The previous thread on this subject involved questions, discussions and the process of discovery of how the Bosch air flow meter ( AFM ) worked and interacts with electronic control unit ( ECU )

The previous thread terminated here:
https://www.k100-forum.com/t10996p50-flowmeter-restoration#131627

Here I will try to condense it to a more coherent account.


Objective:

Determine whether the Bosch air flow meter ( AFM )  is defective; restore functionality; recalibrate rest position if moved during intervention or if recalibration is desired as a function of age, ie. find method to measure wiper position and re-adjust factory wiper calibration.

General description:

The Bosch AFM is based on a vane obstructing the air flow. The air flow into the inlet manifold pushes the vain against a return spring. Change in vane position displaces a potentiometer ( variable resistance ) and provides a voltage output, proportional to the volumetric air flow, to the ECU. This signal is used to control the volume of fuel injected by each fuel injector.  A bypass screw controls the air flow through a bypass passage when the vane is closed under idle conditions.

There is also a thermistor ( temperature dependent resistor )  inserted into the air flow inlet. This acts as an air temperature sensor and forms part of the resistor chain that thus affects the output of the unit. This is a negative temperature coefficient device ( NTC ) whose resistance decreases with increasing temperature.

Construction detail:

The following image shows what it's all about: track wear due to years of a little cursor rubbing up and down each time the throttle is open or shut. How many million times has this moved in 20 odd years? I'm impressed with how long they last. This is a tribute to the rigorous design and top quality construction that german manufacturers are renown for. These are beautiful, precision instruments.


figure 1. Close up of track wear on Bosch AFM.

Though less clear in the photo,  the dark lines on either side of wear mark tracks are a collection carbon power. This is easily blown off. This case of track wear is pretty typical of these units after 20y, not a case of extreme wear.

Note the usually sharp laser cuts in the green resistor blocks have gone fuzzy. This unit was badly stored. This is unusual degradation since on a K-bike they are well protected and usually look nearly new.

This is what the units look like inside. Comparison of early and later, factory modified, models are shown here:
https://www.k100-forum.com/t11043-bosch-air-flow-meter-version-differences


figure 2. General view of AFM internals.

There's loads of folklore out there about super accurate factory calibrations... don't evah touch this bit ... etc. The one thing that is factory precision calibrated at individual production item level is the value of the thick film resistors ( green rectangles ). These can not be mass produced to required accuracy and are individually laser trimmed to give a precise resistor chain providing a series of accurate voltage points along the black carbon strip where the copper wiper blade runs. This is the reference for the output. This calibration will not be lost by stripping and cleaning operations ( unless one is hamfisted and damages the unit ).

The preload of the multi-turn return spring is set to the nearest tooth during calibration. This is unlikely to require adjustment but should be marked if being removed. The preload, should it be lost is about 1.4 to 1.5 N.cm ; or 55 to 60 gram force applied at the lug behind the pinch bolt in the top left of the photo. It will be hard to start the motor if this is not very near the right position.

Because of the pre-load there is a certain amount of air which will pass through the bypass chamber when the vane is butted up against it's end stop in the closed, engine idle state. There is also typically a small movement of the vane possible before the output voltage changes. This is of the order of 1mm at the extremity of the vane. This movement will be referred to as dead zone in the following. 


The ECU is totally analogue circuitry, not a micro-controller or 'computer' as many sources suggest. The circuit compares the output voltage of the AFM ( which is derived from the battery voltage ) with a fraction of the battery voltage. This gives it a degree of immunity from variations in the battery voltage. The inputs are buffered by integrated circuits ( operational amplifiers ) which draw minimal current. This means that resistance of the carbon track can be considerably higher than the low values of the lasers trimmed, fixed resistor chain, thus not affecting it. The carbon track just acts as in interpolation path between the set points of precision resistor chain. The resistance of the path is not critical since very little current is drawn. This is why testing the resistance of wiper contact externally is not an appropriate test. The contact pressure is intentionally very light to ensure longevity of carbon track. As a result the resistance can be very erratic.

Tests should be done with a voltage applied as under running conditions and the voltage at the wiper measured. Resistance measurements will not give a useful indication of the state of the AFM.


There are 12 resistors in the chain between pin 8 and pin 5 ( ground / chassis connection). Between pin 8 and pin 9 there are two more plus the thermistor temperature sensor.  Battery voltage is applied at pin 9. This means that the thermistor modifies the resistor chain and thus the output, directly. This effectively  applies a correction for changes in air density: it is the mass of air ( ie number of molecules ) which is relevant to the fuel requirement, not air volume.


The multi-turn return spring on the AFM provides a very linear ratio between angular deflection and applied torque. However, air flow through the flow meter  varies approximately exponentially in relation to the vane deflection.  For this reason, the resistor chain is arranged to have a logarithmic relation to position, the idea being that the output voltage will then be the log of the exponential: ie nearly a linear function of the volumetric air flow.

The following graph, using a logarithmic scale on x-axis shows that the output follows a log relationship over a wide range, then deviates somewhat towards fully opened positions. It is likely that gas flow is not truly exponential and this description is an over simplification. The unit will have been designed to return a good linear indicator of air mass flow.






Figure 3. Logarithmic plot of potentiometer output against angular deflection. Data show both an unmodified, used unit and a modified unit under test. Output voltage is normalised and given as a ratio of the input voltage.



The bottom line test for the flow meter is the output voltage measured as a function of the force applied to the vane return spring. A physical set up for applying a measured torque is described below. Figure 4 shows a suspect unit under test and a known functional unit. It is readily seen that this unit will give a very bad signal to the ECU and cause very poor injection regulation. However, it does progress in a roughly monotonic fashion and this rather serious default probably would not be obvious without systematic measurements.


Figure 4. AFM output as a function of torque on return spring. Unit under test shows very erratic behaviour.



Recovering a reliable output signal: 

If the carbon track is worn to the point of producing erratic output as shown in figure 4., a clean track must be recovered. This implies moving the copper wiper arm in relation to  the metal plate carrying the resistor network. There are two principal ways to achieve this: a) move the track; b) move / bend the wiper arm.

Several sources suggest bending ( kinking ) the wiper arm to shorten it in order to find a new track. Others suggest raising or lowering the wiper arm height where it is clamped to the spindle. The major disadvantage of both these methods is that there is no control over the contact pressure of the wiper contacts on the carbon track. This will necessarily mean they end up with too much or too little pressure. It can be assumed that Bosch studied this quite closely in order to design the maximum lifetime whilst ensuring proper reliable electrical contact. Any ad hoc messing around with this arm will either be unreliable or lead to accelerated wear of the new track. Rather counter productive outcomes.

Option a) allows retaining the designed pressure by not distorting the wiper arm. The inconvenience is that it will disturb the calibration of the output in relation to the vane position. This can have a notable affect on mixture across the full range and needs to be correctly reset afterwards.


Moving the resistor plate.

Before removing the plate, it would be desirable to set up the recalibration procedure below and get a good calibration curve as a reference. This will save a lot of testing and tuning later. Also a score to mark the position of black plastic base of the wiper arm before removing the spot of thermal glue locking the retaining screw. Once that is removed you are committed to spending some time recalibrating its position. 

Also note the height of the wiper mechanism on the 1/4" spindle. Usually the end of the spindle will be flush with top of the black plastic heel of the wiper. This should go back in the correct position to retain the correct contact pressure on the wiper arm. This is important. The resistor plate can be removed with this in place but it does risk bending the wiper. It is safest to remove the whole wiper assembly by slackening the pinch bolt at the back. The plate may still nip onto the shaft and need easing off.

Remove the electrical connector block from the casting, then the wiper assy, then the resistor plate.


The resistors network is set on a ceramic substrate bonded to a solid metal plate. This plate is attached to the AFM body by three screws. This plate can be removed by first removing the electrical connector plate entering the AFM. This is retained by four screws and is also sealed with silicone. The mounting holes in the metal plate can be slightly elongated to allow positioning the plate closer to the body of the AFM by moving it slightly towards the connector plug. 0.5 to 1.0 mm is enough to get a clean track of virgin material.

Care needs to be taken not to damage the carbon film resistors or the carbon track while working on the holes. Both are rather fragile to knocks and scratched. Extreme caution is needed.



Once the plate is back in place and locked into a position giving a clean track for the wiper, it is necessary to reset the starting position of the wiper in relation to the resistor network. This can be initially done by aligning the contact points radially with the corner of the first green rectangular resistor. It will already be close. This should be close enough to start the engine, though unlikely to run well. The limit of the old wiper wear marks on the carbon track will be a guide.




Calibration test procedure:


The basic method is to progessively apply a known force to a radial point on the wiper assembly and measure the voltage at the wiper connector ( pin 7 ) whilst applying a fix voltage ( 6.0V) on pin 8. This bypasses the variable thermistor air temperature sensor which should be excluded from this test. Pin5 is ground.



Figure 5. showing electrical connections, load attachment and clamping arrangement of the AFM under test.

A single G-clamp can be used to enable rotation of the unit as load increases to ensure that the radius from the axis to the point of attachment remains perpendicular to the line of tension in the thread. This needs to be adjusted about every four readings as the vane opens. Taking two readings at the same load, before and after movement of the unit, provides a check on how much error this is introducing. This can be seen as vertically superimposed points in the derived graphs. Care should be taken not crush the aluminium casting by exerting undue force with the clamp.

A known load can be applied using a measured amount of water ( eg.  adding 2.5 ml at each step ). A recipient weighing less than the load needed to move the vane off the end stop should be used. This is about 50-55g at the lug on the back of the wiper assy.  Water has a convenient density of 1.0 g/ml or 1 kg/litre ( 1.0 ml = 1.0 cc = 1.0 gram of water ). Graduated plastic syringes provide a useful way of applying a controlled amount of water to load the spring.



Figure 6. Applying a load using a measured quantity of water

A nice smooth pulley wheel should be used to minimise resistance so as not to bias the readings.




Figure 7. Bench power supply and voltmeter connections, with load attachment thread running near edge of bench.
Probe is clipped to central copper pickup arm, rather than pin7 itself. Pin 5 is the on the extreme left of the connector block, pin 8 on the right.

A stabilised bench power supply will have over-current protection in case of accidental shorts, which could cause damage. Otherwise a 6V motorcycle battery ( with fuse-holder )  could be used. It is imperative that the supply voltage remains stable throughout test. It should at least be checked at the start and end of the test run. In normal operation, 12V is applied to pin 9. The air temp. thermistor and other resistances between pin8 and pin9 drop the battery 12v to about 7.5V on pin 8 in real conditions. Anything around that value is suitable as a test voltage applied to pin 8, as long as it is constant.



Figure 7b. Showing fine tweezers used as a guide to consistently align wiper plate lug with successive cog positions ( as used to produce figure 3.)



Figure 7c. Showing close up of AFM connector and pin numbers referred to in the article.


Visualising the test data:


Clearly it is very helpful to be able to visualise all these numbers and readings to see how the various calibrations affect the flowmeter. The graphs produced here have been done using some open source ( free ) software called Gnuplot. It is available for all major platforms http:www.gnuplot.info

Windows versions here: https://sourceforge.net/projects/gnuplot/files/gnuplot/5.0.3/ , Linux users will find it in their package management system. MacOS users here: http://www.flapane.com/nix.php

It reads in data from ordinary text files and takes commands typed into its command console. Pasting in the commands below should produce some similar to figure 3.

Gnuplot is a very powerful graphical tool once mastered and takes time to learn, but a few snippets of code will provide a means to reproduce what is presented here without the need for a tutorial on how to master the software. That codecan be reused to achieve a similar result if the same text file layout is used for the data file.


The following commands will recreate the voltage vs angular displacement graph in figure 3.
$0 refers to the line number in dataset, in this case the tooth number from starting position.

Code:

set title "Bosch flow meter: angular displacement test"
set xlab "spring tensioner wheel tooth number ( closed = 1 )"
set ylab "potentiometer wiper voltage"
set key top left Left reverse
vref=5.98

 plot "bosch_flowmeter.dat" using ($0):($1/vref) index 0 w p tit "modified unit", "" using ($0):(($1/vref)) index 1 w p tit "untampered unit"

The following data stored in the file called  "bosch_flowmeter.dat" were generated using the teeth on the spring tensioner wheel as an angular scale. A long, thin tweezer  was used as a guide to ensure accurate alignment on each tooth. Here the load was not measured, simply the angular displacement from the rest position with the vane closed.

Code:

# first data is K100 hacked unit : suspect spring has been moved from factory setting.

# index 0:  K100 tampered unit , using alignment guide
0.85
2.18
3.10
3.87
4.34
4.70
4.96
5.22
5.29
5.39
5.54
5.64
5.73
5.78
5.83
5.87
# two blank lines must separate data runs, re. "index" in the plotting command.


# index 1:  good K75 unit, using alignment guide
0.86
1.81
2.96
3.73
4.23
4.59
4.84
5.14
5.36
5.53
5.64
5.73
5.79
5.83
5.86




The torque test:


The torque test data was logged in two columns : col 1 the volume of added water, col 2 the voltage on pin7. The output was divided by the input voltage to give a normalised result maxing out at 1.0 , the weight of the water recipient ( 26 grams )  is added to column 1 and converted to a torque figure which is independent of the point of attachment. These calculations are done on the fly by the plotting commands.


The following code and data will produce figure 9 and can be adapted to produce similar graphs:

Code:


set title "Bosch flow meter: torque test"
set xlab "spring load / N.cm "
set ylab "noramlised potentiometer wiper voltage"
set key top left Left reverse
vref=6.01
rad= 27.4e-3; g=9.81


plot "torque_test.dat" using (($1+26)*g*rad/10.):($2/vref) index 1  title "unmodified K75 AFM" \
, "" using (($1+26)*g*rad)/10.:($2/vref) index 2 title "spare AFM no dead zone" \
, "" using (($1+26)*g*rad)/10.:($2/vref) index 3 title "spare AFM small dead zone"\
, "" using (($1+26)*g*rad)/10.:($2/vref) index 0 title "spare AFM +two notches tension"




click to see data for figure 9:

Test results:


To assess the effect of dead zone adjustment, the test unit wiper was set so that the output voltage moved off the set base value as soon as the vane was moved off the stop. This leads to earlier injection and a richening of the mixture compared to the unmodified unit with a notable dead zone.

Figure 8 comparing a good, unmodified K75 ATM with a test unit with zero dead zone.

Road testing showed this zero setting was good but less than optimal. Some more experimentation resulted in a setting with a small dead zone that gave good results on the road, rapid snap open throttle response ( for a K ) and good low speed throttle control ( no marsupials ).

Subsequent tests on a gas analyser showed an AFR of 14.70 at tickover. Bang on the theoretical optimal mixture.





Fugire 9. Comparing the effect of adjusting the dead zone and spring pre-tension to that of a standard working AFM.


Resealing the casing:

To ensure the unit is protected from the potentially damp air in the air filter box where it is situated, it much be thoroughly resealed. The factory uses silicone sealant but many such pastes used for sealing casing use acetic acid ( white vinegar ) as a curing agent. This is know to be detrimental to electronic circuitry. It is not know whether this will affect the resistor network but in absence of such knowledge it would been best to avoid its use. There are electronics compatible silicone pastes available or a polyurethane mastic could be used. The latter is much stronger than silicone and should be used sparingly to avoid difficulty in future removal.

A thin bead of sealant should be applied to the recess around the plastic cover and wiped off to be a little lower than full. Similarly, if the contact plate was removed this should be refitted with sealant. To be really rigorous it may be good to flush the air in the potentiometer housing using  a can of pressurised dry air which is available in electronics or computer shops. It would certainly be best to avoid sealing in warm summer air which holds considerable moisture. Cold morning air is preferable. Especially in tropical or semi-tropical regions dry air should be considered.



Conclusions:

Moving the resistor base plate provides a good way of getting a clean, new track whilst maintaining the factory contact pressure on the wiper contacts. This seems to be essential to ensure good contact and long service. The downside is the need ot reset the wiper rest position. The method developed here allows an objective measurement of the output of the AFM as a function of wiper zero position and thus a means of setting it up. It also reveals how the wiper rest position affects air fuel mixture accross the rev range and give some understanding of how adjusting it is affecting the AFR of the bike.

It is likely that there is some instrument drift in these units after 20y of service meaning the factory calibration may no longer be valid. The test procedure allows identification of whether an intervention is needed.

Taking a measurement of the output voltage vs the torque applied to the spring provides a thorough assessment of reliability of the unit under conditions which realistically represent in service usage. This will assess both electrical and mechanical defects that would lead to poor fuel regulation.

This text may get modiified as I spot errors or have them pointed out by others. I hope that this is useful to other brick-pilots. There seems to be a lack of reliable information and understanding of how these instruments work. Hopefully this will help fill that gap.


Appendix:

This circuit layout of the resistor chain maybe useful for reference:


http://www.k1100lt.de/files/flybrick/dkjk/luftmass/Seiten/Luftmassenmesser_2_gif.htm
From 'Dieters K Jetronik Kompendium'

The sum of all the resistors on the left is approximately 300 ohm, this means about half the battery voltage is exposed at pin8. Under normal service this is an input to the ECU which can then be compared to the potentiometer output on pin7, giving a ratio independent of fluctuations in the vehicle battery voltage.

Thanks for this information Beamer. I think it will be very useful to quite a few members who venture down this path.

Below are a few notes I've made. They aren't criticisms, just some suggestions and questions.

**************
I'm just being a little nit picky with this but you seem to be someone who likes accurate information.
The control unit is referred to by Bosch as the Electronic Control Unit (ECU). Most here refer to it as that. More accurately its called the Fuel Injection Control Unit (FICU) because we have two control units in the K100 the other being the Ignition control Unit (ICU).
You use the term EFI which refers to the process of injecting electronically controlled fuel.

**************
Re:Construction Detail Para #8
What is ATM ? A typo maybe

**************
Re: General Description para #2
My understanding is that the thermistor is a separate input to the FICU where its used to evaluate air mass. From your description further down, the Thermistor outputs directly from power to Pin #9. It appears it is not a part of the resistor chain of the AFM which you mention in General Description para #2 and Construction Detail para #10

**************
Can you explain why you have used 6V for testing and not 12V?

**************
Regarding the ECU (L-Jet) being totally analogue. Yes... I've had a chuckle from time to time when people mention they've had their ECU 'remapped'. Not possible.

**************
Re: Calibration Test Procedure para #1
In normal operating setup, power is applied to Pin #8 and returns to GND on Pin #5. Output is thru Pin #7 (back to the FICU to GND).
I'm not great on electronics testing at all but is this the same as how you have connected your test up. The photo you've posted under this does not show what you have connected and where. You've probably connected it correctly, its just not clear imo for others to reproduce.

**************
"A nice smooth pully" is very subjective. The sensor flap spring loading is very light. I suspect there will be some torque lost thru the pully. Would it be possible to mount the AFM vertically so the load apparatus just hangs directly off the armature. I can't see that that would affect the functioning of the AFM.

Is it possible to screw/bolt the unit to a board and then clamp or fix that to the bench. Would make adjusting the loading angle much easier imo.

**************
Using a syringe to incrementally measure the water has the possibility of incrementally introducing a large error. Better to work out the total load needed to deflect the flap to WOT and introduce that to the load mechanism as a total volume of water incrementally via a burette. That would be accurate.

**************
I think you will loose a lot of people with the gnuplot software. Can't this simply be plotted in an Excel spreadsheet?

**************

There seems to be a lack of reliable information and understanding of how these instruments work. Hopefully this will help fill that gap.

There's a lot of information on the AFM on this forum. It may not cover the same area as you've investigated but its just as relevant. Reliable or not, information shared on this forum is given in good faith. Its up to members to discuss contributions and hopefuly correct any misleading information without criticism or derision thereby adding to the collective knowledge. This comment will only serve to aggravate those who have contributed previously on this subject.

Can you explain why you have used 6V for testing and not 12V?

Actual tension under real conditions is about 7.5V at room temp. because the other resistors drop the voltage  I'd put note about that but edited out for brevity. I've added a note on that back in. Others may wonder too. It's not critical but many may not realise that. Good point.


Thanks for all these pertinent comments. I'll get onto it. Thanks a lot.

Reliable or not, information shared on this forum is given in good faith

That comment was not referring to this forum but internet sources generally. There is a whole lot of garbage and mis-information out there. Lot's of youtube vids where people are poking at these things with screw drivers in a way that will just not work or will degrade the unit. They usually avoid saying what happens when they put it back on the vehicle !

Using a syringe to incrementally measure the water has the possibility of incrementally introducing a large error. Better to work out the total load needed to deflect the flap to WOT and introduce that to the load mechanism as a total volume of water incrementally via a burette. That would be accurate.

I figured most people can get their hands on a plastic syringe, who has a burette?  In principal you are correct, it would be better to have something that can hold the total quantity of water in one hit. The 20ml syringe needs filling four times, so the errors at each 2.5ml step are not accumulative.  I get eight 2.5ml squirts for one fill up error which should be less than 0.5ml . There are four such refill errors in a test run. It is not the principal cause of error.

I could have said explicitly that the syringe should be filled but if I have to start explaining how to hold the pencil too, the article will get very long. If anyone does a refill every 2.5ml there will be an accumulative error.

As an alternative using a smaller capacity syringe may be a good idea to get some better low down detail. Smaller syringes will have much better accuracy for small quantities: maybe 1ml each time with 0.05ml graduations. That may be worth looking at.

Re: Calibration Test Procedure para #1
In normal operating setup, power is applied to Pin #8 and returns to GND on Pin #5. Output is thru Pin #7 (back to the FICU to GND).
I'm not great on electronics testing at all but is this the same as how you have connected your test up. The photo you've posted under this does not show what you have connected and where. You've probably connected it correctly, its just not clear imo for others to reproduce.

According to the circuit I have the pin8 is an input to the ECU. Pin9 is 12V. Why do you say power is applied at pin 8? This may need verifying.



AFAICT the two 360 resistors shown outside the ECU represent the resistor chain between pin8 and pin5 which is the point of study here, and the lower 360 is the other two resistors on the ceramic substrate, in parallel with the air temp thermistor.  So thermistor is independently used by ECU but also directly affects voltage applied to the rest of the resistor chain. That is why I applied the test voltage at pin 8 instead of 12V at pin 9.

I don't know the origin of that schematic, and the nominal valued of 360 ohm are not correct. I measured 7.5V on pin 8 and if they were both equal it would give 6V at pin 8 ( which was in fact my test voltage ).

"A nice smooth pully" is very subjective. The sensor flap spring loading is very light. I suspect there will be some torque lost thru the pully.

It would in some ways be better vertically mounted but IMO it would be so difficult to operate like this it would probably end up flying across the room. I have a large window here that I wish to keep  

I used a large diameter, bearing mounted pulley and this gave pretty good repeatability. There is a little frictional 'stick' but this seems to come from the unit itself mainly.

At each reading, after adding the water dose, I raised the load a little by hand and then lowered it gently, so each reading was taken as an opening movement of the vane not a static increment of load.

I found that the biggest repeatability error factor was keeping the thread at right-angles to radius of the point of attachment. I ended up adjusting the position of the AFM about every fourth reading to keep it correctly aligned. This can be seen in the data files , marked as #R, and on the graphs where there are two points vertically aligned. By the time I'd done a few runs I was getting consistent, repeatable results. ( The +2 notch line was done ealier with a little less care. )

It would be possible to improve the technique a little, especially to get more precision and detail around the initial opening of the vane but I considered the results good enough for the purposes of seeing how all this worked and setting up the unit.

At the end of the day it is road testing , not the graphs which perfects settings. The load test is just a guide as where the wiper now is since positioning it and locking it down is a bit hit and miss. An experimental measurement of where it is in relation to vane is needed.

For the purposes of setting it, maybe some kind of screw adjuster could be clamped onto the inlet and the dead zone set that way. I devised this method initially to understand how the unit worked and to assess the effects of shifting the wiper. It also served for setting the wiper later. A full-swing, torque-driven analysis is needed to check whether the unit is giving stable output too.

I think you will loose a lot of people with the gnuplot software. Can't this simply be plotted in an Excel spreadsheet?

Those who are familiar with spreadsheets could do that. Personally I hate them. Spreadsheets are fine for accountants. They are pretty crap for science and engineering and data processing. .

It should take one cut and paste operation to throw my gnuplot commands in and a produce a graph.

If someone wants to spend time clicking around in a spreadsheet, I'm sure we could produce a portable document in Libre-Office spreadsheet that would universally usable too.

Libre Office is a clone of Excel and has copied most of its defects. I find it a royal PITA to try to get it to produce what I want, instead of what someone else thinks I need to do.  I have a heavy metal object on the bench and a large window ....

I can get gnuplot to do whatever I want and produce nice clear graphs. YMMV.

I showed my plot to the garage that did the gas test for me and an he was like: wow how did you produce that? .

Fuel Consumption: 5.4 lit/100klm

I am still well above that , last top-up was 10litres / 150km ( =6.7)  but I have to admit I've been gunning it everywhere doing testing and no long runs. I guess that 5.4 kind of figure is on long run.

I used to get 5.6 out of the Le Mans IV in its hayday on a long run, even if going full bore up and down the box . That has 40mm Del Ortos with accel. pumps ( and two or three times the low down torque of a K )

Now less blogging and more air-miles required.

Beamer wrote:Calibration test procedure:

The basic method is to progessively apply a known force to a radial point on the wiper assembly and measure the voltage at the wiper connector ( pin 7 ) whilst applying a fix voltage ( 6.0V) on pin 8. This bypasses the variable thermistor air temperature sensor which should be excluded from this test. Pin5 is ground.

In this and in other sections you refer to pin numbers. I can see now that you are quoting FICU pin numbers but in your pics you show your test equipment connected directly to the FICU connection. Using the FICU pin number to reference the AFM connections is both confusing and dangerous.  What if some poor noob, while following your procedure, got it wrong and connected his power supply into the FICU

I can see now that you have modified part of your OP without notating any of your changes so now its become a little difficult to follow this progression. As far as I could determine you made no reference to applying 12V to the AFM in you OP but I see now that you have added that thru your OP.

Beamer wrote:According to the circuit I have the pin8 is an input to the ECU. Pin9 is 12V. Why do you say power is applied at pin 8? This may need verifying.

You said you applied a voltage of 6.0V to pin #8 See your quote above at the top of this post.
From what I can determine pin #8 (ECU) is an output of the AFM


The wiring schematics clearly show that the AFM is fed with 12V directly from the battery (thru the fuel injection relay) and then returns directly to ground (Nothing to do with the ECU but the wiring just happens to piggy back of the ECU plug connections for convenience). As far as I can see this applies tension to the AFM and the air temp thermistor which output a modified voltage directly to the FICU.

I've since noticed that the AFM's connector has numbered pins 1 thru to 4. I'd suggest those would be more appropriate to use.

I've been compiling pinout information for all the connectors on our K100. Not done the AFM as yet but will put that together now I have the information and post it here on the weekend. You can use that to reference your setup if you want.

To be honest, I think you have done some important research on this but I feel it needs to be a little clearer so that others will be able to replicate and validate your procedure.

I'm finding it difficult to follow this thread so I think I'll wait till I can get to my own AFM in the coming month or so and take a reccy before I go entering into an indepth discussions atm. It's certainly peaked my interest. There's been a lot of discussion on this forum about the topic so I'll revisit those posts as well I think.

Cheers

I've since noticed that the AFM's connector has numbered pins 1 thru to 4. I'd suggest those would be more appropriate to use.

Where do you see those numbers ? There are five pins in that connector. :?

I have four examples of the AFM. All have numbers as I used in the article : 5 9 E 7 8  reading from left to right. These seem to be taken to match the ECU numbers. I've added a photo of the connector for clarity.


Figure 7c. Showing close up of AFM connector and pin numbers referred to in the article.

Using the FICU pin number to reference the AFM connections is both confusing and dangerous.

You need to write to BMW / Bosch about that

I can see now that you have modified part of your OP without notating any of your changes so now its become a little difficult to follow this progression. As far as I could determine you made no reference to applying 12V to the AFM in you OP but I see now that you have added that thru your OP.

I have made some minor modifications to clarify the text in response to your comments. I have not changed the substance. I did not want to have a mess with strike-outs which would just make it more difficult to follow. Sorry if that has made if difficult to refer back.  I want the top post to be self contained and not subject to any future reader having to read all following comments to get the full storey. I don't intend to replicate the situation on the other forum where you need to read following comments to find out that the power is applied backwards.

I made no mention of applying 12V because I did not do that. I applied 6V at pin 8. I added some more explanation as to why that was done in response to some points you made asking why that was done. Maybe this is still not clear.

Under normal running conditions. pin 9 gets 12V and something like 7.5V is seen at pin 8. This is variable due to the thermistor. To remove this variable element I applied a fixed voltage at pin 8 ( and no connection to pin 9 ). This is similar to the voltage it will get in-service and provides constant conditions for the test. As I said (and you quoted) pin 8 is normally an input to the ECU. Applying 6V here is just the test voltage for the resistor chain.

You will note that I've normalised the potentiometer output by dividing by the test voltage. So if anyone wants to repeat this and they use my gnuplot code, with their own test voltage, they will get output directly comparable to mine irrespective of the test voltage used.

Thanks for proof reading this for me. Is there anything incorrect or unclear about the voltage which was applied or where it was applied?  It seems clear to me but I know what I did and maybe I'm over-looking something.

To summarise the resistor network:

there are 14 laser trimmed resistors on the ceramic plate. Ten of those are within the range of wiper arm, with one extra one at each end going to pin5 and pin8. Those two are why graphs never touch either zero or one on the vertical scale.

The remaining two are between pin8 and pin9 , one of which is in parallel with the thermistor air sensor.

So there is a continuous chain of 14 set resistors between 12V and ground. The tests done here are on the lower 12 between pin 8 and ground.

seems to be working swell...unlocked.

hi a question that might be dumb but i am wanting to no is it possable to make the afm run richer on high end of revs to help with nasty pining if it had a boost spike as am running small turbo at 7 lb but could spike to higher sometimes for some reason.

Don't waste your time stuffing about with an L-Jetronic. have a look at megasquirt
http://megasquirt.info/

A while ago a mate told me he found unpleasant to drive behind my K100 because of the smell of the exhaust gas.
As there is no excessive oil consumption whatsoever, I started investigating the fuelling system but I couldn’t find any evidence of defective components.

After this, the vane-type AFM was for me the first thing to examine in detail.
Based on the info provided in Bertrand Vogel’s comprehensive no-start-troubleshooting document, my AFM was OK.
I was wondering what could be the influence of the return spring inside the AFM and the effect of it’s softening over the years.
Unfortunately, I couldn’t find any absolute airflow figures on the net.
I decided then to throw the damn thing on the flow bench and determine the relation between airflow and vane position.

The AFM was fed with 12V between pins 8 (+) and 5(-) and the output voltage was measured between pins 5(-) and 7(+).
The air bypass screw on the AFM was open for 1 ½  turns, just as it was on the bike.
The measurement device is a PTS flow bench.
A picture of the setup

The lowest measuring range of the bench is 75 CFM, determined by the smallest orifice plate.
Because the readout of the airflow on the inclined scale is logarithmic, measured values below –let’s say in this case- 45 CFM suffer from poor resolution.
I did my best to look very carefully and averaged a couple of measurements.
The result is as expected a more or less logarithmic looking graph.


For later comparative measurements I made myself a smaller orifice plate to overcome the resolution problem but this has the disadvantage of not being calibrated.
Therefore, later graphs are presented with a relatively dimensioned horizontal axis.
This graph shows the very little difference between three different AFM’s, mine, the one of my previous K and one that a friend was so kind lending me.


This shows the effect of different positions of the airscrew.
I was surprised to see how far the influence of this screw reaches.


The influence of the AFM internal return spring preload.



Anyway, all this showed nothing was wrong with my AFM.
Later I could adjust the airscrew on my bike with the use of a Gunson CO meter.
The CO value was at 2.5% and I could bring this back to 1.3% by turning the airscrew counterclockwise by half a turn.
Result is a less smelly K.
No animals were harmed during this experiment

Just my 2p

Thanks for this contribution.

It is certainly interesting to have this calibrated output in CFM.  It would be interesting to see it plotted on a vertical log scale, as I did above, to see whether it really is logarithmic. 

This seems to contradict my expectation, stated above, that the non-linearity of the potentiometer would cancel the non-linearity of the air flow.

Would you mind posting the data for your first graph so that I can look into this?

Exported as csv

Code:

CFM,Bosch p/n 0 280 200 040 - 118.953km
0,1.707
3.75,6.46
7.5,8.52
11.25,9.51
15,10.05
18.75,10.36
22.5,10.63
26.25,10.79
30,10.93
33.75,11.01
37.5,11.1
41.25,11.16
45,11.22
48.75,11.26
52.5,11.29
56.25,11.32
60,11.35
63.75,11.37
67.5,11.4
71.25,11.42
75,11.44


I gave it a try in excel. It's pretty well approaching what you say but with a logarithmic horizontal scale. I guess that's what you ment anyway.

Many thanks.

That is still a long way from straight on a log plot. Doing a second log gets most of it straight, so maybe it's not as contradictory as it seemed, needs more thought. Obviously the flow rate vs angular deflection ( what I effectively used as the controlled variable ) will be very non linear and probably not even as simple as an exponential relationship. Turbulent flow is a bitch.

Anyway it's great to have some hard data from someone with a flow bench.






Actually it's better to plot exp of AFM output since the first point of zero flow does not fit on a log scale.  This starts to look similar to my fig 4 in the original post. We can see what I called the dead zone before AFM vane starts to move. This regime is affected by the bypass screw, which is not accounted for in my graph of the vane angular displacement.



It looks like square of exp fits the higher end and may fit lower end with bypass screw shut.

Force of air on vane being square of air speed ( and hence vol/min ) seems to make sense.  I had fitted log of torque on spring to output voltage, so that all hangs together.

If it's not too much effort, could you post the data for the other graphs where you had better resolution at low flow rates? This is quite informative  as to how this all works.

Since the Jetronic is an analog device, I wonder if some of the non-linearity in the output signal vs. airflow was programmed in to compensate for the response characteristics of the injectors or possibly the flow resonance of the air passages.

It would be very simple to shape this curve via the fixed resistor chain in the AFM, so this is certainly intended to be an analogue of some of the physical effects.

Recently I found datasheets of some Bosch hot-film air-mass meters.           
Apart from the scale, the Output voltage versus Mass air flow graph definitely has a different shape than the one from our vane-type meters.           
IMO not the kind of difference that can be matched for with a resistor chain, unfortunately.

Yes, these K100 flow meters are doing quite a specific function. I would expect more recent hot wire devices are more normalised to  standard mircro controller input voltage ranges and will designed for purely digital devices, not the bespoke, hybrid analogue computer found in the K100 ECU.

Anyone considered these?

http://max.pilotpowersupply.com/product/pilot-vaf-to-maf-converter-with-a-bluetooth-interface/

* Train dogs. Teach people.
Some people are slow to learn  

You cannot substitute a new digital device into a bespoke hybrid analogue computer design.

Do the Pilot Power Supply product developers need a good spanking for writing this?

Thanks Laitch, I did not read far enough. I thought it was just a generic MAF sensor.  This unit has been specifically designed to work with the LE-Jetronic AFM and back-engineered to emulate the voltage / flow profile of the original via a look-up table. It should work better than a worn out OEM flow meter for those who don't feel comfortable with complexity of the repair method I detailed here.

This device reads the MAF sensor and calculates the equivalent volume of air, scaled to the OEM VAF sensor’s characteristic from a special table of conversion.

The main problem is that you need to mount and connect this unit is thorough and robust way that will withstand the tough vibration and environmental conditions of a road vehicle for years and years without the slightest risk of a loose wire or other breakage.

Failing to do so could leave you with a sudden loss of engine power half way around a bend in a mountain road or in the middle of busy intersection !!

Thanks to jcd for posting the link, It's worth knowing about.

Some people are fast to write                          
                                    
The Pilot Engineering converter is a much cheaper item than the Split Second one.        
https://splitsec.com/wp-content/uploads/technotes/TN2_AFM_to_MAF_Conversion.pdf        
                                    
I posted in the Pilot Engineering support forum.                    
Let's wait for an answer. How much material is worth also depends on the support provided 
Meanwhile I'll try to find some time to have a look at how to deal with the IAT sensor in their setup.    
                                    
Regardless of the converter used one will need a MAF, and for sake of simplicity it would be best if we all use the same.
Could we agree upon a MAF to be used for the K100 engine?                
There is s few Bosch AFM listed in a catalogue                     
https://www.bosch-ibusiness.com/media/images/products/sensors/xx_pdfs_1/sensors_i-business.pdf    
Keep a local copy of the pdf because these links appear to change regularly.            
                                    
A 1000 cc 4-stroke engine at 8000 rpm assuming a volumetric efficiency of 100% will pump 4000 litres/min.
At 1.25 g/l air this would be 5 kg/min or 300 kg/h.                    
I would suggest the 0 280 218 440 with a measuring range till 480 kg/h. It's the smallest in the catalogue anyway.
Any other suggestions or remarks?

This is turning into a thread hijack. Can I suggest you start another thread about these retrofit units and link it here?  It's well worth sharing and discussing but not really anything to do with restoration the original Bosch AFM units.

I'll follow any discussion in the new thread. Looks interesting.

Beamer wrote:This is turning into a thread hijack. Can I suggest you start another thread about these retrofit units and link it here?  It's well worth sharing and discussing but not really anything to do with restoration the original Bosch AFM units.

I'll follow any discussion in the new thread. Looks interesting.

You're absolutely right.
I'll do that in the next few days.

jcd06 wrote:

Beamer wrote:This is turning into a thread hijack. Can I suggest you start another thread about these retrofit units and link it here?  It's well worth sharing and discussing but not really anything to do with restoration the original Bosch AFM units.

I'll follow any discussion in the new thread. Looks interesting.

You're absolutely right.

It's here
https://www.k100-forum.com/t18099-replace-k100-afm-by-a-maf#207599