Comment by Animats | Hacker News Reader

Comment by Animats | Hacker News Reader

There are about 15 free-piston generator projects. Toyota had a prototype in 2014.[1] Here's one from the UK in 2016.[2] It's a more reasonable idea today because switching power supplies are so good. You're going to get variable-frequency AC out of the thing, but converting that to something more useful is easy today. Also, using the generator windings to start the engine is easier with current power semiconductors.

You have to have electrical control on all the valves. There's no camshaft and no place for one. Electrical control of engine valves has been a dream for a long time, but the actuators were a problem. Someone finally built a working engine in 2018 with total software control of all valves.[3] The valve people say "Electromechanically, you could look at it and ask 'why didn't you do that 20 years ago? The difference is in the electronics that control it. What's happened in the recent past is that there's now sufficient processing bandwidth at a low price that can tolerate top of engine conditions, so you can actually put real time control on top of these motors."

This will probably all work commercially about the time everybody switches to electric cars.

The parent article is lacking in solid "why ours is better than the other people who already did this" information.

[1] https://www.roadandtrack.com/car-culture/a6326/out-of-turn-t...

[2] https://youtu.be/u4b0_6byuFU

[3] https://newatlas.com/camcon-digital-iva-valve-system/55827/

>"The difference is in the electronics that control it. What's happened in the recent past is that there's now sufficient processing bandwidth at a low price that can tolerate top of engine conditions, so you can actually put real time control on top of these motors."

is something a non EE person would say about another branch of science he is totally unfamiliar with. We had 40GB consumer Hard drives 20 years ago. 20GB per platter, 7200 RPM. Now think for a second what precision of timing you need in a hard drive versus mechanical engine. We are talking ~50 nm/bit, ~250 nm/track, ~100K TPI, >500Mbit internal transfer. The level of precision required to control engine valves more accurately than mechanical means (cams) was achieved in consumer off the shelf ~$200 hard drive controllers around 1995.

That was a quote from the CEO of the company who makes the thing.

Fast precision proportional valves for ordinary compressed air have been available since the 1990s. I once had to price some. Price was over $1000 each. They're now a moderate volume product from Festo and others, and precision positioning of pneumatic cylinders is available. It's still not common.

Automotive valve actuation has all the same problems, plus pushing back against big forces from the combustion process and operating through the entire under the hood temperature range. A solenoid-type actuator for this needs considerable power. Schemes with pneumatic and hydraulic power to drive the valve have been used on some race cars. Plus there are little problems; neodymium magnets start to weaken around the boiling point of water, a temperature easily reached on top of an engine block.

There's now a supercar using this technology.[1] It may filter down to less expensive vehicles.

It's an obvious idea that turns out to be quite hard to make work. Study guide: [2]

[1] https://cecas.clemson.edu/cvel/auto/systems/valve-timing.htm...

[1] https://www.freevalve.com/

I fully agree with everything you said about this being a big electro-mechanical challenge, but the dude pointed squarely at the controller with "processing bandwidth...real time control" quote.

If you approach the control problem the way a modern programmer would - read position sensor, adjust actuator current - you need some control loop running at some speed like 100KHz or higher. Probably doing floating point calculations. So you need a powerful CPU.

If you didn't have that, you might have to do the valve fine position control with analog hardware controlled by the digital system. That's how daisy-wheel printers controlled the daisy position. And you might have to have hardware timers for triggering events, like SOiC systems with PWM drivers. That's what you'd probably do if you were making a million cars.

If you're making 300 supercars, just throw some very fast CPU at the problem and let it do all the work. You'll never see the extra $20-$100 or so per car in the sticker price of a supercar. You can get hardware for automotive environments with that performance now. 15 years ago, you couldn't.

>modern programmer would - read position sensor, adjust actuator current - you need some control loop running at some speed like 100KHz or higher. Probably doing floating point calculations

Modern programmer would make sure to do those floating point calculations in an ISR too ;-). Cars have been fine with lookup tables and interpolation. If you can precisely control spark, you could valves as well.

As for hardware requirements. Afaik control always outraced controlled business end. Lets take Honda for example, all their ECUs from 92 (P05) onward had full facility to drive individual coil on plug, but officially Honda switched away from mechanical distributor in 1999 (S2000?). You can trivially convert 92 car with a small adapter board, ECUs shipped with all the needed hardware sitting unused for 7 years.

I think the issue here is more about the operating temperature of the semi-conductors. It's only recently that chips running at 400 C (?) using silicon carbide or Gallium Nitride semi-conductors has been practical and cheap enough. So these electronics can be integrated right into the motor without extensive cooling.

This same technology is having an even more dramatic impact on High Voltage Direct Current (HVDC) transmission lines where costs have been dropping very quickly.

Again, author of the quote was talking specifically about the controller part. Car ECUs use off the shelf semiconductors. ECU module is usually mounted inside the cabin under the dash.

According to Konigsegg they have a camless valve control system ready to go. They plan to sell 300 such engines in their new Gemera grand touring car this year.

Reply on news.ycombinator.com