Injector Control Options

[Fred]

Great explanation. Now all we need is someone with a scope and bench fuel setup (can't run em dry) to set this up and give it a whirl with some input (555 timer would do, or ms, or ms2 test mode, or freeems proto code). Any volunteers? Abe???? :-) Abe IS the man for this job, I hope he's not too busy fixing his head.

Also the supply rail for U1 should be on the clean supply, but thats just me being bloody picky!

Like I said, if your driving a normal MOSFET then you leave it attached to Vbat - I don't have an autofet in the LTspice simulator so I sim like that. If we're driving an AutoFET then yeah it will be powered by 5V.

I think he means clean 12v as opposed to dirty 12v. I'm not sure if/how much this would matter with an appropriate decoupling cap though.

Fred.

[Fred]

Delta, is Jared right about the heat?

If we had two digi lines from the MCU that drove two injector drive circuits, one with a power resistor and one with out, we could run the FET's saturated, avoiding oscillation issues. The resistor could also be made external to the ECU keeping the heat away.

If you can think of a way to do this WITHOUT burning more pins of a high speed nature that we don't have and/or adding software complexity that we don't need to, I'm all for it. If not, I'd rather go LM1949 and BJT or not support low Z.

Fred.

[jharvey]

If you can think of a way to do this WITHOUT burning more pins of a high speed nature that we don't have and/or adding software complexity that we don't need to, I'm all for it. If not, I'd rather go LM1949 and BJT or not support low Z.

I can think of ways to do it, but it requires more parts. I figured software is cheap, parts cost more, so if it can be done in soft... Not to mention if I push it into software, that's less work for me

You could put some form of an "on pulse" timer on the full power FET, that would automatically turn off after a couple ms's. I don't currently know a chip for this. Perhaps a brown out chip would work nicely for this. Hmmm. I'll have to stew that over a bit.

[Fred]

I can think of ways to do it, but it requires more parts. I figured software is cheap, parts cost more, so if it can be done in soft...

MS2 uses 2 channels per output channel to achieve P&H behaviour and one part of the reason we are all dissatisfied with it to some extent is because of that. MS2 has 30 pins, we have 90, but rest assured, they will run out just as fast as the 30 did and someone will (rightfully) be whinging that we wasted 6!! pins on P&H when it belonged in an IC or outside the box all together etc.

Software isn't cheap at all BTW, software complexity costs a LOT. Just because software is cheap for you to obtain (think KiCAD) don't forget that some poor bastard slaved over that code for weeks/months/years/decades to make it as good as you use it. This site and the code and time put into it all has cost around 50 thousand US dollars so far if you count my time. I dunno about you, but that's NOT cheap to me ;-) I'm OK with putting time into it, but I want to see results, and I want to see good results :-) I also want to put time in where I see that it is most valuable and P&H in software doesn't score on that board.

We also want the software to be simple to read and modify to encourage contribution in the future. That shouldn't come at a cost of complex expensive hardware, but both simple hardware and software should be attainable IMO.

So, post up your ideas and when all is said and done we'll make a good decision with the tools (designs) we have available to us. Group consultation FTW! :-)

Fred.

[Delta]

My design - change the injector model so its 15mH and 16.5ohms - a saturation style injector - you will be greatly surprised

This design only requires ONE PIN for BOTH saturation and peak/hold. Saturation injectors do not draw over 1A so it never goes into limiting, and no oscillation occurs and the current is much lower everywhere, so the heat is minimal.

I designed it that way....it works for BOTH sat and PH without a single change to any component value.

The time averaged power for 98% duty cycle is 8W per injector driver - so yes around 65W if your at full duty continuously 8 cyl. Perhaps a bit high, but nothing that can't be overcome with a good case.

[Fred]

This design only requires ONE PIN for BOTH saturation and peak/hold. Saturation injectors do not draw over 1A so it never goes into limiting, and no oscillation occurs and the current is much lower everywhere, so the heat is minimal.

What about P&H injectors? How much heat are we talking about per channel, and in which devices is it dissipated? If it's just too much heat then that probably settles the discussion about supporting P&H as we really don't want a hot running box with the CPU inside it. If it puts out a lot of heat it belongs outside the ECU IMO. In that case each person is free to use whichever P&H solution they like separate to the design we provide for a main board.

eg.

Resistors (yuck)
Jean's board with LM1949 (cool)
A custom board with your lovely design

In the event that we decide to not support P&H then we should include an IC to buffer the CPU so that the external setups are driven with a good low impedance source.

Fred.

[Fred]

The time averaged power for 98% duty cycle is 8W per injector driver - so yes around 65W if your at full duty continuously 8 cyl. Perhaps a bit high, but nothing that can't be overcome with a good case.

Damn post'n'editers :-p

OK, that's also a fair call. A user that wants on board P&H must choose a suitable case whereas a user that doesn't have P&H injectors can just use a normal case.

We could design the setup to have the capability and instruct people to jumper out those parts if they don't need them and install the board in a good case if they do.

BTW, we only have 6 drivers, not 8 unfortunately. The CPU doesn't have enough ECT pins for that AND a good accurate rpm/engine position input setup.

Fred.

[AbeFM]

ok, have to catch up on this thread later, but yes, I have a couple set of low ohm and a set or two of high ohm injectors so should be able to run some tests. Even got a thermocouple or two and access to oscopes.

My feeling without thinking about it much is something akin to jean's board on ours - support for P&H injectors, and dropping this hold current thing, then there's no issues?

More tonight.

[Fred]

My feeling without thinking about it much is something akin to jean's board on ours - support for P&H injectors, and dropping this hold current thing, then there's no issues?

You can't "drop" the hold current thing. You must have hold current and it does generate some level of heat regardless of which way you hook it up. It seems Delta's method generates less heat than others, but there is no free lunch here. So, heat is still the issue, how much, not sure, do we care, maybe. Perhaps making it the builders responsibility is what's called for here :

Code: Select all

if(requiresP&H){
    useSuitableHeatsinkOrCase();
} else {
    don'tWorryTooMuch();
}

In that way we can support it and make it someone elses problem to worry about the heat which simply won't be generated with saturated injectors. Mint :-)

Or maybe we should ignore P&H and let Jean sell them all boards so that the heat/noise is always kept away/not generated at all.

I really can't decide. We do need a clean solution, but we don't want a huge cost to all users for a feature that only a fraction will need. I'm still leaning towards no native P&H support. Jean already sells a good solution for this. If he just adds a 6 cylinder variant to his range I think it will be a very elegant solution to all those that need it and it keeps the power supply clean as you have discovered really is an issue.

Fred.

[jbelanger]

Just to make things clear, the LM1949 circuit will also drive a saturated injector fine and without generating (a lot of) heat. It will just try to regulate to 1A which will require a fully saturated transistor (or nearly there if the injector impedance is lowish).

I must say I like Delta's design. It does solve the issue of different peak/hold ratios and might make it easier for some people to source the parts since the LM1949 is not easily available everywhere. It is also slightly more efficient from a heat management point-of-view. However, I think it's going to take a bit more board space and I'm wondering if noise will be an issue especially with the very fast switching.

I look forward to seeing the results when/if someone tests this circuit.

Jean