Last article in this series we got the car on the road and tuned the fuel, ignition, turbo boost and EGR. Now it’s time to get closed loop fuelling working, improve the acceleration enrichment and chase-down a full-throttle flat spot.
As described earlier, I chose to initially tune the car out of closed loop. That is, tuning was done using an external air/fuel ratio meter (a MoTeC PLM) and manually setting injector pulse-widths to give the required mixtures at different loads and engine rpm.
Taking this approach meant that the car wasn’t trying to correct mixtures at the same time as they were being set, and also allowed me to have a better ‘handle’ on what was going on at any one time.
But with the air/fuel ratios largely correct (within around 0.8 of an air/fuel ratio) across the normal driving range, it was time to fit the wide-band oxygen sensor and activate closed loop control.
A Bosch LSU4 sensor was fitted and wired to the ECU. It was calibrated in free air and then the car was started. The measured mixture values were close to that previously shown by the PLM – ‘close’ but not identical, since the PLM sensor has never been free-air calibrated.
A Lambda table was then constructed to show desired Lambda values at all engine loads and speeds. This table can be thought of as the ‘aim point’ – the values don’t have to be at Lambda 1 (ie stoichiometric) but instead can be richer at higher loads and leaner at low loads. Because a wide-band sensor is being used to provide feedback, the car will then maintain these air/fuel ratios – even if they’re not stoichiometric.
One useful function of the MoTeC software is a ‘Quick Lambda’ function.
Once the Lambda chart has been completed, pressing ‘Q’ on the keyboard tunes the ‘live’ fuel site to the required injector pulse width to give the air/fuel ratio shown by the Lambda chart. The fuel map indicates the sites that have been tuned in this way, allowing the user to differentiate the sites tuned with Quick Lambda from those sites tuned manually.
When using Quick Lambda you need to be careful that the engine isn’t using acceleration enrichment (more on this in a moment) or is in injector over-run shut-off. However, outliers (where the fuel value is completely wrong) show up clearly on the 3D visual – sudden peaks or troughs in the 3D ‘topography’ are indicators of this type of error.
That said, as this chart of the Honda’s fuelling shows, engines do in fact require more variations in fuel through the range than you might first expect.
Quick Lambda was very effective, especially when tuning alone and/or at higher loads, where on the road, things start to happen more quickly. (Quick Lambda was the only tuning that I could easily do without pulling over to operate the laptop – just pressing a single button on the keyboard is easily and safely done while driving.)
With the Quick Lambda completed for most of the fuel sites, the rest could be manually set in accordance with the Quick Lambda values. With this done, the car then used the feedback from the oxy sensor to maintain air/fuel ratios as requested – ie it worked in closed loop.
The Lambda table was set to:
Acceleration enrichment took a long time to get right. There were two reasons for this:
The best approach to setting acceleration enrichment was to use a chart of MAP pressure versus engine speed. The enrichment applies only when the throttle is actually being moved (in contrast to just engine revs increasing) and the values in the chart are activated as the active load site passes through them. That is, during tuning, you need to watch both the speed of accelerator movement and what enrichment values are being passed through on the map during this foot movement.
I found that acceleration enrichment made a dramatic difference to a whole range of driving behaviours.
For example, when re-applying throttle in fifth gear cruising (eg backing off when driving down a hill and then re-applying throttle), acceleration enrichment at very low manifold pressures and relatively low revs was needed if a ‘lean jerk’ wasn’t to occur.
Another example of acceleration enrichment making a major impact on driveability was in the smoothness of gear changes, especially those made when driving the car hard.
Finally, more as you’d expect, acceleration enrichment made a large difference to the responsiveness of the engine. What I’d considered to be lag associated with the turbo spooling-up after a sudden throttle shift was actually more related to a short-term lean condition that made response doughy. In fact, to improve turbo response, increased acceleration enrichment was added at 100 kPa manifold pressure (ie atmospheric pressure) at 3000 and 4000 rpm to bring the turbo back onto boost faster.
In a way rather like the EGR covered earlier in this series, acceleration enrichment (at least on the little Honda) proved to be far more important to driveability than is often discussed.
Getting rid of the flat spot
After the Quick Lambda tune, the car drove very well, with smooth transitions between the differing air/fuel ratios.
However, there was a clear flat spot that occurred at just under 4000 rpm on full boost and at full throttle. The flat spot was particularly noticeable when passing other cars in third gear – it existed in the lower gears, but because the engine spent less time in that spot, it wasn’t as major a problem. (The car’s gearing is such that 4000 rpm is not usually accessed in 4th and 5th gears.)
Close analysis of the fuel and timing charts showed no major changes in numbers at this spot, while logging indicated that boost momentarily dropped at this point. Thinking it might be a boost control issue, I inserted extra sites at this point in the boost map and increased duty cycle. This stopped boost dropping but the flat spot remained.
People on the MoTeC discussion group then suggested it might be interference between the 100Hz sampling speed of the MAP sensor input and the resonant pulsing of the intake manifold. (At 4000 rpm there are 33.3 intake strokes per second, multiplied by 3 cylinders = 100Hz.)
While the ECU has user-adjustable filtering available on all inputs, the suggestion was that I leave electronic filtering at a minimum and instead fit a mechanical filter in the MAP sensor hose. I sourced a specific MAP sensor filter from wrecker and installed it. The problem immediately disappeared.
So does the car now drive perfectly in all conditions?
What the tuning has shown me is that there’s a good reason why mapping a factory car might take a year and 20 engineers. Even after perhaps 40 hours of on-road tuning, I still occasionally meet situations that the car had not previously experienced where the tune could be improved.
However, what can be said is that in the vast majority of situations, the car now drives like a well-mapped standard car – and vastly better than a standard Insight running without its electric assist functional.
Next issue – the MoTeC electronic dash