Coolant and Inlet Air Temps Transfer Required Accuracy??

[AbeFM]

Boy, hasn't this thread gone Off Topic! For the best. Now we have code to generate thermistor tables, a gui to make it easy, and a plan with regards doing the sampling and lookups. Sweet.

You're lucky I'm not a mod. I'd totally delete this thread it's so far off topic. It's supposed to be a discussion on the importance of maping the curve accuratley, and when to read it. This 'developing a windows GUI for calibrating sensors' stuff belongs in it's own thread, Mr.



(MAP sorta doesn't belong here either, but it's the trickeist. Some averaging is good even on it, but I would almost recommend asynchonos sampling for it, plus averaging. There are standing pressure waves all through the manifold. Where you take your signal, you're going to get (relative to a particular engine rotation/position) an over or under error in the pressure, and this modification will vary with RPM.

You can do this averaging mechanically, with a pressure resivoir, but I think it'd be better to read the map very often, purposefully irregularly, (or at least many times per piston stroke) and average to make something meaningful of it. My father once worked on a project for the dept of defense, and it was all centered around taking samples in a way that was random on the timescale of the events in question so as not to miss resonances or read incorrect data.

[Fred]

It's supposed to be a discussion on the importance of maping the curve accuratley, and when to read it.

Not the when at all, just how much accuracy is required in the output of the transfer. Which turned into the multiple ways of doing it and their weighted benefits and costs, and then into how to implement the best method :

This 'developing a windows GUI for calibrating sensors' stuff belongs in it's own thread, Mr. :-)

Cross platform GUI thanks!!

(MAP sorta doesn't belong here either, but it's the trickeist. Some averaging is good even on it, but I would almost recommend asynchonos sampling for it, plus averaging. There are standing pressure waves all through the manifold. Where you take your signal, you're going to get (relative to a particular engine rotation/position) an over or under error in the pressure, and this modification will vary with RPM.

It does depend on where you take your reading. However the main thing is that at any given RPM it will be consistent with respect to crank angle. Provided it is predictable and not noisy you have what you want.

Fred.

[MotoFab]

I see what you mean about off topic, it's understandable at this point though as many things are related to and predicated upon many other things. Every choice made now affects the possible choices later in the design.

To the thread topic of required accuracy, ±3°C seems adequate.

- Jim

[AbeFM]

Thank you. I agree that a 5 degree window is fine. Again, I'll check this as soon as I can.

As to the "well, measure it wrong and make up for it with tables", that might even kill something like a free-points-in-space approach. Well, maybe not. But, it means your table, instead of being flat, will have waves in it, and it's just a bad idea. You want something representative of the average pressure in the manifold as a whole! Then your table will talk about the resonances of the manifold as they relate to flow - not as they relate to skewing sensor readings.

[MotoFab]

Start with the desired accuracy window, and the temperature range, and work forward.

Say that a ±3°C window is the objective. The range is what? -20°C to 120°C maybe?

That is achievable with a table of 64 or so, and interpolation.

- Jim

[AbeFM]

ok.

Code: Select all

rpm         coolant(f)    MAT(f)     PulseWidth1(ms)   IgnAdvance
5992        145             40             4.752                     9.7
5992        144             40             4.752                     9.7
5992        142             40           ~4.750                     9.7
5992        140             40             4.750                     9.7
5992        135             40             4.750                     9.7
5992        120             40             4.752                     9.7
5992        155             40             4.750                     9.7
5992        215             40             4.752                     9.7

Ok, I'm calling mixture and timing *insensitive* to coolant temps. Clearly a 0.1 degree error is acceptable on an MS-IIx box unless you have warm up enrichment on. If you do, you're not beating the motor to the hairy edge of AFR's.

Let's look at MAT:

Code: Select all

rpm         coolant(f)    MAT(f)     PulseWidth1(ms)   IgnAdvance
5992        145             40             4.750                     9.7
5992        145             39             4.781                     9.7
5992        145             38             4.787                     9.7
5992        145             37             4.785                     9.7
5992        145             35             4.820                     9.7
5992        145             31             4.830                     9.7

5992        145             80             4.603                     9.7
5992        145             79             4.603                     9.7
5992        145             77             4.603                     9.7
5992        145             75             4.605                     9.7
5992        145             73             4.608                     9.7
5992        145             70             4.609                     9.7

5993        145             80             4.583                     9.7
5993        145             81             4.547                     9.7
5993        145             83             4.550                     9.7
5993        145             85             4.583                     9.7
5993        145             88             4.570                     9.7
5993        145             91             4.573                     9.7

Ok, MAT has an effect. It's hard to tell, as RPM drifts, it moves a lot. But the simple thing seems to be that a 10*f change in temp was maybe a 0.15% change in fuel, no change in spark. I was certainly in the noise of the fuel output. Certainly 10*f (call it 5* c) was less than one percent error. Without a better understanding of how coolant and air temp effect fuel delivery, I would say a +/-5*c error is fine. How does that compare to an 8x1 table representation of the curve? I don't know just now. But it certainly motivates a revisit if you're looking to save space and time.

[MotoFab]

If you feel that interpolation consumes too much processor power, some of the processing can be done in the PC before the chip is programmed.

Instead of point values, program the chip with a table of the slope between points.

- Jim

Edit

[MotoFab]

Ok, I'm calling mixture and timing *insensitive* to coolant temps. Clearly a 0.1 degree error is acceptable on an MS-IIx box unless you have warm up enrichment on. If you do, you're not beating the motor to the hairy edge of AFR's.

Without a better understanding of how coolant and air temp effect fuel delivery, I would say a +/-5*c error is fine.

CLT usually represents port wall temp, and as such has a large influence on PW as it relates to port wall fuel and evaporation. I'd have to go with ±5°C as an acceptable overall accuracy. And certainly it's achievable.

- Jim

[AbeFM]

If you feel that interpolation consumes too much processor power, some of the processing can be done in the PC before the chip is programmed.

Instead of point values, program the chip with a table of the slope between points.

- Jim

Edit

Oooo - I like that!

[Fred]

Abe, your graph is invalid because it's your tune, not the possible tune situation. IE, you might pull a bunch of timing with increasing CHT temp etc. You should be pulling a bunch of timing with IAT at some point, but either you aren't, or ms doesn't let you etc. ie, your tune may be sub optimal ;-)

This is the result of trying to run the app on windoze :



Does anyone know if it is possible to get hold of a glade-sharp installer? Or how else it is supposed to work in a true "just run it" cross platform way?

Fred.