rusefi.com

I tend to think this would be a good idea. For a variety of reasons. http://www.digikey.com/product-detail/en/IF-E10/FB104-ND/3724

I personally feel that this is overkill. As we discussed, it might make sense to have available as an option for module linkup on a PHY level but that still leaves a higher level protocol (SPI, I2C, CAN?, ATM?, AL?) to implement over the top.

Why do I think this is overkill?
-differential signalling buses like CAN already have excellent noise immunity and area designed to not create ground loops or other nastiness
-existing networks like CAN are already optimized for handling low latency traffic but there comes a point where certain functions (i.e. cam/crank decoding, ignition events, for instance) need to be handled in a more time critical manner than is possible with the overhead of networked traffic.
-cost is about 4-5x higher than CAN
-Lack of established tools, topologies and methods
-total galvanic isolation for why again? aren't you going to share a car frame? common power source i.e. alternator/battery?

Modularity is an awesome idea but I think modules belong either in-case (i.e. spark, injector, wheel decoder, etc.) or CAN networked with differential signalling isolated by transceivers.

Don't forget, Jared, that the item you linked to is "simplex" so two are necessary for any bi-directional comm.

Either that or their tranciever. Also it depends on the topology. I generally like the concept of a kind or ring network. However Blunder has noted that a CAN network could work nearly as well. I posted to bring up the discussion. I like having a modular approach, such that I can buy a low cost system to play with a small engine, then over time I can upgrade it to a variety of projects. I see some kind of back hall or back bone infrastructure as critical for such an approach. I tend to like the optical approach as it's immune to RF and many common issues under the hood. However I also understand that CAN is already under the hood and has been proven to work fairly well. So optical might not be the best approach, but I can't help but think of it. One item I like about it is that if I have a single injector or single ignition, and if I neglect the feedback, you could have a transmitter on the ECU then a receiver at the injector. The receiver would drive the MOSFET directly, and no brains would be needed at the injector. Think of it like an extended optocoupler. Then later on I could upgrade my injector driver to include some kind of feedback path. I could get something more expensive that does include brains at the injector driver and could report back to the ECU. I could get through the door for a low cost, then I could keep upgrading as time and confidence allows. However with CAN, you need a brain at both ends all the time, and multiple source codes ect. So I kind of see that as extra complications that can be problematic to a new comer.

Any how, is an optical bus a good idea? What are the pro's and cons. I see cost is a bit of a con, but not really that bad in low qty, and it gets better as qty goes up. This particular fiber optic is a clear plastic inside a black plastic, it can be cut with any set of scissors or razor blade you have laying around. No polishing required or anything like that. Plus it's reasonably low cost. The trade off is that it's also reasonably low bandwidth.

CAN is a good standard and used by literally everyone.

If you need a deterministic protocol there is FlexRay.

kb1gtt wrote:...
Any how, is an optical bus a good idea? What are the pro's and cons. I see cost is a bit of a con, but not really that bad in low qty, and it gets better as qty goes up. This particular fiber optic is a clear plastic inside a black plastic, it can be cut with any set of scissors or razor blade you have laying around. No polishing required or anything like that. Plus it's reasonably low cost. The trade off is that it's also reasonably low bandwidth.

I can see the benefits to optical triggering injectors and coils: run 12v from the battery, through a relay of course, up to the engine; run fiber to the engine for each endpoint; trigger away. ECU is electrically isolated, for the most part, from the inductive parts... maybe a little ripple on the ground / power rails, but easily dealt with in the power conditioning ECU circuits.

Downside would be added bulk at the endpoints for the circuitry necessary at the coils and injectors... and then you have to deal with power filtering for those components out there, on top of the engine maybe, where temps can be > 150F.

My thought is that you'll have a little box, near the injector, with + and GND leads running direct to the battery. The box also has the optical connection. The box can leak a bit of current to self power, and can be basically an optically driven MOSFET. We are talking really simple, and all the heavy currents that bounce around the GND signal will be isolated on wires that run direct to the battery. Heat is away from the ECU. I see it as a win win. I also like that if you go Direct Injection, the little black box can be configured to generate the high voltage, and if something goes wrong, the brain is protected as the optical link keeps it isolated.

I can visualize things a bit, but the physical topography escapes me.

Either each device has something near it and the optics are the branches out to devices, or optics travel to a central distribution point where all the stuff is, and then things traverse outwards, like tentacles on a squid.