[nitrogary]

Originally Posted by Sutripta

Exactly my thoughts. But to go from philosophy to engineering, one needs some concrete data. So what is the production spread of these values, how are those compensated for, how do those vary with time/ age/ usage, and how is this drift detected.

Regards
Sutripta

good question. will try to answer. this phenomenon is known as injector drift. due to wear and
hard deposits like soot, tar and gum, the nozzle and needle behave differently.
here's by how much:
i talk about a 2 litre engine. the values i quote are not standard but a comparable to an
engine of this size. you may apply a multiplication factor based on engine size. hence for the
1.3 litre FIAT engine, the values may be 60% of what i write. i dont know for sure. this is just for reference.
at low idle, the injection pressures can be as low as 250 to 500 bar and the needle is also
lifted for a small time. the time for which the injector remains open during this time is
about 400 to 750 us (microseconds) and the injection quantity can vary from 3 mg to 10 mg. at
the end of something like 2 lakh km on the engine, this quantity may reduce by as much as 80% so instead of 5mg, it is sending something like 1mg only. a correction is necessary and is incorporated over the life of the injector in the form of "Zero Fuel Calibration" so i will leave the injector open for something like 900us as compared to the 400 when new.
the percentage variation in injection quantity is a function of rail pressure and the relative
time for which the injector remains open and is pretty much inversely proportional to both.
think of this analogous to your common house water tap. we want to fill a bucket of water. the
there are three variables:
1. percentage of tap opening
2. time for which the tap remains open
3. the location of water reservoir / tank. the higher this is, the faster the water will fill.
people staying in a tall building with an overhead tank will know this. the residents at the
ground floor have water coming out with most force. the ones staying on top floor suffer from
low pressures.
point 1 is the injector lift. point 2 is injector lift duration (in microseconds; a derived
unit can be crank angles) and point 3 is rail/injection pressure. simple hold on... it gets better form here!
mind you, the injection does not happen instantaneously so there is a finite time lag which begins from the ECU sending the signal to the injector opening initiation. say this is 100 us. the injection pressure build up is also not instantaneous. say it is 50 us its called ramp up time. similar 150 us during ramp down as well. thats 300 us of nothing. this value changes over the opeating range
so what i am saying is, the percentage of drift will change over the rev-range and thru the torque map; coming as low as 20% worst case to 80% as discussed.
the drift will be detected by one of the learnings incorporated in the ECU. the ECU knows that for a certain engine operating condition, an 'x' amount of fuel needs to go in. it will also know, that if this amount of fuel doesnt go it, the driver will depress the accelerator pedal further. so there's a mismatch. this is one of the ways to figure it out.
every engine also has a friction map. meaning, the amount of energy lost to friction is available in a table form across the rev range and over time. this amount of energy after a lot of processing sublimates into amount of fuel. so this finds its way into the drift calibration.
i could go on because the ECU has thousands of such maps and corrections. the higher the emission and durability targets, the complex this gets!
good food for thought eh? and imagine getting paid for this!! thats what gave me the jollies

[nitrogary]

Originally Posted by DerAlte

What's a 'tone wheel'?

There is nothing on the cam shaft (too little space) usually; sometimes the distributor shaft in petrol engines is the host to the disc. The 'missing tooth' wheel on the crankshaft is what gives the crankshaft angle. And the missing tooth signifies TDC of cylinder 1, or the main index of 1 rotation. Not sure whether it is the leading edge or the trailing edge. The ECU sees this as an event for further processing.

.

one camshaft usually comes with 4 lobes which are of unequal size of the opposite side of the gearing (gear drive or chain drive).
the other often comes with a vacuum pump fitted to it with a woodruff key.

the missing tooth by iteslf will not tell the TDC is occuring at that time. there is a fixed relation of the missing tooth and the TDC. say 75 degrees BTDC. but what you cant figure out is that whether the TDC is arriving at the end of compression or at the end of exhaust. you can waste spark, but not diesel by firing every time TDC comes.

for this, the cam phase sensor is used. for one TDC, both the valves will be closed and for the other, the exhaust valve will be open. you can realistically have the lobes and pump mounting any which way. both are possbile.

[bzr77k]

Originally Posted by Sutripta

^^^
Yes. Whichever way you look at it (and implement/ engineer it), you have to know where in the 720 deg cycle you are. (Think of wasted injection!). So you HAVE to know the cam position also. So cam has to indexed.

Regards
Sutripta

Thats true... consider a 4-cyl inline, at a certain given point, two cyls will be at TDC and two will be at BDC... to identify the cyl which is in the power stroke, u need an additional cam sensor...

by simple math, a missing tooth can only tell u a specific position (of course within tolerance) in a 360 deg cycle. To detect the position in a 720 deg cycle, you will need an additional position sensor... hence the cam sensor

Originally Posted by nitrogary

one camshaft usually comes with 4 lobes which are of unequal size of the opposite side of the gearing (gear drive or chain drive).
the other often comes with a vacuum pump fitted to it with a woodruff key.

the missing tooth by iteslf will not tell the TDC is occuring at that time. there is a fixed relation of the missing tooth and the TDC. say 75 degrees BTDC. but what you cant figure out is that whether the TDC is arriving at the end of compression or at the end of exhaust. you can waste spark, but not diesel by firing every time TDC comes.

for this, the cam phase sensor is used. for one TDC, both the valves will be closed and for the other, the exhaust valve will be open. you can realistically have the lobes and pump mounting any which way. both are possbile.

u will 'waste' sparks... but does this cause any unwanted fuel burn in the wrong stroke??? just wondering!!

[star_aqua]

Originally Posted by nitrogary

sutripta: the least count of the crankshaft sensor is about 3 degrees at full speed.

star, correct me if i am wrong. i suppose a tone wheel is situated on the cam shaft and not the crank. i am not sure what the crankshaft wheel is called. i used to call it missing tooth wheel. maybe i missed some comments in between where this is mentioned.

Hi Nitro, Tone wheel is actually a generic term. same is used for finding vehicle speed as well. the missing tooth will be exactly 2 numbers for crankshaft tone wheel.

Originally Posted by Sutripta

What is the resolution of the crank position sensor?

6 degree which will be effectively 3 degree for one complete cycle(4 strokes)

Originally Posted by nitrogary

one camshaft usually comes with 4 lobes which are of unequal size of the opposite side of the gearing (gear drive or chain drive).
the other often comes with a vacuum pump fitted to it with a woodruff key.

the missing tooth by iteslf will not tell the TDC is occuring at that time. there is a fixed relation of the missing tooth and the TDC. say 75 degrees BTDC. but what you cant figure out is that whether the TDC is arriving at the end of compression or at the end of exhaust. you can waste spark, but not diesel by firing every time TDC comes.

for this, the cam phase sensor is used. for one TDC, both the valves will be closed and for the other, the exhaust valve will be open. you can realistically have the lobes and pump mounting any which way. both are possbile.

to add more, Missing tooth will actually point to a TDC of a particular cylinder. (reason is for backup mode in case of cam failure) if not exactly, then by a fixed angle depending on the sensor mounting position at the flywheel which varies with engine to engine and OEM to OEM. The angle will be specified during calibration.
The cam sensor is the one which points to the TDC. it generates two triggers per every cylinder. one indicates pilot and other indicates main.

[vina]

Originally Posted by nitrogary

...
at low idle, the injection pressures can be as low as 250 to 500 bar and the needle is also
lifted for a small time. the time for which the injector remains open during this time is
about 400 to 750 us (microseconds) and the injection quantity can vary from 3 mg to 10 mg. at
the end of something like 2 lakh km on the engine, this quantity may reduce by as much as 80% so instead of 5mg, it is sending something like 1mg only. a correction is necessary and is incorporated over the life of the injector in the form of "Zero Fuel Calibration" so i will leave the injector open for something like 900us as compared to the 400 when new.
the percentage variation in injection quantity is a function of rail pressure and the relative
time for which the injector remains open and is pretty much inversely proportional to both.

...

point 1 is the injector lift. point 2 is injector lift duration (in microseconds; a derived
unit can be crank angles) and point 3 is rail/injection pressure. simple hold on... it gets better form here!
mind you, the injection does not happen instantaneously so there is a finite time lag which begins from the ECU sending the signal to the injector opening initiation. say this is 100 us. the injection pressure build up is also not instantaneous. say it is 50 us its called ramp up time. similar 150 us during ramp down as well. thats 300 us of nothing. this value changes over the opeating range
so what i am saying is, the percentage of drift will change over the rev-range and thru the torque map; coming as low as 20% worst case to 80% as discussed.
the drift will be detected by one of the learnings incorporated in the ECU. the ECU knows that for a certain engine operating condition, an 'x' amount of fuel needs to go in. it will also know, that if this amount of fuel doesnt go it, the driver will depress the accelerator pedal further. so there's a mismatch. this is one of the ways to figure it out.
every engine also has a friction map. meaning, the amount of energy lost to friction is available in a table form across the rev range and over time. this amount of energy after a lot of processing sublimates into amount of fuel. so this finds its way into the drift calibration.
i could go on because the ECU has thousands of such maps and corrections. the higher the emission and durability targets, the complex this gets!
good food for thought eh? and imagine getting paid for this!! thats what gave me the jollies

Great post and thanks

What I'm thinking is:

1. How do the ECM remap guys do all this? Nothing will make me believe that they have anywhere close to enough resources (or if they have the resources, they use them) to find out so many parameters and their interrelations for a given engine.
2. Which ones are the components that drift 80% over the lifetime of the engine and whether changing those periodically may help (e.g. injectors every 50k kms). Also if ECU already knows about them then isn'te there a way for the ECU to give recommendations on such maintenance to the user/mechanics.
3. How many continually operating feedback loops are active (not calibration loops that can work from time to time) in the engine? This is a full fledged control system now with parameters being tweaked in real time - e.g. going by what Nitro wrote above, the common rail pressuren the injector timing and aging - all have to be taken into account at the same time to get the amount of fuel injected. Now all of these variables will change with rpm etc. and many will change non-linearly. Which means they will also come from their own maps. Enough to make any sane head go
4. With all of this going on, whether drive-by-wire makes way more sense.

[vina]

Originally Posted by star_aqua

...

6 degree which will be effectively 3 degree for one complete cycle(4 strokes)

So does this mean 360/6 = 60 teeth on the tone wheel? or is it fewer teeth but from the profile of the teeth and further processing more information is extracted?

[star_aqua]

Originally Posted by vina

What I'm thinking is:

1. How do the ECM remap guys do all this? Nothing will make me believe that they have anywhere close to enough resources (or if they have the resources, they use them) to find out so many parameters and their interrelations for a given engine.
2. Which ones are the components that drift 80% over the lifetime of the engine and whether changing those periodically may help (e.g. injectors every 50k kms). Also if ECU already knows about them then isn'te there a way for the ECU to give recommendations on such maintenance to the user/mechanics.
3. How many continually operating feedback loops are active (not calibration loops that can work from time to time) in the engine? This is a full fledged control system now with parameters being tweaked in real time - e.g. going by what Nitro wrote above, the common rail pressuren the injector timing and aging - all have to be taken into account at the same time to get the amount of fuel injected. Now all of these variables will change with rpm etc. and many will change non-linearly. Which means they will also come from their own maps. Enough to make any sane head go
4. With all of this going on, whether drive-by-wire makes way more sense.

1. it will happen stage by stage during development. Not everything is done from the scratch. each person is responsible for one system and its calibration.

2. ECU do recommend the user via SVS lamps(service vehicle soon). drifts are applicable for turbo, air mass, intake manifold pressure, pressure control valves etc..

3. could not understand what you meant.

4. ALL diesel CRDIs are drive by wire and only MPFI with electric throttle flap is drive by wire.

Originally Posted by vina

So does this mean 360/6 = 60 teeth on the tone wheel? or is it fewer teeth but from the profile of the teeth and further processing more information is extracted?

yes its 60 teeth minus two. few OEMs deviate from this.

[vina]

Originally Posted by star_aqua

1. it will happen stage by stage during development. Not everything is done from the scratch. each person is responsible for one system and its calibration.

I'm not sure you understood what I meant (my bad) - what I meant was that how do the after-market ECU remap guys do it. My guess is that they do a trial and error job but there is no way they can ascertain long term reliability or correct for drift etc.

Originally Posted by star_aqua

3. could not understand what you meant.

What I meant was - control system can be open loop or closed loop. Open loop is faster and simpler but has accuracy limitations - in some cases these can be corrected by periodic calibration (like the case for injectors).

closed loop is slower but more accurate (e.g. measure O2 levels via lambda sensors and use that information for fuel injection) - in a way drift due to aging/temperature change etc. is corrected for in real time.

so I was wondering whether the engine is being controlled via a full fledged closed loop control system that also calibrates all of its parameters from time to time.

Originally Posted by star_aqua

yes its 60 teeth minus two. few OEMs deviate from this.

I could figure at least this one out


By the way, is there any modern text on engine control - I have a copy of Heywood's book but nothing modern, and I guess the techniques you guys are mentioning were not possible when Heywood wrote his book.

My level (a reference for you when you suggest a text): I have good enough knowledge of sensors physics, control systems behaviour analysis and design, and electronics and have been teaching students Physics and Maths for years.

[DerAlte]

Originally Posted by Sutripta

... Tone wheel is the generic name given to any encoding ring ...

Aah! Another one of the colloquial terms, like the "commutator" as used by the motor winding guys (the set of copper wedges arranged around a shaft) which I confusingly used somewhere else

Originally Posted by Sutripta

... The 4 stroke cycle is 720 degrees, so just knowing position of crank will not do, position of cam has to be known. ...

Originally Posted by Sutripta

... you never know with the (we are not happy till its complex) Germans. ...

You said it! Somehow we are never satisfied without complexity. Mazaa nahin na aata otherwise. The Germans (and the Japanese) are true believers in the KISS principle.

With the internal relationships of the engine components 'cast in iron' so to say, why wouldn't we be able to keep track of what is the current position of each of the components from one signal? (Assumption: engine rotates in only one direction all the time ; ignore back-fire in petrol engines?) The missing teeth only confirm that the signal is 'plausible', and give a certain positional reference. Beyond the first recognizable reference during starting crank, wouldn't everything appear (to the ECU) as a set of repeating events strung out over time? The only time variance would be RPM dependent.

Most of the timing relationships actually start with data written in configuration (ROM). This is done on paper first, looking at the design of the engine, and then configuring on the test bed. How else would the control guys adapt the same ECU for 4/5/6 cyl engines? If not, one would have to write software all over again for every new engine, and incur - instead of incremental cost - one complete s/w dev cycle cost every time.

Originally Posted by Sutripta

... ?!?! ...

For example, to know whether the engine has started firing during the start-up cranking, one looks for the (angular) acceleration of the crankshaft for each cylinder cycle. Using a conventional encoder is one (rather costly due to cost of the encoder) method. The cheaper (and low cost) method is to look at the changing time relationship of the pulse leading and trailing edges of the pulse train coming from the tone-wheel sensor. A set of continuous 'successes' signifies engine can be switched to normal mode of operation (idling).

Originally Posted by vina

... so I was wondering whether the engine is being controlled via a full fledged closed loop control system that also calibrates all of its parameters from time to time.
...
By the way, is there any modern text on engine control ...

Fully closed-loop control, but the adaptive part as you write is done in very limited context.

And, if someone manages to write a tome on modern engine control, he / she would be violating a lot of NDAs. IP, baba, do you want to bankrupt the companies or what?

[Sutripta]

Originally Posted by star_aqua

6 degree which will be effectively 3 degree for one complete cycle(4 strokes)

to add more, Missing tooth will actually point to a TDC of a particular cylinder. (reason is for backup mode in case of cam failure) if not exactly, then by a fixed angle depending on the sensor mounting position at the flywheel which varies with engine to engine and OEM to OEM. The angle will be specified during calibration.
The cam sensor is the one which points to the TDC. it generates two triggers per every cylinder. one indicates pilot and other indicates main.

Could you explain all this in a bit more detail please. Instead of starting of with a barrage of questions, let me try and understand what you are saying.
The 3 deg (or is it 6?) uncertainty in determining TDC is something I'm finding astonishing.

Is 60 ppr as standard as 48/ 96 for ABS?
For 2 missing tooth, are these non symmetrically placed. (If so, obviates need for the second quadrature channel)

Any progress on decoding the injector code?

Originally Posted by nitrogary

good question. will try to answer. this phenomenon is known as injector drift. due to wear and
hard deposits like soot, tar and gum, the nozzle and needle behave differently.
...
good food for thought eh? and imagine getting paid for this!! thats what gave me the jollies

Thanks for the actual variation figures. What you have described (I mean the cause - effects) is pretty well known. I can add two more to it: fuel temperature (density compensation), and at the pressures involved, modeling treats diesel as compressible. And I'm sure there are tons more (eg combustion chamber temp, which cannot be accurately determined from coolant temp).

If one were to model the injector, what would the model be? And as it ages, would the model remain valid, with just a change of parameters,
or would we have to change the model.

Further questions (if any) after Star's reply.

Regards
Sutripta

[Sutripta]

Originally Posted by DerAlte

You said it! Somehow we are never satisfied without complexity. Mazaa nahin na aata otherwise. The Germans (and the Japanese) are true believers in the KISS principle.
IMHO, not the Germans.

Beyond the first recognizable reference during starting crank, wouldn't everything appear (to the ECU) as a set of repeating events strung out over time? The only time variance would be RPM dependent.
Theoretically yes, but think of the cost of missing a pulse from time to time. You need to synchronise from time to time. Since the sync (TDC) signal is always there, use it every cycle.

Most of the timing relationships actually start with data written in configuration (ROM). This is done on paper first, looking at the design of the engine, and then configuring on the test bed. How else would the control guys adapt the same ECU for 4/5/6 cyl engines? If not, one would have to write software all over again for every new engine, and incur - instead of incremental cost - one complete s/w dev cycle cost every time.
For engines, timebase is almost always expressed in terms of crank degrees, not time.
Programming one of the standalone ECUs (megasquirt?) would be highly educative. Don't know of any such for diesels.

For example, to know whether the engine has started firing during the start-up cranking, one looks for the (angular) acceleration of the crankshaft for each cylinder cycle. Using a conventional encoder is one (rather costly due to cost of the encoder) method. The cheaper (and low cost) method is to look at the changing time relationship of the pulse leading and trailing edges of the pulse train coming from the tone-wheel sensor. A set of continuous 'successes' signifies engine can be switched to normal mode of operation (idling).

Even the incremental (as opposed to absolute) encoder is called an encoder! In fact, shaft encoders invariably mean the incremental encoder. If you mean an absolute encoder, you have to specifically specify it. Check the products of companies like Accucoder and Baluff, to name two out of a zillion.

Regards
Sutripta

[DerAlte]

Originally Posted by Sutripta

IMHO, not the Germans. ...

Oh, they most certainly do. Both the principle and the (lowercase) practice.

Originally Posted by Sutripta

... Theoretically yes, but think of the cost of missing a pulse from time to time. ...

What cost? No cost at all (thinking as an engineer). Adaptation, sir. As a singer, how would you 'adapt' to the percussionist missing a beat? Botch up your song? No, you wouldn't (though it will invariably show up in your expression LOL). It is the same principle that is used there. The 2 missing pulses are a designed event occurring twice in every cylinder cycle, like a programmed pause in music. 2 missing teeth because it is next to impossible for 2 other pulses going missing one after the other. If this happens out of turn, it might cause a small hiccup. If it is constantly happening, it is time to 'limp home' with the Check Engine lamp on. Switching off the engine might be catastrophic (think autobahns).

Originally Posted by Sutripta

... For engines, timebase is almost always expressed in terms of crank degrees, not time. ...

Correct, but for human understanding only sir, not for programming purposes. Like time and frequency domains for signals (the closest analogy), for the engine controls programming world it is the time domain and events.

Originally Posted by Sutripta

... Programming one of the standalone ECUs (megasquirt?) would be highly educative. ...

Megasquirt is a couple of generations behind in algorithms for production vehicles, but may be worth a try (depends on how well the code is commented). The website says "MegaSquirt-II™ adds ignition capabilities (including missing tooth crank wheel support)" so one can imagine. Since there aren't too many DIY diesel afficionados, I haven't heard of such ECUs for diesels either.

Originally Posted by Sutripta

... Even the incremental (as opposed to absolute) encoder is called an encoder! ...

Sir, what I meant was any encoder, whether absolute or incremental, would be pointless since it will add to the cost of controls without giving overriding benefits. The difference is you are expecting mandatory presence of an encoder for knowing c/s angle (c/s angle domain reference), whereas I as a programmer am not (time domain reference).

[tsk1979]

The Safari diesel engine has 2 sensors related to timing
1. Crank Angle sensor
2. Cam sensor

On the early 3.0s, when either of them died, the engine would not start, leading to many stranded owners on the highway. Tata fixed it with a more robust model in later iterations. Therefore the 2.2 has never had the dreaded "Cam sensor/Crank angle sensor failure issue".

I suspect, the ECU programming would not be time driven. It would be more of "interrupt driven" with timekeeping purpose just as a "failure checker". For example, if you interrupts are coming at the wrong time, you could put the engine into shutdown or limp home mode.

[DerAlte]

Oh shoot!!! <running away to find a buyer for the 3.0l DICOR>

You are right, @tsk, the programming IS interrupt (incl. non-interrupt Event / soft interrupts) driven. The allusion to time hear is not about clock-driven real time, it is relationship with the edges arriving from the c/s and cam sensors. The edges arrive at RPM dependent times relative to the previous one - and interrupts are raised by the module that is monitoring the edges. The main software just keeps count of how many edges passed to figure out where the c/s is (angularly). The edge-monitoring module also figures out things like edge or pulse not arrived on time (comparing with short term history, since there are no sudden changes expected one pulse to the next) and makes the other modules handle exceptions as applicable.

Not allowing the engine to start on failure of either sensor is rather drastic. It is possible to allow engine start and limited operation (limp-home; pollution standards can wait) under such circumstances. This is one example of difference in techniques between one injection system vendor and another.

[headers]

Originally Posted by star_aqua

The cam sensor is the one which points to the TDC. it generates two triggers per every cylinder. one indicates pilot and other indicates main.

@Star:
IS there only 1 pilot injection? I heard the famous Quadrajet / MJD / DDiS has 4 injections per cylinder? And thats why the combustion is better than market and it is able to meet BS ?? norms

Originally Posted by Sutripta

Even the incremental (as opposed to absolute) encoder is called an encoder!

@Sutripta:
Sir, What IFF the encoder fails, I've seen many encoders fail over time in a different environment and we have accidents [not in a car of course] Plus they are bloody expensive and very sensitive to dirt!! I doubt any automobile manufacturer uses them!!

Originally Posted by DerAlte

Not allowing the engine to start on failure of either sensor is rather drastic. It is possible to allow engine start and limited operation (limp-home; pollution standards can wait) under such circumstances. This is one example of difference in techniques between one injection system vendor and another.

@DerAlte: You a programmer! - Great

Ok - Now that we have great minds here - let me start with my silly questions :

1. How do the Remap guys tweak the values? What values are tweaked?

2. How does piggy back ECU affect the vehicle - I want to understand the disadvantages.

3. Can we erase the ECM - Chances?