DIY-WB Output Voltage Scaling

[bcroe at juno.com]

Now that the WB project is off the ground, a potentially
even bigger problem looms: How is it best used?  Here
are some thoughts about that.

Say I'm changing fuel injector pulse width (changing the 
fuel added to a volume of air).  Lets say I make a 1/10 ms
change.  I would expect the output format of my WB unit
to change a certain amount.  If I keep making 1/10 ms 
changes (same amount of air), I would expect the WB 
output format to keep taking equal sized steps across 
the lean to rich scale.  In other words, a linear response.

I would be annoyed if the WB output showed a huge
step in the lean area, but a small step in the rich area,
for the same size change in injected fuel.

If you agree with the above, then how might that be 
achieved?  The WB output voltage curve faithfully tracks
the sensor, with a continuously changing slope and an
extra bend at stoich. 

What about a direct A/F readout?  The very same 
amount of fuel change required to make one A/F division 
change at the rich end, will make TWO A/F division changes 
at the lean end.  Thats because A/F is NOT directly 
proportional to the change in fuel.  A/F has been a handy 
handle in the past, but just what IS the linear tuning 
function I (and perhaps you) want?

I have cooked up an output display for the WB project
which (I hope) displays a LINEAR input fuel change to
output display change.  Just how does this work?  

Imagine you have this box of air, and you add a drop of
fuel, say 1 % or 1/100th by weight.  Then more and more 
drops.  The ratio of fuel to air (by weight) is increasing
LINEARLY with the number of drops.  Just plot 0/100,
1/100, 2/100, 3/100, 4/100 against 0, 1, 2, 3, 4 etc drops, 
and you get a nice straight line.  

So what about A/F?  WAIT A MINUTE, the above is a
description of FUEL/AIR, NOT AIR/FUEL.  F/A is the
inverse of A/F.  OK, lets plot the inverse series, 
100/0, 100/1, 100/2, 100/3, 100/4, 100/5  etc against 
the number of drops of fuel.  Gee, the first number is 
infinity, then 100, then 50, then 33 1/3, 25, etc.  This
has a TERRIBLE curve to it, not linear at all.  When 
we get into the practical range of 100/5 to 100/10, the
curve is still pretty bad.

So from the perspective of  tuning aid, what is the
answer?  Here is my proposal.  I have this moving
bar graph of LEDs which is perfectly linear, as 
described above.  Along it is a scale of A/F.  Since
A/F is NOT linear, the A/F divisions are twice as far 
apart at the rich end, compared to the lean end.  We
have linearity for tuning AND an A/F readout.  Oh
yea, the sensor has a different level of output on 
each side of stoich.  That's been taken care of too.

Is this a good solution?  I'm asking you.  The first 
display is already operational at my house.  A
couple of you already have a picture; sorry, I don't 
have a scanner.  The sensor limits are not centered
around stoich, so the scale is lopsided.  If we use 
a practical operating range of 19:1 to 10:1, the rich 
end from stoich will be twice as long as the lean 
end.  OR, we could make the scale twice as 
responsive at the lean end to equalize the length,
but forget about linear tuning response from end to 
end.  Your thoughts please.  This display does not 
prevent you from attaching any other output format 
devices you like.

For those interested in the nitty gritty details of this
display device, write me and I'll return the preliminary
(lengthy) writeup.  If you would like to get a better 
feel for this, get out that WB output voltage vs A/F
curve.  Calculate the inverse of A/F (divide 1 by the
A/F), and plot these numbers (F/A) against the
voltage.  You will get a straight line, with a bend at S.
The sensor is LINEAR after all, if we could fix that
bend at stoich.  The display design fixes that too.

Bruce Roe



On Fri, 24 Aug 2001 19:43:50 -0500 jll at edge.net (John Lamb) writes:
> Are you saying that the Display has an embeded PIC that allows you to
> program a segmented curve fit?
> That's what is needed here to linearize the output.
> Another method is to use the an A/D to address memory locations that
> linearize the output.
> Sorry if this is old discussion.
> 
> JL
> 
> Peter Gargano wrote:
> > 
> > It should be pointed out that the current incantation of the
> > DIY-WB unit has a non-linear voltage:AFR transfer function.
> > The actual transfer function is described here:
> > 
> >   http://www.diy-efi.org/diy_efi/projects/diy_wb/
> > 
> > See "Output voltage/AFR reference". A graphed version is here:
> > 
> >   ftp://ftp.diy-efi.org/incoming/DIY_WBafrVolts.gif
> > 
> > The curve can be thought of as a series of straight lines
> > for simple interpolation, but it means that a simple digital
> > panel meter connected to the WB output with a simple scaling
> > circuit will not be able to show accurately a large range of
> > AFR ratios. This is why I suggested the Jaycar kit/Silicon Chip
> > kit with 16F84 PIC and 3 digit display that comes with source
> > code that could be modified to scale the DIY-WB's output to the
> > above transfer function.
> > 
> > If you're logging the DIY-WB's output and post-processing the
> > logged voltage, then this isn't too much of a problem.
> > 
> > For the DIY list, here's the information I posted to GMECM
> > yesterday.
> > 
> > Product Name:        DIGITAL FUEL MIXTURE DISPLAY KIT
> > Catalogue Number:    KC5300
> > Price:               AU$62.95  (less than US$35) + freight
> > 
> >   http://www.jaycar.com.au/home.htm
> > 
> > Peter.
----------------------------------------------------------------------------
To unsubscribe from gmecm, send "unsubscribe gmecm" (without the quotes)
in the body of a message (not the subject) to majordomo at lists.diy-efi.org