Setting the bar high

Part Three

By Frank Massey |

Published:  22 April, 2021

Before I begin part three I have somewhat of an important admission, right up to the closing paragraph of this instalment; I still don’t know the actual cause of the incorrect fuel pressure during warm up.
I hope that part two showed a methodical approach to data acquisition to determine how the fault occurred and a clear path towards further evaluation. I also need you to accept that a great deal more testing behind the scenes had been carried out, but for the flow and purity of the topic I have cherry-picked the more interesting elements within the logic timeline. In other words, I have not cheated you with the facts, as presented to me.
So, what do we know? GDI fuel pressure is reducing in a predictable, non-random manner under PCM control. We have not yet discussed Lambda feedback. We did monitor this much earlier in our evaluation, but I decided to introduce it within the topic, in a way that is logical, allowing me to explain fully, and in detail, the diagnostic process with component functionality.

It is not possible to accurately diagnose any fault without fully understanding how the system responds to data input. With any fuelling fault evaluation, you must observe request and corrected data in order to understand if the PCM is responding in closed loop or attempting to correct a fault condition. Our PCM is in closed loop but appears to be causing not correcting the fault.
Most sensors in Europe tend to be 5-wire Bosch, the remainder fall into the 4-wire DENSO type. The 5-wire ID is as follows: Grey, NBV; White, pulsed heater ground; Yellow, reference low 2.2v; Black ref high, 2.8v; Red signal milli/amp, voltage.
The early Bosch variant carries a zero current on the signal wire there will also be a voltage transition between 2.2v and 2.8v, if AFR = Lambda 1. An excessive oxygen condition will cause current to go high of zero and oxygen deficiency would cause current to go low of zero (+/- 5ma). Voltage response on red is similar, lean above 2.8v, rich below 2.2v. The two reference voltages, black/yellow do not change.
I am mindful to avoid the rich/lean description as it can lead to incorrect diagnosis especially without noting fuel trim characteristics, air leaks and dribbly injectors for example, as our problem vehicle clearly demonstrates.

Now look in your pocket. I previously mentioned the GDI system storing pressure, unlike common rail diesel. If you rev the engine hard and cut the ignition at peak RPM, you should reach approximately 180 bars. Cycle the ignition back on and observe for any pressure decay. Pressure will hold semi-indefinitely over time.
The later Bosch broadband sensor as fitted to our 1.8 engine is somewhat different in circuit response. Both high and low reference voltages are a little higher, the red signal wire does respond to current in a similar way, however both reference voltages do change in the opposite direction. Sorry if this is confusing, I did warn you.
We need to confirm what the Lambda current is doing on the Audi A3 at the instant of the fuelling anomaly, i.e., when the fuel pressure drops, and more to the point what the PCM is doing about it.
To be sure of our findings, we obtained an identical engine management system fitted to a SEAT LEON FR. Please refer to Fig.1, our first Pico image, which shows current dropping below zero red trace, with both reference voltages rising symmetrically in response to low exhaust oxygen content {rich} From left to right, red trace, initial current at zero, open throttle, load enrichment, overrun fuel off, high exhaust oxygen, repeat test. All normal responses.
Now, please refer to Part two Fig.2 in the March issue, which is conveniently also Fig.2 here. This shows serial data during warm up, taken at a similar time as in the previous data from the faulty Audi A3.
It is obvious the fuel pressure taken from the FR is dramatically higher than the AUDI A3, so what would cause the PCM to adopt a lower GDI pressure? Answer that in the privacy of your own mind. This is the moment that defines the essence of a diagnostic technician. Assess data, do not guess, measurements are essential, prediction is the mother of all mistakes.

Back to the PICO scope, take a close look at the Lambda current at the point the fuel pressure is reduced (please refer to Fig.3). Red trace, lambda current, cursor set at zero = Lambda 1. Blue trace, rail pressure, cursor set at 45bar, which is too low? Green/ black = reference voltages, normal response to current change.  Bank 1 sensor 1, red trace, suddenly outputs a negative current which theoretically represents a rich condition. The PCM obliges by reducing fuel pressure still further from 45bar to 38bar. This is the essence of the problem. The pressure was already too low. Looks like a faulty sensor. However, replacing the sensor had no effect on the fault condition.
So, we went back to look at fuel trim characteristics, when 38 bar pressure was set, the pcm adopted between -25-32% trim.  With a reduction of mean injector quantity from 2.5m/s at 19mg/s to 1.6m/s at 12mg/s. remember the mean fuelling value is taken from two injection events per cycle.
At this point, and based on the absence of any obvious sensor deviation or cross-reference variation, I suggested that cloning the PCM from the LEON FR would confirm or exclude any internal PCM error. My thoughts here were based on the PCM adopting a rich fuelling correction without any input request from a sensor, Lambda error accepted. Diagnostics can be defined by a series of negative results leading to the eventual successful conclusion. So long as it has discipline and a logical process, coding the donor PCM from the LEON FR did not solve the problem. That was a big positive for me. We now know for certain that the error is within the engine fuelling system or an obscure sensor input deviation.

Endgame or Infinity War?
Endgame, we hoped, arrived at ADS Preston, with David G, and me. Earlier interruptions did not help continuity of thought! Today David G and I were given uninterrupted time and access. I suggested we limited the scope observation to lambda current observation only, as this was the critical instant of the fault occurrence. Concentrating on focused blocks of serial data using VCDS.
Fuel trim correction was selected with all the following group data:

  •  All temperature sensors
  •  App and throttle angle
  •  Camshaft angle and MAP values
  •  Mean fuel injection value
  •  High, low fuel pressure actual and corrected.

There were no obvious changes with sensor inputs, except the puzzling rich Lambda response. However, we did note that bk1 Lambda did respond first, in other words appears to be driving the fault. There is a clue here. We then went back to Pico very carefully checking voltage and ground at the PCM, looking for any current surges, voltage spikes or background noise at the instant of Lambda reaction. No errors found.
The fountain of knowledge was running dry, but my determination was growing stronger, although we could not see the wood for the trees. The vehicle must be lean?  Yet the Lambda current indicates rich. It’s right in front of our eyes, yet we cannot see it. So, we cut the red Lambda signal wire at both ends, spliced a direct temporary circuit and started the vehicle. The Lambda current instantly indicated lean, i.e. positive current. This is good, as the PCM trim was previously in rich trim. The fuel trim very quickly corrected to 0% and the vehicle ran clean.
Not daring to claim success David G went for lunch while I continued to monitor near perfect trim and perfect fuel pressures, 5bar priming, 160bar GDI, a figure not previously reached.
My conclusion was that current appeared to be decaying between sensor and PCM, thus driving the fuel trim in an ever-spiralling rich correction.  Yes, mistakes were made in the process and in my opinion due entirely to interruption and workload constraints. The real lesson here is an incredibly old and well understood one, you must separate diagnostics from the general workshop demands and environment, in both space and time.
Remember, I am writing this in real time, and all of this happened yesterday. Since then, despite our efforts the fault has returned. David G has opted to remove the GDI injectors for the second time. If you want to guess how it ends, answers on a postcard please, and see you here next month for the exciting conclusion.

Related Articles


Click here to submit an event


©DFA Media 1999-2021
Terms and Conditions