In-Range Failures

By Richard McCuistian

After an engine swap, a vehicle is supposed to run better, not worse!

1996 Mazda 626

2.0L Engine

CD4E Transaxle

122,653 miles

Unit vibrates at idle; doesn’t run good on the highway.

Sick Six-Two-Six

At the risk of offending somebody out there, let’s begin by making the assertion that any technician who is worth his salt will finish what he starts.  I can remember a few technicians I’ve known in my tenure who really enjoyed fixing cars… to a point.  They liked the nuts and bolts aspect of engine work but once all the fasteners were torqued and it came time to get the engine running right, they’d “call for the calf rope,” so to speak, hoping somebody would waltz by, turn a magic screw and put a hose in place to make the car start and run right, effectively bailing them out.    Isn’t it nice to be paid the flat rate hours it pays to swap an engine and not have to worry about ironing out the kinks?  As fuel, emissions, and ignition systems become more complicated, it seems that more and more customers come to us and begin their story with the words

“I had the engine replaced, and…”

This 626 was a classic example.  To begin with, my phone conversation with Debra (the lady who owned the car) wasn’t as detailed as it should have been.  She and her husband had paid a local shop about $1500 to install a salvage yard engine because the original engine had developed a knock and after the job was done and the bill was paid, the car didn’t run as well it had before the previous engine started knocking. 

Debra described the problem as a “vibration” and I immediately began thinking of driveline or wheel balance problems.  When we finally spoke face to face as she was filling out the paperwork, it turned out that the vibration to which she was referring happening when the car was sitting still.  When the students started the vehicle and showed me the idle problem, I had to admit that it was pretty excessive, in spite of the fact that even a healthy 2.0 platform has a tendency to send a mild jiggle through the steering wheel at idle.  The MIL light was on (although I didn’t remember her mentioning it), so before we did the highway test drive, I had Jimmie plug the Hickock NGS into the DCL and have a quick look at what was stashed in the memory.

Rich Problem or Lean Problem?

The P-codes we received showed that the fuel control system had reached maximum compensation for both lean and rich conditions.  While such a combination of codes sounds peculiar, most of us have seen this little syndrome before.  One possible cause of this concern is an oxygen sensor that drops offline long enough during closed loop operation for the PCM to ratchet the Short Fuel Trim ‘way up into the +20’s percent range and stay there long enough for the Long Fuel Trim to follow its more sensitive little brother to the rich side of the control band and file the lean code while coating the ceramic zirconia O2 sensor bulb with a nice layer of hydrocarbon soot.  Then just as suddenly, the O2 will decide to start telling the truth again, but now the Long Fuel Trim numbers will have been corrupted enough by the dead sensor that the suddenly truthful O2 sensor will instantly begin to reflect a rich-running engine, and the problem remains persistent enough while the soot is burning off that both Short and Long Fuel Trims will drift deep into the – 20 percentages, and now the PCM has stashed the opposite code in its folder.   Since this phenomenon can be terribly intermittent, it’s not at all unusual to see normal datastream numbers at the time of testing, along with a normally running engine.

But the Mazda wasn’t running right at all.  The idle was far too rough, and a quick peek into the datastream revealed numbers that were anything but proper.  The Fuel Trim Readings were dancing around the -25% mark, but everything else looked fairly normal, with the exception of a MAF voltage that seemed a bit too high.  Ordinarily, MAF voltages will run in the .7 to 1.0 volt area, but this MAF was reading nearly 1.2 volts.  Checking the idle air bleed screw on the throttle body, we discovered that it had been backed almost all the way out; somebody was desperately trying to smooth the idle by giving the 2.O a little more speed with a closed throttle.  After running the idle speed screw back to its normal position (650-750 rpm with the Air Control valve disconnected and the cooling fan not running), neither of us was satisfied with the idle quality, (let alone the scan tool readings; MAF still seemed a bit high) but we opted for a highway test drive.

Anemic Acceleration

Chrysler hung on to the Manifold Absolute Pressure sensor and stayed away from the annoying MAF, so Chrysler/Dodge/Jeep techs haven’t had to worry with the bucking and jerking that can be caused by a crack in the air inlet tube.  How many times have those of us with experience seen a buck-jerk or serious surge problem due to an unmetered air leak that would shift fuel delivery to an intolerable range as soon as the engine torqued back on its mounts?  With the engine rocking into and out of a hard pull due to power that came and went with the opening and closing of an air tube crack, the situation can feed itself to the point of bouncing your head off the headrest.  The Mazda ran a lot like that, but the problem only seemed to occur at about ¾ throttle.  If Jimmie pushed his foot past the ¾ throttle point, the car would hunker down like it was starving for fuel.  Right at the ¾ throttle level, the 626 would buck and jump, and below that, apart from the rough idle, the engine ran normally, albeit a little weak on takeoff.  The TP voltage was sweeping smoothly all the way from .50 volts up to about 4.60 at WOT, but the O2 would go lean, even though the PCM should have been delivering full enrichment at that load and throttle angle.

“What do you think, Jimmie?”  As student and instructor, we were about to enter opportunity for a rapid learning curve, and Jimmie was eager to make the call.

“I think we need to check the fuel pressure,” he postulated.  It made sense.  We headed back to the shop.

Fuel A’plenty; Still No Power

            Fuel pressure that drops in a hard pull is a dead giveaway, and will usually show up on the gauge even under no-load snap of the throttle in the service bay.  With the fuel pressure gauge connected (it takes a special adapter) and taped to the windshield, we test drove the car and found a smooth 45 lbs of fuel pressure even the heaviest acceleration.    With that piece of data under our belts, it was time to look a little closer at the scan tool readings.

           With an under-load stallout problem, it’s only natural to suspect fuel pressure or quality.  While the gauge might show tattletale pressure loss on a heavy no-load service bay accelerator snap, the smart way to check the pressure in a situation like this is to record the numbers with the pedal to the metal in a hard pull.  And unless this gauge was reading air or water pressure (which, as it turned out, it wasn’t) this number looked good, even under load.

  At wide open throttle acceleration, as we peered through our window into the system, we reiterated the discovery that both O2 sensors were reading lean, Short Fuel Trim was at 33% and that the Mass Air Flow (MAF) sensor was reflecting only 2.78 volts. 

            When I teach my Power Stroke diesel workshop, one of the principles I hammer on is something I learned twenty years ago when fuel injection was still relatively new.  We’ve all applied the principle, but sometimes it helps to verbalize it.  

A technician who carefully studies and learns to recognize normal readings on a particular vehicle or system can usually pick out an out-of-the-ordinary set of signals and find the source of a concern in short order.  Understanding the nuances of a particular system and being able to compare one signal in a system to another signal that should respond in a particular way is one of the defining qualities of a technician’s knowledge base and becomes a powerful tool that enables the Top Guns in this industry of ours to surgically repair vehicles that have whipped other technicians for days.  What we have to learn to do is interpret the inputs and outputs in their correct correlations the way the PCM does.  There are some correlations that are fairly standard.

This is what the MAF should look like at Wide Open Throttle (WOT).

For example, if the PCM commands an IAC percentage increase on an idling engine, the MAF doesn’t reflect increased airflow, and the rpm doesn’t increase, the PCM has the idea that the IAC isn’t operating.  When a PCM knows there isn’t an electrical concern but can’t detect a physical change with correlating sensors after commanding an input, a “performance” code is set, and it will be defined that way on the scan tool when the code is retrieved.  We as technicians can think the same way when we learn how.

            In this case, while we had no MAF code at all, it was becoming evident that the MAF wasn’t tracking with implied throttle position the way it should.  Any of us who have seen scan tool graphs are aware of the fact that MAF and Manifold Absolute Pressure (MAP) generally respond in direct correlation to Throttle Position. 

             Also in this case, the MAF was responding fairly normally up to 2.78 volts (see above - this MAF reading was taken at Wide Open Throttle), but was unable to rise beyond that.  And since the PCM was seeing the MAF respond to TP changes, it stored no code.  It was a beautiful example of what service engineers call an “In-Range Failure.”  My student Jimmie, with his limited experience and understanding of scan tool readings, had no clue that the MAF wasn’t performing as expected. For my part, it took me longer to recognize the problem than it should have. When we disconnected the MAF sensor, the PCM defaulted to a strategy that operated based on its other inputs, and the Mazda accelerated smoothly and powerfully.

MAF Stuff

            MAF anomalies can be extremely hard to spot.  In many cases, I’ve personally seen an engine with an MIL light, rich or lean P-codes, and double digit fuel trim readings (either positive or negative) and no other apparent problem, then discovered that a known good MAF was reading just a few measly millivolts above or below the existing part.  For example, the difference between 0.760 millivolts (which we’ll call the normal reading on our hypothetical vehicle) and 0.830 millvolts (the actual reading on our hypothetical vehicle) can cause the PCM to ratchet fuel trim figures deep into the -20 percentage band.  Only a tiny difference on a 5 volt scale, and not enough to cause a noticeable driveability concern, but plenty enough to cause the PCM to store a code and illuminate the MIL.

            Another important input provided by the MAF is Barometric pressure, and on this type sensor, it reads in hertz, like the old Ford MAP sensors.   This sensor likes to rest at about 153 hertz, and anything below or above that is suspect.  A MAF sensor can also respond sluggishly to changes in airflow and the engine will run that way when it does.

            On older GM cars, a white box aftermarket MAF might give acceptable hertz readings, yet still toss an MIL if the PCM doesn’t like the sensor for one reason or another. 

What about those wonderful Asian vehicles? Well, their MAF sensors go bad too, and I had to replace a Japanese-made MAF a few years back that destroyed a dozen of the customer’s Ben Franklin notes when the bill was paid.  (It was a $1200 part).  A MAF sensor for an American car costs a small fraction of that.  We were told at one class I went to a few years back that it would cost more money to buy all the fuel injection and emission replacment parts on a Mitsubishi vehicle than the full sticker price of that same vehicle when it was new.

Conclusion

On our Mazda 626, the MAF range was a little high on the low end, and a little low on the high end, causing the engine to run rough and rich at idle, then to run lean and weak with the throttle and actual airflow above what the MAF was able to indicate.  This explained the rich and lean P-codes in our case. There was no problem with the O2 sensor.

            Debra gave us permission to replace the $175 sensor, and I would have retained the old part except for a $75 core charge.  The 626 drove a whole lot better when we were done.