10. Commissioning and tuning step by step

This chapter takes a new map, with the pump connected for the first time, to a stable tune.

Follow the order: each step depends on the previous one. Steps 4 and 5 already use provisional values, which are refined later: a provisional target in the step 6 table and the initial gains from the table Starting values by number of lobes (field experience), at the end of this chapter.

Each step says what to do on screen, what to look at in the log and when it has passed.

"Pass criteria" describe what you should see.

"Field experience" values are field-tuning starting points to validate on your engine, never factory defaults.

The screens are in Other Functions > High pressure fuel pump (HPFP) (tablet and panels: Functions card over Other), on the Adjust and Settings tabs.

The page names below are the PC ones; on the phone, the tablet and the panels the same fields are on other pages, without the group titles.

Before step 1, leave the pump channels recording and visible in the datalogger (on the FT470 with a new map, check Record): see Reading the log.

#Step 1: Sensor and scale

On screen:

• In Sensors and calibration > DI rail pressure (button Pressure sensor DI #1 or DI #2), check the sensor and the Sensor offset of the DI1 input (Rail #1) or DI2 input (Rail #2).
• In General settings, point the Rail pressure sensor reference to that input: HP Rail #1 (DI1) or HP Rail #2 (DI2).
• The last point of the sensor table is the full scale.
• The signal has to stay between 0.05 and 4.95 V; outside that, the sensor fails.
• Target, guards, minimum pressure and cut-off limit are all based on the DI Rail pressure #1 reading, with no atmospheric correction.

In the log: DI Rail pressure #1 (or DI Rail pressure #2), GDI - Rail 1 sensor fault and Fuel pressure (low side).

Pass criteria: with the key on and the engine stopped, the DI reading is stable at the low-side pressure, the sensor fault stays at 0 and the full scale is the real one.

#Step 2: Protections

On screen:

• In Protection settings, group Protections (guards), set Guard 1: entry-level, Guard 2: latch level and Guard 2: release level (hysteresis) below your sensor's full scale, with the release lower than the latch.
• Adjust all three before the first start: the defaults (180, 190 and 184 bar) only act with a sensor whose full scale is above them (except Guard 2 through the condition target plus 20 bar, with the release below full scale).
• Field experience: set Guard 1 8 bar below full scale and the latch 2 bar below full scale. For the release, keep the default difference: 6 bar below the latch.
• 150 bar sensor: Guard 1 at 142, latch at 148 and release at 142 bar. The example only shows the calculation against the sensor's full scale.
• In the group Low rail pressure injector cut-off, leave the Injector cut-off rail pressure limit below the lowest operating pressure and below the lowest target.
• With the cut-off on, the injectors only come back above the limit plus 2 bar (32 bar by default), even with the start lockout already released (see Cranking and protections).
• Leave Enable low-pressure injector cut-off off for now: step 4 measures the operating pressure, and then you set and enable the cut-off.
• In Feeds bank A / B / C, mark only direct injection banks, and never the same bank on both rails.

In the log: DI Rail pressure #1 and GDI - Overpressure cut rail 1.

Pass criteria: the guards are below full scale and, in the following steps, the maximum pressure in the log stays below Guard 1 (field experience: at least 10 bar of margin) and the overpressure cut does not trigger.

#Step 3: Output and polarity

On screen:

• In General settings, assign the output in Output (Rail #1) (Output (Rail #2) only with Dual rail) and choose the Output behavior: Normal or Inverted.
• In PID settings, turn on Enable feedforward (PID as trim) already in this step. The start routine (valve held on) uses no target, gains or G(W) curve: for this test, they can stay at zero.

• The valve is only driven after the VCU synchronizes the engine (with a phase sensor, after confirming the phase).
• Crank the engine with the starter.
• Between 50 rpm and the Crank out RPM (default 400 rpm), the VCU holds the valve on, for at most the Lockout timeout (from start of cranking) (default 5 s), counted from synchronization.

In the log: Engine Sync, Duty Rail #1 - HPFP and DI Rail pressure #1.

Pass criteria: Engine Sync goes to 1, Duty Rail #1 - HPFP stays at 100 during cranking and the DI Rail pressure #1 rises above the low-side pressure during cranking.

The rise is slower with the direct injection injectors injecting. Field experience: about 50 bar per second on a 3-lobe pump at about 125 rpm, and about 6 bar per second with those injectors injecting at low pressure.

If the pressure does not rise (it stays at the low-side pressure):

• Change the Output behavior, write the map, power-cycle the VCU (off, then on) and only then crank again.
• Output (Rail #1), Output behavior, Rail pressure sensor reference and Rails count only take effect after writing the map and restarting the VCU.
• If both polarities fail, see Diagnostics: symptoms and causes.

#Step 4: Synchronization and window

In a new map the target, the G(W) curve, Kp and Ki come zeroed: with them zeroed the duty stays at 0 even with the output, synchronization and window right.

Before running the engine in this step, enter the provisional values (recommended practice); the G(W) curve can stay zeroed until step 7:

• Provisional target: a low value in the step 6 table, which the pump can sustain with the consumption of the condition. Field experience: a single provisional target in steps 4 and 5, for example the idle target value.
• Margin: the maximum pressure in the log stays at least about 10 bar below Guard 1 (field experience, as in step 6).
• Initial gains: Kp and Ki from the table Starting values by number of lobes (field experience), as in step 9.
• With the G(W) zeroed, the pulse comes from the trim (limited by the PID maximum trim), the kick and the pressurization. The step 5 sweep compares the pulse that sustains the target and needs the PID active. In the sensor safe mode there is no trim: with the G(W) zeroed, the pump has no pulse at steady state.

On screen, in General settings, group Rail #1:

• Before touching the angles, confirm the Crank index position with a timing light: correcting it later shifts the pump window by the same value.
• Number of lobes per cycle, Lobe #1 opening angle and Lobe #1 closing angle: crankshaft degrees, from 0 to 720, counted from TDC of cylinder 1 (see The lobe angles).
• The width (closing minus opening) must be above 0 and below 720 ÷ number of lobes (240° with 3 lobes).
• Outside that, the VCU switches the rail off, and the map diagnostic list shows an error.
• Start with the width from the table Starting values by number of lobes (field experience), at the end of this chapter.
• Leave Cam dependence on None.
• Turn it on only with a valid VVT and cam deviation: before that, record Intake cam position #1 (or the one for the pump cam) and check that it reads.

In the log: Engine Sync, GDI - Rail 1 window: effective opening, GDI - Rail 1 window: effective closing and Duty Rail #1 - HPFP.

The effective opening is the start of the real pulse, not the angle entered.

Set up and enable the low pressure cut-off:

• With the cranking pressure and the lowest operating pressure known, adjust the step 2 limit so that the limit plus 2 bar stays below them and below the lowest target.
• Turn on Enable low-pressure injector cut-off on each rail (on the PC: Enable on Rail #1 and Enable on Rail #2).

Pass criteria: Engine Sync is 1, the effective window is different from 0/0, the duty is different from 0 with the engine running, the map diagnostic list shows no pump error, Enable low-pressure injector cut-off is on for each rail and the limit plus 2 bar stays below the cranking pressure, the lowest operating pressure and the lowest target.

#Step 5: Phase

On screen, the same angles:

• Edit with the engine running and the VCU connected: the value takes effect immediately.
• The Lobe #1 closing angle is the phase: it is where the pulse ends (see How the pump delivers fuel).
• Field experience: the sweep-calibrated closing usually lands 9 to 13 crankshaft degrees before the pump lobe peak.
• Recommended practice: on a new engine, start about 10 to 15 crankshaft degrees before that point, when the cam is known; without the data, ask the pump or cam manufacturer for it.
• Sweep (recommended practice): fixed speed, first at idle and then near 2500 rpm.
• Move opening and closing together (moving only the closing also changes the width and the duty), in steps of 10 crankshaft degrees, from −20° to +20° from the starting point, 20 to 30 s per step.
• To retard the window, move the closing first.
• To advance, move the opening first.
• This order avoids the invalid intermediate value, which switches the rail off immediately.
• With 2 rails, each rail has its own angles: repeat the step on Rail #2, starting from the Rail #1 angles plus the engine V angle (field experience in The lobe angles).

In the log: GDI - Commanded pulse rail 1, GDI - PID trim rail 1 and GDI - Pressure error rail 1.

Judge by the pulse in ms and by the trim, not by the duty.

Pass criteria: one step sustains the target with the lowest GDI - Commanded pulse rail 1 in ms and the pressure error is near zero.

Recommended practice: if the best step is at the end of the range, repeat the sweep centered on that step.

Valve delay and cam:

• Repeat the sweep at a higher rpm.
• If the advance needed grows with rpm across the whole range, use Valve delay (phase advance).
• Otherwise the phase is fixed: shift the angles.
• Field experience: 0.33 ms was enough on a 3-lobe pump.
• With Cam dependence on, the effective closing must vary 1 to 1 with Intake/Exhaust cam position ..., with the sign reversed (retarded cam, later window).
• Change one thing per test: phase and dead time never in the same map version.

#Step 6: Target

On screen:

• In General settings, the Fuel pressure target map field chooses the load axis: TPS (%) or MAP (bar).
• On the Adjust tab, fill in the Fuel pressure target map (TPS) or (MAP).
• Columns: RPM; rows: TPS in % or MAP in relative bar (negative vacuum). The cells are edited live; switching the axis requires writing the map (see Pressure target).
• Target is the pressure the pump sustains with the consumption of the condition.
• Never use the pressure left over with the duty at 0: it depends on the conditions of the day.
• Keep the maximum target below Guard 1 with margin (field experience: at least 10 bar between the maximum pressure in the log and the guard).

In the log: Fuel pressure rail target - HPFP (smoothed by the VCU with feed-forward on), DI Rail pressure #1 and GDI - Pressure error rail 1.

Pass criteria: at steady state with a pulse greater than zero, the pressure error is near zero and the maximum pressure in the log keeps the margin below Guard 1.

#Step 7: Feed-forward

Stabilize first: with the step 4 initial gains, the loop must not cycle before populating the G(W) curve (field experience).

On screen, in PID settings:

• Group Feedforward (physical model): Enable feedforward (PID as trim) stays on, since step 3.
• Injector rated flow and Injector rated flow preset reference: if the G(W) curve was built with other injectors, fill in both on the same basis.
• With one of the two flows at 0 the ratio is 1.
• Chart Capture curve G(W): axis W (demand), data in crankshaft degrees.
• Until now the curve can be zeroed: with no model volume, the pulse comes from the kick, pressurization and trim.
• Calculate the W of each point as in Feed-forward (physical model) and trim.
• Where the trim parks, the curve is wrong.
• Correct W point by W point, only with samples of pulse greater than 0 (duty 0 is a zeroed pulse, not noise).
• Extend the W axis to the maximum W in the log: beyond the last point the curve holds the end value and goes flat.

New G (crankshaft degrees) = current G + trim (ms) × 0.006 × rpm of the point

0.006 × rpm is the crankshaft degrees the engine turns per ms.

Positive trim increases G; negative trim decreases it.

Example: a median trim of +0.5 ms at 3000 rpm calls for +9 crankshaft degrees in the curve.

• Do not copy the G(W) from another engine: it belongs to the car, not to the family.
• Start zeroed or modest and let the trim teach.

In the log: GDI - Feedforward rail 1, GDI - PID trim rail 1, GDI - Trim at limit rail 1 and the injection time of the direct injection bank (for example, Injection time B).

Pass criteria: the median trim per rpm range stays within ±0.25 ms (field experience) and the duty does not alternate between 0 and the ceiling.

#Step 8: Fill factor

On screen, the chart Fill factor (RPM): default 1.00 across the whole axis (no effect), range 0.50 to 1.20.

• Change it only if the trim, at points of the same W, shows a trend with rpm.
• Change one rpm point at a time, in small steps, and check in the log whether the trim gets closer to zero.

In the log: GDI - PID trim rail 1 per rpm range.

Pass criteria: the trim has no trend with rpm.

#Step 9: Gains

On screen, in PID settings:

• Proportional gain Kp (RPM) and Integral gain Ki (RPM): 8-point rpm charts (default axis 800 to 7500 rpm).
• Start with the initial gains from step 4.
• A high, flat Kp (2.0 across the whole axis) makes the duty alternate between 0 and the ceiling (field experience).
• Adjust one rpm point at a time.
• Before changing a gain, classify the frequency of the error cycle (see Diagnostics: symptoms and causes).

In the log: GDI - Pressure error rail 1 and GDI - Trim at limit rail 1.

Pass criteria: no sustained error cycle, and GDI - Trim at limit rail 1 does not stay at 1 for long.

Field experience, if you calculate statistics from the log: in 1 s windows, the peak to peak and the fraction of time with error above 15, 25 and 35 bar stay within what the application requires.

#Step 10: Transients and dead time

On screen, in PID settings, group Pump model: Valve base dead time (t0).

• Adjust t0 last, and on its own.
• t0 gives no volume and grows with rpm in degrees: 1.5 ms is 27 crankshaft degrees at 3000 rpm, 54° at 6000 rpm.
• Sign of a high t0: error above the target remains, growing with rpm, and GDI - Commanded pulse rail 1 stays at 0 more than 50% of the time and, otherwise, near t0 (field experience).
• In that case, reduce t0.
• After changing t0, check the trim.
• If the trim has moved away from zero, adjust the G(W) again.

In the log: GDI - Commanded pulse rail 1 per step, GDI - Pressure error rail 1 on acceleration and pedal lift-off.

Pass criteria: no residual error above the target grows with rpm and the pulse stays at 0 less than 50% of the time.

#Starting values by number of lobes (field experience)

Values from field tuning: a starting point to validate on your engine, never a factory default.

All the angles in the table are crankshaft degrees.

The G(W) curve is not included: it belongs to the car.

The closing of lobe #1 comes from the step 5 sweep.

Gains per rpm (800, 1500, 2500, 3500, 4250, 5000, 6000 and 7500 rpm), with feed-forward on, one set only, with no separation by lobes:

• Kp: 0.70, 0.70, 0.55, 0.45, 0.40, 0.36, 0.31 and 0.27.
• Ki: 0.35, 0.35, 0.28, 0.22, 0.20, 0.18, 0.155 and 0.135.

Leave the Fill factor (RPM) at 1.00 and the Leakage coefficient (beta) at 0 until there is a measurement.

#When the tuning does not improve

Field experience: if three attempts do not move the metric, stop and return to the best measured point (log saved).

Before the next change, look at the limit signals to see which ceiling holds the pulse: with the rail below the target and a ceiling at 1, the limit is a configuration one (see Reading the log).

With no ceiling acting, follow the Diagnostics: symptoms and causes table by the symptom.

If it persists, save the log and map and contact FuelTech technical support with the FuelTech Vision and firmware versions.