Why does plotting nans in 3D cause problems with time and ability to save the plot?

The following script illustrates my problem - when iFloorRem=0 there is no problem with running the script and when it is finished I can save the graph as a *.eps file. When iFloorRem=1 then it takes significantly longer to run and then a long time to actually produce the graph. While that is happening Matlab is locked up. When the graph has been finally produced it takes an extremely long time to save it during which time Matlab is also locked up. For the problem that I am working on, which is more complicated than the script below, it takes so long that I have not been able to complete the save action. I have to use task manager to shut down Matlab to regain control of the computer. I have a Toshiba Qosmio with the I7 chip.

Here is the script:

% Example Script illustrating effect of plotting nans

close all

clear

iFloorRem=1;  % set this to 1 to remove the floor, otherwise set to 0

Fs=44100;  % sampling frequency

h=1/Fs;   % sampling interval

tMax=1;    % duration of sample

t=0:h:tMax;   % time grid

nt=length(t);

noKeys=50;   % number of sinusoids

y=zeros(noKeys,nt);  % reserve some room

f=2;  % initial frequency

delF=2;  % increment in frequency

%===generate the noKeys sinusoids

for k=1:noKeys

      y(k,1:nt)=sin(2*pi*t*f);

      f=f+delF;

end

if iFloorRem==1

      %=== remove floor

      [rr, cc]=size(y);

      II=y<0;

      y(II)=nan;

end

%===plot the sine waves in 3D

for k=1:noKeys

      kp=k*ones(nt,1);

      plot3(kp,t,y(k,:),'k')

      if k==1

          hold

      end

end

grid

title('A Bunch of Sinusoids')

zlabel('amplitude')

xlabel('sinusoid number')

ylabel('time (s)')

view(37,82)

[Walter Roberson]

Could you experiment with setting the locations to -inf instead of nan ?

[David Koenig]

Yes, I tried that suggestion and the performance was pretty much the same. Thanks anyway.