D-Cal

Could be a failing fuel pump. It can't deliver the volume the car demands at WOT and starts leaning out costing power. Tie in a fuel pressure gauge and zip-tie it somewhere on the windshield that you can see from the drivers seat. Note the fuel pressure under WOT.

See this thread: http://www.ultimatesubaru.org/forum/showthread.php?p=389217

Yeah, if it's hard wired, great. I only mention it because I use a tiny racing battery and since the heater draws so much current, if I'm doing anything like programming the ECM in the garage with the key on run, it will kill the battery in minutes. I have mine setup with an extra cigarette lighter socket I wired in under the console so I can move the 2A0 to another car easily, or disconnect the LSU while I program. The gauge is a 4 digit digital unit that is velcroed to the dash for the same reason. All I need to sniff another car is a bung in the exhaust. Yeah, new. But as the sensor ages is where the recalibration comes in. Also accounts for different altitudes, etc. Perhaps your brain box does it's own compensation and calibration internally. The TechEdge guys are really knowledgeable about WB units. Their site contains a lot of excellent information, even circuit diagrams of all of their stuff. The Bosch LSU page: http://www.wbo2.com/lsu/default.htm The one negative thing I can say about the TE units is that the logging software is crude and clunky, but I can write my own. They've published the protocols. It logs 8 channels including 3 thermocouple/EGTs and 3 0-5V inputs, RPM and WB. They also give excellent after-sale support.

What sensor does it use? Bosch? NTK? I haven't heard of a sensor that doesn't require calibration. Also, don't run the car without the UEGO heater plugged in, it'll foul the sensor in 10 minutes. The heater draws about 2.5 amps cold and levels off at about 1 when running.

It's also an excellent way to check for even fuel distribution. The reddish tinge on a new set of plugs is a fairly good indicator.

I have a Tech Edge WB unit I bought as a DIY kit. www.wbo2.com Was fun to put together. It's installed in my turbo Dodge. Also has 8 channels of datalogging. Haven't gotten much use out of it yet, snowed in til next year. That's a fairly lean cruise. Good for fuel economy, but bad for emissions if your area tests for that. You've free air calibrated the unit etc? My TD car cruises within 0.1 of 14.7 (still using the stock O2, not emulated). How close to the head did you install the bung? If it's too close the sensor can overheat and give false readings. They want to be at a steady 900F. Mine is about 20-25" downstream of the turbo.

This is an excellent (but long) read on MMT: http://healthandcleanair.org/mmt/icct.pdf Canada banned MMT in a roundabout way around 97 and was sued by Ethyl Corp. under the provisions of NAFTA for damages and it was eventually allowed again. 95% of refiners are voluntarily not using it, so it is not widely used in Canada but you have to check where you fill up. The US EPA was also sued by Ethyl and MMT was federally allowed again in the US starting in 2004, except where banned by the state (CA). Some small refiners use it, again, check where you fill up. If your plugs or combustion chamber have a reddish tinge to them, your regular fill up refiner is using MMT. If you've ever used an octane booster while tuning a high-pressure turbo engine, quite often they leave the same deposits. It can foul your O2 sensor(s) quite badly. Toluene and xylene are quite often used as octane boosters as well, but both are highly carcinogenic. Not great alternatives. Most gas stations in Canada that I have personally observed are using ethanol up to 10%. Husky/Mohawk fuels is building a massive plant near their refinery on the Alberta/Saskatchewan border to produce it from corn and other local produce. With regards to gas mileage (the real topic of this thread, lol), you will obtain the best mileage by supplying the engine with the octane it is tuned to require. Try advancing the base ignition timing a few points and try again with the premium, you will probably notice a small increase in low end power. Also, to run the test accurately you should datalog for knock to know how the ECM is regarding your fuel.

KYB GR-2 struts/shocks are nice.

Canadian Subaru filters seem to be completely different part numbers. I just bought this Legacy at a dealership a week or two ago and they changed the oil before releasing it. It's not leaking at all. The brown crap is dried mud. The writing on the filter is: "SIX STAR SERVICE Subaru Canada, Inc. SCI530001 C41733" Here's some pics. Maybe an interim solution would be to order some filters from a Canadian dealer until SoA gets their act together. People at the end of that thread were reporting that the "new" filters were still leaking.

Any word on your fuel economy after refilling with the Redline?

Good records are definitely important, however controlling all of the variables is far more important. You must correct for things like ambient temperature, barometric pressure, type of grade travelled, etc. You must be mechanically anal and make sure your test car isn't suffering from rubbing brake pads, low pressure tires, change the oil regularly, etc. Hardly anybody actually does things like this, so I keep my skepticism handy. A 10% change isn't really statistically signifigant. 17-28% change is, and you're the first person I've seen to claim such gains. You must halt acetone use to really prove the point though. I wonder about the injector cleaning properties of acetone. These claims beg someone to run a car on a dyno in an absolutely controlled situation to prove or disprove this stuff.

The 97 Legacy GT I just bought had a non-functional rear defroster. Dealership fixed it before I bought it. They found a broken connection in the rear under the trim and resoldered it. They didn't say exactly where unfortunately, but he did say he had all of the rear trim apart looking for it.

It's great that there is enthusiasm for this. I can't really help you guys much because a) I'm a total Subaru noobie, and I don't have an older EA82 model to work on or test. You need a couple knowledgeable guys to step up and take charge. First, set yourself up a forum to work in. Either a message area here on the board, or better yet an entirely new forum with file and picture download/upload capabilities. Next, someone needs to pull the calibrations from their computers and send them in. In the TD arena anyway, some MY cars used several revisions of calibration, and there was a pattern of code evolution from the beginning to the end of the 84-94 MY roadmap. The OBD-II stuff is probably going to be an order of magnitude harder to decipher, so I would set your sights on the easier pre-ODB stuff first. Code based on the 8-bit ECMs. Someone skilled with electronics can take a complete ECM and get it running in a test-bench situation. Hook up LEDs to the outputs and run some test code to identify what output is hooked up to what memory location. Figure out where the ADC/DAC input and outputs are. Use your FSM wiring diagrams to translate these into "OK, this is the ADC that measures the air intake temperature sensor". Find a way to get test code into and running on the ECM. If you can build a flashable setup with a serial interface (maybe through the diagnostic/scanner port?) then that makes future tuning that much easier. Worse come to worst, you can get a device like the Romulator (www.xtronics.com), remove the stock EPROM and emulate it with a laptop. Then you can send in whatever you like, in addition to being able to monitor what memory areas the ECM is reading, in realtime while the car is running. Take pictures of the differnet MY ECM boards and send them up. Examine the tops of the chips and write down the part numbers there. Use the internet and white papers to try and identify what chip they are. Manufacturers typically order chips with custom part #s on them to obscure what they are. You may have to guess using information like "I know the ECM has an 8 channel ADC, circa 1988 and it's probably Motorola stock with 28 pins". If you know what I/O package is installed, and you know the locations of some of the binary outputs, you can sometimes guess where the rest of the package is mapped into memory. Examine the bus lines going to each chip. Sometimes you can guess the memory area the chip is mapped to by the configuration of the bus lines (though not always if it has an internal address decoder). All of this hardware information feeds into the symbol table for disassembly I talked about. Once you start substituting all of the hardware ports with names like "ALTERNATOR_FIELD_CONTROL_BIT7" it gives context to the assembly language and becomes easier to understand what the code is doing. Once you know how the code works, you can work on editing software. It can sometimes be as simple as an Excel spreadsheet with some custom macros. You can write your own from scratch like I have, or you can contact some of the people who market existing programs for other platforms, like www.tunercat.com from the GM arena. I am told some of those guys are receptive to expanding their market, you just have to give them the specs of how your ECM software works. They need to know how your table lookups work so they can program support for them. Sometimes it can be a simple (X,Y) table lookup, or it can require more complex interpolation math like the TD ECMs use. They use configuration files to map out where the editable data is, so you just fill in this table file and you get support from a fully functional editing program. What editing software you need will depend on how complex your code is. Won't know til you get there. This is pretty much a roadmap to hacking your ECM. It's not going to be news to experienced hackers. Like I said, what you need is a couple of techie EA82 owners to step up and take charge, although there is room for anyone to contribute in terms of finding spare computers, scanning out ROMs, etc.

Well, we've done it for the turbo Dodge cars. The editing program I publish is free for download. I have several hundred hours of development time in it now and I give it away. There are 600+ members in my Yahoo group learning how to use it. Many other guys contribute by deciphering calibrations, posting 3 bar custom versions, modifications that light CHECK ENGINE when the ECM sees knock, and so on. Early on we formed a talented group of individuals, opened an exclusive Yahoo group for the purpose of cracking the code. It can be done, you just need people interested in and capable of tuning their own cars, with a side of electronics background. It was never about the money, it was about being in control. What value would you place on being able to change your injectors to any size you want? Being able to set your boost goal anywhere you want? Not having to jimmy-rig your fuel system with RRR + AFPR to fool the computer into working with stuff it was never engineered to? If that stuff has no value for you or you guys are too lazy to go after it, I'll just shut my mouth right now. I'm really surprised that guys who drive K-cars have more initiative than this bunch.

Actually the stated reason on the site that the manual is broken into small files is because most of the world is still on dialup, and mechanics accessing repair info only need small sections. $200US is a blatant ripoff. I've bought complete FSMs for other cars for $75US.

Have you ever reverse engineered anything? What we found was that once you had mostly decoded one model year's code, you could take advantage of the similarities in other years to get a head start on other codebases. It's not really hard work, just time consuming. But it can be fun too, like detective work, or a very complex puzzle. That particular CPU can only address 64K max, and probably only has 16K or 32K of ROM in the address space. It's not really that hard to decipher code that small once you get a few addresses mapped. It will snowball. I believe the 6502 was memory mapped, but it's been about 20 years since I've done any assembly for that processor.

If you get a disassembler that uses a symbol file during source generation, you can feed in everything you know about the memory map. From there you start labelling routines based on what you think they are doing with the memory locations you know about. Repeated cycles between observing the code and updating the symbol file for generation will eventually get you to fully commented source.

Am I capable? Yes. Do I have time? Unfortunately no. But if a computer savvy individual wants to take a whack at it, I can give some basic pointers on how best to proceed.

If you got a web snake program that understood java script properly and was able to continually keep re-trying failed downloads, you might do it. I did it manually. I ended up with about 550-600 files and it took 2.5 days of downloading from 9am to 11pm. It really sucks. I wish they would offer a complete file to download for an increased price, like $75 or something.

I work on Dodge turbo cars 84-94. We have a similar problem in that there is almost zero aftermarket. As a group we've managed to reverse engineer the ROMs to the point where we can convert them from a 2 bar MAP to a 3 bar, rescale for larger injectors, change the fan schedule etc. I make a program called "D-Cal" (surprise) that is free to download and allows you to edit your ROM files. It supports flashing a modified computer, and I am (slowly) working on support for a ROM emulator (www.xtronics.com) that will allow you to alter fuel and spark in realtime, on the dyno. Pull the ROM from the computer, scan out the contents and disassemble the code. Look up the white papers on the processor and make note of the addresses of the hardware bits. A chunk of code addressing the area that deals with the crank sensor can be marked off as having something to do with engine timing, etc. This gives you a clue as to what a particular chunk of code is doing. Slowly, over time you can build up enough knowledge about the code to allow you to intelligently alter the contents to achieve your goals. There are many good reasons to keep and modify the stock computer. Emissions appearance in some states like CA for instance, maintaining the "stock" look, not wanting to do a lot of cutting and splicing of the wiring harness, many more. Many times, like the Dodge ECMs, they are actually very capable systems and once the code is cracked they can do anything an aftermarket ECU can do. There are a couple of Ford mustang guys running Dodge computers with my editing software for that reason. The 6502 processor is what was used in the Commodore 64. While this is kinda humorous, it also means there are a lot of tools out there for assembling and disassembling your flavour of code. Many hobbyists still program and work with the C64. Newer ODB-II cars are trickier to work with, but there are published protocols that would allow you to do things like datalog sensors etc. without completely disassembling the code. So it's quite possible to develop your own ECM editing tools, it just takes a couple of computer-knowledgeable people to spend some time disassembling and sharing the results. If such a group exists, I'd be interested in talking to them.

Thanks Murdoc, that actually makes a world more sense. If I ever have a serious engine failure, I'll be looking at this very closely.

I'm all for scientific tests, but I can't ignore that I've seen 3x as many people say they tried it and it made no difference.

That's better quality than my phone has, LOL. These cars look so sharp in black. I need to pull the fog lights off and polish the lenses, they're sandblasted all to hell. Will sand and polish them back up to new.

To be a valid test, you need to soak it in a gasoline/acetone solution of the ratio discussed, not pure acetone. Failing pure acetone wouldn't necessarily decide the issue, but passing it probably would.

They're retreated to near pre-Katrina prices around here.