I have had many conversations with both technicians and trainers about the issue of misfires caused by intake tract deposits on Gasoline Direct Injected engines and what causes the misfire. There are several schools of thought, from compression loss across the intake valve seat from deposits, to intake valves sticking in their guides from carbon, to airflow disruption in the combustion chamber. Whatever is causing the misfire, many technicians have encountered this problem and corrected the misfires through intake tract and valve cleaning.
It is this problem that I was contemplating when a 2013 Chevrolet Captiva showed up at my shop. The customer was another shop that had worked on the vehicle and decided to get a second opinion concerning a misfire on cylinder #2. The shop stated the engine misfired during warm-up and had a noise from the engine. They replaced the ignition coil and spark plug with no improvement and were confused by the symptoms the engine exhibited. This compact SUV has a 2.4-liter GDI, naturally aspirated engine with only 45,000 miles. The most interesting symptom was a sharp popping noise that seemed to come from the intake system and could be clearly heard around the air intake throttle body. The noise accompanied the misfire and made me believe there was a mechanical problem with the engine such as a valve sticking due to intake deposits. I recorded the noise with my cell phone when the engine ran but cannot include it in the article, but trust me it was there and very apparent. The scan tool confirmed the problem and can be seen in Figure 1.
|Figure 1 - Scan tool capture showing greatest misfire counts on cylinder 2|
My diagnostic plan was simple: I wanted to perform a running vacuum test with a pressure transducer in the intake manifold and a transducer in the exhaust to confirm a sticking valve on cylinder #2. I would then sell the shop on pulling the intake and cleaning the valves, sounded simple to me anyway. The results were not what I expected. With the scope set on a slow timebase I ran the engine at idle and waited to see a disturbance in the pattern. Both a running vacuum waveform and a tailpipe pressure waveform should be a series of similar pulses with four pulses for each 4-stroke cycle. With my Pico scope connected to the cylinder #1 ignition coil trigger signal, the Pico WPS500 pressure transducer connected to the intake manifold and a Sen-X Technologies 1st Look transducer in the tailpipe I should see a pattern like the one in Figure 2 when no misfire is present. I will note here that there is software filtering applied to both pressure transducer channels to make the pattern easier to view.
|Figure 2 - Pico scope pattern with engine idling and no misfire present. The middle waveform is intake vacuum and the top waveform is exhaust pressure pulses.|
Soon I began to see irregularities in the exhaust pattern as seen in Figure 3. The arrows indicate where the pattern will be zoomed into for a closer look in Figure 4. While the exhaust pattern indicates a misfire, there is no upward pulse in the vacuum pattern that would indicate pressure pushing back into the intake manifold if the intake valve stuck open or the valve leaked across its seat.
While my initial theory does not seem to hold water at this point, the noise from the engine was driving my diagnostics toward a mechanical problem. I continued to perform many more engine mechanical tests including cranking current and vacuum waveforms and some in-cylinder tests on cylinders #2 and #3. One cranking test is seen in Figures 5 and 6, both the whole test and a zoomed in portion. After careful analysis of many waveform captures, I did not uncover a single problem on any waveforms other than the misfire indication in the exhaust waveforms. At this point all I can say for sure is that the engine appears to be mechanically sound.
|Figure 5 - This 15 second cranking current and vacuum test is textbook perfect. No problems are present.|
|Figure 6 - This zoomed in view shows very consistent vacuum pulls and compression peaks in each waveform.|
After deciding the problem is not mechanical in nature, I began to shift my focus to other potential problems. The scan tool can perform two different fuel injector tests, an automated pressure balance test and a user-controlled injector kill test. The automated test actually increases each injectors on-time and the ECM measures the rail pressure change with the rail pressure sensor. The engine will misfire during this test due to an over-rich mixture in the cylinder. This test produced consistent normal results as seen in Figure 7.
|Figure 7 - This is a screenshot from a Snap-on Ethos of the automated injector balance test. The displayed pressure drops are very even. I have not seen printed specs for pressure variation for this test but over 2PSI would be suspect in my opinion.|
After seeing very even injector pressure drops, I began to wonder if the spray pattern from these GDI injectors could be a problem. Using the scope and scan tool I decided to scope upstream oxygen sensor operation while I used the scan tool to shut off each injector. I expected each time I turned off one injector to see a flat-line on the oxygen sensor voltage. The actual test results were not quite what I expected to see. The scope capture is seen in Figure 8. As the callouts show, when the injectors for cylinders #1 and #4 are turned off the oxygen sensor voltage flatlines low confirming no fuel was delivered to the cylinder. But when cylinder #2 injector is off, there is greater voltage from the oxygen sensor and cylinder #3 is even higher. I was not sure how this could be and wondered if there was some leakage from the injectors or if there is residual fuel reaching the intake manifold from possibly the purge control solenoid. I decided to tell the shop I would like to replace all four injectors, and they said it was OK to do so.
Because a complete set of injectors were not in stock at any of my suppliers, there would be a delay in proceeding with the repair. This gave me time to ponder the tests I had done so far and look at each test in deeper detail. While I looked at this scope capture something unusual was noticed. This engine uses a conventional oxygen sensor, not an air/fuel sensor, and while the range of sensor voltage output is normal at about 900 millivolts, the sensor voltage is switching between 2.1 to 3.1 volts, which is a large offset from ground. I was not aware if GM was supplying a bias voltage to the oxygen sensor signal, but after doing some research after the repair was completed, I found out that the oxygen sensor signal is offset from ground on this computer by 1 volt. I will pay more attention to this in the future. When I looked back at some of my earlier captures, I also noticed the #1 coil trigger signal was offset from ground on my scope.
|Figure 8 - Scope capture of oxygen sensor voltage while turning off each injector. The scope time-base is very slow, 10 seconds per division. The green pattern is the #2 injector control side voltage so you can see when that injector was turned off.|
I decided to scope a computer controlled, 12-volt solenoid to see if there was a voltage drop on the ground side of the computer. I scoped the canister purge solenoid and #1 coil trigger and saw both signals were offset 600 – 700 millivolts off ground as seen in Figure 9. There has to be a ground problem on this vehicle!
|Figure 9 - This scope capture of the #1 coil trigger signal and charcoal canister purge solenoid show the offset of the ground signal measured with scope cursors. The canister purge solenoid ground level is 700 millivolts above battery ground.|
After consulting a wiring diagram, I saw the engine computer was grounded through both the X2 and X3 connectors with a black/white wire at terminal 73 in each connector, terminating at ground location G109. Both Mitchell and ALLDATA service information showed G109 located at the rear of the cylinder head towards the driver’s side. After cleaning this ground another test was done but the ground offset remained. I realized the ground wires at the terminal I cleaned did not have a white trace and knew there must be another ground location. After looking the vehicle up on the General Motors service information website, I found the correct G109 location. The ground I had cleaned was G112, ground G109 is on the front of the block behind the A/C compressor. Figure 10 shows the two ECM connectors, and Figure 11 shows the GM service information ground location illustration.
|Figure 10 - ECM connectors X2 and X3 with the ground wires identified.|
|Figure 11 - Illustration from GMSI showing correct location of G109 ground. The callout in the upper right shows G112 which was labeled G109 in Mitchell and ALLDATA.|
When I accessed the G109 ground, I noticed first of all that the bolt was not very tight and also some sort of shrink wrap sealer had oozed out onto the ring terminal. Figure 12 shows what the ground looked like before cleaning.
|Figure 12 - The sealer for the wire shrink wrap can be seen on the ring terminal of G109. This was cleaned off and the bolt tightened securely.|
After cleaning and tightening the ground wire, the scope test was repeated and of course the ground offset was gone. Figure 13 shows the ground levels for both cylinder #1 and #2 coil trigger signals and the ECM ground wire terminal 73 on the X3 connector. All ground levels are below 50 millivolts. The more amazing fact is the misfire and popping noise are gone as well!
|Figure 13 - This scope capture shows all three waveforms have under 50 millivolt ground offsets.|
I decided to repeat the injector kill test while monitoring the oxygen sensor voltage and found the results were quite different from the first time. This time the oxygen sensor flat-lined each time the injector was turned off and the oxygen sensor swung between 1-2 volts.
|Figure 14 - Scope capture showing oxygen sensor during injector shut-off. The green trace is injector #1 control signal. Injectors were turned off in order, 1, 2, 3 then 4. The purple trace in the background is coil #2 trigger signal.|
The vehicle is fixed and not a single part was replaced. The order I placed for the injectors was cancelled and the shop informed that the Captiva was repaired. There were some other items I noticed after the repair that were now looking more normal. I had noticed that the load PID on the scan tool was about 47 percent at idle before the ground repair and the load now showed 23 percent at idle. I had graphed spark timing during the initial scan test and the timing moved between 1 degree BTDC to 6 degrees ATDC, which I thought odd at the time. The timing now stayed around 11 degrees BTDC. The ground problem clearly seemed to cause the ECM to incorrectly control or calculate several functions for engine control. Although I am not totally clear on what caused the engine to misfire and why the popping sound occurred, I do know for sure the Captiva is fixed. This vehicle was a really important lesson on just how important good power and ground circuits are to late-model engine control systems. This case study also reinforced the fact to me that I cannot make a car have a problem that I want it to have, you must trust your tests and follow a logical diagnostic process to flush out the “weird” problems. Keep an open mind and try not to fall into the rabbit hole during a tough diagnosis.