S2R 1000 More playing - Written 10/07
Summary: More S2R 1000 based confusion with Lambda sensor unplugging and engine lights and dyno results for Moto One header, Staintune mufflers, std air filter and Dobeck unit.
I’ve played with a couple more S2R 1000 since doing our demo bike earlier in the year and coupled with responses from posters on the DML forum it seems there’s some inconsistency in the way the bikes respond to having the Lambda sensor unhooked. Our demo has been running with the sensor unhooked since it had around 1,200 km on the clock. Now it has around 4,000 km on it and we haven’t had any issues or apparent shifts in tune nor has the engine light come on at any time.
However, I’ve unhooked a couple of Lambda sensors recently and both bikes have bought the engine light on. I don’t see this as any sort of problem, except that the light may annoy some. As all the Euro spec S2R 1000 share the same ECU I don’t understand why some do this and some don’t. This appears to be mirrored by DML forum postings, with similar inconsistent results. Maybe you need to unhook them with the engine running I’ll have to try that.
A poster on the DML forum also commented that the tuning may shift richer with time once the Lambda sensor is disconnected. If you power down the ECU to clear the adaption table after the sensor is disconnected, there’s no reason for the tuning to move anywhere, as the ECU has no reference to go by in terms of either live feedback data (Lambda sensor) or previous experience (adaption table). The first graph shows the air/fuel traces for two runs on our demo bike, done 3 months and around 2,500km apart. Bike configuration is Moto One header, Staintune mufflers, no airbox lid with std paper filter, std ECU with Lambda sensor disconnected and no additional tuning. As you can see, they’re very much the same.
A short aside on the adaption table clearing bit. Duane Mitchell from Ultimap asked me if I actually knew that this worked (as he does sometimes when I say stuff without backing proof) and I had to reply that it was speculation based on what I know of BMW Bosch Motronic systems. So I must say that this bit is actually speculation. There is no facility to clear the adaption tables on 5.9M / 5AM ECU that I can find through either the Ducati Mathesis or DDS testers, or the Aprilia / Moto Guzzi Axone tester (it will also talk to Ducati models with these ECU).
Anyway, I’ve seen the procedure for this quoted quite a few times and every time it grows a little and gets a bit more involved and abstract. All this procedure requires is powering down the ECU. Meaning removing the constant power supply the ECU has that maintains the RAM and hence any data the ECU has stored itself. To do this you can either disconnect the battery or disconnect the ECU. I tend to unplug the ECU as that way I don’t have to reset the clock afterwards. I don’t know how long it takes, so I give it 5 to 15 minutes for superstition and plug it back in again. It’s usually dependant on how much of a rush I’m in. Thinking about it the BMW manuals talk about removing fuse 5 (from memory) so maybe you could just pull the ECU fuse. If you’ve got the time leave it overnight.
So, onto issue 1 inconsistency:
When I disconnected the Lambda sensor on our demo S2R 1000 I checked the idle mixture as we usually do and found it was a bit rich. So I leaned it off by winding out the air bleeds ie, allowing more air to bypass the throttle blades giving a leaner mixture as the fuel being injected doesn’t change. You can’t adjust the idle trimmer as the ECU doesn’t allow access to change the trimmer via correct S2R 1000 model connection using the DDS (Ducati Diagnostic System I think). And if you use another model that does have an adjustable trimmer you can change the number, but nothing actually happens to the mixture.
However, the customer bike I played with last week was very lean with the Lambda sensor disconnected. I had plugged it into the 4 gas analyser with the temp gauge showing 55 or so degrees (Celsius) and it was about right CO % wise. But as the temp came up to 65 or so the mixture suddenly leaned off dramatically, like a switched response. As this is the temp at which this model goes closed loop it’s obviously just the way it works. So I had to richen it up somehow. Given the air bleeds were between a quarter and a half turns out, that made very little difference. The richest I got it was 0.6% CO on the horizontal cylinder and 1.3% on the vertical. I have heard people say disconnecting the sensor made their bike idle poorly, and this result gives a clear indication of why. Although there’s every possibility that it could go very rich when the sensor is disconnected, giving the same symptom of poor idle. I’ve had Sport 1000 and ST3 models do this.
But, a couple of things worked in my favour here. One was that I had a Dobeck / SFI (see http://www.moto-one.com.au/performance/staintunetuningtechlusion.html) unit set up on the bike. The second was that the higher, ECU regulated idle speed of these bikes was above the green pot ‘on’ RPM. This meant that at idle, the green LED was lit, instead of all 3 LED flashing to indicate nothing is happening. So I could adjust the idle mixture as required using the Dobeck, which was very handy indeed.
With the Dobeck on setting 3 and one of the air bleeds wound out to balance the mixture for each cylinder, we had 4.5% CO at idle. As an aside - normally, on any of the older open loop bikes, we use the air bleeds to balance the idle mixture for each cylinder, probing the mixture from each header using the little take off ports. We do this instead of balancing idle vacuum, and it can make a real difference to low throttle performance.
One other point here is about how the Dobeck works. It adds fuel in increments of pulse width, not %. This is quite important as it gives much less comparative increase at higher throttle openings. It’s actually the same as the way the idle trimmer on the std Weber Marelli ECU works. So all in all, it’s very handy in this case. Setting 3, being 2 increments over 1 (zero added fuel) adds 0.32 ms of pulse width. This is probably about 10% at idle, whereas at WOT it’d be maybe 3% or less. Setting the red pot to 3 as well makes the whole map 0.32ms richer, just as it would if you’d adjusted the idle trimmer in an open loop ECU.
The above inconsistency (from a sample group of two bikes thus far) also brings up another issue with owners trying this stuff at home you may get a result that you don’t measure and therefore can’t determine the cause of. Sort of like when Doug Lofgren says “If you don't check your work, you can assume it's perfect!”, the fact that most don’t have access to a 4 gas analyser means that the end result will depend on where on the inconsistency bell curve your bike lies and that’s the end of it. So if you are going to try the Lambda sensor disconnection thing, I’d try finding a local bike or car workshop with a 4 gas analyser to have a look see.
Some of you might be asking at this point why disconnect the Lambda sensor. The answer is the same reason people are fitting things like the Rapid Bike Lambda emulator kit stability.
Issue 2 Stability:
The intrinsic nature of a closed loop system is transient. That is, it is always changing as directed by the specific nature of the closed loop parameters. In broad terms and in this instance, the stored adaption system means the areas of the map that are controlled by the closed loop system will constantly be open to change. In normal use with consistant fuel composition and no abnormal engine conditions, the adaption will approach a zero point, so to speak, where the closed loop system has to do very little to keep the feedback within the desired range. Which, in the case of the closed loop exhaust gas management system running narrow band Lambda sensors, is a low amplitude variation cycling around the zero point of 14.7:1 air/fuel ratio.
But, if you do something to change the tuning, the adaptions will then change as required to again approach that same zero point. If the adaption system works over most or all of the fuel map, this means any mapping changes you make may or will be overridden by the adaption system as it sees fit, within a time frame dictated by the adaption system’s definitions and running conditions. So to us, the punters outside Weber and Marelli who have no specific information on how the adaption system works, the system is for all intents and purposes unstable. And the only easy ways to make it stable are:
1/ to supply misleading feedback information (feed the ECU a signal that is both valid in terms of the internal ECU error sensing capacity and tells the ECU no action is required) that will keep the system happy. E.g., Lambda sensor emulator.
2/ directly impair the feedback system (unhook the sensor) and deal with whatever effects that brings as required.
Either way, you end up with a system that is stable and which you can then modify (via an add-on box, etc) as required to give the desired result. Or as close to the desired result as the systems allow, as the case can sometimes be.
The third and definitive option is replace the ECU, but I haven’t tried that yet due to free time at hand.
Back to the bike at hand. With the idle mixture set via the Dobeck unit we headed to the dyno. This bike was running our Moto One header pipe, Staintune mufflers and std airbox lid. I had dyno’d this combination with our demo, but with the sensor connected and without Dobeck. The first graph shows the power comparison between the two. Green is this bike with the Dobeck green pot on 3 and the yellow on 1, red is the same excepting the yellow pot set to 4. Blue is our demo with the same exhaust setup, but connected Lambda sensor. Not sure why there such a big power difference, but the curve is much smoother.
The next graph shows the air/fuel traces. Going by the difference between the green and red lines (3 Dobeck increments) and the fact the green line is theoretically 2 Dobeck units richer than the blue line, maybe there wasn’t that much change due to disconnecting the sensor with this exhaust and airbox combination. The rich swing at the top end that the previous report showed is still there though, but only comes in around 8,000 RPM.
Next we’ll compare the Dobeck tuning runs for this bike. Alterations are to the yellow pot only the RPM switch was set to 11 for the dyno runs, so the red pot was never activated. As the titles show, green is yellow pot set to 1 (no added fuel), red is 4, blue is 7 and yellow is 11. Given I like to run high 12 to low 13 air/fuel ratios we went with a final yellow setting of 6, lowered the RPM switch to 8 and set the red pot to 4. Power first, then torque and air/fuel.
At this the bike seemed to ride quite nicely, and the owner was happy with the improvement. We were lucky to get around the unexpected lean result of disconnecting the Lambda sensor though the inconsistency and not knowing exactly what’s happening with these bikes when keeping the original ECU is really annoying me.
Of course, after I’d written all of the above this bike came back for some other work so I checked the idle mixture again. It was now about 8% CO! Why, I have absolutely no idea. I wound the Dobeck unit green pot back to 1 (zero added fuel) and it was still over 6%. Winding the air bleeds out a quarter of a turn each lowered it to just under 6% CO higher than I’d really like, but acceptable. Winding the air bleeds out further started to lift the idle speed.
Maybe the engine temp has an effect initially I’d set the idle mixture with an oil temp of around 70 degrees, but when I checked it this second time it was over 100. Whether this is the issue I don’t know. I have played with a S4RS and found that the idle mixture after unhooking the Lambda sensor initially went quite rich, then came back to very lean. So it might be a time issue as well. Something to keep in mind anyway.