ULV Pictorial Development Diary (56k die)

[Fred]

Finally, ULV has been in real world use on my machine for some days working with real files for real reasons. Here's the original uncropped screen shot of the 4 cylinder engine using J Series decoder. Uploaded May 4th, yesterday!

[Fred]

Tonight I thought I'd play with windowing functions and added three basic types. First pic is 5 sample size, second and third are 25 sample size. Third illustrates the step response of each. Note triangle smashes flat in both step and smoothness. Flat sucks.

[Fred]

To steal a term from Ben, I've been "dogfooding" all day, and it's been a great success! It's kinda crappy to use in some ways, and obviously unfinished in others, but by using it you get a real feel for what's most broken, and what to fix next. Here's a shot taken "just now" of a log loaded some hours ago while debugging a decoder change I was making:

[Fred]

The min/max display was annoying me a bit, so I added a quick control for it! Finally, at last, the "control panel" is actually a control panel, at least, a little bit :-) Taken "just now":

[Fred]

I decided to do something easy this morning, after only 4 hours sleep... so I added fill mode to the min/max rendering. It makes this log of Sim's RPM vs Battery Voltage pretty obvious. At lower RPM there is a lot less noise. Uploaded "just now":

[Fred]

Now controllable, and with distinct edge and fill colours:

[Fred]

Because we need some good data for TPS based transient enrichments, I thought I'd work on an automatic calibration function to assist Andy in getting his tune SOLID ASAP. The first step is obviously to be able to display the VE table before and after modification, and to represent a 3D table in memory. So I completed the core table drawing part today, and will try to get it more functional with axes and various other stuff pretty soon. Eventually the colour scale will be fully configurable, however in the short term I plan to provide saturation adjustment on a slider so that the arty types don't whinge. As per usual, development takes a few steps. The first image was the first time I even tried to execute it (except unit tests, of course), and I had a loop condition slightly wrong. The latter three are with a better colour generation setup, hard coded in for now.

[Fred]

Simple histograms are now real, except it's always RPM, if available, or the first record if not. Colour is cycled from a list. Obviously needs bin range labels and scale and counts and title and so forth. But if you know what it is, then the shape is all that really matters.



Edit: The title was hard coded in the previous "implementation", now corrected to be real.

Edit2: My theories on what you're looking at:

the highest bar is idle
the other bump that is half bell-curve is likely "cruise"
the flat bit "between idle and cruise" and the taper off "enthusiasm"
that bump below idle is the "nearly stall" bit, I think
i think this log started while idling
could be wrong
if wrong, the first bit includes cranking, AND the near stalls
no way could the first bin be that big without the near stalls IMO

[Fred]

A few subtle enhancements to it, ie, wider bars, configurable background, 50 bins instead of 20, reasonable min size, etc.

[Fred]

Had to lose a 2 week ulv-uptime. Running the same version for all of my log viewing needs for 2 whole weeks. Killed it so I could change a few things. Fixed a little bug in the window menu stuff, unnoticed by anyone else, annoying to me. Histogram now has selectable field and bin count: