Maximum matrix size for surface plot?

[thoult]

Morning,

I've streamed data from an experiment to TDMS, and would like to display it in DIAdem. I have M channels of N data points and would like to display them in a surface plot.

Usually, I'd expect to do the following:

- Build a X axis (Generate Numeric Channel - N points at dX interval)

- Build a Y axis (Generate Numeric Channel - M points at dY interval)

- Build a Z matrix (Convert Matrix from X axis, Y axis and the M channels of N data points)

I can then plot the X Y Z triplets to a surface plot.

Alternatively, I can build the X and Y arrays and plot X Y Z in matrix mode. This has the added benefit of not forcing interpolation, which *always* crashes

Now the kicker. For my dataset, M = 192, N = 290 000 and I'm unable to plot more than the first 20% of the data in X. I suspect that Diadem's limited to displaying 2^16 in surface plots, but no error is returned. Does anyone know if this is the case?

I can decimate my data to fit the limit, but it's never nice to have to throw away data!

(For clarification, the data is a raster scan of a sensor over a surface. I could drop the sampling rate and/or increase the speed over the surface, but that's a secondary concern)

[Brad_Turpin]

Wow,

If you broke that down into 200 x 200 windows, you'd still have over 10,000 windows without decimation.  That would be a LOT of windows to look through, say if the were arranged side to side on separate REPORT tabs.  Are you really interested in looking through all that data interactively, or are there events you'd ideally want to identify automatically and then have ahuman just interactively review the discovered events?

Brad Turpin

DIAdem Product Support Engineer

National Instruments

[thoult]

Actually, it's the creation of an intensity plot that I'm more interested in, for two reasons. I can split my data into smaller segments and then compare measured data against my CAD model, but it's also handy having the global intensity plot available for human overview. Amusingly, almost any way I try to split the data besides decimating means that the split between datasets will be in some region of interest, which makes it awkward for non programmatic review 🙂 Meanwhile, the problem with decimation is that I have four channels of data for each of those slices: channels 1 and 2 are analogue signals and averaged to give one plot, channel three is another analogue voltage that is plotted in another, but channel four is a digital waveform with some pretty short pulse widths. Decimate too much and you'll barely see any meaningful data when plotted on an intensity plot for this.

My instinct before starting was that DIAdem would handle the large dataset better than LabVIEW, which it did if I only wanted to plot YT data. However, after changing tack and heading back to LabVIEW I found that it was able to churn through things a bit better.

I'm happy to send an email to describe what I'm doing in a bit more detail, as well as provide some of the resulting plots, as it's a sensitive project. Understandably, even if I could upload the data for someone else to work on, I'm not sure how much use 1.7 Gb of TDMS is to anyone else! 🙂