Boost demand is influenced by the torque model, so the target moves with operating conditions. That makes Boost Target Error the useful measure: how closely actual pressure follows what the ECU is requesting.

A steady target makes diagnosis easier, but the car needs to track changing demand properly too.

How the target is built, and the PID tables as they stood before this work, are in Boost Control on the Emtron KV12: Where the Target Really Comes From. That post describes refining the initial position table by no more than 3 percent at a time. The changes here are much larger, and they are provisional.

What the original logs showed

The earlier logs contained periods where boost remained around 10–25 kPa below target, alongside substantial changes in wastegate duty.

Three features of the original calibration stood out:

• An isolated 10.00 proportional-gain cell, at −25 kPa error with manifold pressure rising at 5 %/second, compared with 0.50 in every other cell at that error.
• Integral gain dropping to zero at ±10 kPa error, limiting its ability to correct a sustained deficit outside the narrow near-target region.
• Derivative output making sizeable contributions to duty during some transients.

Those observations justified investigating the tables, but they didn’t establish a complete replacement calibration. Some logs also had gaps exceeding a second and channels updating at different rates. That limits how confidently individual output changes can be tied to events happening a few hundredths of a second apart.

The first PID revision

The first revision removed the isolated proportional-gain spike, extended modest integral action further from target and reduced the positive-error derivative gains.

These were trial values, not a finished calibration.

Proportional gain

The only change from the original was the 10.00 cell. With it gone, proportional gain no longer depends on how fast pressure is rising.

That made the gain surface consistent. It did not establish that overshoot would improve. That cell was the overboost catch: boost well above target and still climbing was the one situation where the controller was allowed to hit hard. Removing it removes that correction too.

The table is also asymmetric, which was inherited rather than introduced. Gain is 0.20 at −5 and −10 kPa, against 0.50 at +5 and +10 kPa. Proportional correction is weaker when boost is above target than when it is below.

Integral gain

The near-target values stayed unchanged. Adding 0.010 at ±10 kPa allowed integral action to extend beyond the original narrow region, tapering towards zero at ±50 kPa.

The integral limits remained ±5 duty percentage points.

Derivative gain

The gains at +10, +20 and +25 kPa were halved from 3.00 to 1.50. The negative-error end was retained.

The output limits remained 5–80% duty, and the initial-position table was initially left unchanged.

That last decision became important.

Why boost status mattered

Some of the earlier duty changes coincided with changes between Emtron boost status 15: ON–Mass Flow Limit and 1: ON.

Those transitions could change both the final boost target and the initial duty selected from the table. A sharp duty change therefore wasn’t automatically evidence of a PID problem.

The distinction matters on this car because the final boost request is influenced by the torque model. Looking only at the base target, or only at the PID outputs, misses part of the control sequence.

Boost system status and boost PID status are also separate channels. Their numbers cannot be interpreted interchangeably.

Status names are from Emtron’s Boost Status reference.

The sixth-gear log below shows how large those transitions are. Between about 3,000 and 3,100 rpm, the status flipped between 1: ON and 15: ON–Mass Flow Limit four times. Each flip moved the target between 290 kPa and roughly 190–200 kPa, and moved commanded duty between 80% and about 31%. None of that was the PID.

At a target near 190 kPa the initial position also read zero, because the table holds zero at 200 kPa and below. From about 3,100 rpm the status stayed at 15 for the rest of the pull, so the whole overshoot happened with the mass flow limit setting the target.

The run that exposed the overshoot

The next useful log contained a sixth-gear pull from approximately 3,000 to 4,300 rpm, with no shift recorded during the pull.

It showed a clear overshoot:

Positive error means actual pressure is below target. Negative error means it is above target.

This wasn’t simply an apparent overshoot caused by the target suddenly falling. During the initial pressure rise above target, the target itself increased from approximately 261 to 271 kPa absolute.

Actual pressure climbed much faster.

The maximum logged error and maximum manifold pressure occurred at slightly different timestamps. With independently updating channels, those figures should not be forced into an exact same-row calculation.

The initial-duty table explained much of it

At approximately 3,680 rpm, boost was almost on target:

• Boost Target Error: +3.3 kPa
• Initial duty: 75.5%
• Final commanded duty: 74.6%

The engine was already reaching the requested pressure while the controller was still commanding substantial spool duty.

The original initial-position table held 100% duty through 3,500 rpm for target rows from 220 to 340 kPa. It then dropped sharply towards the values at 4,000 rpm.

For the 260 kPa target row, interpolation produced this curve:

The table tapered spool assistance according to engine speed, without directly accounting for how close actual boost had come to target.

In this sixth-gear pull, boost reached target before RPM had advanced far enough through that taper. Feedback then had to remove a large amount of duty while pressure was already rising rapidly.

Between target crossing and the pressure peak, about half a second, the taper itself took initial duty from 75.5% to about 51%, and P removed roughly another 26 points. Pressure kept rising anyway.

That strongly implicated the feedforward transition. It did not prove that every aspect of the overshoot came from that table.

What P, I and D were doing

The individual outputs helped separate the causes.

P was correcting the overshoot. It eventually removed approximately 27 duty percentage points. That correction arrived as negative error developed, but it did not prevent the pressure peak.

It was also starting from its weakest settings. Between −2 and −10 kPa the gain was 0.10 to 0.20, reaching 0.50 only at −25 kPa. This overshoot passed −25 kPa with pressure still rising, which is the region the removed 10.00 cell used to cover. The log can’t show what that cell would have done, but this run went through the conditions it was there for.

I was not winding up and driving the event. Integral output was approximately zero at target crossing and slightly negative at the peak. Increasing integral correction would not address the main cause shown here.

D had a gap around target crossing. The revised table had zero derivative gain from +5 to −40 kPa error, as the original did.

As error approached zero from the positive side, derivative gain tapered away. It then remained zero through the first 40 kPa of overboost, approximately 5.8 psi. Gain only returns beyond −40 kPa, reaching 3.00 at −50. With a worst logged error of −57.6 kPa, derivative gain was available only over the last 18 kPa or so of the overshoot.

The log shows what that did. Above target, derivative output was zero from about −13 kPa to beyond −50 kPa. Its only negative contribution there was −2.1 points, at the peak. As pressure began to fall it switched to +6 and then +12 points, with boost still 30 to 50 kPa over target.

That is derivative doing its job, resisting the rate of change. Here, though, it was adding duty to an engine that was still overboosting, and final commanded duty briefly climbed from 17% back to 46% in that window. On this run, the −50 kPa end of the D table added more duty during recovery than it removed at the peak.

The combination was the problem: high initial duty approaching target, with weak proportional and no derivative correction through target crossing.

The first D revision had retained that gap. The later log showed why it could not be treated as a validated overshoot-control calibration.

Adding resolution where the problem occurred

Seven engine-speed breakpoints were added to the initial-position table: 3,250, 3,600, 3,700, 3,800, 3,900, 4,100 and 4,250 rpm.

The existing 3,500 and 4,000 rpm breakpoints were kept. The 4,000 rpm values stayed as they were. The 3,500 rpm values did not: for the 220–280 kPa rows they came down from 100% to between 50 and 57%.

Adding columns alone changes nothing if the new values simply reproduce the old interpolation. The purpose was to create enough resolution to reshape the transition deliberately.

The revision reduced duty earlier and brought it closer to the existing 4,000 rpm values before the previous target-crossing point. It covers the 220–280 kPa rows from 3,250 to 3,900 rpm.

These are provisional trial values, chosen to reduce the excessive duty carried into target crossing. They were not calculated from measured equilibrium wastegate requirements.

At 3,700 rpm and a 260 kPa target, initial duty falls from approximately 72% to 38%. That is a substantial change, with a possible cost in spool response.

The 80% at 3,250 rpm changes less than it looks. It matches the 80% output limit, so while boost is still below target the commanded duty there is the same as before. The real reduction starts at 3,500 rpm.

The 4,100 and 4,250 rpm columns are there for later shaping. For now they hold interpolated values only.

Checking the entries mattered

The first screenshot of the edited table showed that the 280 kPa row had been missed.

That would have affected operation below 280 kPa too, because the ECU interpolates between target rows. EMtune had filled the new 3,700 rpm cell in that row with 72.5%, the original interpolation rounded to the nearest 0.5. At 270 kPa and 3,700 rpm, that would have produced approximately 55.25% initial duty, rather than the intended 39%.

The row was corrected before considering the revision complete.

A separate limitation remains above it. The 300–340 kPa rows retain the original aggressive spool profile, and the step into them is steepest at 3,500 rpm:

That matters in sixth gear. With the 10 kPa sixth-gear target offset, 290 kPa is the full-demand target in that gear, and this log requested exactly that whenever the status returned to 1: ON. The revised 220–280 kPa region does not establish that the rest of the table is suitable.

Where the calibration stands

The revised initial-position values have been checked in the calibration screenshots. There is not yet a subsequent log demonstrating their effect.

P, I and D are being held unchanged for the first comparison so the effect of the feedforward change can be assessed separately. Three closed-loop issues remain open: the derivative gap from +5 to −40 kPa, the weaker proportional gain above target now that the 10.00 cell has gone, and a −50 kPa derivative gain that adds duty during recovery from overboost.

The next assessment centres on Boost Target Error:

• Peak negative error during boost arrival.
• Time spent above target.
• Recovery after the peak.
• Any additional positive error or slower spool introduced by the reduced duty.
• Behaviour when the torque model changes the target or boost-control state.

The useful result will be a smaller, shorter overshoot without creating an excessive boost deficit elsewhere. Until that is demonstrated, this remains a calibration under development.

The lesson from this stage is specific: before asking PID to correct a boost error, examine the initial duty it is correcting around. In this run, boost was arriving on target with roughly three-quarters duty still commanded. That was the first problem to address.

More to come.

The post Tuning the Supra’s Boost Control: PID Gains Were Only Part of the Problem first appeared on BlownByTwins.

The post Tuning the Supra’s Boost Control: PID Gains Were Only Part of the Problem appeared first on BlownByTwins.

This is how I worked outwards from the symptom until the log told me where to look, and what the arithmetic said once I got there.

Ruling things out

Status: Verified

I expanded the log window out and walked the closed loop items one at a time. First question: was idle control intervening during the downshift? It was not. Verified and set aside.

The next thing I noticed was that air mass was going erratic, and that is worth being precise about, because the air mass model on this car is a blend. Below 100 kPa it runs the throttle mass flow model exclusively and blends towards speed density above that. In the region where this fault appears the air mass estimate is therefore a function of throttle effective area and pressure ratio and nothing else. No speed density term diluting it, no trapped mass term in the estimate.

So erratic air mass in that region means the blade, or the area characterisation, or both. It cannot mean much else.

Pedal throttle demand translation was smooth through the whole event. The driver demand table was not doing this. That left the DBW loop, and the DBW PID was the gold seam.

What the loop was doing

Status: Verified

Three things stand out.

Proportional output is saturating at minus 100 percent repeatedly, and topping out near plus 63. That asymmetry is not the axis autoscaling being unhelpful. It is the loop reaching its closing clamp over and over while never reaching the opening one.

Integral output is sitting around plus 24 percent and climbing slowly across the window. An integrator that keeps accumulating while the proportional term is pinned at its clamp is not doing the job an integrator is for.

The motor command is swinging roughly minus 65 to plus 100 percent while the measured blade position only moves a few percent either side of 15. The blade is inertially filtered. It physically cannot follow a command oscillating at that rate, so the engine never sees the full excursion.

That last point is the one that connects back to the shift. The engine does not see it, but the model does. Torque estimate comes from air mass, air mass in this region comes from the throttle mass flow model, and that model reads blade position directly. Every percent of blade wobble goes into the torque estimate whether or not the engine produced any of it.

The gearbox was reacting to torque noise that did not exist at the crankshaft.

The arithmetic

Status: Verified

This is where the log stops being suggestive and starts being conclusive. At the cursor, 38.138s in the screenshot above:

• Servo Position Main: 3.7 percent
• Proportional output: 0.9 percent
• Integral output: 24.3 percent
• Derivative output: 0.0 percent
• Motor command: 6.2 percent

My feed-forward table runs minus 30 at 0 percent and minus 15 at 5 percent. Interpolated linearly, feed-forward at 3.7 percent is minus 18.9.

Two things fall out of a match that close. Motor command is the summed controller output rather than a measured position, and feed-forward interpolates linearly between breakpoints.

But the finding is the third thing. At a near steady 3.7 percent blade position, the integral had wound to plus 24.3 for no purpose other than to cancel a feed-forward of minus 18.9.

The feed-forward in the 0 to 5 percent band is asking for roughly 25 points of closing duty that the throttle body does not need. The integral is spending its entire authority undoing it. That is a large stored state sitting permanently in the loop, and every time the proportional term saturates and the target moves, that state has to unwind before anything else can settle.

It is a far better explanation for a sustained oscillation than any single gain cell.

The feed-forward step, and what it actually is

Status: Provisional

There is a 25 point change in motor duty for a 1 percent movement in target between 5 and 6 percent, and the log crosses that region immediately before one of the instabilities. My first instinct was that this was simply wrong. It is not, and this is the part most likely to be copied badly, so it is worth saying plainly.

The return spring on a DBW body pulls the blade towards its limp home position from both directions. Below limp home you need negative duty to hold the blade down against the spring. Above it you need positive duty to hold it open. A feed-forward table that changes sign between two cells is not a tuning error. It is the table telling you where limp home sits.

Delete the step and you have not removed the discontinuity. You have removed the compensation for it, and handed the spring preload to the integrator.

The real problems are different. First, the magnitude of the low band values, which the arithmetic above says are too negative. Second, a 25 point change interpolated across 1 percent of travel behaves like a gain of 25 through a region the blade is oscillating in.

The proportional gain table

Status: Provisional

The low position P gains are aggressive, and lopsided towards the negative error side.

Reading negative error as blade above target, the closing correction gets close to twice the gain available for an equivalent opening error, at every row.

The spring makes this more interesting than a simple symmetry argument, because the asymmetry it creates flips across limp home. Above limp home the spring already assists closing, so the loop needs more authority to open and less to close. Below limp home it reverses.

A table that favours the closing side at every row is therefore directionally right in the 5 percent row and wrong in the 15 and 25 percent rows. The blade sits at 15 to 17 percent through this event.

One caveat that anyone copying this needs to take seriously. The whole reading above depends on the error channel being target minus actual. If your ECU signs it the other way, the argument inverts and so does the revision. Confirm it on your own log before you touch a cell: find a step where the blade visibly lags a rising target, and read the sign.

What I have not proved yet

I changed two tables at once. Until they are separated, no outcome can be attributed to either.

I have not measured the true limp home position. The right way is to unpower the driver and read the TPS, then check whether the feed-forward breakpoints actually straddle it. If limp home sits at 6.5 percent rather than between 5 and 6, the compensation is being applied in the wrong place, and that alone would explain the instability without a single gain change.

I have not confirmed the error sign convention from a step.

And I left the 0 and 5 percent feed-forward cells alone, which are the two cells the arithmetic actually points at.

I’m not declaring this as “fixed” yet. However, the gear change is smoother, particularly the upshift. The lower speed acceleration is cleaner and everything just feels much more refined. Always remember the first few percent of opening on a throttle blade are extremely non-linear, and even more so on a 82mm throttle body.

The test that closes it

With the loop stable, park the blade at steady positions across 0 to 25 percent, in 2 percent steps through the 0 to 8 percent band, and log feed-forward and integral at each point.

Wherever the integral is non-zero at steady state, the feed-forward is wrong by exactly that amount. That lets the whole low band feed-forward table be built by subtraction rather than by iteration. The position at which feed-forward crosses zero and the integral stays at zero is the true limp home. Any step within the sweep settles the sign convention.

Run it on the same road and the same coast down and you also get a genuinely matched before and after, rather than two logs taken a week apart under different load.

The principle

The data is not transferable, but the principle is.

A feed-forward table and an integrator are solving the same problem from opposite ends. If the integrator is carrying a large standing value at a steady operating point, the feed-forward is wrong by that amount, and you can read the correction straight off the log instead of iterating towards it.

The throttle area table that the whole torque model sits on is covered in Throttle Body Effective Flow Area, and the same feed-forward versus PID argument applied to traction control is in Traction Control on 1,114 whp. Full specification is on the Supra project page.

If you run a DBW loop and have taken a different view on any of this, particularly the limp home reasoning or the gain asymmetry, I would like to hear it.

The post Erratic Shifts and Idle Overfuelling on the 2JZ: Finding a DBW Feed-Forward Fault first appeared on BlownByTwins.

The post Erratic Shifts and Idle Overfuelling on the 2JZ: Finding a DBW Feed-Forward Fault appeared first on BlownByTwins.

Saturday, 27th July 2024, what a long day, but totally worth it. I had to drop the R34 GTR at The Exhibition Centre in Liverpool for The Ultimate Show, then my mate Big Paul (ironic name) came to collect me and take me down to Nottingham to collect the Supra. We arranged a collection time of 4pm, and then I head south to Watford Gap to meet an old friend for a coffee and give the car a run before heading home for 10pm. 220 miles of blissful driving in my new toy. I forgot just how good they are, and how much stronger it pulls above 4000rpm. I can’t wait to start the modification route with SRD Tuning.

When I arrived at the previous owners house, I was immediately greeted with the extensive history, and the receipts for pretty much anything that has been done to the car in the UK ownership. We went through this when I went to view it anyway, but it’s reassuring to see. Although the sight of a certain “respected” tuners name in the file sent a shudder – although it was from the import and not for work. And it was a long time ago. Even though, their name is on the license plate. Suffice to say, a new plate was ordered immediately and is in the post as we speak. Wye aye, I’d not let that place tune my piano, pet. And I don’t have a piano.

Once we’d done the paperwork, and had a look around the old owners new car, it was time to leave. Or was it? No. Battery. He’d had it on trickle charge and forgot to turn the plug on! Mortified, but we got a laugh out of it. Jump leads to the rescue. 5 mins and it started first time. We left it running, shook hands and both agreed it was a pleasure to deal with each other – see, I can be nice!

I head off to Watford Gap to meet Peter, and found a Shell petrol station around 30 miles away. Plenty of distance to put a good charge in it. The fuel was low when I arrived at the Shell, and I totally forgot about the 70 litre tank. That was an expensive, but equally brilliant stop.

The 56000th mile clicked over on this journey and all was going well until the traffic built up just before the Watford Gap services turn off. That gave me 5 mins to play with the stereo. Yep. Definitely upgrading that.

After leaving Watford Gap, leaving the car there for just under 2 hours, chewing the fat, it was time to go home. Two words: very smooth. The car is only a 4 speed auto for now, with 390bhp. But it can delivery it nicely and it makes it a really nice motorway car, albeit quick and relatively thirsty. But it drives well, no noises, very few squeaks and rattles – nothing that some soundproofing won’t cure.

As night fell, the lights came on, I think. The stock bulbs in the Supra are ferociously crap. That’s already been remedied, with LED bulbs on order all around – thanks to Auto Bulbs Direct for relieving me of £295. But, I remember from my last car, just how much better night time driving became. And this will be the same – but it says a lot when I ordered the license plates first!

I got home and I was still giddy.

Isn’t she lovely. Side view, no license plate!

The post Mk4 Supra – It’s Home! first appeared on BlownByTwins.

The post Mk4 Supra – It’s Home! appeared first on BlownByTwins.