I am currently helping a high school engineering class rebuild a Firepower home edition. They are using an Arduino Mega to scan switches and fire the solenoids, while feeding events over a serial connection to the Pi to trigger audio/video.
Did you find I2C sufficiently fast to pass events?

Thanks for the (admittedly offtopic) advice! I will start a post with the school project once they've made sufficient progress!

RE: Arduino speed and loops
Here's some interesting reading material.
http://www.paleotechnologist.net/?p=1808
http://www.instructables.com/id/Arduino-is-Slow-and-how-to-fix-it/

Also, a shoutout to Hugh for his help with the power filter. I made a simplified version for the demo I built for the kids, and it worked great(I cleaned out the capacitors at a Radio Shack that was going out of business). I'm now working on the MPF platform. A bit of a slow go, since there is no docs. My approach was to figure out what the minimal platform code would be(init serial, send/receive messages), and get that working before fleshing out the full message protocol.

Slowly. We are in the midst of a move, so all of my projects got packed up. Happily, we're moving to a bigger place, so once the chaos is finished, it should be more conducive to experiments.
In the meantime, I'm slowly working on getting the Mega and my mac to talk via MPF. For the hardware side, I built a simplified driver and power filter system and a little flipper and bumper demo for the kids. We are still trying to get time in the school shop.

That is 100% the truth

Yeah, I think you can get pretty close to the bare metal by ditching the Arduino libraries. Sketches are just C code that is hiding some stuff for simplicity. I imagine that someone is working on a "real time optimized" version of the core Arduino libraries.
For my project, given the scope, I think that spot fixes, like replacing the digitalWrite calls, is likely good enough, and keeps enough of the Arduino libraries to ease the kids into it.

Well, due to a change in security at the high school, the original project pin(a home edition Fireball) is being relocated to my workshop, so I expect real world testing to expedite the code I'm working on.
I'm also going to write a cut down version of the Arduino code to work with an Uno or Micro, for use with the Zizzle machine I donated.

I'll take some shots this afternoon.

Ok, had a chance to go through the strip down that the kids did while it was at the school.
One flipper coil was cooked. A new one is on order.
Swapping lights out for neopixels.
The playfield has some damage to the particle board. Debating on a fix there. I'm inclined to grind out damage, fill with putty, sand down the whole playfield, add a custom piece of vinyl artwork and cover it over with a playfield protector. But, if I'm not going to do the retheme, I might just leave it ugly and let the kids do a theme next year.
No back box. Building a new one
Don't trust the power supply. Replacing with my switched 24v and a power filter board.
I think I'm going to leave the switches alone and attached to the big circuit boards instead of mounting on the bottom of the board.
So, first order of business is to finish the solenoid driver code and the switch matrix.

20170712_234508 (resized).jpg

20170712_234417 (resized).jpg

My project topic is here: https://pinside.com/pinball/forum/topic/fireball-home-arduinoraspberry-pi-rebuild

Nicely done!
Are you releasing your code anywhere?

I generally go with a beefy switching PS for the solenoids, coupled with an OPP power filter board to add capacitance. Then, an ATX PS for the logic side. However, I also have a project coming up with a DMD at 12V, so I will either need to add a 12V supply for it or add capacitance to the 12V rail from the ATX.

I was running straight off of the 12V rail with my tests, and I would occasionally trip the PS breaker and have to do a reset on the PS. This generally happened on frames with a lot of bright pixels, so I interpreted it as an instantaneous load issue. I tried multiple supplies, and saw it on both.

Yeah, I was quite surprised by it. Initially, I thought I had fried the DMD somehow with a short. So, I grabbed a brand new, beefier PS that promptly did the same thing.

I agree. I lucked into a free case of Antec ATX supplies a few years ago, and I've still got a pile of them. Aside from the DMD, they work great.

I just ordered a hand full to test.

I opted for a Teensy 3.5 for my main controller, and I designed some i2c i/o boards to go with it. It was my first PCB, so I made some mistakes, but they work. I chose the 3.5 because it is 5V tolerant, it had 3 i2c busses and runs fast. I can run 24 of these boards from it, which comes out to 384 direct I/O ports or 1536 ports in a matrix form. That's enough switches and drivers for just about any configuration I can think of.
I also started thinking about boards to support slower Arduino forms, so I'm working on some secondary boards as well that can take some of the processing load from the main controller and reduce latency.

48385500_2291332590910935_6779409008320053248_n (resized).jpg

Also, for a sister project, I got pinmame running on a Pi and drawing to a physical DMD. Running X, it is a little slow, so I'm rearchitecting my pinmame hack and working to minimize the load on the Pi, so that it can run pinmame at full speed. My first experiment was getting a Scared Stiff ROM to draw. I have some video of before I cleaned up the signal. I need to make a new one. This one also uses the Teensy based system I'm working on.