« Demolishing "scientific concensus" | Main | House temperature analysis »


DPlot program

Every once in a while I will buy some software and the DPlot program that I purchased about a month ago is one of the best pieces of software that I've come across in decades. It's well worth the $195 US price tag and comes with very reasonable license terms -- install it on as many of your computers as you like if you're the only one using it. It's available for online purchase at DPlot.com and there is a free trial version.

I spend a lot of time crunching data from my various personal research projects and the greatest bottleneck that I've run into is creating graphs from the data. What I've been using, when I need a nice graph, is CricketGraph for the Mac. To run this program I have to launch my Mac emulatorBasiliskII, get the data transferred to a "disk" in the emulator, run CricketGraph, paste the resultant PICT file into M$ Word for the Mac and then export it as an rtf file which then needs to be copied from the Mac emulator "disk" drive and pasted into Open office. Needless to say, this is an incredibly tedious way to creating graphs and one of the reasons I'm constantly writing crude data display programs which are fine to look at data but leave much to be desired aesthetically.

My first attempt to deal with this problem was to use the MSChart control which comes with VB. For anyone who's thinking of doing this my advice is DON'T!!!! MSChart is an incredibly complex control which, after one has learned the arcane details of how to program it, doesn't allow you to print what you've created on the screen!!! This control is one of the worst POS that I've encountered from M$.

Anthony Watts put up a rave review about DPlot on WUWT in December 2010 and I was intrigued enough to download the demo version. Once I started using it I realized that I finally had something almost as good as CricketGraph. I say almost as good as one can't shift data up and copy it to another column to create Poincare plots like CricketGraph can do, but OTOH, DPlot will do FFT's on data and produces incredibly nice looking graphs. Now I can spend my time writing fast programs to manipulate data and, while I haven't tried it yet, DPlot is scriptable from VB. DPlot easily handled a 24 hour 3-axis accelerometry graph with over 3 million points which I found very impressive.

One application of DPlot I did today was to calibrate the USB temperature monitors which I bought from the Weather Shop (link at WUWT). These are really neat units that sample temperatures from once every 10 seconds to every hour or so up to 16383 readings. 2 of the 3 original monitors I bought 2 years ago still work (the onlly complaint I have about them is that they use a very hard to find 3.6 V Li battery)

Last night did a calibration of monitors 2 and 3 and results are quite impressive. Monitor 3 has some electrical tape and wire around the rear end and the effects of this on the monitor thermal parameters are seen during calibration process!

The monitors were both restarted after download of data and, after a short period at downstairs room temperature, were placed in downstairs fridge freezer for about 1 hour. Sampling frequency was 1/minute. After being removed from freezer, I had the bright idea of putting them on Athena keyboard as this is above room temperature. This was a bad idea as the keyboard is not at a uniform temperature as there was 4 F degree difference between the two monitors when they were read out. After spending about 5 hours on keyboard, they were placed on my table behind futon which is not heated by any computers. They were left there all night.

Since the keyboard was a non-uniform temperature, this section of data was deleted from the data finally used for calibration. There was hysteresis noted when USB3 was plotted against USB2 and will have to re-run the calibration to see if this was due to the electrical tape on USB3. There is some deviation at the high end of the range, but the least squares fit equation was:

USB3 = 0.998*USB2 + 0.321

Correlation coefficient was 0.9989 for R^2 of 0.9978

The two monitors are highly correlated as expected but I find it hard to believe how close the two units are to each other. I would have been ecstatic if I'd had a calibration curve this good when I was building my own DAS's in 1982. For all practical purposes, the temperatures are identical.

Standard error about the line is 0.558 so every temperature reading is +/- 0.56 F degrees. This is about what I'd expect from a thermometer with 1 degree F resolution. Considering that each unit was only $60 US this is very impressive and demonstrates how good off the shelf electronics have become in the last 30 years.

Calibration temperature range was from 13 to 84 F degrees and should try to get to 120 F to see if there are any non-linearities at the extremes.

The hysteresis indicates that the thermal time constant of the thermometers is >1 minute and have to keep this in mind when using them to monitor temperatures. There's not much point in using the 10 second resolution of the thermometer except for rapidly varying temperatures (like hanging from air-conditioner vent) where one doesn't get an absolute temperature but rather a partially integrated temperature. One would need a much faster thermal time constant to sample this, eg my thermocouple on DMM.

And here's the calibration graph described above.


Posted by Boris Gimbarzevsky at 23:34.07
Edited on: 24/01/2011 1:03.37
Categories: Computers