Flinkly

it's nice to add the bit about AFR, catylitic converter and alternating rich lean (make sure everyone reading knows), but what would be the preferred AFR for MPG without CAT? just a suggested addition to cover the bases. an awesome start to a summary post. compliments the thread well. keep up the good work with logging and analysis!

Elf is still selling Shark Tank on the idea to get the funding. But he did get the website up and is spreading the word across the forums!

Subaru oem. parts counter will get you the right part numbers/parts. Sent from my XT1028

yeah, for winter or cruising, grill louvres would be nice. keep your engine bay "sealed" up during warmup, and only open when you actually need any cooling, and only enough to get the coolant below 200. otherwise you just have increased warmup time and drag. i'd just put winter grill blocks in, but would be afraid of temps being too high after warmup. and to me, making a homebrew grill louvre mechanical system would be easy. making it respond to coolant temp, and electronically open and close in response to coolant temp - difficult.

I've got a windows 7 10" laptop in my car at all times. when others are riding with me, it's usually under the seat. otherwise it's my passenger and logging. I use it sometimes to analyze data, but a larger screen and faster graphics processor are sometimes needed for big logs.

well, I've only had time to look at the base ROM data. looks like, with zero'd comp tables, (an older) stock Subaru should have a big stumble at 2400-2800 RPM, at 8ms IPW and above ("high" load). there are two smaller spikes (occurring at "low" load, 2ms IPW+) at 4.4k and 6.4k. single injector compensating up to +10%, with a range of up to 10% (at one point, it's 0%, 3%, 6% and 10% for each cylinder at the same point in the table. should see quite the stumble. but this only happens at 8ms IPW and 2.4k rpm, and a little less so at 2.8k rpm for the same IPW. it quickly drops off to a 4-6% range max for the same RPMs and higher IPW. other than that, it only goes to about 3% compensation max (although it varies per cylinder by 6%). there are two of these occurrences. there is also an odd +6% at 800 RPM and 2ms IPW (lowest IPW and RPM in the table). looks like there's some low flow oddities in the system too. makes me wonder if the older subarus have injector "knee" tables, or if this was their way of correcting for that. So for the "high" load spike, with zero'd tables, i'd expect a +5% correction factor, with two cylinders being off by +/-5%, and two being off by +1% and -2%. enter stumble. Same sort of occurance at "low" load at 4.2k and 6.2k, but with fueling being off by +/-3%. And of course, any fuel system changes would move these spikes, and might make them better or worse. could also expose spikes not currently seen. I suspect the STI regulator reduces spikes, making them easier to compensate for with learning and correction factors. now to see what the logs have to say about the "high" load spike, since I assume that was the only one captured (due to how I wanted the log). hopefully there were some steep hills...

i'm not sure anyone has really tuned it out (not to say they aren't awesome at tuning). I've heard you can mess with MAF and a few other tables to try to smooth it over, but that's just masking the problem. IF it is fuel harmonics, you'd need bungs in your header at each cylinder, and to log AFR at each, simultaneously. to be able to re-create the correct IPW vs. RPM comp tables. short of that, you could zero your comp tables, log where and how big your stumbles are, then try to adjust your comp tables globally to correct. would probably be better than running OEM comp tables, but could be worse for the engine than running zero'd comp tables. so overall, best bet would be to get a few feet of hose... p.s. i'll admit i'm surprised no tuners have attempted a 4 AFR sensor dyno tune to look at this problem at the source. expensive, but if they could fix it, worth it I think. then again, Cobb is getting people to pay $250 for the STI regulator (and maybe they have figured out how to fix it from there and just don't share...).

Nightmaresmk, there's other things that could do it too (i was being a little shortsighted in my earlier post): fuel rail, headers, intake, major maf changes (tune), deleting/changing your ipw per cylinder comp tables, etc. I mean, if it is why the per cylinder tables are there, then anything that changes fuel flow or major airflow thru the engine, would foul up the stock table. I actually haven't yet, but i should zero mine tables and redo my datalog. Covert: engine rpm, ipw, afr, correction, maf (just to remove non-steady state), learning (if you have it enabled) , and before that, send over your rom so i can see the comp tables. And yeah, stumble seems to be low ipw accross the ful rpm range. So closed loop, steady state or slow acceleration/hill climb. Need to keep it steady state enough for nearly constant afr. The stumble is a two parter, from what i've seen. The actual "engine bumpiness" is due to two injectors needing to stay open as much as 6% longer than the other two, while the overall harmonics seem to drop fuel flow vs. expectations causing hesitation while its correcting. Both extremes happen at the same time. The regulator or hose fix is just to minimize the %'s listed above, and maybe move them. When plotting correction vs. rpm, the overall harmonics are visible, and i'd guess that they correspond nicely to what was in your tables (assuming you don't already have a bit of stumble from mods already). But i'm speculating from only my experience and what i've read. No real data. Sent from my XT1028

i'm not sure anyone knows what exactly causes it, but my current assumptions are injector changes (fuel pumps usually get changed then too). so stage 3+. From what I've seen and read, I think per cylinder PW compensation tables (or really their reason for existing) are what's causing the bulk of what we see. once we change injectors (and sometimes pumps) we "change" this table, and I don't think we know enough about the table to be able to adjust it back to where it needs to be (or some just zero it out). Actually, i'd be interested to see a datalog from you of fuel correction vs. RPM on your current ROM (would be better if fully stock, but you're close), and depending on what that looks like, doing the same datalogging with your IPW compensation tables zero'd out. Steady state logging, across the full range of RPM (closed loop). zero-ing the tables should only induce a 10-12% fuel correction where you have harmonics, and a 2-3% difference in cylinders/injectors. not a great thing to cause, but i'd be interested in seeing it. and you should be fine with "slow" driving, plenty of time . after all, we all seem to be running with incorrect or no injector tables ourselves.

Need to roughly estimate volume available for air in the tire, psi you want, and expected heat while driving. Or just good ol checking. Check pressure before you start (and temp of ambient) then pressure after driving. Do this a few times. Then guesstimate from there! Sent from my XT1028

Here are my two (Stock) AVCS (cruise/non-cruise) and Timing tables. I read them as cruise=emissions optimized and non-cruise=power optimized. find it interesting that both have a 40 AVCS bump for a few cells (but below "normal" cruise speed). I also removed (didn't show in the picture) everything beyond 4k rpm, as the cruise/non-cruise maps are identical (and 4k+ really isn't cruise anymore...) I also cruise at 65mph, so about 2600rpm and .4-.65g/rev I think i'm going to start lowering the cruise map, so first rev will be reducing all 40's to 30's. I'll do some coarse changes like that until my 65mph highway cruising maximizes MPG/minimizes load.

The green test connector for flashing new code, opensource style. And yeah, for some reason my car has two wires in the connector. And theres a little bit of discussion on the last page of my build thread about the design, and a cool alternative to that connector (that link was my last post there, only linked a single post though). Sent from my XT1028

too lazy to write duplicate post, so follow link for my rev. 1 Aux Panel for easier ECU flashing and datalogging: http://legacygt.com/forums/showpost.php?p=5027876&postcount=50 still some work to be done if anyone else wants one, but feel free to write me and we can work something out for files to discojon (or whomever)...

very interesting. so AVCS set to a high number (for cruise) seems to NOT be a good thing for MPG (or at least for your setup). I like the method for testing these. only taking data at cruise, as opposed to an entire trip. did you remove datapoints where the engine wasn't under constant load (like when it hits a little dip in the road and "comes off" the accelerator peddle...)? really gotta fix my tip-in so I can follow along.

yeah, been reading the same threads. now if only E85 was locally available! makes me think ELH's (or anything better than our crappy OEM header) would be better too, pulses not colliding with one another constantly. if only I had the funds to test... keep up the good work.

yeah, VE has always seemed to be the thing to optimize, but optimizing for minimal engine load might be best, and is what is most often discussed/compared in MPG tuning threads. so like what you do already, changing AVCS angle and timing, but then use engine load as your output (at least for cruise scenarios).

have you ever looked into tuning VE? at least in the cruise range. might even be worth "poking" beyond what seems logical, and by that I mean not just tuning into the first VE "spike" you see/find, but seeing if there are others that might be better. in engineering it's usually easiest to test only 2 or 3 "levels" of a parameter, and then extrapolate from there. which is easiest and probably will find a pretty good answer, but this approach might miss the better sweet spot. as in, if you test only two points, you can only assume the results are linear, where it could be a crazy 3rd order polynomial... speaking of, I wonder if DOE analysis software would be useful for tuning... (I've got access to quite a lot of engineering analysis software tools) also, VE doesn't always correlate to MPG directly, as seen in some of the tuning for MPG going on with E85. just some thoughts I have that I figured i'd share. after I tune my tip-in and mess around with trying to fix my fueling harmonics, I will be starting to tune cruise VE (or MPG actually).

Is there enough room between the vents and ducts in a regular cubby to mount? cause then we just need to redo the face of that dash piece, and re-use the vent, duct, etc. does anyone care if the clock goes away? I mean, a nexus or a new radio both have clocks... and it'd probably be better if we just made it dual din size, and then people could adapt from dual din to whatever flavor. that all sounds pretty easy to me, if it fits between the vents/ducts. just get a $20 dollar dual din cage, redo the face to accept the cage and OEM ducts, some paint and screws, voila.

can you log it? if you log fuel corrections vs. rpm (and remove non-steady state data), any harmonic inherent in the fuel system will show up clear as day. you'll even learn where it/they happen, and how much your car is having to account for the problem. but since you say yours only happens on light throttle, it almost makes me think your tip-in is off. either way, logging the data and graphing it will show you if you've got harmonics. PM me if you can log and I can assist in parameters and data mining. in all honesty, i'd love to see what your harmonics are, and then see what happens if you replace the regulator (since you've got the same system that I do). If it works for you, I'll be ordering parts after seeing the data.

k, I just found this thread, and as a Mech Eng design engineer with access to a full shop (and machinists to make my desires) I've got a few things already done, but are fairly specific. 2 gauge cubby pod: to hold 2 52mm gauges (specifically designed to PLX guages) and still close the cubby door (for parking in creepy places). I can't promise it'll hold anyone elses gauges though. I also made 2 other revisions that have usb ports (one revision with 1 port, 1 with 2 ports) built in for logging/flashing. it seemed a good idea, but having cables coming out the dash sucked, and I used some "expensive" usb port parts available on Mouser. Clock housing redo: I hated the reflection off my clock face, it was super scratched up, and the mounting parts were cracked/broken. I got a 1/8th thick piece of smoked plexi and a printed part to mount it to and all was well. I also removed the front panel "clock set" buttons, so a little rewiring is needed, and have to remove the dash piece to reset the clock for daylight savings or after battery replacement. Also, I only have a clock, so i'd have to modify the part for anyone with the other version of clock that has temp too. - In thinking about it right now, could add a spot to hold a cellphone in landscape for GPS navigation or something... AUX/USB/test connector center console panel: This is to give me a spot to have a usb plug into for datalogging/flashing, and a different way to connect the green test connector for flashing. it's all located with the radio AUX port in the center console. a lot of extra parts needed to make this all work, but I've got it listed in my build thread (as with the others too). I will admit the USB port part is pretty specific, but I can link the specific cord you need. Let me know if anyone's interested and I can supply the files to the community. I can also do a more in-depth writeup of the other items that need purchased, and how it all goes together. ALSO: If anyone knows of a good way to get a small digital readout/display for oil temp and/or pressure (or whatever), i'd be up for modifying the clock housing to mount gauge readouts... i'm also working on a AVO TMIC to stock splitter adapter, so I can remove my engine cover and still have a seal with the stock hood splitter.

one thing I do want to repeat (for clarification) is that there really isn't an "eliminated stumble" solution, only "reduction" or "moved/shifted" solutions. Does that make sense? so you either reduce the amplitude of the hump/peak/stumble (seems to be what the STI regulator does), or you shift it in the RPM band (left or right, which is what mostly happens with 3 extra feet of fuel line). also, there is and always will be more than one hump, even if you don't notice it. I've found 3 between idle and 4k so far with my car. I can't speak to fixes, as I've not tried anything but ECU tuning (which hasn't worked so far, since it seems to be a physical harmonic inherent in the fueling system).

I just wanted to attach this chart of my "stumble". you can faintly see one at 1200 rpm, and "the" big one at 1700. I have another above 3k, but this data is for fuel injector fine tuning and so I'm mostly after cruise and idle data. I have been going slowly thru the 1700 rpm range to capture this though. No matter what I change for injector tuning, it always tends to be about +10% at it's peak, and is always at 1700. stupid harmonics.... For reference, I have a Whalbro pump, DW850 injectors, and OEM fuel components otherwise. And I still have my injector comp tables intact, but adjusted for bigger injectors. I did zero them for a few runs and the stumble continued to exist.

from what I've read, the wires break and the rubber accordions are fine 95% of the time. so you've got to pull the accordions out of their holes and tug on the wires to see if you've got broken ones. or alternatively wait for electrical items to start failing.

damn. after really considering spending the money to do this, I find out that the 07+ topfeed fuel system is "totally" different. this thread, at the moment, is really only relevant to 06 and previous systems... I've got dampers where y'all seem to have your regulator. and my regulator is somewhat built into my hardlines at the manifold. not as simple, but I am going to look into how I could make it work. EDIT: looks like a Turbosmart TS-0402-1001 is needed ($25) to fit where the current regulator is, then similar hoses to connect it up. looks like 1 less quick connect, but 1 threaded fitting for the turbosmart adapter.

yeah, it's just a closed loop "compensation" (or additive?) vs load table A and B. B was 14.7 idle, but A had a 14.4 idle. I think it's better to have it lower though, for MPG. no need for moar power at idle. could probably go lower... now for tuning, I did bump it up to 14.7, so that I could do a better job of tuning. the change from 14.7 to 14.4 at idle has been messing with me.