[Diy_efi] Q regarding placement of a pressure sensor

[Chris Conlon]

At 10:31 AM 7/15/02 -0700, Bob Moon wrote:

>  This leads to my question:  why not run a pressure
>xducer before the turbo (say, in the stock MAF
>location) and just read the vacuum that is created as
>the turbo/engine inhales the air?  This would provide
>a reading from -15 to 0 psi (the further negative, the
>more air is flowing)

This could be done, but it is probably not a great idea.
You run into 2 basic problems.

To start with you need some kind of geometry that will
generate a pressure drop as flow rises. If there is no
pressure drop, you can't measure the pressure drop to
estimate flow. Ideally you'd like to keep the pressure
drop low, so you're not choking the engine, but then you
need very sensitive pressure transducers, noise becomes
more of a problem, hysteresis becomes a problem, and so
forth. So from a design point of view there are competing
goals, one to keep the pressure drop low to maximize power,
the other to make it higher to maximise resolution and SNR.

The second problem compounds the first. The 2 common
airflow measuring geometries (pitot tube and venturi)
both generate a pressure signal proportional to the
square of the airflow. This makes your resolution worse
in the low airflow areas than in the high airflow areas,
which is pretty much the opposite of what many other
systems strive for. (AFMs and MAFs tend to have more
resolution in low airflow areas than in high, so that
idling, tip in and light cruise can be dialed in well.)
This provides more of an incentive to use a high pressure
drop, to regain resolution on the low flow end.

There is probably some exception but my crude understanding
of aerodynamics is that most any fixed geometry will generate
a pressure drop proportional to the square of the airflow
speed, at least within some range of Reynolds numbers.
This is the basic principle behind pitot tubes, venturis
and orifice airflow meters.

Flap type AFMs use a variable geometry and so the pressure
drop is roughly flat over the design airflow range. But then
you need to measure flap opening as well... sounds like you
just built an AFM.

Someone clever could probably build a variable sprung
geometry which resulted in a pressure vs. flow curve that
was less than N^2 and closer to N^1. That might work.

You might consider other types of variable geometries,
such as a larger venturi blocked with a butterfly which opens
at higher loads. Basically you're building a carb here,
and so lots of what has been done with carbs will be
applicable.

If you can find a precision, very low noise pressure sensor
that reads in a very low pressure range (say 0-1psi) you could
probably make a system like this work, if not super well.

You might also get away with 2 venturis in parallel, one much
larger than the other, but with the smaller one towards the
center of the flow and the larger to the outside. If you sized
them right you'd get a usable reading from the large one just
about the time the small one was going nonlinear. It might
take more than 2 different sizes to cover the whole flow range
needed, especially on a high boost turbo engine.

Don't forget to compensate for airflow density at the inlet.
(BAP and ambient temp will allow for this.)


Along these lines I've generated flow curves for a couple
Bosch style flapper AFMs with a crude flowbench, using sharp
edge orifices and a low range pressure sensor. I needed to use
several different sized orifices to cover the AFM's full range
in sufficient resolution, even allowing up to ~2.5psi drop
across the orifice. I would think your current Karmann meter
does not generate anywhere near 2.5 psi of drop. Certainly
the design is capable of doing that, anyway.

Picking one AFM at random, I see that the highest airflow it
can measure is ~120 times the lowest it can measure. Let's
round down to 100:1 since the factory design doubtless includes
some margin beyond what is really needed. An airflow range of
100:1 will give a pressure drop range of 10000:1. That's
between 13 and 14 bits of a/d. You should go to 16 to have
some margin to deal with noise. That by itself is doable, but
the basic problem here is that you need a pressure sensor
ranged for (say) 0-1psi, which is about 1 in 10000 parts
accurate, or 0.01% accuracy. If you find one even close to
that anywhere close to cheap - let me know! (I'm dead serious
here, I'd love to have one.)

If you can come up with a geometry such that pressure drop
is linear in airflow, you only need ~1% accuracy which is
vastly more doable.

AFMs and the MAFs I've seen deal with this by compressing
the higher airflow ranges into much smaller signal changes than
the lower airflow ranges, basically companding the signal
so you get more effective resolution down low where you need it.


   Chris C.


_______________________________________________
Diy_efi mailing list
Diy_efi at diy-efi.org
http://www.diy-efi.org/mailman/listinfo/diy_efi