Most Stage 2 overboost cuts on MED17 are blamed on the turbo. Most of the time it is not the turbo at all. The monitors that watch boost become the limiting factor long before the hardware does, and unless you know where to look, you will keep raising the same threshold that was never really the problem. This case walks through a Mercedes GLA 180 where our first two Stage 2 attempts both cut throttle under load, what the log actually showed, and what it took to clear it on the third try.
The Car
The car is a Mercedes GLA 180 running the M270 DE16 AL, a 1.6 litre turbo with direct injection and 10.3:1 compression. Stock output is 122 hp and 200 Nm with around 1.1 bar relative peak boost. The ECU is a Bosch MED17.7.2 with project code 3472T. If you want to see what is on the hardware side of the Mercedes lineup, the GLA family page lists the ECU variants we have mapped for this platform.
The interesting part of this car is the turbo itself. The IHI AL0069 is the same unit fitted to the GLA 250 and CLA 250, where it makes 211 hp from the 2.0 litre. Same compressor, same housing, same actuator. On the 1.6 it is oversized, which is why realistic Stage 2 on this platform sits comfortably between 180 and 195 hp with 1.30 to 1.40 bar relative. Stage 1 files we have seen on this engine settle around 175 hp and 310 Nm at 1.25 to 1.30 bar relative. The ceiling is nowhere near the turbo. The ceiling is in the software.
The Symptom
Our first two Stage 2 attempts on this car produced the same fault. Under load from low RPM the ECU would cut throttle hard, the car would drop to reduced power for 30 to 60 seconds, and it would only recover fully after a key cycle. No limp lamp, no DTC stored, just a throttle that closed itself while the driver was still asking for power. Each rework after the first simply raised the absolute overboost threshold in the hope that would be enough, and each rework after the first cut again.
So we put the car back on the road with a Vediamo logger running and reproduced the cut on demand. That log is what told us this was not an overboost problem in the way everyone assumes it is.
What Was Logged
We pulled a Vediamo datalog at 50 ms resolution while reproducing the fault, covering 135 seconds and just over eleven thousand samples. Eight channels were recorded and grouped into five panels so the relationships between them would be obvious at a glance. This is the layout we use on every MED17 diagnosis, because the story of the cut only makes sense when you can see pedal, throttle, boost, load, and wastegate together on one time axis.
Vediamo Datalog · 8 channels · 50 ms resolution · 11,125 samples
Panel 1 · Engine Speed DT_Motordrehzahl · 1/min
Panel 2 · Boost & MAP Pressure DT_Ladedruck · hPa
DT_Saugrohrdruck · hPa
Reference lines: 1100 hPa stock · 2200 hPa Stage 1 · 2500 hPa monitor wall
Panel 3 · Wastegate & Engine Load DT_Wastegate · %
DT_Motorlast · %
Panel 4 · Driver Input DT_Drosselklappenwinkel · deg
DT_Pedalwertgeber · %
Panel 5 · Lambda (AFR Correction) DT_Lambdaregler_Abweichung_Mittelwert
Reference line: 1.00 stoichiometric
Everything the diagnosis depends on sits in those five panels. Pedal tells you what the driver is asking for. Throttle angle tells you what the ECU is willing to give. When those two separate, the ECU is overriding the driver, and the reason is almost always a monitor that has tripped somewhere else in the calibration.
The Failure Event
The cut happens fast. Between the moment pedal load stabilises and the moment the throttle snaps closed, about one and a half seconds pass.
t (s) | RPM | Pedal | Throttle | Load | WG duty | Boost (hPa) | MAP (hPa) | |
|---|---|---|---|---|---|---|---|---|
46.0 | 1568 | 46.0% | 78.8° | 85.8% | 90.3% | 1165 | 1187 | |
46.5 | 1830 | 51.1% | 78.8° | 93.5% | 90.3% | 1268 | 1281 | |
47.0 | 2035 | 55.2% | 78.8° | 111.5% | 59.3% | 1489 | 1506 | |
47.5 | 2344 | 54.6% | 78.8° | 151.7% | 50.7% | 1881 | 2001 | |
48.0 | 2784 | 54.6% | 78.3° | 191.4% | 34.8% | 2481 | 2465 | |
48.3 | 2974 | 62.6% | 20.6° | 108.6% | 9.0% | 2275 | 2184 | |
48.5 | 2928 | 69.3% | 9.6° | 57.7% | 9.0% | 1360 | 876 |
Peak boost reached 2481 hPa, which works out to roughly 1.48 bar relative. Engine load hit 191.4 percent, pinned against the sensor ceiling. At t=48.3 the throttle collapses from 78 degrees to 20 degrees while the pedal is still held at 62.6 percent. That is the ECU shutting itself down, not the driver lifting.
Here is what made the first two attempts so frustrating. The absolute overboost threshold in the failing file sat at 2500 hPa. The measured peak was 2481 hPa, only nineteen hPa under it. That gap is small enough that the absolute path could plausibly have tripped on its own, either on a sample our 50 ms log missed or on sensor noise sitting right against the threshold. But it was also small enough that we knew raising the absolute number further on its own was not going to give us a clean fix. Something else had to be in play alongside it, and the deviation path was the obvious place to look next.
What MED17 Is Really Checking
MED17.7.2 does not rely on a single overboost check. Instead, it runs two checks in parallel, both of which trigger the same fill clamp flag if either one is activated. The first path is the more familiar one: it compares actual boost to a fixed pressure threshold. If this threshold is exceeded, a fault is latched and engine power is reduced.
The second path is less obvious but often more restrictive. It continuously compares actual boost to the commanded boost and triggers when the difference exceeds a defined tolerance for a specified debounce period.
The tolerance on that second check is not a single number. It sits in a table whose allowed deviation shrinks as the boost setpoint moves further above base boost. At stock Stage 1 targets the tolerance is generous enough that normal spool overshoot passes through without complaint. Push the setpoint into Stage 2 territory and leave the tolerance table alone, and ordinary turbo behaviour becomes enough to trip the monitor on its own. There is no warning, no DTC, no rich mixture, nothing visual in the log except the throttle closing.
When either path fires, the ECU clamps maximum cylinder fill. The only way for the physical engine to respect that clamp is for the throttle to close until filling matches, which is exactly what the log shows at t=48.3. And because the clamp is latched, it stays active until the engine is switched off. That is why the customer reported a long reduced power episode that only cleared on a key cycle. The engine was not recovering. The fault was sitting there, holding the clamp, waiting to be cleared.
The earlier attempts had raised the absolute threshold and left the deviation table at stock values. That combination left the file balanced on the edge of both paths at once. The logged peak was nineteen hPa away from the absolute wall. Based on how much the actual boost was overshooting commanded boost in that RPM band, it was also sitting directly on the deviation wall. One or the other was firing. From the outside they look identical, and that is why the fault kept coming back no matter how high the absolute threshold was pushed.
What We Changed
The first move was lowering the boost target itself. Our earlier reworks had been asking the turbo for too much pressure, and the way a turbo responds to a sudden pedal demand is to overshoot the target before settling back down to it. That overshoot was a big part of what was tripping the alarm. We pulled the target back from roughly 1.45 bar to between 1.30 and 1.35 bar relative. The car loses almost nothing at the wheels because the power curve on this turbo is already nearly flat between those two pressures, but the lower target means a smaller overshoot, and a smaller overshoot gives the alarm room to breathe. Most tuners do not want to take this step because on paper it looks like going backwards, and on a customer car it is usually the step that actually makes the file work.
From there the real work was on the deviation alarm itself. Inside the ECU sits a table that decides how much extra boost the engine is allowed to produce above what the ECU asked for, and only past that does the ECU count it as a fault. We widened that allowance across the rev and load range where a Stage 2 file actually spends its time. This was the change every earlier attempt had missed, and without it no amount of raising the absolute number is ever enough to make the cuts go away. We also gave the alarm a slightly longer fuse, so a brief pressure spike during spool can come and go without the ECU treating it as a real fault and latching anything.
The torque side of the calibration had to move alongside the boost side. The ECU does not only watch boost. It also watches how much torque the engine is producing, in absolute numbers and gear by gear, and it cross checks the engine output against the torque the throttle position says it should be making. All of those numbers were calibrated for a stock 122 hp engine, and they have no idea a Stage 2 file is now sitting on top of them. If you push more boost in but leave the torque limits stock, the torque side of the ECU fires its own protection and closes the throttle in exactly the same way the boost side does, and from the driver seat you cannot tell which one tripped. This is one of the quieter reasons aggressive files feel nervous on the road, and it is one of the first things we touch on any rework.
On top of all that, we did raise the absolute boost ceiling too, with proper margin above the new peak instead of the token bumps the earlier attempts had used. We softened how hard the ECU clamps the throttle if this protection ever does fire for a real reason, so a genuine fault leaves the car drivable home instead of stuck at almost no power. The knock detection settings were realigned to the new load points so the ECU is still listening for real knock in the right places at Stage 2 load. And the VIN, immobiliser and security regions were verified byte for byte against the stock file, so there is zero chance of a non start or a coding mismatch after flashing.
Result
File flashed on car and tested back to back with the old one. No throttle cuts, no stored codes, no reduced power events across a long mixed road test through the same load points that used to trigger the cut.
Not every Stage 2 fault is a hardware limit. On MED17 specifically, the deviation path is the one that trips most Stage 2 reworks that look like overboost but are not, and it stays invisible until you know what to look for. If you are running a file that cuts under load and the peak boost is sitting close to but not clearly over the usual threshold, this is the first place to look. If you want us to handle the rework, the tuning services page lists what is available for the M270 family, or you can check the GLA page directly to see whether your exact ECU variant is in our database.