow line the pink bent is up of the horizontal red line. The red line is a limit below of which the servo velocity is not enougth to complete the shifting. We estimated it around a value of 100 mm/second. When the resistance force is higher than a critical value, probably when the car is accelerating and maybe wheel-sleeping, the velocity reached is too low and the gear finishes not to be disengaged. In order to validate our hypothesis we tried to change the servo arm in length (increasing it) and we obtained what we expected: the number of missed shiftings increased.

Thanks to Apalrd we have another confirmation of our problems. You have just said that you oversized your actuator about 8 times the maximum static torque and this was our main failure.
Is it also an issue of velocity or is rather a question of torque, according to you?

@Apalrd: What kind of actuator do you adopt?
Thank you for your time.

[federico_raceup]

P.S. I hope the picture is clear, let me know if you have some problems to look at it.

[tromoly]

It is impossible to read your picture, it shows up on my screen about three inches wide.

Why are you running a servo? Why wouldn't a pneumatic cylinder work better? Or an electrical solenoid? In no way criticizing your choice, looking to get your justification and number results to back-up your choice.

[Jay Lawrence]

Federico,

Velocity isn't really the issue as you can just increase your ignition cut times, but at that point you are slowing your shifts and the whole point of the system is lost. What engine are you using? If you've done your gearing appropriately and have a reasonable power spread then it's likely that on a typical FSAE track you will only use 2-3 gears. Personally I don't get why people bother with complicated shifter systems when a simple hand actuated stick does the job with much less headache.

[Swiftus]

...I don't get why people bother with complicated shifter systems when a simple hand actuated stick does the job with much less headache...

Because money-shift.

There are a number of positive reasons to complicate shifting, but nothing beats a hand-actuated stick for reliability. It comes with closed-loop control and on-the-fly programming! :)

[federico_raceup]

I agree with you, Jay. But I think the driver has a lot of other thougths during a competition and I find quite disappointing to obligate him to change gear manually.
The image above is a part of data acquisition in which I show the velocity of the servo. I realised that every time the velocity is lower than a specific value, the gear isn't disengaged. As I said before, our problem should be a lack of torque.

I would like to make this discussion a clear schedule of the prons and cons of different actuators. What do you think? In this way everybody can have an organic vision of the topic, without headache in searching for dispersed information.

Hoping to help people in answering their questions,
Federico

[tromoly]

Federico, your image is very small and unreadable. If you would like anyone to review the data and give their input, please upload it in a larger size.

Many successful teams run a hand-operated shift stick, such as Monash, UTA, Missouri S&T, Maryland, and many others I am forgetting. And looking at professional racing there are many series that use a hand-operated shift stick (BTCC, WTCC, WRC, many lower formula categories, shifter karts, etc.). Don't rule out a simple sequential shifter.

[SomeOldGuy]

The manual option is not a bad option.

We here at WWU have, in the past, run with a mechanical paddle shift utilizing a push pull cable meant for boat throttles. We have since gone back to a lever though to reduce hand fatigue.

Many have also run a variation of this setup, integrating a clutch into the downshift side of the paddle with the ability to pull the clutch separately. Oregon Institute of Technology ran one like this in 2012 that was really simple and very functional. I mention this one as it is one that I have seen myself, but I'm sure there have been many others like it.

I know that this does not get at the specific question of OP, but mechanical shifting should not be thrown. It is a reliable solution that will not hamper the preformance of the car in any large way.

[Homemade WRX]

If you're driver is having an issue operating a sequential shifter, you have driver problems.

If you can go faster because of an electronic shifter, and it's cost effective, then you have reason to improve.

[MCoach]

Because money-shift.

Have problems at some point, Jay?

[Swiftus]

Have problems at some point, Jay?
Only on Chump teams with some chumps :).

[federico_raceup]

I have upload a bigger image, it is clear enough to understand what I mean:
449
The picture shows that in correspondence of the vertical yellow line the pink bent is up of the horizontal red line. The red line is a lower limit, below of which the servo velocity is not enougth to complete the shifting. We estimated it around a value of 100 mm/second. When the resistance force is higher than a critical value, probably when the car is accelerating and maybe wheel-sleeping, the velocity reached is too low and the gear finishes not to be disengaged.

There is no doubt as far as the reliability of an hand actuator. From the other side it is very difficult to predict if an electro-actuated shifter is relevant in terms of seconds per lap, but it is also true that the electronic one makes a more precise driving.

I want to do anything but to say that one is better than the other, it depends on drivers' habits, it is only a question o design, the result of a choice.
There isn't the perfect actuator, but a perfect actuator for every team.

I think we solved our troubles, at least we recognized the main problem.
Thank you everybody for having actively partecipated in the discussion

[VFR750R]

Not to digress, but the 2014 Rolex 24 hours of Daytona overall went to a car with manual sequential shifting in a series where pneumatic shifting is allowed. The car did have flat shifting capability (no clutch up or down), but those transmissions are basically oversized motorcycle transmissions.

I've become familiar with pneumatic (megaline) which is used throughout Motorsports but I have been interested in electronic options which should be a bit more precise and less physically complex (no separate compressor, valving, lines etc).

I don't see much in your data, but drum position is probably the most important shifting feedback. It's critical information to recognize when the current gear is not being released(typically not enough power reduction, but also a function of drivetrain stiffnesses), when you hit dog to dog of the next gear and when full engagement occurs. That information enables easier fine tuning of timers and adjusting the logic to recognize the next gear. I will say that downshifting is probably more complex than up shifting, especially with FSAE and rules regarding the throttle.