T-leg Piggyback ECU R&D

[jake15]

pics

92-94 N/A injector-red- by V6 STi injector-yellow-


V6 injector in the 92-94 N/A rails (still on the 92-94 intake manifold)


STi FPR by a T-Leg FPR


sorry about the pic quality, the camera didnt want to focus.

[Legacy777]

The only thing I'd be concerned of is that if you look at were the screen is on the sti injectors....it's down the stem of the injector further. Depending on how rails are setup, this may create a restriction.

[jake15]

well, my injectors and douglas's injectors have the screen in the same place, and he's using the same rails, so i think it will work ok. plus i saw that when i was test fitting them and checked it out, and it looks like the screen is almost even, if not maybe a little below, the actual fuel lines where they connect to the fuel injector holder. i took pictures of it, but they all came out blurry..... dumb camera

[jake15]

looks like this is going down tomorrow (saturday) morning, finally. i have a couple questions though. should i reset the ECU after i swap the injectors? and how do i go about doing that?

[free5ty1e]

Remove the underhood fuse for it (forget what its labeled but its relatively obvious once you look at them) for about 30 mins, then reinstall it.

If you don't care about your radio station presets and are impatient, remove the negative cable from the battery and press the brake pedal for a good 30 seconds to discharge any leftover capacitance in the system, and reconnect it.

And yes, you should reset the ECU after any modification that would alter its adjustments to the default stock settings. Theoretically you don't have to reset between mods, it is always gathering data and adjusting. But keep in mind if you don't reset, it has all that data it collected before the mod and you will have to drive it enough for the ECU to collect enough data to average over to the new optimum tuning parameters - basically it will take much longer without a reset for the ECU to properly tune its settings.

[jake15]

cool thanks, i'll do that. injectors are in, the driver side took me about 30 mins to do. the passenger side took me about 1hr 50 mins, (theres so much crap in the way!!!) but now i'm going to reset the ecu and go for a drive. hopefully i can do a write up later after work.

[jake15]

just got back from the first drive, it runs really rich, and feels like there is no acceleration. it blows black smoke (rich) i reset the ecu, will it learn how to control these injectors? how long will it take? if not what is a cheap way i can control the injectors?

[free5ty1e]

hmm, no idea how long it might take for the ECU to reach that level of fuel trim, but if the injector size is within the capacity of the ECU's trim adjustments it should just be a matter of time before you're running the standard level of Legacy richness once more. Be sure you've got a a working O2 sensor installed for any of this to work.

Congrats on your higher capacity fuel system! I'm jealous.

[jake15]

thanks, i just checked and the o2 sensor is working. i'm resetting the ecu again (i'm just unhooking the battery for a little while) and we'll see if its any better. i was thinking and apexi afc or what doug did with the pentiometer thing, but i'm not that good at wiring things we'll see how it goes, hopefully it will learn how to control them within a day or two. Vikash said earlier in this thread that the ECU will control injectors up to 20% larger, and i figured that 440 is just below the 20% so hopefully the ECU will learn

[vrg3]

Well, I don't think it'll ever learn to run correctly. But after enough driving (maybe a couple hundred miles?) it should start behaving somewhat reasonably at low-to-moderate loads.

[jake15]

i cant wait for a couple hundered miles...... what about an apexi s-afc? i think i'm going to switch the injectors back tomorrow or monday, and put the 440's back in when i get some fuel management stuff, because the way its running now makes me sad, when you press the gas to accelerate or just rev from idle, it stumpbles and then revs, then its okay until when your driving untill about 2500 rpm then it just bogs down anywhere above that and barely accelerates. yay, 2hours and 20 mins of more work, hehe. maybe i'll do it faster this time

[douglas vincent]

Mine has started to run much better, at least milage wise, off boost. I am getting close to 20 mpg now when cruising on the freeway. But when i go happy fast, then milage drops accordenly. But I AM using the simple AFC too.

[free5ty1e]

whats the s-afc do exactly to adjust the injector pulse width - what inputs do you have to hook up to it? throttle position, o2, injector control wire tap, crank position sensor...?



--if I were to write some simple firmware for a 24MHz microcontroller - to monitor the o2 signal and injector control wire tap (anyone know if this signal has a pull-up or pull-down resistor anywhere on it?), and while the o2 sensor is too rich it takes the injector signal from the ECU and pulls the duty cycle down step by step until the o2 signal reads stoich - would that be useful?

I think this is possible. It'd be a chip that should automatically lean out the a/f ratio for a proper burn when you've got problems like we're describing here. I'm new to tuning so let's dissect this issue and I bet I'll be able to make little cheap auto-tuning devices to correct for oversize injectors.

I'd probably need to take the tps or cps or possibly even the maf or map signal to reference what o2 reading is actually considered "stoich" at a particular point in acceleration.... or is the desired o2 sensor voltage the same at idle and WOT?

Depending on the injector signal, I might have it go through the chip, which would directly output to the injectors instead. Intercepting the signal is probably more realistic than altering it (and potentially fighting the ECU depending on the signal's drive mode!).

Just thinking out loud.

[vrg3]

Jake, you could try putting on an AFC and scaling the signal a little. It's a lot to spend on a very crude workaround, but it might help these injectors work.

Chris, the S-AFC basically scales the airflow signal according to a table indexed by RPM. It's got an input for the TPS and oxygen sensor as well... I think you can set a TPS threshold so you can have two profiles, and I'm not sure what it does with the oxygen sensor signal. Maybe it just displays the oxygen sensor voltage on its whizbang vacuum flourescent backlit display that makes up most of the cost of the unit.

Modifying the injector signal's a little more complex than simply tapping into it; you'd need to build some kind of repeater circuit. The injector's flyback voltage is several dozen volts, and I believe the ECU needs to see that flyback voltage so it knows the injector is working correctly. The injector itself forms the "pull-up" in the circuit. Even if you built a repeater and whatnot, you'd have an ill-posed problem when it comes to shortening pulse width. Extending pulse width is pretty straightforward, but how do you know how much to shorten a pulse if you don't know how wide the pulse is to begin with? Simply delaying the pulse until you do know could cause driveability problems since injection is sequential.

Those complexities are why APEX'i, and most other piggyback manufacturers, chose to manipulate the airflow signal input rather than the injector drivers' outputs.

Also, under any appreciable load, the oxygen sensor signal isn't really gonna tell us what we want. The oxygen sensor's signal is very nonlinear; it pretty much tells you "richer than stoichiometric" or "leaner than stoichiometric."

Stoichiometric mixtures under load aren't desirable; you need to run somewhat rich.

You should note that the functionality you're describing (a rich signal reducing pulse width and a lean signal increasing it) is exactly what the ECU does under light load. That's what closed loop running is.

So what you're describing is a good idea... So good an idea that all carmakers already implement it for the conditions under which it would be useful.

I think the cheapest way to get this stuff to actually work correctly (without opening the ECU's case) is actually to use an aftermarket ECU. I was playing earlier (and will play again) with setting MegaSquirt up as a piggyback ECU to control fuel. I think that holds a lot of promise.

[free5ty1e]

lol... thanks V

Oh well. Yeah it'd delay the signal, having to first detect the pulse width. Anyway I bet I could imitate an s-afc type device and offer them real cheap to the group, that doesn't sound too terribly complicated a function to write firmware for.

Is the rpm table you spoke of constant in the s-afc, modified by a potentiometer or anything like that? Or is it programmable via computer or something? Perhaps it forms a closed loop by reading the o2 for auto and real-time adjustments...? (if it doesn't, it should...)

[jake15]

well, since i had nothing to do yesterday, i started looking at some fuel management stuff, the most promising looks like either the mega-squirt or the greddy emanage (apexi SAFC and HKS SAFR were in there, but these other two are so much better) i was reading about them and i dont know much about the MS, but it seems to be the favorite over on usmb. the only bad thing about the emanage is the cost. i'll post the links to all the info i found when i get home (i'm at school now) but which one would you rather use, apexi SAFC, MS, or the Greddy Emanage?

[vrg3]

Chris - Yeah, you could easily build an AFC equivalent for peanuts. The RPM table is tuned electronically. Like I said, most of the cost of the AFC is in its interface. Why use potentiometers when a complicated UI would do?

You could make a really simple AFC without even using computers. An LM2917 frequency-to-voltage converter feeding into an LM3914 bar graph driver feeding into a multiplexer or a bunch of analog switches which select from an array of noninverting amplifiers whose gain is each controlled by a pot.

Or it could be even simpler if you indexed off manifold pressure instead of RPM -- scrap the 2917 and instead feed the 3914 the pressure sensor signal.

Jake - Between those three? It's not really a fair comparison because the e-Manage has some ability to do ignition control. Then again, it costs way way more than the other two, I think. If it were me I'd choose MegaSquirt.

[free5ty1e]

megasquirt would be badass but I don't believe they're easy to come by. Waiting lists are involved, last I remember. Also need someone (hint hint Vikash) to get one up and running with a usable baseline fuel map for stock injectors, and then everyone can just get a copy and start from there.

S-AFC based device would probably be good to support quite a significant amount of additional safe boost with the proper amount of fuel available to it. I don't know how expensive they are but if I can program a microcontroller to do the same thing then that would become the cheapest option for everyone here...

I'm talking even if I were to put an LCD display on the bitch. Digi-Key carries 8x2 LCDs for like $8 and it still would just be one chip. It wouldn't be fancy but it'd tell you what you wanted to know and shit I could have it selectable which source to read from for the input. The microprocessor I use only costs a couple bucks, but I have plenty lying around anyway...

It's settled then. I'm gonna focus on that after I get my intercooler pipes to like each other enough not to split up when I'm boosting.

Vikash, any idea what that rpm table might look like? Got any graphs handy?

[vrg3]

There haven't usually been any waiting lists... Both times I ordered a board from Bruce he got it to me within a week.

The new revision of the motherboard's coming out very soon, though, and it's definitely worth waiting for. It has a lot of cool stuff.

A base fuel map would be pretty easy for anyone to come up with. There are certainly lots of people who'd be better at it than me anyway. I don't really know anything about tuning.

The only difficulty with having one person making these AFC dealies and then passing them around is that they'd effectively have a static tune since not many of us have the ability to reprogram them. I don't think any of us should place the responsibility of mangling the signals "correctly" on someone else.

Do your microcontrollers have DACs on 'em?

I don't know anything about what most peoples' AFC tables look like. You might want to try Googling for AFC tuning advice.

[free5ty1e]

Oh, they could be adjusted with three buttons and the LCD display. Sure, I've got ADCs, DACs, filters, even basic switched capacitor blocks for custom modules. I'm just wondering what would the base table even look like that comes with the s-afc? That'd be what I'd start with, and hey I can even have the interface give the user a way to write the adjusted settings into EEPROM so's it remembers them each time. All one chip. I'm tellin ya.

I'll do some googling for the AFC tuning advice, thats probably what I'd end up doing anyway. Thought perhaps you had the answer, most times you do lol

[vrg3]

Cool.

I'd say the base table would be zero trims. And then if I had to start tuning one I'd just start ramping them down slightly.

The thing that's frustrating about the idea of AFC tuning Legacy Turbos is that the range of our airflow meter is so limited. All the really interesting airflows exceed our sensor's capacity to measure.

[free5ty1e]

but.... isnt that a moot point when my chip would be controlling the signal fed to the ECU that is supposed to be the airflow? I mean... we're not actually reading air flows here - I thought all it'd do is, if you're at a certain point in the RPMs and have the table set for a nonzero trim value, adjust the voltage coming from the maf sensor accordingly and output it to the ECU so's it thinks there's less airflow than there actually is and will think it needs less fuel and lower the injector duty cycles as a result.

Or do I still have this wrong...?

[vrg3]

But your chip won't know the difference between, say, 250 grams per second and 280 grams per second.

[free5ty1e]

Right... is there some reason the chip needs to know the actual airflow reading? Thought it just needed to read the maf voltage and then scale it according to the current point on the table to elicit the proper reaction from the injectors - thereby never actually letting the ECU see the real airflow signal (since it can only use that to tune the stock injectors or at least injectors in that range). Isn't it in this way that the S-AFC allows cars like ours to see higher airflows than the maf can sense and spray the proper amount of fuel to compensate, even when the maf signal is pegged? Still the same theory, just different trim #s in the RPM table... right?

If not, I must be missing something...

[vrg3]

My point is that 250 g/s and 280 g/s both produce the exact same voltage from the MAF sensor. As far as the sensor shows, there is no difference between the two situations.

The S-AFC can kind of let the ECU work with higher airflows if you just assume that the signal's gonna be pegged and then assume that airflow is gonna be dependent only on RPM. That's really not true, especially on turbocharged cars, though. But you basically end up tuning it just to run right-ish at wide open throttle and full boost.

The stock MAF sensor is actually fairly close to being ideally sized for the stock injectors. It's a little small, but not by much. That allows for good resolution in airflow sensing, but doesn't leave much headroom for power increases beyond the limits of the stock injectors.

Now, there is one sensor on the car whose signal would change between 250 g/s and 280 g/s: the pressure sensor. You can actually strongly correlate airflow with manifold pressure. In fact, if you know ambient air density, manifold pressure, engine speed, and the volumetric efficiency characteristics of an engine, you can determine airflow. That's precisely what speed-density ("MAP-based") ECUs do.

So maybe you could rig up something that makes use of the pressure sensor signal as well. Something to think about...

But then you're well on your way to designing an ECU. Sigh. Whenever I start thinking about what to do with this stuff I end up on a slippery slope that lands me at the conclusion that the ECU's gotta change.