Tag Archives: Diesel

Log Graphs 1

Specification

Car: 2009 Impreza 2.0 Turbo Diesel, 110 kW / 148 hp / 150 PS, European domestic market, Euro 4 spec.

Important: Newer Boxer Diesel generations (Euro 5/6) may show different behaviour!

ECU firmware: patched ROM for unlimited logging, otherwise stock.
Protocol: SSM2 via CAN
More than 120 items had been logged, plenty of RAM variables (*) plus standard SSM2 items at roughly 170 ms interval.

Graphs

I prepared two graph images plotting some interesting parameters, click picture for full resolution:


Image 1/2

loggraphs1_p1


Image 2/2

loggraphs1_p2

Additional details

… at specific time positions:

#1: State before active regeneration

Time @ ~ 33 ½ min
Cruise Control: active
Vehicle Speed*: 110 km/h
Engine Speed: 2300 rpm
Gear: 5
Coolant Temperature: 93 °C
Injections: 2 (pre + main)
Soot Accumulation Ratio: 64 %
DPF Pressure Difference: ~ 5 kPa
Exhaust Gas Temperature (EGT) Catalyst Inlet: 345 °C
EGT DPF Inlet: 370 °C
Intake Air Amount: 430 mg/cyl
Mass Air Flow: 33.8 g/s
Manifold Absolute Pressure (MAP)*: 126 kPa
Inlet Air Temperature: 25 °C
Manifold Air Temperature: 48 °C
Fuel Temperature: 58 °C
Throttle Opening Angle: 79 deg
EGR Valve Opening Angle: 38 deg
Final Oil Dilution Rate*: -1.9 mg/s (evaporation)
Oil Dilution Amount: 282.0 g (4.6 %)

#2: Soot 65%, preparing for DPF regeneration

@ 33 ¾ min
Soot Accumulation Ratio* reaches 65%, this triggers active regeneration preparations. Note: I am referring to the actual RAM value, diagnostic parameter may indicate 65% earlier due to rounding.
Apparently ECU now does 3 injections (pre + main + after) when power demand is high enough, no post injections yet

#3: Active Regen ON

@ 34:04; = 20 seconds after #2, DPF Regeneration Switch turns ON
EGR valve closes instantly, 0 deg
Manifold Air Temperature dropping
Boost Control opens VGT immediately, from 52 to 25%
Manifold Absolute Pressure* dropping
pilot-injection kicks in, 2 injections (pilot C + main)
post-injections begin to fade in but not active yet

#4: Post-Injections

@ 33 seconds after #2, post-injections A + B become operational (injection amount > 0)
oil dilution rising
EGTs climbing
Manifold Absolute Pressure*: 75 kPa (~ 25 below ambient), varying, stays below ambient most of the time during regen at low power demand
EGT Catalyst Inlet: 340 °C
EGT DPF Inlet: 320 °C
injections: 4 (pilot C + main + post A + post B)

#5: Coasting Fuel Cut-Off During Regen

Example @ 35:12
EGR valve opens instantaneously, 70 deg; EGR behaves rather digitally during regen – either fully closed (0 deg, for max EGT) or max opened (70 deg, less fresh air)
Throttle Opening Angle: rising up to 31 deg
injections: 0
Fuel Consumption*: 0 mm³/s
oil dilution going down slowly due to estimated evaporation
EGT Catalyst Inlet: decreases fast, EGT at DPF inlet (> 600 °C) follows with a delay
Engine Speed: gear change from 5th to 6th to reduce engine braking effect

#6: Idling with Active Regen ON

@ 45:30
Engine Speed: 800 rpm
Post A Injection Amount*: 0
Post B Injection Amount*: ~ 8 mm³/st
Apparently while idling the ECU prefers the 2nd post-injection (“B”), otherwise during driving it’s rather mixed
Final Oil Dilution Rate: ~ 20 mg/s (medium)
Boost Control: 25 % (VGT fully open)
Manifold Absolute Pressure*: 74 kPa
Throttle Opening Angle: 5.5 deg
EGR Valve Opening Angle: 0 deg
Intake Air Amount: 370 mg/cyl
Mass Air Flow: 9.8 g/s

#7: DPF Regeneration OFF

@ 46:15, decision is based on elapsed time from #2, achieved soot level does not matter (!)
DPF Regeneration SW had been ON for 12.2 minutes
all post-injections off
Oil Dilution Amount*: 299.6 g (4.9 %) = + 0.3 % during regeneration
EGR back to normal operation
Boost Control: 65 % instantly (max speed, spooling up turbo)
MAP rising
EGT Catalyst Inlet: 270 °C
EGT DPF Inlet: 490 °C

Injector Codes

Denso Injector DCRI107890 (Denso part# 095000-7890, Subaru part# 16613-AA020) for Euro 4 Subaru Diesel

Introduction

Quote from a DENSO document incl. picture below:

Replacing a Diesel Common Rail Injector:
When replacing a DENSO Diesel Common Rail Injector, marked with a compensation code, it is necessary to register the ID code, printed on the upper part of the injector, with a genuine OEM – or DENSO diagnostic tool, into the electronic control unit (ECU). The injector compensation (ID) code is used to compensate injector production tolerances.
Some vehicles also require Small Injection Quantity Learning.

Denso Injector ID graphic

Subaru’s term for injector compensation or ID code is “injector code“. Dealership Subaru Select Monitor (SSM-III or later) software has menu items like “Injector Code Display” and “Injector Code New Registration (SSM to ECM)

First off, the engine control unit has no way of measuring actual injected fuel amount, such technology would be expensive for these kind of small fluid quantities. Due to high common rail pressure, even tiny production tolerances result in unwanted injection quantity variations. Obviously, each injector must be registered using its correct cylinder number so the ECU can apply individual adjustments when it is calculating injections. Basically, the ECU accomplishes desired (target) injection quantity by adjusting the duration of injector drive signal.

Since programmed injector codes as well as any other important data is being saved into an extra EEPROM chip, there is no risk of data loss having the car battery disconnected.

Consequences of Wrong or Missing Injector Codes

According to DENSO:

  • Knocking noise
  • Unstable idle
  • Wiggling during driving
  • MIL (Check Engine Lamp) on

Cylinder numbers – Quick Reference

As for a quick reminder, looking at the front of the car into engine bay, cylinder numbers are:

towards back
transmission
  3        4
  1        2
car front (radiator etc.)

Getting Injector Codes via QR Code

Although normally this should not be necessary, injector codes can be read from the actual injector parts even when mounted on the engine and inside the engine bay, therefore not easily accessible. Taking a picture using a mirror tool is relatively easy, no need to disassemble any parts:

Injector QR Cyl2 small

While the injector code label itself is hidden by the (white) electrical connector, its QR code is visible by default.
Make sure the QR code on the photo is as sharp as possible and has sufficient resolution. At the car I usually just try to get high quality pictures. Later on my computer I simply select the best pic, then scan QR info straight off the computer screen using a smartphone.

Using an image manipulation program (i.e. GIMP, PhotoShop) in order to improve the QR code area can result in much better QR detection. In my case, this was not needed as the app can also detect inverted QR (light code on dark background).
Android app tested: “Barcode Scanner” from F-Droid repository. Check settings → Invert scan.

As an example, the following pic is the extracted and improved QR code portion from above picture. I used these steps in GIMP: crop, perspective correction, grayscale, invert, brightness & contrast . You should be able to scan this:

Injector QR Cyl2 processedScanning captured QR code results in a line of text containing 49 characters:

7890AA0200809118736B30000000000E9EBEBECF300000045

First 19 chars:

7890AA0200809118736
  • First four chars “7890” match Denso basic part number.
  • Possibly contains production date “2008-09-11” ?

Injector Code Format

Remaining 30 chars is the exact injector code needed for ECU. These chars must be in hexadecimal (0-9, A-F) form as they are transmitted as 15 bytes to and from the ECU:

B30000000000E9EBEBECF300000045
Byte index [0..14] Content
0 const, sort of generation ID, verified by ECU:

  • Euro 4: B3
  • Euro 5: B2
  • Euro 6: B6
1..12 12 payload bytes containing correction data, type int8. Each signed byte must be within range [-125, 125]. In other words, the following five bytes are not allowed: 7E, 7F, 80, 81, 82.
13 const 00
14 simple XOR checksum

Resultant properly formatted injector code (left to right, 4 chars per block):

B300    0000
0000    E9EB
EBEC    F300
0000    45

Notes

Invalid codes will be rejected! Euro 5+ will return NRC 0x31 (request out of range).
Injector code data is not compatible between common rail system generations due to differing correction points! Euro 4 uses fewer points (10) than E5+ (12). Therefore, for example, programming a Euro 5 injector code into Euro 4 ECU will not work and vice versa.

Software

AFAIK, Denso’s own PC diagnostic software has the ability to read from a QR scanner device, Subaru OEM application SSM-III does not – need to type in codes manually.

Protocols needed for injector code display & registration:

  • Euro 4: SSM2 via Serial
  • Euro 5/6: Extended OBD-II

As far as we know, there is no free/open-source software for this yet.

EcuTek tool probably supports injector coding up to Euro 5.

Updates

  • 2017-04: additional notes and details
  • 2016-04-03: Euro 6

Fuel Consumption

Consumption Display Correction

It is the ECU software that estimates fuel consumption as it calculates all individual injections (multiple per stroke) anyway. Based on injection amount demand it will calculate injector drive time, taking conditions like rail pressure and fuel temperature into account. It cannot measure fuel flow directly as this would require expensive equipment – tracking fuel flow + return flow accurately, all temperature compensated.
In reality most drivers notice some trip computer offset vs. actual consumption calculated at gas station fill-up. Usually the displayed value is too optimistic. The offset may depend on driving conditions and fuel quality. Often enough it is roughly constant for an individual driver so it is still useful information.

We offer fuel consumption display correction by patching ECU software.
Simple solution is to derive a constant factor.

E.g. indicated 6.6 [liter/100km] vs. 7.0 actual → correction factor = 7.0 / 6.6 = 106.1 %.
After modifying + reprogramming ECU ROM, fuel consumption CAN value [mm³/s] (ECU output) will be adjusted accordingly, to be picked up by combination meter computer. This one calculates all trip values, finally sends display control commands to center display over serial connection.

Injecting complex ECU subroutine f(torque, temp, rpm, …) and/or table data is also possible, requires measurement data in the first place, though.

DPF Active Regeneration

As you may have noticed, active DPF regeneration has a big impact on instant fuel consumption. All injection parts are included in CAN consumption value, therefore post-injections are noticeable in center display when it is set to current consumption mode.
Collected CAN data showed ~300 % actual fuel consumption at idle when active regeneration is in progress, mainly due to post-injections and intake air throttling in order to heat up the DPF, compared to normal idling with regeneration off.

Idling at 800 rpm, coolant temperature ~90°C both times:

Item Unit Normal Active Regeneration
Injection Amount mm³/st 3.3 12 1
Fuel Consumption mm³/s 88 319
cm³/s 0.088 0.319
liter/h 0.32 1.15

1) Calculated from CAN value, SSM2 item “Final Injection Amount” would indicate 4.0 mm³/st (!). Logged data suggests this SSM2 parameter matches CAN consumption value as long as post-injections are not occurring (DPF active regeneration). To get fuel consumption, best way is to parse the CAN fuel flow value instead.

Formulas

Oil Dilution Graph

Shown graph contains a complete DPF active regeneration process – the rising section (750 seconds duration = 12.5 minutes).
Basically, oil dilution is driven by:

  1. post-injections during active DPF regeneration
  2. estimated diesel fuel evaporation

According to internal algorithm and logged data, normal operation does not increase dilution. Oil dilution slowly decreases as the ECU estimates fuel evaporation out of engine oil. (On Euro 4 at least, this algorithm is also running if ignition is on, engine not running but coolant temperature warm enough).
Active regeneration however uses one or more post-injections (small additional late injections – during exhaust stroke) in order to heat up the DPF, raising oil dilution amount, at much higher rate therefore winning over evaporation.
Notice these short intermittent steps during regeneration – these are caused by coasting – ECU suspending all injections, including post-injections.
Wouldn’t coasting cool down the DPF then by pushing rather cold air through the system? To mitigate this, the software fully opens EGR valve (70 deg). As soon as injections resume, EGR valve is being closed again. Normally, during active regeneneration it is in fully closed position (0 deg) helping to increase exhaust temperature (more oxygen).
If you look carefully, you can spot more evaporation going on after regeneration had finished compared to before it started. This is mainly due to higher engine temperature, having reached normal operating conditions of around 90 °C. For evaporation to get going it needs temperatures beyond 30 °C, the higher the better.
Also take a look at post “Estimated Distance to Oil Change” for additional information.

DPF Management

Some parts of Denso’s complex Diesel Particulate Filter (DPF) algorithms are somewhat known by now. Will update with new results from time to time.
Link to page

RomRaider Logger Status

Kudos to RomRaider for including diesel definitions!
Especially interesting for people driving both petrol and diesel cars, no need to use separate definition files.

Links to RR pages:
Latest app version: RomRaider 0.5.3b RC9
Latest definitions: (2011-04-04) Experimental Logger Definition Updates

The application can only do SSM2 via Serial, no CAN protocol(s) support for Euro5 diesels I’m afraid. TCU support has been added, too, but then again, diesels only have MT.

We’ve spotted some extended diesel parameter errors, easy to fix in XML, though. Still safe to use, won’t hurt ECU, just expect some weird data. Will upstream any corrections to RR source of course…

Diesel ECU Patch v1

Finally, now that we are able to reflash diesel ECUs it makes sense to do some changes – or improvements as we think of it.
Couple of things already implemented and working perfectly:

  • Added SSM2 getter function Engine Load, using standard definition 0x000007, x*100/255 [%] – unused in stock diesel ROMs. As engine load value is being calculated internally anyway, also available as OBDII Mode 1 PID 0x04, it makes sense to provide this in SSM2 protocol, too. Very easy to do actually, just a single function pointer change – 4 bytes.
  • Added SSM2 getter function Gear, standard def 0x00004A, x+1 [-], unused in stock ROMs. Tiny function required because of +1 offset.

Btw, as with all SSM2 functions, the physical layer does not matter – any SSM2 function gets called via both serial (Euro4) and CAN connection. Diesel ECUs typically have SSM2 virtual address space 0x000000-0x00034F. That’s 848 function pointers, plus same amount in SSM2-write vector table. You could hook up a lot non-standard things as there are plenty of unused addresses…

  • Modified SSM2 init capability bits accordingly. That way above additional params do show up automatically on ALL SSM2 capable software (RomRaider, FreeSSM, SSM3, DashDAQ ?, …).
  • DPF light active regeneration flashing mode. Applies to closed-type DPF models only. Stock behavior (steady on and error flashing) is untouched, we managed to append own additional code. New code part checks if any stock DPF light mode is currently active. If not, it checks DPF regeneration flag, and only if true it performs this new flashing mode. It’s particularly useful at low engine speed – warning driver that additional turbo lag can occur. Remember, engine management reduces boost and manifold air pressure – sometimes a lot – when active regen is in progress.
    Update: got very positive feedback on this one. Even on a bright sunny day looking through the windshield, that blinking light on dashboard gets noticed immediatly. Whole thing is only a screen page of easy C++. Kudos to GCC, v4.6.x generated code is awesome – highly efficient.

Right now this is all internal, experimental, Linux-centric stuff. Availability of our services to the public will be discussed. It meant many months of full-time work to get this far, call for open source collaboration (less work for everyone) did not work…