I 3D printed a small enclosure for the Arduino and connections to sit in to. I've struggled to fine a nice way to connect the wiring to the arduino in a way that won't shake loose but I think with this enclosure it will stop the wires vibrating and working loose.

https://cimg7.ibsrv.net/gimg/www.mia...dbe3569d43.jpg

I'm using a single 18 pin circular connector which is just enough pins for the shift controller. I might still split the CAN, power and ground wires out of this connector as I still might need a separate wire for the autoblip and I don't want the pulsed signals interfering with the analog. Should I be trying to run shielded wires? I'm fairly useless when it comes to electronics.
4 x PWM outputs for the Pneumatics for the shifter
1 x PWM output for the Pneumatics for the clutch
4 x PWM inputs for the magnetic sensors
1 x analog input for the steering buttons
1 x analog input for the 150PSI pressure sensor
1 x PWM output for the shift cut
2 x data outputs for the shift display
2 x CAN connectors
1 x ground
1 x 5v output

I also managed to fry my shift display controller by accidentally feeding it 12volts. I was just busy connecting stuff together and realised I hooked it up to the wrong power rail. Oh well I've ordered another one ($5) but it will take a few weeks to get here.

The compressed air is fairly valuable... I could use it to open MAC valve or something like that though. But there are nice compact electric motors to do that sort of thing... and I don't want to have to have the air on in the car to start the engine.

I am not sure about an idle valve which is both fast and flows well.
The 4 wire stepper type idle valves are usually worm wheel drive which are not particular fast as far as I know.
The 2 wire idle valves like on the Miata engine usually do not flow very much.
Idle control needs to be stable: it does not make sense to use both a big and fast valve as it just makes the system unstable.
The supercharger valve looks usable, but it is kind of bulky.
So I am curious what valve is best to use.

Using compressed air would allow for reasonable flow with a small valve/solenoid. I would not touch the throttle and idle valve in that case: just add another solenoid to the IM for the blipping.

Or just do the traditional way and have a short throw pneumatic cylinder to kick the throttle open a bit.

Yeah I think this might be the best way. It will be easy as I have a couple of spare pneumatic cylinder I can use and I can hook it up to the throttle body easy enough.

I did a bit more work on the final mounting plates on my gearbox. I still need to make the last lever arm and trim up the back one to final length but I can only do this once I receive my final pneumatic cylinders which should be sometime next week.

https://cimg3.ibsrv.net/gimg/www.mia...a6d2ddf60d.jpg

I'll add some more lightening holes to the steel plate and the aluminium base once I have it all running. I also still need to make a seal up for the shift rod to stop oil coming out of the box. Everything else is done... I hope it fits into the transmission tunnel! I know the PPF clears at least.

https://cimg5.ibsrv.net/gimg/www.mia...24f9e0c395.jpg

Wow it's been nearly 4 weeks so it's time for another update. I've been waiting for parts to rock up and I've also been away on holiday for a few weeks so I'm keen now to get this finalised and running in my car.

It's all getting a bit more serious now!
https://cimg8.ibsrv.net/gimg/www.mia...c94fb5e08c.jpg

I ordered 2 x 3 position air cylinders and a single cylinder to operate the clutch (not shown). I'll machine up a billet connector so that a second clutch master will directly bolt to the pneumatic cylinder. I'll then run this second clutch master cylinder inline with my existing pressure line just like a hydraulic handbrake setup.

https://cimg9.ibsrv.net/gimg/www.mia...1f11639e36.jpg

You can see the nice little magnetic sensors on the image above. The LED light turns on when they are in range of the magnetic cylinder ring inside switching the input to the arduino. I have the gear checking code totally separate from the pneumatic code so that I can fire off a shift and then check for achieved gear afterwards. The sensor code also drives the LED display. This way if a shift is not achieved the system will error and won't allow further gear changes without my intervention.

https://cimg0.ibsrv.net/gimg/www.mia...4ab469a569.jpg
I am using a common manifold to connect all the control valves (2 for each shift direction and 1 for the clutch). Although I just realised I need two more valves to effectively control the 3 position air cylinders without slowing them down. It's so hard to find information on these sorts of cylinders... even the local experts have no idea! Oh well I'll just hang a couple more off one side.

https://cimg1.ibsrv.net/gimg/www.mia...349e34e62e.jpg
I'm going to run a 4L aluminium tank air tank driven by an ARB 4x4 air pump. It only takes around 15 seconds to build up to 80PSI. The tank gives me a decent buffer of air which should be enough for around 20 -30 shifts without the pump turning on. I'm using more air than a normal pneumatic sequential as I have to run multiple shifts axis to move a single gear position. With a sequential box you just move it up or down.

Most of my time now is in finalising the wiring loom. It's tedious work with now over 24 control inputs / outputs running to various sensors, relays and my ECU. It's all a bit crazy but will be awesome when it's working... if it works!

https://cimg1.ibsrv.net/gimg/www.mia...1539116231.jpg
https://cimg4.ibsrv.net/gimg/www.mia...cf9a40cc63.jpg
https://cimg1.ibsrv.net/gimg/www.mia...73aa45ceb5.jpg
https://cimg8.ibsrv.net/gimg/www.mia...3dc4b2e30e.jpg

Minor update:

I have finished the mechanical side with the pneumatics mounted to a cradle that directly push and pull the shifter cables. The gearbox levers system some final trimming and the cable mounting plate needs painting but other than that it is fully functional. I'll loctite everything on final assembly.

The arduino is fairly crazy niw with over 42 wires connecting various sensors and outputs. To fire these 3 position cylinders I need 3 x 2 port valves per cylinder meaning I need 8 valves total with clutch and idle blip control. This all requires some mad coding to fire the required valves to acheive a move. It's all gotten a bit complex but I'm pretty sure its doing the right things and it wirks great on my test bench.

All I need to do now is wire up the valves to the relays and apply some air pressure. I have to be a bit careful though because these air cylinders might have enough grunt to bend shift forks etc if the dogs aren't in alignment. My plan is to hook up an electric drill to drive the input shaft at speed so that the dogs can bounce and engage. I can even run the drill from a relay so that I can test shift cut outside the car.

One more week and I should have a video of it all running.

Great stuff!
Did you install this after all?Any videos?

I'm in risk of this thing becoming vapourware. Nearly 12 months late but here is a quick video of it running on the bench.



I've been busy with work so this thing got shelved for a while. It still needs some tweaking and the whole shift order needs flipping 180 but I can cycle through the gears. In this video it's running at 25 PSI to reduce loading but eventually I'll push it up to full pressure and speed. I' need to sort out the code and start tweaking the timing.

I promise the next update won't take a year!

First time seeing this thread and I'm in awe! Great job and I can't wait to see it finished and running in the car!

yeah i dont even know where to begin. You are amazing with this stuff. sort it out, then help me set one up for BMW trans/K24 miata. Then Ill have to build that engine just so I can hear it shift at 10K

I think a mechanical sequential would have been easier. There are some difficulties with paddles and pneumatics in that they are basically digital. On or off.

Getting the car to engage 1st or reverse is actually pretty challenging due to the dog teeth. The faces of the dogs are fairly large and flat so if they aren't in the perfect spot they won't mesh. Normally you just put some load on the shifter and gently release the clutch to get the input shaft rotating and when the teeth align it clunks into gear. I'm going to have to find the balance between stalling and launching the car forwards.

In hind sight I should have kept the gear shifter so that I could manually select 1st and R then use sequential when running.



The reason the WRC cars jump fowards when selecting 1st is due to this reason. The clutch never fully disengages so the input shaft is always turning. I have to do something similar.

Wow, nicely done! You could experiment with a bushing instead of a bearing for the pilot bearing. With a little more drag than the bearing it might keep the input shaft turning enough to let the dogs engage properly.

Would you consider producing these setup for regular shift patterns? This is insane! Great work man!

Holy shit, that is awesome.

That is fucking amazing. I have just enough experience to know how huge an effort this was but this is so far beyond anything I could do myself.

I don't even think about it before I start... I just dive right in there and learn as I go! If I sat down and thought out all the issues I probably would never attempt it.

I did have another play with the code and I've sorted the shifting order and fixed a few bugs. One thing had me stuck for a while until I realised I was using the same arduino pin for the shift cut and the one of the pneumatic solenoids so it was being turned off by another function. It meant it wasn't quite engaging the gear properly. My code is pretty messy and I as I've added functionality I've made it even worse. Once I know what I'm doing I might just rewrite it from scratch.

The box is cycling up and down through gears smoothly whilst being driven by the drill. When running it I realised I need to change the way I'm shifting out of neutral because one bump of either of the paddles engages a gear. I've accidentally tapped the paddles a few times whilst adjusting something mechanically and those pneumatics pack a punch... plus it's probably not good to engage a gear by accident when in the pits. I'll make it so I need to hold the shift paddle 2 seconds before engaging 1st or Reverse.

The next stage is to bolt it into the car and sort out the CANbus integration, shift cut, startup procedure etc. Based on the bench testing I think I'm at the point I could safely use it to go up through the gears no problem. I've done it a hundred times now without an issue. Stepping down through the gears I will still need to use the clutch, hit the down shift, manually rev match and release the clutch oldschool way until I build something to autoblip. I just need to be confident that the gear shifting works and selects the appropriate gear in order without any chance of a miss-shift. If I get 3rd instead of 5th or even R I could easily end up with destroying a motor or putting the car into the wall.

Well this is pretty fucking bitching.

Shame the cam design got ditched. Few moments on a CNC and we'd all have sequential shifters.

What do you estimate the total weight of all the components to be?

I haven't ditched the mechanical sequential idea... Last week I actually dreamt up a new concept which is even better than my first design with only a single main moving part. The 3D designing of it is pretty complex but is really only the first step. The hard part is getting it to function without issues. I think the new design will work really well as there are less wear surfaces and less overall mechanics. In my head the geometries all work... I think.

Here is another test video showing the shifter cycling up and down gears whist the input shaft is driven by a cordless drill (I have issues with a power drill causing emf through the earth... not sure how to fix it). I've added code that requires a 2 second hold to move it out of neutral either into 1st or R. I have also added the functions that only enable the next gear if the magnetic sensors detect that the pneumatics successfully entered the target gear. If a gear fails to engage, then the shift controller will set the corresponding selector paddle to retry the same gear instead of moving onto the next. This is now all the code implemented that can be tested on the bench. All up it's looking fairly resilient to errors which is good so I think the next stage is to bolt it into the car and sort out the shift cut and CANbus integration.

Please use a tripod :D

Looking great.
Is it weird that neutral had the prop spinning?

Just had a realisation that whilst wonderful all this comes from the engine bay which is quite tight for space. The boot is pretty empty.
Would recommend a dip switch to enabled reversion of the cables.

Measurements of the gearing would be interesting to all I'd think.
Distances and angles from center point sort of thing.

The output does spin in neutral, but will stop if I hold it. There must be some friction from the input shaft onto the main output shaft that spins it unless there is some load.

I think the distances I measure on this box are different to a standard 5 speed and also a 6 speed. The gates are all similar but how far the input shafts move in and out depend on the sizing of the dog teeth which allow for different movement to the synchros. Also the straight cut gears are much wider and everything is shuffled around in terms if gear placement.

Part of the tweaking has been adjusting the amount of movement as R and 1st to some extent requires a longer move than 2 to 5th.

I've decided to build up a spare engine to test this box. I'm worried about damaging my race engine which is fairly unique. I'm also going to use this 2nd engine build to experiment with a couple of ideas. So I'm going to run my lightened crank, spin it up high and see if I swiss cheese the block. I'm also going to run my ported head with SR20 valves and Astina cams. It's going to be an interesting build in itself.

Hydraulic coupling is why it spins in neutral

Great work!
I'm looking at doing something similar for my 6R4 project car as the Ford Sierra Cosworth gearbox in in the back, the gearchange is set too far back in the cabin. I have a mechanical converter to revert the change direction to standard but it is a bit clunky and a sequential conversion would be ideal.
One question, how do you stop the cylinders holding pressure on the change forks after a gear is selected?
I see the MME motorsport kits have a linear pot on the cylinders presumably so the system can see the amount of throw achieved.

My system uses magnetic sensors on the cylinders to test if they achieved a successful shift. They are also used to display to the LED gear indicator. Once successfully in gear, the pneumatic valves shut so that there is no load on the shift forks, but I've adjusted them so there is some wriggle room on each gear.

Basically the logic goes like this:

Detect shift button UP
Send signal to ECU to cut ignition, wait 5 ms
Fire pneumatics to move out of gear into neutral wait 10ms then fire horizontal move if required to the next gear plus another 10ms wait
OPTIONAL: Wait for revs to drop to within rev threshold via CANbus rpm signal
Fire pneumatics to move into gear
Check magnetic sensors to see if gear was achieved
Disable ignition cut
Turn off pneumatics

Detect shift button DOWN
Fire clutch pneumatic
Fire pneumatics to move out of gear into neutral,
Turn off Clutch
Fire Throttle opener
Wait for revs to raise to within rev threshold via CANbus rpm signal
Disable Throttle opener
Fire pneumatics to move into gear
Check magnetic sensors to see if gear was achieved
Turn off pneumatics

I need to get this box installed in my car, the only thing holding me up is that I don't want to run it on my race engine in case I hit the wrong gear... so I'm building up a spare engine. The spare engine is a bit ambitious as I'm running a chopped crank, DIY ported head with cams out of whatever I could find. Plus I've had a few side projects on the go that has kept me busy for the last 3 months. I'll get there soon.

Give yourself a pat on the back, pal. Pretty impressive.
I've some experience myself with Arduino when I built a dual sonic height sensor for the air suspension on my van. Great to work with.
Mines a syncromesh box so would be a little easier than your dog but I think I'm going to struggle with space in the gearbox tunnel to fit the linkage. I'd have to make a plate to bolt onto the top of the gear lever mounting holes inside the car to mount the linkage and cylinders, that should work.
Do you have a parts list so far? I'm going to start sourcing parts ;-)

I don't think pneumatics are gonna play nice with a synchromesh engagement, they work well for MadJaks dogbox cause dogs have to be shifted hard and fast. Synchromesh would likely just jam and not go into or out of gears cleanly, too many teeth in the way, not to mention synchros themselves.

I'd suggest looking into how the BMW SMG gearboxes worked if you wanna pursue the idea. Similar to what you're talking about, they were a regular synchromesh 6 speed with actuators fitted to shift gears, but because of the synchros I believe they had to do it all with the clutch fully disengaged--which to me kinda defeats the purpose, it just becomes a clumsy automatic.

Plenty of cars use ASMs (auto-shift manual) which are just that to drive, a clumsy auto but that is still better than a clumsy manual becuase my lever is in the wrong place.
No good me looking at the BMW system, not going to help my Ford gearbox. My application is more about gear selection than clutch disengagement and I wouldn't need necessary engine rpm match needed for dog changes.
My current gear selection is a problem and the syncros should slow the gears to allow the change even with air change. I don't need a dog box as my syncro box is already strengthened for 300bhp. Anymore power and I would have to go for the dog but certainly not for the gear change quality unless it was dog and sequential :-)

I don't understand--what vehicle is this in that you can't reach the shift lever?

As for the BMW trans, I just meant to look at it for setup guides, not to actually use parts from it. From what I've read the BMW trans is unique in that it really is just a manual gearbox with servos added, and can be converted back and forth to a proper H pattern box. The ASMs you're talking about are just selectable automatics, and don't share any internals with a manual transmission.

Ah. You've answered a post without reading the thread :-)

You _might_ be able to get pneumatics to work on a synchro box by running lower air pressure or some sort of spring mechanism in the linkages that applies a softer constant force.

My setup uses 2 x 3 position air cylinders each controlled by the 3 x 2way valves. This is a fairly complex way to do it and I found it really hard to get any info on how to control it. It is far simpler though to use 4 cheap ebay air cylinders and piggyback them in pairs. Not only is the valving more simple but the logic is easier to code.

I spent some time tonight putting together the bottom end for my spare engine. It is a crazy build with parts recycled from all over the place plus being a test rig for a few ideas. Here are the specs:

NA8 block, oil squirters removed
Oil pump with oil pressure relief valve blocked up
2000 pistons and rods (higher compression)
lightweight bladed crank
Billet water pump bypass plate
BP05 head, ported, cnc chambers, decked 1.5mm
Nisssan SR20DE valves (+2.15mm I and +1.65mm E)
Protege intake cam (2Y)
VVT exhaust cam
SuBs, Eibach springs

Most bits are either left overs from my race engine, machined by myself or picked up cheap / free. I think all up it will cost me less than $300 including gaskets, rings etc. It will probably also cost me 200 hours too.

Oh now I see, you started referencing arduino controls on a van of some sort so I got confused.

I really think the pneumatic setup for a synchro box is not going to work and is vastly overcomplicating the system. like MadJak mentioned, you'd need some sort of constant, gentle pressure to allow the synchros to do their job before the next gear will engage. Why not just run a FWD shifter setup wtih cables? Surely that will be more effective from both function and cost perspectives, even with adding some bell-cranks like MadJak built.

The pressure and speed of the actuators is easily controlled with inline adjustable valving.
I could make a manual cable system which would maybe give a better feel than the rod linkage system I have but ideally a H pattern to sequential converter is what I'm after. If I had a spare £15k I could just buy a Quaffe sequential box, but alas no. A mechanical drum system converter would likely be better too, but I've not found one yet.

This sounds like it needs a build thread.
Ive always wanted to build my own "Group B" 6R4

https://cimg3.ibsrv.net/gimg/www.mia...f05193c213.jpg
Took me 4 years, raced at 6 events last year.
https://cimg9.ibsrv.net/gimg/www.mia...f8236e46dc.jpg
Gear change much improved when I put the reverse gear lock-out slug back in the box but could be better hence my interest in the thread.

Maybe a 12v electric linear motor may be better for me, mounted directly to the shaft through the end of the box.

Thats probably going to give you some really slow shifts unless you get a bunch of torque in them and use really short linkages.

A company in the UK called proshift sells some beefy solenoids for motorcycle shift kits. We're trying that on our FSAE car this season.
They are huge but fast.

These shut down solenoids are plentiful and have two circuits, pull and hold. Not sure if it is useful in this application but I thought I would mention it.

https://www.ebay.com/itm/New-Fuel-Sh...9/281632465737

I Guess if you're going to act directly on the shift rod you really only need a pretty short 3 position linear setup to push in and out on the rod, and a 3 position rotary setup that you could probably do with a permanent magnet motor, timing belt/pulleys and 3 limit switches.

https://cimg5.ibsrv.net/gimg/www.mia...09db8f0cef.jpg

I've been thinking about throttle blips on downshifts and came up with a fairly simple idea. Since I already have compressed air onboard I may as well use the small air cylinders I ebayed when I started the project.

I'm using the air cylinders to push the throttle cable outer sleave backwards to open the throttle plate. They have enough strength to punch open the throttle nice and fast and will return to a closed position under the spring power from the throttle. This should work so that if the shift controller loses power or there is air pressure loss, I'm not going to be ontrack with a throttle that is stuck open. It also means I can have my foot on the throttle and the blip will just add some extra throttle without too much feedback through the pedal... I think.

The next step is to hook it up to the arduino shift controller and start decoding the CANbus so that I can test how responsive it is. I'll wire up a button on the dash that will trigger a throttle blip up to a set rpm, say 6000 and try to hold there by cycling the pneumatics on and off. This will test both the speed of the cylinders and the delay from reading the CANbus rpm. (It should be 20ms)

Nice idea, just make sure the cable end doesn’t unseat from the gas pedal when it suddenly retracts

yeah, I'll just have to see how it works in car. This design relies on the throttle spring pulling it closed as soon as the pneumatics are turned off which should keep tension on the cable the whole time. The s90 throttle body has a really strong throttle spring so I think it should work ok.

My other design was a rotary cam sitting over the throttle wheel that engages a pin that sits in a slotted hole in the throttle wheel. When the cam rotates it pulls the throttle open with it but allows the throttle to rotate independently. I could then drive the cam wheel via a thottle cable and remote pneumatic air cylinder, but this would have created the loose throttle cable problem you mention and also the potential for it to bind and jam. So I went back to my first idea which looks fairly inelegant but it's nice and simple and safe. The only issue I can see will be rubbing the throttle cable on things as it moves forwards and backwards.

Use a set screw to secure the cable housing to the mount/socket

SAR
 

Where did you bought those 2x3 position air cylinders?

Well, that's one way to revive a 4 year dormant thread....

Pneumatic
 

When dormant it like you saying dead, it has risen :)