"too many points" error - FFT 1

[SomeYahoo]

I have a rather large (110K points) set of acceleration data that I must analyze in Mathcad Plus 6.0. I have managed everything up to plotting the FFT data in frequency domain. I can plot with a simple n=1..32768 for the x axis, but if I try to use the frequency data (n=.004,.008..128) for the x-axis I get a "too many points" error.

Any thoughts? Is .004 just too small to be on a log-scale x-axis?

Thanks in advance for the help!

Note: numbers rounded to protect the significant

 

[GregLocock]

I think that's the graphics engine having a whine. You are only trying to plot 4.5 decades. Try creating a new x axis with log(x) and see if that works.







Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[SomeYahoo]

Hope I understood your suggestion correctly:

I set up a new variable d := log(.004), log(.008)..log(128) and attempted to plot with d on the x-axis. I had no success plotting (on a standard scale); same "too many points" error.

Any other suggestions?

 

[jghrist]

Could it be that you have 32768 points and the range n=.004,.008...128 has only 32000 steps? log(.004),log(.008)...log(128) is only 7 steps.

 

[SomeYahoo]

I'm showing the rounded versions here, actual numbers have that many data points. Shouldn't matter either way though, MathCAD will plot the number of matching points that it has.

I'm confused on how "log(.004),log(.008)...log(128) is only 7 steps." I get -2.398, -2.097, -1.796... 2.107.

 

[jghrist]

I agree that it shouldn't matter.
I divided by the wrong number - should be 16 steps.

 

[pstuckey]

On the graph itself you can divide (or multiply) n to rescale the x data points. So if it plots OK with n:=1...32768, and [n] in the x placeholder, it will also plot with [n/256] in the x placeholder, but the xaxis will now be scaled by 256.

Another approach is to just plot two vectors. I got in the habit of doing that when I used Axum for my plots. In this case you would use your range variable to make a vector with you FFT amplitude, and one with your frequency (for instance: frequency[n-1 := n/256 (assuming ORIGIN=0)). Then you could plot those two vectors (no subscripts).

Peter

 

[SomeYahoo]

Thank you pstuckey! A star for you!

I used the n/256 on the x-axis for now, but will look into using vectors in the future.

Now if I could just smooth the plots I'd be all set... but that's a new thread if I can't figure it out.

 

[IRstuff]

The plot function has a limit of 500,000 points

TTFN