A few comments:

Stratified combustion, at least in piston engines, is always at very low loads. It's a fuel economy thing and an emissions thing during warmup. The engine switches to conventional combustion at higher loads. I'd be surprised if a rotary were any different.

Earlier injection timing allows more time for mixture formation as we've already discussed. The rotary combustion chamber moves, so early injection requires injector placement in the top of the housing. Mazda has patented engines using port injection plus two direct injectors, one in the middle of the housing and one near the spark plug.

To control the fuel system, you need to control the high pressure fuel pump spill solenoid valve. The timing of the solenoid determines how much of the delivery stroke on the pump lobe will be used. For solenoid type injectors, you need 60 volt peak and hold drivers. The current control signal affects the atomization at part load, so ideally you would tune the peak portion and the dither/hold portion of the current profile. Solenoid type injectors are typically multi-hole injectors, kind of like DI versions of Bosch EV14/ID injectors. The targeting of the holes is a lot more important on DI than PFI.

Rail pressure comes in 120 bar, 150bar, and 200bar systems most commonly. 200 bar is the most recent, something you'd find on a new BMW engine. 150bar is like your mid 2000s Audi engines. 120bar is like a Mazda 2.3 MZR turbo engine.

So here's what you minimally need to control the DI fuel system, beyond a normal PFI system:

In addition to low pressure fuel system control, you need a high pressure fuel pump spill valve delivery angle control logic and a solenoid driver for that valve. You need 60V peak and hold drivers for the injectors. You need drive by wire throttle which requires H bridges as C. Ludwig mentioned.

 

The experiment I'm planning on wouldn't be full time DI. I would only do DI below 3k for emissions and fuel economy. Then I would transistion over to the older primary and secondary style port injectors for the mid range and upper rpms. You gotta get creative with this sort of thing.

 

There is no need for an expensive ecu to control the direct injectors. The mazda mzr engine uses an external box to control the injectors that can be sourced cheap. It takes 12V and a regular injector signal to trigger it.

This is the unit your looking for:

http://www.ebay.com/itm/06-07-Mazda-Speed-6-Speed6-Turbo-Fuel-Injection-Control-Computer-2006-2007-3248-/140951771176?pt=Motors_Car_Truck_Parts_Accessories&hash=item20d1615828&vxp
http://www.ebay.com/itm/BB1699-06-07-MAZDA-SPEED-6-TURBO-FUEL-INJECTION-CONTROL-MODULE-L3K9-18-941-/400562483005?pt=Motors_Car_Truck_Parts_Accessories&hash=item5d43626f3d&vxp
Or for european buyers:

Ebay link 3


These pdf's show how it works:

Mazda Speed 6 DI PDF 1

Mazda Speed 6 DI PDF 2

Mazda Speed 6 DI PDF 3

As for the injector, I've used this one in the past:

http://www.bosch-motorsport.de/media...6067211pdf.pdf

It's oem equipment in a lot of cars (Volkswagen or audi FSI of TFSI engines for instance) and can be sourced used for cheap. In the engine I did I mounted the injector as close to the apex seal as possible to reduce blowby, The clearance between the seal and injector was only about 0.5mm or so.


I'm not really sure how to control the cam-driven DI pumps found on regular DI engines. The engine I did uses a cylinder filled with gas, one side of the cylinder was filled with gas, and connected to the injectors using a high pressure hose. The other side of the cylinder was pressurised using nitrogen. I wonder how well a regular hydraulic pump will hold up when pumping fuel instead of oil, maybe it's not so bad if you premix.

 

http://www.motorauthority.com/news/1033290_patent-application-reveals-new-direct-injeaion-rotary-engine-from-mazda

If this is legit its shows mazdas 16x engine patents with di at 12 oclock pointed towards the intake by roughly 20 Degrees. But it doesn make more sence to me personally to have the injector mounted facing the plug and more at a 1/2 oclock posotion. To try and keep the sweeping of the rotor from collecting fuel on the back of the rotor face

 

Do some reading, and prepare to be surprised!
It is true that SC offers little advantage at full power, but full power is not where fuel economy is of primary importance. It is optimal to operate on SC type combustion in a wankel engine much farther into the load range than on a piston engine. Once again this is mostly due to geometry: the long narrow wankel combustion chamber has a lot of quench area, and the sweeping motion of the rotor tends to trap some unburnt charge near the trailing apex seal.

Take a look at:
http://www.rotaryeng.net/Direct-inj-strat-charge.pdf

Not the best reference, but the only one I could find which is posted somewhere publicly available. There is a lot more good reading on the mazdamaniac site, but either the password I used to have doesn't work anymore, or I forgot what it was. If it will help I can post the citations to some papers and you can go dig and find them.

Also take a look at:
link

In particular, look at the BSFC shown, you can see that BSFC is both low and fairly constant across the curve, which is the advantage of SC operation. The John Deere/Curtiss Wright engines mostly used a single high pressure injector rather than the dual injector strategy of the Mazda engine in the first paper. Also do not be confused by the prechamber ignition, there have been engines with and without this and both are operable. Also note that one goal on the NASA engines was multi-fuel capability, so most of these engines were designed for operation on both gasoline and heavy fuel (DF2 and/or JP8). I also thought that guy's idea about air injection was interesting.

In reference to the spill valve, this can be set up as a rail pressure bleed valve with a simple pressure feedback PID control, which many ECUs can accommodate.