EGOR Output Signal Survey

[bcroe at juno.com]

Hi Dig,

On Thu, 17 May 2001 06:47:10 -0700 (PDT) Dig <turbodig at yahoo.com> writes:
> 
> > Date: Wed, 16 May 2001 04:12:24 -0500
> > From: bcroe at juno.com
> > Subject: EGOR Output Signal Survey
 
> > Next is the output voltage.  A curve seen in some 
> > equipment reads like
> >     A/F             Volts
> >     22:1           3.5
> > 14.57:1         3.0     stoich
> >   10.5:1          2.0
> > 
> > A circuit from the list runs about 0.5 V lower
> > than above, but otherwise the same.  My ckt
> > runs 0.2 V lower than above.  
> 
> Can I ask how you know that? Are you calibrating 
> from a known source? Not trying to back you in a 
> corner here, just curious.

All circuits are referenced from stoich, where the
Ip current is zero.  Just opening the Ip circuit will
reveal the output voltage if you haven't already 
calculated it.  Moving lean or rich from that point
is where calibration errors start to build up.  I chose
to make stoich 2.8 volts (with Ip = 0) to give good
accuracy with simplicity of design, but this point
can be shifted.  The rest of the curve will just move
with it. 

 > This is one of my biggest questions about EGOR-
> How do we *know* what the voltage to AFR curve is
> without calibrating one or more against a standard?
> True, you can use the known characteristics of the 
> sensor, but with part variances and tolerance, you
> could still be off a fair amount. (especially on a
> 0-1v range) 

The question here is, how close is the sensor to
its Rcal predicted response, when new and after
some use.  That can only by known by using some
gas method of calibrating.  I plan to have an air
calibration feature which at the least, can show the
sensor has aged/reduced output.  More extensive 
checks are up to you guys in actual use.  

The issue of parts tolerance can be reduced to
almost negligible.  I am using op amps with lower
offset voltages than an LM324.  More important
is resistors.  In most cases pairing off the critical
resistors is more important than the exact value.
By taking a dozen 1 % resistors and finding pairs
that are very close, this error will drop out of the
picture.  I will be writing up which units need to be
paired to maximize accuracy, if a circuit is released.

> Personally, I'd like to see the following 3 ranges:
> 
> 0-1v, inverted; rich = 1V  (for the guys doing WB
> control of the stock ECM)
> 
> 0-5v, normal, for the guys with dataloggers and for
> general purpose use.
> 
> 0-16v (0-12 acceptible), for use with stock ECMs 
> datalogging via ALDL. 

The polarity of the output signals could be set
with a pair of jumpers that either cross, or not.

The basic output is reduced to the range of 
about 1 V to 3.5 V by losses in the Rcal calibration
circuit, when placed to minimize noise.  However
the second buffered output could have any voltage
swing between 0V and 5V, set by a few resistors.
It would be inverted from the above to save parts.

An output of 0V to 16V would require a special 
power supply, and I don't think most people
would need it.  However, I could draw up a fairly
straight forward circuit to generate the power 
and the signal for those who wanted to add it.

You might reduce noise (always a problem in
a car) by tying your ALDL ground directly to the
ground coming out of your ECM

> The rationale for the last one is based on some wide
> band testing I've done recently via the ALDL. Using
> the 5V A/D inputs, you end up with .1 Lambda being
> .2 volts or so. The noise floor in my particular ECM
> seems to be higher than that, so I'd like to use the
> 0-16v Fuel Pressure A/D to get better resolution (and
> lessen noise-induced errors)


> Maybe a better solution would be to design it 0-5v,
> then stick an Op-Amp divider/multiplier on the output
> to do the scaling. Or isn't that feasible with the
> design?

Taking the Rcal circuit near signal ground (to minimize 
noise from the cable) takes the 0V to 5.5V swing, and 
reduces it to more like 1V to 3.5V.  But a buffer can 
restore the full 0 to 5.

 
> Personally, If you were to add any display, it'd be
> nice to see a numeric one like the Horiba has. Nice
> to, but not necessary.

I haven't seen the Horiba, what is it?  For those that
want a digital display, I would recommend buying
one of those digital voltmeter modules and driving
it with the proper voltage range.  I don't think it's
worthwhile to build one of these from scratch.


> > Allright, there is one more factor.  That almost 
> > perfectly linear delta fuel to delta voltage
> > relation, shifts as you pass stoich, to a slightly 
> > different very linear relation.  I propose that the 
> > gain (number of divisions per volt) BE SLIGHTLY 
> > DIFFERENT on the 2 sides of stoich.  So now a 
> > given increase in fuel will cause the SAME 
> > number of divisions movement of the bar, from 
> > one side of the display to the other.  Circuit wise 
> > this function is almost free. 
> 
> Are you saying you want to "linearize" the AFR such
> that a linear voltage is linear AFR? That'd be a 
> *very* good idea, provided you can be sure you know
> that your calibration is true.

I'm saying I could easily build a display that that has 
a linear relation between the fuel you add and the 
display change, across the entire range.  This seems
reasonable to me.  But understand the above is F/A,
NOT A/F (see my original post).  Think about it,
when fuel goes to zero, A/F goes to INFINITY!  You
are not going to linearize that, though part of the 
curve could be done.  F/A goes to zero, makes 
more sense to me.  I think linearizing A/F would 
not be as useful, although easier to read off.  

The circuit should be linear, even if the calibration
has the scale factor off.  Once again, liinearization
of A/F would be more subject to calibration.  

I would use a different circuit to linearize a voltage 
output than a display, either one costs about one 
amplifier for F/A, several for A/F.

> I'm working on a way to do this in the ECM (via a 
> table lookup -> out to ALDL AFR commanded), but it
> requires a lot of table entries to get it accurate.
> Anything to get the voltage more linear would be a
> big help.
> 
> > 
> > Something else we could add is a switch for
> > free air calibration.  This would normally be 
> > off scale, but it would reduce gain to put an
> > indication at 22:1.
> 
> This would be very nice... like hold a button down,
> and 2.8 V (the ref) = free air.

I'm thinking hold the button after complete warm up
in air, twist the fine adjust to read 22:1.
 
> > There is a warmup time approaching a minute
> > for the sensor.  
> 
> If you're lucky. :)  I've had to "Reboot" WB circuits
> several times before it "takes" and sez it's warmed.
> Could be a quirk of the circuit, though.

Never happened to me.  Probably their circuit.  I'm
running 1.5A warmup current, which initially takes
a dozen seconds to ramp up and a minute to run.
I found 1.2A would not push some sensors over 
the top to 10V heater operation.

> >I suggest the display be blanked
> > (turned off) during warm up.   Besides giving an
> > indication of operation, it allows the considerable
> > LED current to be "stolen" from the sensor 
> > heater circuit, saving power, heat, and heat sinks.
> > The cost is one transistor. 
> 
> This would be fine. Would be nice to have some 
> indication for the user, though, maybe a single
> "warming" LED? 

A "warming" LED could be added, most useful if
you have no display.  But I'd prefer an "ON" LED.  
That and a blank display would say you're warming.


> > One more output that might be useful would 
> > have a 1 volt output range, reading 0 to 1 volt as 
> > we moved from 15:1 to 14:1 A/F.  The idea is to
> > look like a conventional narrow band OX sensor.
> 
> Even more so... have stoich be .450, and let 'er
> swing full range (AFR-wise), as I mentioned above.

I think you are asking for the 3 volt stoich point 
shifted to 0.45V, and the rest of the range is reduced
to fit between 0 and 1volt.  No probolema.


> The integrator/BLM rates could then be modded to have
> the thing hit stoich based on EGOR... then maybe some
> more code to do full-tilt wide-band control. (Much
> work to do there)

There certainly is application work to do after the thing
is working.  

Thank you for the suggestions and ideas.

Bruce Roe

 
> Thanks for your work on this, Bruce and gang. The 
> ECM world is a much better place to live. :)
> 
> Later,
> 
> Dig
> turbodig at yahoo.com
> '91 Syclone, with Wide-Band Option.
----------------------------------------------------------------------------
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