This is just an idea, but there's plenty of smart people here so I wanted to get some feedback on the best hardware to build this.

The goal would be to build a shifting system that controls the mazda 6 speed transmission. It would need to shift faster than a human can to justify it adding weight/complexity to the car. It would also need to shift reliably, no missing gears or breaking down when in use. Reliable enough to daily drive without any issues. Must be able to upshift/downshift from 1st to 6th. Would be push button or paddle shifted.

The part I want to ask about, is what would be the best way to move the shifter? Since it's an H-pattern, you really only have 2 axis of movement. I'm thinking a push/pull solenoid on each axis, coupled with a linear potentiometer on each axis to measure the position.

Electric push/pull solenoids are considerably more reliable than a pneumatic system, and simpler/lighter too. With the potentiometers to measure their position, you could give them power long enough to make a half-movement (example going from 2nd to 3rd, you need to go forward to neutral, then over, then forward again).

A 2nd to 3rd gear change would be:

Clutch in

Power front/back solenoid forward to 1/2 way point, then shut it off once potentiometer shows it's there.

Power left/right solenoid right until potentiometer says it's in the middle.

Power front/back solenoid forward until it's in 3rd.

Clutch out

Keeping the movements electric allows for precise on/off so that all movements can be made with just 2 actuators. Downside is you'll need electric push/pull solenoids that are very strong to shift a transmission.

For the clutch, I don't know how I would control that yet. I'm afraid it would have to be pnuematic, I don't know if anyone makes an electric solenoid strong enough to smash in an ACT Extreme.

Thoughts? Is there a better/easier/lighter/more reliable way to build an automated shifting system? I know enough to do a basic micro controller to control all of this, just curious what the best hardware would be to do this with.

 

Interesting idea. Haven't heard of it being done before.
Easy button would be a quafie sequential and pneumatic paddle shifters.

What are your shift times. I have a hard time believing that there are electronic actuators that move the gear stick fast enough to make a noticeable difference over shifting by hand. I assume there is a mechanical limit to how fast you can shift gears.

Ideally you would want to be able to downshift too. Which would require throttle control for rev matching.

 

I'm sure it's been done, but I haven't found anybody do it with electronics. It's been done with pneumatic systems before. Those are the fastest, but I don't want that complexity/weight in my car.

Not interested in spending thousands of dollars to get faster shift times. More like a few hundred, maybe.

My shift times vary. From full power in one gear to full power in another gear, about 0.200 seconds is the best I can do, but that creeps up to 0.300 seconds sometimes. I think my left leg with the ACT extreme is what slows me down the most. I'm not consistent, and I'm not that fast either.

The system will have to upshift and downshift, that's a requirement. Rev matching would be nice, but not focusing on that right now. That can be a project after the shifting system exist and actually works as designed.

 

I had some friends that developed something like this for the r53 mini. M7 Paddle shifter install - North American Motoring

There is more information on how they did it if you start google digging.

 

Ouch

 

Yeah I was never interested. TBH they should have marketed it to the porsche guys as no one would ever pay for it. The electronics and solenoids were what made it expensive way back when. All of which is cheaper now.

 

I've thought about it before and don't think it's worth it for a synchronized OEM gearbox, but it's fun to think about anyway.

What about using two actuators in line? Both actuators retracted is first gear, one actuator extended is neutral, both actuators extended is 2nd. Then two more actuators for 1st, 3rd, or 5th. Or I guess 3 if you want to use reverse on a 6 speed. Seems like it would be easier and more reliable in the millisecond time ranges you're talking about. You would just need extended/retracted switches on each actuator instead of pots, and you could fire the actuators at full speed without having to worry about overshooting a position before you could stop the actuator mid-stroke.

There's also the possibility of designing a ratcheting drum selector like what's used on real sequentials. It would probably be difficult to implement with the anglular changes of the gear lever, but it would be the fastest and most reliable option.

 

Thanks for the link and info. Yes, today the parts to build a system aren't that expensive. For the shifting part, I can weld and fab most of the mechanism that would be required, and for solenoids, I found some 14 lb force, 1.5" movement solenoids for ~130 each. If I could get away with only one per axis, that's 260 for actuators. Linear pots are cheap, and the rest of the electronics could be as simple as an arduino, a few transistors, and two push buttons.

Of course I don't know if a 14lb force actuator would cut it. Maybe.

This one: Murphy Swich RP2309B 12 Volt Push Pull DC Solenoid Rack | eBay

 

Interesting idea. I guess the downside is more actuators, but no overshoot. I don't know if overshoot would be an issue yet. I think with the linear pots, i could program the controller to compensate for that.

Say if 0% = 2nd gear, 50% = neutral, 100% = 3rd on one axis, then if shutting it off at 50% lets momentum carry it to 65%, shut it off a bit early until it has minimal overshoot. I don't think this would actually be a problem though, and if it were, could be solved in software since with pots you have feedback.

Also just thought of this, but you could also design the controller to handle overshoot by having it reverse the current to stop it from moving. Example, from 0 to 49% it's forward at 100% speed, and at 49-51% it shuts off, and if it overshoots to 52%, controller applies reverse current to bring it back.

So that problem is solvable no doubt.

 

There is (or at least was) many many threads and quite a bit of info on that system available online. They built quite a bit of very obvious yet smart logic into it. If memory serves you were able to select the next gear and it would wait for a signal from the clutch switch to initiate the shift. It would show the next gear as flashing on the gear indicator until it completed the shift... you could cancel it by clicking once on the opposite shifter.

Ultimately they designed a rally style bump shifter which was fun to play with. It was capable of very fast shifts and had to be toned down in order to not destroy synchro's. These are all obviously computer science 101 problems to solve with some very simple code.

I got to drive one of the cars and came away impressed. I expected it to be pretty hokey and pointless but it was pretty fun.

 

It wouldn't make sense to do any of this unless the clutch was also controlled by the computer. I don't know anything about designing computer controlled hydraulic systems, but I imagine you'd need something to generate pressure and then a series of electrically controlled valves to route that pressure towards the various pistons that move the shifter. That would be pretty quick and pretty reliable I imagine.

I think you could get big performance gains in terms of shifting speed by removing the stick interface entirely and just directly actuating the gear selector forks. Better yet, instead of having pairs gears, each of which share a selector fork, why not just have a bunch of forks/gears that are like a couple of milimeters away from engaging. It would feel retarded to have a 7 wide transmission gate that only had gears on the top instead of both top and bottom, but that's only because of the stupid stick interface. This is a better approach because you can just slam the gears out of engagement without having to worry about overshooting and accidentally engaging the opposite gear on the fork.

Of course if you somehow accidentally engage two gears at the same time, you're completely boned because there's no mechanical lockout (ie, on a stick based system, you can only engage one gear at a time and you can only go from one selector fork to another when they are all disengaged. Directly manipulating the gears with only software lockouts could lead to the shedding of tears. Obviously some experimentation would be required.

Anyway, this would obviously cost dollars.

Edit, you'd also need to convert the car to drive by wire so that it could rev match up and down shifts.

 

Agreed on the clutch, that's the slowest part of the system on my car, trying to smash the clutch down quickly. A simple push button that instantly smashed the clutch in would cut the shift times down a ton on my setup. That would have to exist for this to work and be faster.

Good post on the rest too. I don't think I would take that approach, though it could be faster/better, it would be wayyyy more work I think, and the possibility of accidentally hitting 2 gears at once I don't like. I think I could make this work by keeping the shifter in its current position and be ok with it not being as fast as directly actuating each individual gear with its own actuator.

 

Subaru did exactly what you're describing.

Cost them 70 grand, but it was bulletproof.

 

Swap a powerglide. Profit?

 

Building on AlwaysBroken's thoughts:

Ditch the shift lever and work only with the shift rod that runs into the trans. With that lever out of the way you only need one point of attachment for any mechanism to operate the rod in a combination of push-pull and rotation movements. There are rotational as well as push-pull solenoids, so perhaps a push-pull type mounted in a housing that is turned by a rotational type solenoid. Or the housing could be rotated with a simple servo motor and gear drive engagement.

There is a hydraulic TOB fork replacement for Evo8/9's already made (basically a hydraulic piston) that you might be able to adapt to work in the Miata trans, and then you could control the clutch engagement with a dedicated hydraulic system, perhaps one poached from an Evo9 with an ACD transfer case.

Edit: TOB pic

 

VW also built something very similar to this idea for their VW Lupo 3L/100km hypercar (I am not sure that thing was ever sold in the USA).
They also automated their manual tranny with a complex system.
You will find the technical description of the whole system here (it is in German though, but I can read that if you really want to know some details):
http://www.daffieproductions.nl/pdf/3.pdf

Mercedes also had some automated manual boxes back in the early 2000's.
I remember Smart cars having them and my father owned a Mercedes Sprinter van with an automated manual box. If there is one thing these boxes have in common, it is going to be the terrible slow shift speed. As increasing shift speed is your main goal: it may not be so easy..
Automating the current box will most likely not yield any improvements, I think the mechanical design needs to be suitable for easy/quick/stable electro mechanical control. The efforts I have seen from OEMs were not worth the trouble.

In the end I think it will not be worth the effort. I think a better start would be to look for a different transmission.

 

You know they make something called an "automatic transmission", right?

LOL

Reading this thread, My mind was exactly in the same boat as matt's.

Stop thinking like a human and start thinking like a machine. Why are you worried about controlling the shift lever? You need to focus on controlling the shift rods.

This requires four servo-solenoids, each being of the push-pull type. To shift from 2nd to 3rd, the solenoid on the 1-2 rod shifts from down to neutral while the solenoid on the 3-4 rod simultaneously shifts from neutral to up. Reverse doesn't even need to be a servo-solenoid, just a plain-jane limited in/out.

Set your shift computer up so that it prepares for the next gear within 1k rpm of redline. Anytime you are within 1k rpm of redline and you push in the clutch, it does an automatic gear change. Your gear shifts then simply become "push in and release clutch as fast as possible" If the solenoids are quick enough, then alter your clutch pedal such that you can push it in with your foot and then immediately slide your foot off to the side allowing it to slam back into position.

Your problem is going to be some overwhelmingly fast syncro wear, I think. Gently starting the engagement with our hand operated "fast shifting" is one thing, but throwing a machine into the mix that is going to try to mesh gears nearly instantaneously is going to be a different beast.

 

I don't think the synchro wear will be a problem, as long as the clutch is not dragging/can fully disengage (which kills synchros in any trans), and the shifts are timed properly so that the clutch is not re-engaged until the selector ring on the requested gear is fully seated.

The two automated manuals transmissions that I work on (EvoX TC-SST and GTR GR6) rarely have worn synchros, as in one synchro per 20 trannies, as contrasted to our traditional manual units where 2-4 per transmission are common. The automated manuals are certainly far more complex devices than anything garage-engineered and retrofitted to a manual 'box will be, but the basic principles to prevent wear should be the same.

 

Aren't both of those gearboxes of a staggered-gear twin clutch design?

While I know nothing about how the software on those transmissions is configured, with such a design you could have the synchros engage very gradually. It wouldn't matter if it took a full second for the engagement to occur, since synchro engagement time is not a limiting factor in how quickly such a transmission can shift.

The exact opposite is true here. With a single clutch design, you need to slam the synchros together as fast and as hard as possible if you want to speed up the shifting process, regardless of whether a human or an electronic device is moving the forks.

 

If the computer is matching revs on up/down shift, why would there be synchro wear?

I figure you could probably do some sort of half-*** robo-shifting solution for the miata but I wouldn't bet on it being quick. The R&D to do it the "right" way will cost more than just buying a sequential box. And even then you'd probably be at parity with ferrari, circa 20 years ago.