How to improve performance on CNiGraph when plotting realtime XY data at input rate of app. 30Hz and maintain a chart history of 20000 points visible.

[mko]

I've used ChartXY method in order to plot 3 traces at a rate of 10 samples/second/trace on a CNiGraph. I've used a fixed chart history of 20000 points in order to have all points visible in the chart. The ChartXY is done on point to point basis(ChartXY(double,double)) and is executed approx. 30 times /sec.
Autoscaling is off, and the only manipulation of the graph is the possibility to select chart X and Y maximum values.
I know that the ChartXY is only to be used in "slow processes" and thats probably why I end up with an increasing cpu usage on my GUI thread.
I've also decreased the data input rate
to 1/3 or 10 Hz and the GUI threads CPU usage of course decreased, but it means that 1/3 of the data point would be lost.
If I use PlotXY method in point to point basis I loose all previous points or if I use a Vector instead of that, I end up plotting the entire Vector at each occasion when a new data sample arrives?
I suppose there are many ways to solve this, which would be the most effective?

OS:NT4.0
Cpq AP550 733MHz

[Chris Matthews]

What takes most of the time is the redraw of the graph area. What you should do is chart all the data, except in less function calls to ChartXY. You can do this by charting multiple points at a time.

First, you need to know if the X values for all the plots are the same. In other words, do all three plots have Y values corresponding to the same X values. If so, that means you can plot the lines together.

Instead of using CNiGraph::ChartXY(double x, double y), using CNiGraph::ChartXY(CNiMatrix& xy). The matrix will be one row of X values and 1 or more rows of Y values.

So as data comes in, don't plot each point immediately. Buffer a few points, say 5 or so, together instead. So let's say you just got three new data points, one for each plot. Instead
of charting them immediately, put them into a CNiMatrix as:

CNiMatrix data(4, 5);

data(0, i) = xval;
data(1, i) = yval1;
data(2, i) = yval2;
data(3, i) = yval3;

Then do this 4 more times incrementing i with the next 3 data points for each plot. Then make one call to m_graph.ChartXY(data) and reset i to 0. This will combine 15 of your current ChartXY calls into one call every 1/2 second and you will still be charting all of your data.

If your X values are actually different for each plot, then you won't be able to chart multiple plots at once, but you can still buffer the data points befor plotting.

Best Regards,

Chris Matthews
Measurement Studio Support Manager