Re: [gpm] History graphs - issues and possible improvements

[Richard Hughes <hughsient gmail com>]

Sorry for the delay, been quite a busy few weeks.

On Mon, 2007-01-08 at 21:28 -0500, Phil Sherman wrote:
> A recent upgrade to FC5 and a failing battery have encouraged me to
> attempt to use the history graphs to determine what is happening to the
> battery. The graphs are difficult to interpret because the X axis
> labeling is confusing. All three graphs appear to use the same scheme to
> show a time value. 

Agree.

> My interpretation of the X axis is that the the graph shows the most
> recent 10 minutes of power information. The x axis zero point appears to
> be the last power on time. Graph data propagates from left to right with
> the most recent information located at the right end of the data. Once
> the data reaches the right side of the graph, additional data pushes all
> plotted data to the left by half the graph width then new data continues
> to grow the graph to the right. I was unable to discover any information
> about this in the documentation. The location of the zero point also
> requires continuing updates to the x axis labels.

Well, it grows from the left to the right, until it no longer fits, and
then it re-plots the data squashed by about half. The procedure is
repeated, and redundant data is freed when precision is no longer
required.

> These are very difficult graphs to interpret and don't provide enough
> data to fully evaluate a battery. The graph would be easier to interpret
> if the new data were inserted at one side of the graph; forcing all
> prior data to move to the opposite side. Unfortunately, this requires
> significantly more overhead to maintain than the current scheme. It
> should also require significant changes to the existing code.

Pahh, I'm not worried about changes to code :-)

> There is, however a different display scheme that would improve the
> readability and should use almost all of the existing code. This format
> would do the following:
> 1. Divide the graph horizontally into two sections. The left hand side
> would represent history, the right hand side would represent the future.
> The dividing point would be labeled with the time of the past/present
> division.

Hmm, isn't the future data from a few minutes ago just the past data?

> 2. Label the x axis with positive or negative displacements from the
> division time or use that information to compute the actual time for
> labels. Displacements don't need to be updated while time values are
> easier to interpret. Labeling can be simplified by using integer values
> and an initial displacement for the first data point to be plotted.
> (ie.If the division time were 8:01 and recording started at 8:01:30 then
> the first data point would be displaced 0.5 minute on the future side.)
> 3. An interesting option of this is to display history in hours and
> future time in minutes. This would show the last five hours of history
> with the most recent zero to five minutes shown in fine detail.

Sorry, I don't really understand why we need a past-future split... when
does the future become the past?

> Keeping four to 10 hours of history would provide a full log of a charge
> or discharge cycle.

Wow. 10 hours is a lot of data.

Can you provide more details to your past/future idea? PNG images
knocked up in GIMP might be a better way to share your ideas.

Thanks.

Richard.