Air flow is a main input parameter for the fueling calculations. One would think that measuring airflow directly (instead of the indirect Speed-Density) is the best way. Surprisingly it's not that simple. It has benefits and drawbacks.

----

Technology

TODO: links

* flap type (mechanical)

* hot wire

* film type

* pressure difference

The 'flap type was the old technology, slow, restrictive, and everyone likes to forget it.

Hot wire heats a (very thin) wire to target temperature (measuring its temperature from the resistance), and finds out the airflow from the needed heat (current*voltage). Algorithms very similar to those used for WideBand O2 could be used to drive a dummy wire. From time to time the wire is heated to higher temp to remove deposits.

Film type is an advanced version of hot wire. Usually very expensive (many build cars for less :-)

Pressure difference (between too points of a special pipe) is not used too much in gasoline engines as far as I know, except in airplanes and industry.

----

Why MAF - should we split this comparison to a separate page of fuel-calc strategies (adding alpha-N) ?

There are some reasons for MAF:

* redundancy - extra input besides MAP

* EGR (the inert gas makes MAP-alone reading less useful: MAF is measured before the inert gas added). Alternative is to measure MAP and O2 content of intake with WideBand after EGR valve. (so WideBand in both the intake and exhaust is the best for cars equipped with EGR)

* people who cannot tune (MAF based system needs less work on fuel tuning)

* agressive CAMs can disturb the MAP measurement (see AlphaN)

* controlled CAM (VTECH, VVTI*, VANOS pneumatic valves, etc..) changes VE, MAP is hard there

* MAF is already there. Especially if we want to support an asis harness like PlugAndPlayEEC

What are the problems with MAF?

Normally you can make better resolution, speed and reliability with Speed-Density (MAP). Because MAF is

* slower (responsiveness: quite obvious, measure distance from valves)

* worse precision (surprised? read below)

* inversion of airflow (similar issue to agressive CAM above)

* price (5 .. 10 times)

* reliability

Resolution is quite obvious:

with Speed-Density the injector pulsewidth is almost proportional to MAP (not completely, because of VE, lambda and injopen).

With MAF you divide measured airflow with RPM (than consider lambda and injopen), so you need very precise path for the analog signals to maintain a given resolution (or will have inferior resolution with similar technology). The box-box signals are often nonlinear (but close to logartihmic), so not much precision is lost there (that's just the signal transfer!), but still much precision is lost during the measurement (very precise and sensitive circuits come up with a relatively unprecise output because of the nature of the measurement). The division by RPM means that it is common to have unprecise result at lowRPM and rail out before max power.

Reliability is simply the result of complexity. It was not very funny when recent Audi A4 misbehaved and the shop could not tell the (MAF) problem with the standard Audi diagnostics procedures. When it became obvious, the price of the MAF sensor was a bit higher than a complete assembled VEMs ECM.

At one point a friend had to change MAF sensor twice in 2 months on a Volvo (I don't meet too many cars, still seen many problematic MAF failures - not a single problematic MAP failure: once MAP sensor was unconnected, trivial to diagnose in 2 minutes). The Volvo sensor was at least not that expensive, still above the price of a diy v3.x ECM.

If one finds that MAF shows better results for him than Speed-Density, he should look into his MAP (maybe broken in some way, or not tuned well).

----

MAF off the shelf

links to engines, shops, particular MAF-sensors, output signal curves, etc...

outputs:

* frequency type

* analog voltage type, appr. logarithmic scale

PlugAndPlayEEC

:[Tools and more information]

----

best of both worlds

Best to have both MAF and MAP and good limphome functions...

Calculate MAF and MAP confidence, check the difference between fuel values, use one or the other for calcs (or weight them?) and report diagnostics (especially if one goes bad).

Ideas?

----

bits from a discussion

feel free to clean up...

<code>

<mcell> it requires a lookup table (MAF signal is usually logarithmic) and a division in fuelcalc

<mcell> pw = const * flow / RPM * lambdacorr

<hackish> well I knew it would need to be divided by RPM.

<mcell> but other than that division, it's similar to MAP

<mcell> however MAF is not a good idea: slow

<Jorgen> If we make an alpha-N like setup but that use the MAF instead of the TPS as an input it will work.

<mcell> yes, of course: with the table (MAF, RPM ) it's easy

<mcell> I think we can make WBO2 adjust even when EGR is applied (EGR is nice both for economy, emissions and ALS-substitute or ALS-helper)

<hackish> as I understand there are 2 main reasons for MAF in factory cars...

<mcell> hackish: yes: "lame" "design"

<Jorgen> With MAF RPM is optional.

<hackish> a) more accurate b) better transient response c) more adaptable to manufacturing differences (mods)

<mcell> they get more money when servicing the car

<hackish> er I guess that's 3

<mcell> a) is bullshit

<mcell> b) bullshit

<mcell> c) true

<mcell> MAP has much better transient response, and better accuracy

<mcell> what we have in MAP (25..250 kPa) and RPM (500..8000 RPM) is stuffed (multiplied) into one MAF signal

<mcell> 12500 .. 2000000

<mcell> notice the dynamic range difference

<Jorgen> MAF is used to allow the engine to work even when it is aging and to allow for engine tolarances.

<hackish> good point jorgen.

<mcell> yes, but the WBO2 is better than MAF in that

<Jorgen> Definately.

<hackish> yes, most cars do not come with a stock wideband.

<mcell> so MAF is a dinosaur from the time when WBO2 did not exist

<hackish> interesting.

<mcell> I would only use as a veryfication of speed-density (+limp-home)

<hackish> I'd like to get some limp code in there too...

<Jorgen> Limp home must be mapped, but if we are willing to do that we can look at Autronics application.

</code>

----

See also

* AlphaN