| Author |
 Topic  |
|
|
dldudley
10 Posts |
 Posted - 02 Jun 2010 : 13:42:12
|
One thing.... I'm asking a whole lot of questions..... and you always have good answers....;-)
Ok, now I've got a chart on the display with Logarithmic scales (not sure I care for the "linear scaling between decade intervals", really wanted a true "logarithmic" scale).
I have RT_AUTOSCALE_Y_MINMAX enabled, and on the NumericAxisLabels, I have setAxisLabelsDecimalPos set to 3. Initially, the graph starts out OK, but when the autoscale rescales the graph, all the decimal places disappear.
Sure I've just forgotten something... |
|
|
quinncurtis
1586 Posts |
Posted - 02 Jun 2010 : 14:02:47
|
We can't really follow what you describe. In particular, we don't know what you mean by "linear scaling between decade intervals". If you are using LogAxis as the axis for the scale dimension (x or y) set to log scaling, there is no linear scaling for that dimension. See our log scale examples (Charting Tools example LogPlots) and describe where you see linear scaling in a log axis.
Are you a commercial user of ours or are you using a trial version of the software? |
 |
|
|
dldudley
10 Posts |
Posted - 02 Jun 2010 : 14:41:56
|
I am a commercial user, previously with the City of Corpus Christi, since promoted to Brookhaven National Labs.
I eventually want to purchase a commercial license for the lab, but for right now, I have to prove the capabilities of the software to get it approved.
Looking at Page 133 of the QCChart2D manual, it states:
Use the LogAxis class to create a logarithmic axis with logarithmic spacing between the major tick marks (1, 10, 100…), and linear spacing (2, 3, 4, 5…) between the minor tick marks
On that point, I was just making a comment that that's not Logarithmic scaling, from a scientific point of view, that's "decade" scaling. But I'm not particularly concerned by that...
I'm drawing a historical trending chart. The chart has Time as the X axis, and Radiation intensity has the Y axis. I want to autoscale the Y axis based on the past level of radiation.
I'm showing 3 decimal places, with a 0-1 initial range. This displays with no problem, up until the "autoscale" begins. Because most of the time, the radiation levels I'm reading are below 0.1, the scale automatically changes. That's great.
Problem is, once it scales, it removes the 3 decimal places, and all I see as a scale is "0", where I would expec to see "0.001", "0.010", and "0.100". I don't see ANY decimal places, or the period.
|
|
|
|
quinncurtis
1586 Posts |
Posted - 02 Jun 2010 : 15:34:48
|
It sounds like you need to get your own copy, since we list the City of Corpus Christi as the owner of the software you are using.
Our log charts are scaled properly, and in accordance with scientific use. The spacing between the minor tick marks is linear in the scaled physical units (i.e. 1, 2, 3, 4...9), but it is logarithmic in the actual display on the screen, as seen in all of our log graph examples.
You should not be scaling the y-minimum value to 0.0. That is an invalid value for a log plot. Scale it to the lower decade of the range you want to display, i.e. 0.001.
You have identified bug in the real-time auto-scaling of the log graphs when used in the scrolling plots which prevents the decimal precision of the log axis labels from updating properly. We will fix this and update the software within the next couple of weeks.
 |
|
|
|
dldudley
10 Posts |
Posted - 03 Jun 2010 : 08:36:18
|
I know I'm going to eventually get a copy for the Lab. The package I'm using I bought for the City, while I was there. I was the only person that used it, and the programs written with it were dropped as soon as I left (for lack of someone to support them, since I was the author of them). I've been developing programs here with this copy in order to demonstrate the capability of the software, in order to prove a need to purchase it.
The Lab currently uses JFreeChart, which is useful, but not quite as robust as the QC tools for doing RT displays. Having used the QC software previously, I've been working on demonstrations to show the advantages of the QC Toolkit.
----------
As I said previously, I forgot something. I didn't do setTimeScaleTransforms on the TimeCoordinates transform I was using. Now, things look great, with the exception of the auto-scale bug that causes the labels to loose their significance.
If everything works out properly, I should be working on a purchase for the Lab's use very soon.
|
|
|
|
dldudley
10 Posts |
Posted - 03 Jun 2010 : 09:47:06
|
Is it listed anywhere in the manual, that a LogAxis shouldn't have a Y-minimum of 0?
I've been chasing a problem with spurious readings for a while, which I didn't understand. Turns out, that if the value in the ProcessVar that you're charting with a LogAxis goes to zero, you get an invalid location (usually a 1) on the graph. Forcing a zero value to become a tiny value (I used 0.00001) straightened out the display.
Are their any plans on being able to use a zero minimum for the Y-minimum?
|
|
|
|
quinncurtis
1586 Posts |
Posted - 16 Jul 2010 : 17:13:00
|
| A value of 0 for a point on a log axis is mathematically invalid. A log axis will plot the value of log(y) where y is the current value. The value of log(0) is undefined and cannot be displayed. If you are generating those values it shows a flaw in your data collection. You should clamp your values to a non-zero range, or not use a log scale. |
|
|
| |
Topic |
|
|
|
AutoScaling a chart and keeping decimal places
