TravisR

iTrader: (9)

So we all pretty much know that you can get by with a lot of things, but since we're approaching 200hp/l in the Miata market and now exceeding it... Maybe the standard ECUs out there aren't enough?

I've had a lot of time to think about this, and I've talked with a lot of big name tuners. These guys are the best of the best in their regimens and in our power level they are doing some very unconventional things to keep motors together. For instance on speed limiting during pit stops they implement an ignition retard and spark/fuel cut combination feed back algorithm to reduce valvetrain shock. They'll do a multi-staged injection tuning that deals with both a 3D angle injection map, your standard map, and EGT+WBO2 feedback loops. On the ignition side they use EGT feedback loops per cylinder to keep knock, and EGTs at bay. There are alot more things going on in the high end performance sector then at the low end where you have a fuel and ignition map and thats it.

Has anyone else went to a motorsport class ECU that supports special functions?

I am starting to workout the role-out on the 1280 ECU which is Adaptronic's Motorsport ECU. I'm just curious if anyone out there has cited in the electronics to be the problem vs the assembly builds.

http://adaptronic.com.au/downloads.html (software downloadable here under 1280) You can compare this to the 440 ECU which is much more cut and dry.

hustler

iTrader: (6)

I think 100% of the Miata tuners aren't running "on the ragged edge" to get every possible last drop of output because we're not in a spec class with a detuned engine. I've had my nose in the Motec and patec stuff and it's impressive but it's going to take some coin to use the special features we don't have already.

Reverant

iTrader: (10)

I've had my fair share of Motec tuning - minding you only on the lowly M4 and it's sibling, the M2R. I can honestly say, I didn't like it ONE BIT. Maybe I'm biased, but I did find the AEM better and the PowerFC far worse.

I haven't played with an M400 or an M800 of course, largely thanks to their price tag - no customer really wants to spend that kind of cash on an ECU unless its completely necessary. Two 400+ hp Miatas I've tuned so far are both happy with the Megasquirt.

18psi

iTrader: (76)

I think our cars are not technologically advanced enough to really need all the crazy gizmo stuff that the high end ecu's offer.

I mean maybe for the hardcore track people that need per-cylinder knock/afr/egt monitoring and all that crazy stuff it would be beneficial, but for the average street miata its just not a necessity.

Especially since more than half the setups on here are cheap diy using the bare minimum required to reach desired power level and another 30% are entry level kits from begi/fm/etc.

Our engines (except for the 01+ of course) don't have vvt, don't have any sort of vtec/vanos/etc. They are as basic as it gets.

I think the adap 420c and the DIYPNP are doing a well enough job of controlling our cars to accomodate about 99% of the owners needs and power goals.

TravisR

iTrader: (9)

So specifically, I know there are quite a few 400WHP cars out there. What are you doing to keep the motors together? Are you looking at EGT, WBO2 per cylinder? I've tuned quite a few, but only one in that power range and it made me a little nervous to be honest. At ~400 there is just something uneasy about the whole tuning process. On most tunes I can send them away with full confidence, yes you will enjoy this car for thousands of miles and have no problems. On that 400whp engine?? Who the hell knows? Cylinder pressures are off the map, EGT's are searing even with good tuning, and usaully they are turning a fair amount of RPM.

Any insight would be appreciated.

Reverant

iTrader: (10)

I have per cylinder EGT and WBO2 on my own car and I have found that the differences per cylinder are very small. Most Miata customers are too cheap anyway to go for individual WBO2 per cylinder and the placement would be an issue. I'm using my own LM-2 to tune (unless the customer has a brand new LC-1 on the car) and a Phormula Knock Analyzer Pro to listen for knock.

So far I can honestly say that EGT is USELESS unless you have a very thorough understanding of each and every parameter that affects it. I had one 400whp car burn a piston on #1, and the EGTs were like 820*C after the turbo - not something alarming. The piston melted because I was watching for boost, AFRs and EGTs and didn't notice that the coolant temp had gone up to 124*C - oops. It only took 4 seconds of pinging to get 0 compression. No, I did not have the Phormula at that time.

So I would say that it's pretty much a matter of listening to knock with your own ears while tuning, give it a little room for error, have a decent knock detection system once you hand over the keys and hope that the customer will respect his own car and use quality gas.

JasonC SBB

The problem with high end features is the TIME or COST of the tuner or cost of DYNO TIME and SKILL to tune them.

Just look at the # of rows and columns the typical ECU has. How many people in this forum have optimized the timing every 300 or 500 RPM and every 20 kPA between say, 5 psi and 15 psi? I frakking changed my RPM breakpoints to jump by 800, because I didn't want to pay dyno time to tune every 500 RPM. Besides, how much error could a map optimized with 800 RPM jumps introduce, vs. 500 RPM jumps? 1*?

How many people here have tuned the timing retard vs. IAT curves, by say, turning the dyno fan off and checking for knock?

And just look at how many people have actually optimized timing in off-boost areas? Um, like zero? I have posted at length, of a method to tune off boost numbers - do MAP sweeps with the dyno at fixed RPM, and have the dyno plot torque divided MAP on the Y axis, vs. MAP on the X axis... Look at my threads, and tell me, how many people even understand what I'm saying, much less, agree with my suggestion?:

https://www.miataturbo.net/ecus-tuning-54/dyno-tuning-idea-part-throttle-dynos-can-do-46160/
http://forum.miata.net/vb/showthread.php?t=373617
http://efi101.com/forum/viewtopic.php?t=5581

The efi101 forum has pro tuners, and when I asked how they tune timing retard vs. IAT, I got lame answers. They were long on the theory, but short on the doing. All this stuff can be done if you have an OEM engine dyno cell where you can datalog runs and have control on various variables such as IAT and CLT... Either that, or the efi101 boys weren't sharing.


Now look at the really cool feature of the Adaptronic, the timing autotuning - how many have used it? Sheesh, the Adap boys even took it out bec nobody was using it!:

https://www.miataturbo.net/adaptronic-63/anyone-play-ignition-auto-tuning-yet-42882/
http://adaptronic.com.au/forum/index.php?topic=252.15

The fancy features you discuss - understanding the theory of how to optimize tuning is all well and good, but actually doing it, the time and cost, is another.

I guess I'm ranting pointlessly. Hell, 10 years ago there were no affordable widebands, the FM log turbo manifold with the dongled Link ECU was hot ****, and very few load-holding dynos around...


Instead of adding advanced features that require more tuning time, why not make tuning easier and more bulletproof?

I would say that the most important feature one can add to the typical ECU today would be knock sensing with Mitsu-EVO-levels of sophistication. I'm talking, per-cylinder, time-windowed, pattern-matching DSP correlation filters type of detection. The hardware to do it isn't expensive anymore. Hell, cheapie used PCs probably have the computing horsepower, and can do speech recognition... recognizing knock is a lot easier.

Or how about electronic boost control that autotunes its feedback loop parameters, like the OEMs? Or long and short term 3D fuel trim correction tables? Autotunes idle feedback parameters? Injector voltage trim autotuning?

The AEM is my favorite ECU - one of its major strengths is its datalogging capability (200 Hz on 8 variable internal, 20 Hz and >20 variables via serial cable), and its excellent powerful datalog viewer. But, it is missing several glaring things such as:
- hi idle target for a/c on
- closed loop idle control
- no 'D' parameters in any of its feedback loops, only PI
- etc


Cheers.

hustler

iTrader: (6)

I have a buddy with an M800 in a 600+whp Eclipse and I discussed the computer compared to MS because the guy who tuned it is a god in most circles and he get's my respect. They run a wideband, no EGT, a bunch Motec stuff turned-off except for knock sensing, the power distributors (really cool), and diff controllers are switched off. They use Motec for it's "hub" functions because it controls the engine, gauge panel, data logger, gps data logger, and AMB transponder even works with it. He seems to think MS1 is fast enough for any car. I'm too dumb to have a real opinion on this.

TravisR

iTrader: (9)

I completely understand where your coming from. These guys that tune 32X32 maps are crazy, realistically you could probably get away with a 16x16, and my gut says an 8x8.

What I'm saying is that if you could take one of these motorsport ECUs, implement a control protocal to automatically adjust somethings for you could you save money and add reliability in the long haul?

On the 1280 for instance, you could setup a logic pathway in about 5 minutes, that could add/subtract timing or fuel in the event of a high EGT reading, and do it per cylinder. That could all be done automatically. This safe gaurd reduces the need to tune a ton of small tables for the sake of controling the actual variables that you want to control which are just a few outputs.

The dependent variables we actually worry about with the engine are only really EGT, Knock, and WBO2. The inputs into the engine are really only fuel angle, spark angle, and fuel duration.

So we can pretty quickly build up feedback loops into the system in order to save your motor incase insane things happen.

Stay with me, the possibilities are more fun then they would seem at first:
Lets say we're using a motorsport ECU class. Here are some of the things you could do with the extra 750 dollars you would spend on that ECU:

• You can setup logic to say that if oil pressure goes to within a range and stays there for longer then 250ms there is a complete engine cut.
  
• If EGT exceed maximum value for over a specific amount of time you get engine cut.
  
• If knock counts exceed maximum value for a specific amount of time you get engine cut, or maybe boost reduction, ignition retard, or any combination of these.
  
• If WBO2 goes out of balance with the rest of the cylinders on the lean side for more then 2 seconds then you automatically add 10% fuel to that cylinder.

So basically what I'm saying is that there is A LOT of protection that you can get from an upper class ECU that you simply can't get out of a performance ECU. I'm not saying the lower class is bad, but the value in protection could be worth it for a high end car/race team.

By the way Jason, you need to check out the 1280's advanced mode. You can literally build any logic/control loop you want into the 1280 with infinitely configurable maps, PIDs, and logic blocks of a multitude of types. It will make your head spin.

Best,
Travis

TravisR

iTrader: (9)

99% of the people who run motec, do it for data aquisition, and data feed back to the driver. I actually got to sit down with the owner of MOTEC for an hour at SEMA and talk shop. He explained their business model to me and Andy, and we really don't even compete for the same market share. Their integrated steering wheel displays and all that, thats the reason you run MOTEC, its not because their engine management is so much above the rest in ability even though I'm sure its very very good.

JasonC SBB

I've had my eye on the 1280 for a while, but I have yet to play with the s/w.

The features may be awesome and all that, but I still say the most impt features are:
- robust per cylinder knock detection, not just a simple bandpass filter and RPM-varying level threshold
- robust sensor failure detection and safe modes ... e.g. if your WBO2 develops an offset and long term trims suddenly jump or have slowly reached some limit, you should get a warning light and a message when you connect your laptop
- stuff that *saves tuner, tuning, and dyno time* such as
- autotuning wizards which tune *feedback loop parameters* and things like cold-start for you

Given the amount of computing hp available and how well developed auto feedback loop tuning and adaptive tuning is for the past *decades*, having to spend 2-3 hours tuning boost control, then an hour tuning idle feedback, 2 hours tuning VVT, not to mention dyno tuning time... gets old. I honestly think that at this point if I were to embark on another big ECU project, I should just buy a Porsche or a Corvette and just drive the damn thing, instead of re-engineering a car.

TravisR

iTrader: (9)

Definitely well heard,

I'm not making this discussion about the Adaptronic but-
-We do per cylinder knock detection looking at engine angle on the 1280.
-You can set this up maximum trim limits in the 1280
-PID controls on the 1280 can be linked to other variables and calculations. I would think that these could autotune a map much like tuning a fuel map although you would need to figure out how to intended correction. Zieger Nichols equations can get you close but this requires a first run and then correction.

I also think you wouldn't be doing this stuff unless you had some kind of sick love for it haha. I guess in a perfect world we'd all drive supercars :X I still think I'd like to play with ECUs and do tuning though. Its challenging, and a lot of fun when you make it happen.

JasonC SBB

OK so I couldn't resist playing with the s/w. I looked at the different tabs, and for now stopped short of studying the Advanced View stuff.

There's a lot of stuff I like. Some random comments:

I like the 3D VE setup, but:

The fuel calculation should have the following inputs in the setup screen:
- base fuel pressure at 101 kPa atmo/mani
- fuel pressure that injector flow is rated at (maybe this is a standard value thus not needed)
- check box for FPR - [ ]atmo ref'd [ ]mani ref'd

The injector flow curve vs. MAP is different for an atmo ref'd FPR and a intake mani ref'd FPR. For my AEM, I wrote a spreadsheet to calculate the VE correction curve due to the 99/00 mani ref'd FPR.

Without the proper calculation that differentiates an atmo ref'd FPR vs a mani ref'd one, one or the other's VE table tuning will distort from reality.

I like the idea of the 3D VE table with TPS as one axis, as I think a turbo motor's VE changes more with TPS than with MAP (for an NA motor MAP may be better). However, at lower RPMs the "action" of TPS is compressed at the smaller throttle openings, so you may not have decent resolution in that area. Because of this MAP as the Y axis may be a better choice. In an ideal world we'd have a 4D map with TIP (turbine inlet pressure) as the 3rd axis because the motor sees MAP, and exhaust manifold pressure. In our 3D world maybe a first 3D map with RPM and MAP as the axis, then a 2nd, 3D, VE trim map, with RPM as one axis, and TIP as the 2nd axis. This would also obviate the need for a baro sensor fuel trim.

JasonC SBB

There should be a max and min limit for each of the P, I, and D outputs.
For example in the miata VVT, I would want the I to be limited to +/- 5%, and P to +/- 20%.

The VVT should have a "if angle target is < x degrees, set duty to 0". In the miata a < 1* target should just set duty down to 0.

I like the MAP prediction concept, but is that enough to do throttle enrichment?

TravisR

iTrader: (9)

Yea, and you can create all that in advanced mode. You can physically remap and add new logic and tables to modify how fuel is calculated in the advanced mode. On the PIDs all you need to do is setup 3 separate PID controllers and then limit them using some logic and constants in advanced mode. Then the UI will automatically populate the new constants (your max correction) or you can have it use a multiplier to just multiply it by a fixed amount if your PID's are linked to a table. So you have the open loop value, the 3 PID controllers (P=some value I=0 D=0) and likewise for the rest, then some logic that says if open loop value is X then multiply P by .2 if P is greater the X*1.2 Then its .2 If not then value. (Or close to this!)

We have other ECU manufacturers and big big names buying these ECUs because they can test control algorithms on it in a couple minutes that it would take them days to code into their firmware. The best part is that it's patented for the next 20 years

With the 1280, your ECU is as powerful as your imagination and your will to do logic.

Also MAP prediction is all you need to do throttle enrichment. Whatever is left over is done by the asynch pulse which is basically dictated by your main loop speed. The faster the main loop the less asynch you need.

hustler

iTrader: (6)

1280 doesn't sound too bad with some of those features.

JasonC SBB

What does async do? AFAIK it adds an extra injection pulse for when MAP is rising, if the injector has already closed but the intake valve is still open.

MAP prediction won't be perfect in a turbo car because there isn't a one-to-one correspondence between TPS and MAP. The relationship is affected by the speed of the turbine shaft. IOW at steady state 50% throttle and 4000 RPM, I may have 150 kPa. If I lift and step abruptly the turbine will not have lost much speed and returning to 50% throttle I may get 145 kPa. So for this case the MAP prediction table should produce fuel be at 140 kPa. But if I'm off the throttle then punch it to 50%, the turbine is turning slowly and I may only get 105 kPa, and you'll get way too much fuel... I think a possible solution is to add throttle enrichment based on MAP rate of rise.

TravisR

iTrader: (9)

A synchronous pulse adds 1 batch fire injection at the moment of tip in based on dtps/dt.

Add your own control algorithm with that correction built in for the nonlinearities. Shouldn't be hard.

JasonC SBB

The MAP and RPM breakpoints aren't adjustable? I like to have a little extra resolution around idle, for example.
If one had MAPxTPS for MAP prediction, it would be good to have extra resolution at lower TPS values for the reason I described above.

The setup of the MAP sensor is a bit odd. I use a 2.5 bar Freescale MAP sensor. Do I use the 3 bar, then tweak the numbers to get the right slope? The breakpoints don't follow.

TravisR

iTrader: (9)

Yea they are adjustable and arbitrary. Just right click on the table and click insert column or row, then enter that column or rows value.

For a three bar you would just have to change the way the chart is setup, and the sensor cal. The ECU can accept any 0-5v voltage pressure source. It just has to know what ADC value corresponds to what pressure value so what goes into your brain makes sense. Ok, hang on I see what your saying, let me look at the advanced mode and see what is really going on...