Its a software problem; really!

As of about mid-Feb 2001 all of the parts had arrived, but the overheating problem was getting worse.  The car went into the shop, the dual aftermarket Integra fans were installed, the rad was flushed, the thermostat replaced (with a unit 10 degrees colder than stock), a 70/30 water-to-coolant mixture was used, Redline Water Wetter was added, and the exhaust pipes were wrapped.  This has completely cured the overheating problem.  I can now drive around in stop-and-go traffic, get on the boost, and use the A/C all at the same time without causing the temperature gauge to so much as twitch.  This finally cleared the way for the MAF and injector ECU work that I have been planning for over a year now.
 

Attempt #1 -- Injectors

In March I finally had a chance to get into the shop and try my hand at the ECU.  I decided that the best course of action was to switch to the high flow injectors first and make sure the car ran properly in that condition.  This is a simpler change to the ECU than the airflow sensor conversion, so it seemed the right way to do things.  I went in on a Saturday and we pulled the already-socketed ECU, plugged in my emulator and downloaded an unmodified program into my ROM emulator.  The car ran fine.  Next I replaced the stock program with one that had my slight code tweak but which was calibrated for the stock injectors.  The car ran fine.  Then I made my first mistake:  I did not put the car on the dyno in this condition and test to see how the A/F mixture behaved.  Instead I was overly eager and we swapped the injectors -- this takes 1 - 2 hours (my mechanic gets a lot of distracting phone calls).  With the new injectors in I recalibrated and started the car.  It idled great -- better, in fact, than it did with the stock injectors.  Excited we burned a chip and took it for a spin around the block with the wideband A/F meter (the ROM emulator and notebook both need AC so I can't drive around with them).  The car leaned out seriously as soon as we put any load on it.  I put the emulator back in and richened up the mixture.  We burned another EPROM, but once again the car leaned out under load.  By this time I was not thinking clearly and it was late in the day.  We agreed to pick it up the next day.

The next morning I had a couple of ideas and we put the car on the dyno (should have done that right away -- no need to burn EPROMs).  After several hours of fiddling and experimenting I gave up and we swapped the stock injectors back in.  Out of curiosity we tested it under load... only to discovered it behaved exactly the same.  My injector recalibration was completely successful, the problem lay elsewhere!  Frustrated I packed up and went home.  I left the MAF with George so that he could wire it up and make the intake piping for it.  He wired it exactly as I specified in a carefully drawn diagram based on VAF pin information from the Haynes manual -- he did a beautiful job and it looks OEM quality.
 

Revelations

The next day (Monday) when I went down to my car on my way to work, it wouldn't start.  It cranked and almost caught, but would just die immediately.  Puzzled, I called George and then had the car towed back to the shop.  Since I didn't really need the car, I left it with him for most of the week while he did the MAF's wiring and intake plumbing -- and tried to figure out why the car wouldn't start.  It turns out the car would start... about every other day.  Once it was running, however, it wouldn't run well at all.  There were also signs of serious flooding.  Finally, on Thursday he swapped in a spare ECU and this magically cured the problem completely.  Unfortunately it didn't cure the idle-rich / loaded-lean problem we had observed on the weekend.  He also installed an under-hood fuel pressure gauge so we could keep an eye on the fuel pressure more easily.

My best guess is that somehow, despite the care with which we worked, we had subjected the ECU to a static discharge and killed part of its circuits.  With my original ECU obviously malfunctioning I bought a spare ECU from a fellow ECU hacker ex-Probe owner -- Scott McClements.  Scott's PGT was stolen a few months back just as he was starting to get into this whole ECU hacking thing, and that had left him with an extra ECU that he'd bought, and a bit of EPROM making equipment.  Since he had no use for this ECU I bought it from him, and had him solder in a socket from the ROM.  I'm glad I did too because not only is he clearly better with a soldering iron than I am, but he used a beauty of a socket too.  It took a couple of weeks for doing the soldering work and shipping, and then another two to get another opportunity to go into the shop on a Saturday.  It was 2/3rds of the way through April before I got in.  This time we were going to do the airflow meter.

While waiting to get into the shop I finally got around to doing something that I had meaning to do for a very long time -- measure the voltage of my stock airflow meter (the VAF) at idle.  It is supposed to read 3.2 volts, and this reading will drop as the airflow through the engine increases.  I was out washing my car (also long overdue) when I remembered I wanted to check this, so I grabbed my handy digital multimeter and backprobed the VAF's connections.  I did this repeatedly because I simply could not believe what I was seeing.  At idle my VAF reads 1.3 volts, which the ECU will internally interpret as almost six times the normal airflow at idle!  I can only interpret this to mean that my VAF has failed, and has probably been like this for quite some time.  If it is actually stuck at 1.3 volts then it would clearly idle very rich, but the O2 sensor based closed loop operation would attempt to lean it out as much as possible and this fits with the 11 - 12 : 1 ratio we observe.  As the load on the engine increases the airflow reading would become more and more correct until it was exactly right, and then it would become less and less correct as the engine continued to lean out.  This is also supported by our observations with the wideband A/F meter.  When the boost finally kicked in the FMU would raise the fuel pressure, richening the mixture and somewhat correcting the lean condition.  This too is supported by the puff of black smoke I get when I hit the boost.  At some boost level, however, it could not correct enough and it would start to lean out again -- which has also been verified on the dyno.  I don't know the source of this problem with the VAF, but it could easily be due to how the intake was rebuilt to accomodate the turbo.
 

Attempt #2 -- Mass Airflow Sensor

The next shop day finally arrived.  The advantage of doing the airflow meter is that it takes all of 2 minutes to swap in the new sensor.  The hard part is getting access to the ECU.  George pulled apart my dash and gave me access to the ECU, and then he left me to my own devices.  First of all I adhered to a new and more rigourous static suppression plan.  I used a grounding strap for me and the car, and I pre-installed the emulator into the socketed ECU.  I plugged in the new ECU, downloaded the program, and started the car.  And it died.  Didn't run at all.  Wanted to.  Tried to.  But didn't.  Damn.  Since I wasn't in the mood to fiddle around in the shop (it is uncomfortable and distracting) I decided just to put the dash back together but leave the computer sitting outside the dash in the passenger foot well where I could get at it easily.
 

Attempt #2.5 -- Home Brew

When I got home I hooked up the MAF on the bench and tested it.  It worked great.  I then plugged it into the car, applied power, and measured the voltages on each line.  They didn't make sense -- it seemed that the MAF was reporting 1.6 volts when it should have been reading zero.  After exchanging a couple of emails with Mike Palmer, he suggested that my wiring was "baffed".  I went through my wiring carefully, and it turns out he was quite right although he claims that he just wanted to use the word "baffed".  What happened was that I had based my wiring on the Haynes manual diagram which is not only mislabeled but also ambiguous.  I had interpreted the VAF pins from the left to right instead of right to left.  Woops.  On that Tuesday I picked up some more connectors and wires from Active Electronics and spent the evening correcting my wiring faux pas.  By the time I was done it was dark, limiting how much testing I could do, but I did verify that the MAF did now output zero volts with no airflow and a higher voltage with some air moving through it (even bad breath).

Wednesday I managed to get away from the office while the sun was still in the sky and I raced home to take another kick at the can.  I hauled out the notebook and the EPROM emulator, removed the VAF intake, stuck the MAF on, swapped the computers (very convenient having the computer in the passenger compartment!), downloaded the software, and fired up the car... and it started and ran!  The idle was beautiful except for a continuous 50 RPM up / down behaviour about once per second.  This may be due to the lack of any upstream intake piping on the MAF.  But it behaved better than I've seen it since before the turbo went on, and while I don't have an A/F meter I'd bet that the mixture was bang-on 14.7:1.
 

Pipe down

Flushed with success, I got George to make the new intake pipe and I borrowed the wideband A/F meter. After a couple hours of working with it, however, it became obvious that our first attempt at an intake just wasn't going to work. The pipe changed diameter 4 or 5 times, coming from the 2" compressor outlet up to the 3.5" MAF entrance, and then down to the 2.75" throttle body. The MAF readings were bouncing up and down at random, making it impossible to get the engine to run smoothly. I managed to fake it out so that it would idle passably, but I couldn't drive it without the engine stumbling continuously and badly. On the upside, the 50 RPM up / down oscillation was gone completely, probably because I had plugged the BOV hose. Our next attempt will be to make a single straight 2.75" pipe from compressor to throttle body, and install the MAF sensor element onto this pipe directly, avoiding the need to step the size up to the 3.5" bore that came with the MAF. In addition we will install a honeycomb mesh from a Toyota Supra's MAS intake. With the pipe having a 4 or 5 to 1 length to diameter ratio, and the honeycomb having about a 10:1 ratio, this should generate an almost laminar flow through the intake, and will hopefully result in much much more stable readings. The decrease in pipe diameter will change the sensor's calibration, but should still leave enough range to measure the full range of airflow that I need measured -- and improve the precision of the sensor in the process.
 

Ignition Interlude

While waiting for the new, improved MAF intake a problem developed -- very harsh stumbling and misfiring. It got progressively worse as I drove home from a movie one night. It was so bad I thought it was going to do serious harm to the engine's internals. When I got it home I called a tow truck and had the car taken down to the shop. Unfortunately the shop was really busy for a week, then the problem couldn't be reproduced, I was sick for a week, and finally I went to the shop and drove around for 10 minutes until the problem returned in a very minor way. It was almost certainly ignition related so we swapped to the original plug wires; the problem persisted. We opened the distributor, and lo'n'behold there was a dime-sized hole burnt in the insulation. This was an aftermarket cap and I'd heard that the plastic used is thinner and of a lower melting point than the Ford part, so I had George price Ford caps and rotors. Since I have an MSD I need to have the cap modified, and if this was going to be a problem then I wanted a couple of caps as spares.

George managed to negotiate a discount if I bought 5 sets of caps and rotors (about Can$24 off of each cap/rotor) and he would modify them all so I should never need to worry about running out, especially if these Ford units last longer. Unfortunately the parts guy at Coastal Ford on Main is an idiot and not only did it take a month (literally!!!) to get any part, he kept saying that it would be there "that afternoon" or "tomorrow"! If it is going to take some time, then at least let the customer know so it can be planned for! But nooOOOOoooo, this guy kept claiming it was "on the truck", "on the way down", "coming right over". They even sent a guy over, but when he arrived he didn't have the parts, he just wanted to use the dyno! When my shop sent somebody down to pick up the part, he arrived to find they didn't even have the damn thing despite being told over the phone that they did have one! And this is for a wearable part which is supposed to be replaced annually.

Finally, after sitting for 5 weeks we finally got a new cap and rotor in and working. The cap and rotor are both definitely superior to their aftermarket counter parts, the extra money is worth it (the inexcusable, exasperating wait aside). I drove the car to work, and everything seemed fine. When I drove home, however, it stumbled and misfired twice. Uh oh. I called George and made an appointment for the weekend. The next morning I swapped in new MSD plug wires I had ordered just after the problem had started. On the way to work, once the car had warmed up, the problem returned in spades. It was almost as bad as it had been that first time... what the hell?! At lunch I took the car straight back to the shop, stumbling and misfiring all the way. This time it was trivially reproduced and George agreed it was definitely ignition related. We replaced the spark plugs (they were 2 years old and a dark brown colour). The problem got worse... and we were running out of ignition system parts to replace. I had George call Ford to source a new ignitor module, which is prone to failures in this engine. It turns out that Ford knows it is failure prone, and has a kit specifically to fix the issue; only it is designed to rip off the customer. It costs Can$490, and replaces the ignitor, cam sensor, and crank position sensor all at once. The new integrated module is a different design than the original '93 parts, so hopefully they have addressed the reliability problems. I am using an external coil & MSD, which removes the main heat source for the distributor and reduces the electrical load on the circuits so this should increase its reliability and life time. With this new kit installed the car runs beautifully.
 

A Misadventure in the Interior

I ordered and received a pair of Corbeau GTS-II seats (from Dale at Corbeau Canada, a very helpful fellow) to replace the aging stock ones, and to finally give me a headrest that makes me feel secure. After much cajoling and persuading I managed to shoehorn the passenger seat into the car, only to discover that the Corbeau seat doesn't drop nearly as far. In fact, sitting in it makes one feel "perched" as if sitting in a highchair. Even at the lowest mounting position the occupant would be at least 3" higher than in the stock seat. The front edge of the seat is about the same height, but the stock seats drop radically toward the back of the seat which results in a low position (and consequentially lots of headroom). I had specifically asked Dale if the seats would raise or lower the seating position and he had claimed it usually lowers it by 1/4" - 1/2", so I can only assume that the PGT seating is usually "deep". Unfortunately this means that I had to send these seats back and eat the restocking fee. I may continue to look for used MX-6 seats on the theory that they probably have the same mounts and design, but I'll have to find them locally because I want to verify how they fit before buying.
 

Emissions

While waiting for the ignition problem to be solved, a fellow member of ProbeTalk was selling off various car parts, so I bought a used VAF (airflow sensor) from him. As I mentioned before, my original VAF has been reading 2.0v at idle which is a fair bit too low. When we fixed the ignition problem I also swapped in the new VAF, and examined the original. Because of the modified intake there is a clamp around the end of the VAF, and that clamp had been originally tightened too much (in fact the hose was too small and tight). In mounting the new one we used a larger size of hose, and avoided tightening the clamp too much. The emissions gas analyzer reports that NOx emissions at idle went from 10% (WAY too high) to 0.4% (about right). The VAF reads about 3.2v (just right), and the car now operates in closed loop... well, it did once we replaced the fouled oxygen sensors. The catalytic converter was also fouled and has now been replaced with a new "high flow" model (mostly because it was cheaper than the OEM part).  The emissions testing that followed is a classic example of why emissions testing in its current form is more of a tax than pollution control -- read about it here.
 

Passing the time...

While waiting for my mechanic to build a new intake for the MAF, I also assembled a two-stage manual boost controller.  I'm using the boost valve from Ric Gillis, along with a 3-way solenoid valve I got from eBay for US$15 and a bunch of brass fittings from The Gauge Store.  The brass fittings are pretty cheap, but the shipping and broker fees were brutal -- no doubt I could have got the parts a lot cheaper.  As it is the whole setup will end up costing about Can$150.  It'll be pretty big and ugly, but ought to work pretty well and will not have any bleeds in the system.  There will be 4 hoses total, two of which are quite short.  When it works I'll post a diagram and explanation.  When the solenoid is off (i.e. unpowered) the system will behave as if it were just the simple wastegate-based system that I currently use (3-4 psi of boost).  When the solenoid it will operate as described on Ric's website, except that instead of a bleed there is a check valve to relieve pressure between the WG and boost valve.  I also have a couple of small LCD meters which are powered by a 12v line and which display the voltage of a 0..1v signal.  Which I may connect to the engine's two O2 sensors as described on the AutoSpeed website recently.  The meter is pictured at the left here.  While not useful for measuring the A/F mixture under load, it will be useful for showing whether the engine is properly operating in closed loop.  I intend to wire these in using a socket so I can install or remove them quickly.

For proper monitoring of the mixture I have found a wideband A/F meter design that uses a Honda wideband O2 sensor (~Can$215). The kit designers prohibit this design from being used to build a commercial product, but they will supply the PCB and the complete design. All of the parts (aside from the sensor) come to about ~Can$40 if you just go and buy them from Digi-Key like I have. I bought enough stuff for a couple of kits in case I really botch it the first time around, but if I don't then I might be pursuaded to sell the extra parts after I've built mine.

The wideband A/F meter kit just outputs a 0-5 volt signal. This is not very useful directly, so I have re-awakened my little HC11 project based on the Adapt11. If successful, this will let me log a great deal of sensor data to a laptop (and/or watch some of the values in real-time on the little display). Since I'm hoping to have both MAF and RPM inputs this is enough information to calculate load (assuming no vacuum leaks), which will let me do a fair bit of tuning off the dyno. At least that's the hope.
 

Progress at last!

After about a year and two weeks the new MAF intake is finally done!! I've snapped a couple of pictures of this long awaited hunk of metal and plastic. The plan changed slightly from what was described above: George (my mechanic) took a plastic 3" MAF housing from a BMW, which includes a flow straightener, and built the longest, straightest pipe he could in front of the MAF. He installed the VAT sensor opposite the MAF tube, and put a small deflector shield at the end of the MAF's sampling tube to avoid problems with reversion pulses coming out of the engine (my idea, his welding talent). I'll now try to get this unit working, and if successful we'll transfer the BOV over to this intake. The straight pipe in front of the sensor element is about 13" long. The flow straighter is slightly deeper than the size of the small square openings (about 1.2:1 I'd guess). Since the tube's diameter is 3", this gives a length:diameter ratio of about 5.5:1 which is much better than the first intake. Pro-Flow's website says laminar flow develops at low velocities in about 6-8 diameters, and 4-6 at high velocities. Hopefully the new tube will be long enough (it ain't getting any longer!!).

As mentioned above, I'm concerned about the MAF's calibration. When I bought the unit originally I asked for a standard Mustang calibration for 19# injectors, which should be sufficient for about 300 hp. The problem is that the calibration is done by modifying the sensor's resistance until the measurement is correct for the sampling tube and the sensor body that it is mounted in. We have changed the sensor body from the big metal 3.5" monster that it came in to this new 3" plastic unit, which is a reduction in area of about 26.5%. The area of the sampling tube, therefore, is now 36% greater in relation to the new tube than it was to the original tube. This means 36% more of the air goes through the sampling tube, and so its calibration is off by the same amount. The 300 hp worth of airflow it would have originally measured is now down to 220 hp. Woops. This may be mitigated by the flow deflector we have put at the end of the sampling tube. It might reduce the flow across the sensor and offset this change somewhat. Also, the original tube wasn't a uniform 3.5", it tapered and bulged in odd ways down to 3" at one point. I don't know enough about this stuff to predict the effects of that. I'll have to try it to find out.

Fortunately the Pro-Flow MAF sensor is identical to a Ford unit in size and shape, and I know that a local wrecker will have stock sensors available. The stock Mustang sensor comes in a 65mm tube (2.56") with a 19# calibration -- which if transplanted to my 3" tube will give a maximum reading of 412 hp. This will give me plenty of "head room" but pushes the reading at idle lower, which may cause accuracy problems due to the flatter slope of the MAF output curve at that point. Or it may not -- the stock VAF has a very flat curve near the bottom and the ECU handles that just fine.

Another possible concern is the calibration of the VAT. Pro-Flow claimed it was a drop-in replacement for the unit built into the VAF, but I suspect that they know nothing of the KLD's VAF and were talking about the VAF used in a few motors (2.3L turbo, in particular). The calibrations for these temperature sensors are very similar, but may not be identical. This isn't that big a deal, actually, since I do know how to recalibrate and the engine may not be that sensitive to the small differences in calibration. I'm hoping my logging tool will help me monitor this as well.

Trying again...

Immediately after getting the new intake I had briefly tried it with the modified ROM. I ran into some issues with my ROM emulator (which have since been worked around) and so I just burned an EPROM and tried that. After starting it would jump to 2K RPM and then immediately die. Since I didn't have the emulator I couldn't try any tuning so I went back to the program and walked through it instruction by instruction. After a few minutes I discovered that the calculated load value was in the X register but needed to end up in the D register... but that was normally done by some code I had removed. Woops. I put in an instruction to move X to D and it should (hopefully) now be correct.

Due to being busy with non-car related things I didn't go back to the car for about a month. This project has been going on so long no without any real success that I'm a little nervous about trying this again since if it doesn't work I don't know how to proceed from here! Finally I bit the bullet and tried again...

This time, with the corrected program burnt into an EPROM (the emulator is hooked up to my data logger project, which is coming along nicely BTW), I plugged it all in and started her up. She started flawlessly and immediately went to a completely normal idle, and then continued to idle perfectly through the entire warm up until the RPM was down to normal. The voltage reading was 0.71v at normal idle, which is somewhat higher than I had predicted so I expect that I'll need to find myself a used MGT MAF sensor element. I may also try to see if I can adjust the Pro-Flow unit's calibration myself (will need to talk to an expert before attempting this though). I also gunned the engine a few times while it was sitting idling on the MAF, and it responsed beautifully... none of the issues I had noticed last year were present. I don't know what the mixture was (no wideband A/F meter yet), but there was some moisture on the tailpipe so it couldn't be that far wrong. I'm now jazzed about taking the tuning to the next step.

The current status...

I may have been jazzed about the next step, but it has taken a very long time to get there. It has been ages since the last update so I thought I should add something here. First of all I've been rather preoccupied by non-car projects, one of which means I'm moving in the next couple of months and will have a closed garage for a change. Changes at work also mean I won't be driving to work every day, which in turn means I can be a little more aggressive about mods and leaving the car in an undrivable state for a while (intentionally, not like the unintentional times in the past!).

I have acquired a laptop with a working battery (a Toughbook), and have written some software for the A/F meter. Looking back I notice that I haven't mentioned that I decided to buy the FJO Wideband O2 Oxygen Sensor Controller. It has serial port output, is professionally built, and has RPM and an "extra" 0-5v input (for the MAF in my case). I bought the bare bones controller since I already had the sensor and can write my own software. The unit seems to work very well and is built tough, including high quality connectors. I've wired my engine with the connectors and can now install/uninstall the unit in about 5 minutes. I'm tempted to run a serial cable permanently from the engine bay to the passenger compartment, but haven't bothered so far.

I played around with various options for recalibrating the MAF I already have, but it pretty much impossible to do the resistor modification without the right equipment and expertise and hacking away at the sample tube size doesn't make enough of a difference -- and it probably does nasty things to the laminar airflow. I picked up a couple of other used sensing elements from other cars at a local wrecker, but their calibrations aren't close enough either. I finally decided that I would have to buy another, and then swap the element into the housing we've made. Then fate intervened and Pro-Flow came to the rescue with the Univer Plus.

This unit will soon be available (after a several month wait Jim at Pro-Flow promises me they've got the production under way and my unit should be built and calibrated Real Soon Now; Update Dec/03: Jim continues to have production problems, but I'll wait since I've got more important things going on). I've ordered a 3". The new design is (going to be) cheaper than the old Univer, and they've eliminated the large annulus (ring) around the main body of the unit, so a 3" unit is 3" in diameter. It is also shorter than the old unit, down to 5" from 6.5". And they have improved the unit's performance, resulting in a more stable and accurate signal! It is one of those engineering solutions that improves a product in all parameters -- I liked it so much I ordered one on the spot with the calibration I need. I've been waiting a couple of months now, but during that time I've been busy with non-car items and a few car maintenance issues (dead alternator, dead coil, paint problems on the hood vents; Update: blown head gasket and engine rebuild [ouch!]) so I haven't really noticed. When the unit arrives and I've got the garage then I'll get back out there with my new laptop and A/F meter software.

One other change to the car this year has been to use the new 3" intake pipe with the VAF so that I don't have a clamp on the end of the VAF (hence distorting it and ruining it). I very highly recommend that people don't clamp directly to the narrow end of the VAF.

As an idle distraction from other things I've been continuing to toy with the idea of converting the car to a direct ignition system. This has been ongoing for over a year now, and last year I actually tried to do it using the eDist from F.A.S.T.. Unfortunately they don't support this with anything except their own aftermarket ECUs, and they don't publish their signal diagrams. Not one to be discouraged easily I have been pursuing other options, at least as a thought experiment. One possibility is to use an OEM unit from a Buick V6 or a Ford Thunderbird -- both units are self-contained and with a bit of signal conversion could be made to listen to the Probe's sensors. Unfortunately that requires a bit of a conversion circuit. According to Mike Palmer (who has gotten out of the Probe scene, by the way) this circuit is only slightly more complicated the building a microcontroller circuit to do the whole job. As a result my current line of thinking is to use a PIC or an HC11 (both of which I have several of and compilers/programmers to go with) and build a small unit which takes the signals from the ECU and sensors, and passes the timing signal on to one of three inputs to a device like this. Crazy as it may sound, this doesn't seem like a hard project... but actual implementation will wait until the MAF & injectors are working.

Check out ProbeTalk...