ULV Pictorial Development Diary (56k die)

[Fred]

It's a lot less useful with other patterned types of data, though, as shown here, and uploaded 20th April 2013. Clearly for this to shine some real data was needed; more on that later.

[Fred]

Up until now all of the trace adding had been done on an individual basis. There was no log object, all smoke and mirrors, enough to test with, enough to grow from. But it wasn't enough. It was time to do a first cut of the trace adding/removing/management dialogue. Although hideously ugly, it's fairly efficient to use as you can control click and shift click to select multiple traces. The coloured ones are already in the graph. On the left means staying in the graph, or if not coloured, about to be added. On the right means available, or if coloured, to be removed. The worst part is the way it's laid out and displayed. I need to improve that. This one was uploaded April 22nd 2013:

[Fred]

The next update to this was effectively a bug fix, though I'd not thought of the behaviour in the first place, so it was an enhancement of sorts, too. And that was to exclude already added traces from the right hand side list. Here it is, all traces in use, uploaded Apr 23 2013:

[Fred]

I decided it was time, so I pulled in my CODEC library that I wrote some months ago, and wired it up to the log container in a simple "proof of concept" kind of way. Bingo! Real data, on screen for the first time, uploaded Apr 23 2013:

[Fred]

I was a bit concerned about the obscuring of traces by the min/max lines, however I came across a file that had a clean trace in it, and that's shown below. The reality is that FreeEMS logs are quite noisy right now. The point to note here is that this shows that noise in full scale, so it'll be a useful tool to eliminate it. Not that I need this, I know what's wrong/missing, but in converse, it'll highlight when a user has a problem, as the traces will look bad if noisy and good if not. Uploaded Apr 24 2013:

[Fred]

I was still excited about the min/max/mean stuff for a few days, this picture, entitled "ULV.your.data.like.youve.never.seen.it.before.png" was uploaded Apr 25 2013:

[Fred]

About this time I remembered that the displayed values were far from meaningful for non zoomed in data. On principal, and because it wasn't and isn't a priority, I simply crossed out the values if zoomed out. Image title "ULV.never.lie.to.your.users.png", uploaded Apr 25 2013:

[Fred]

Then I was bored one night, and in 3.5 hours, added a scatter plot view, I'll split the images over a few posts, these ones are from Apr 27:

ULV.first.real.scatter.plot.png shows the interference patterns present from granularity of data combined with pixel spacing and no averaging:




ULV.second.real.scatter.plot.png




ULV.third.real.scatter.plot.png




ULV.100.point.moving.average.png 100 points of averaging is too much, and the weighting is currently even, but all the same, it's an impressive picture:

[Fred]

This is battery state. The cranking, firing, ramp up of RPM, excitation of alternator and subsequent ramp up of voltage are all clearly visible here. With a smoother regulator, the band across the top would be thinner, but it's from an old Volvo, so we can't expect too much. Uploaded 3rd of May:

[Fred]

Why I didn't bother writing the scatter plot until after I had real data loading completed and working. IE, me bored and experimenting with random and generated data in the scatter plot, uploaded may 3rd 2013: