diff --git a/contrastBeforeAfterAmplitude.m b/contrastBeforeAfterAmplitude.m
index 81b1b0d1680dbcebf3c70eda38914b51670ba698..14aeb1b63621ac9bde8c79fced51390317814443 100644
--- a/contrastBeforeAfterAmplitude.m
+++ b/contrastBeforeAfterAmplitude.m
@@ -13,19 +13,29 @@ afterRaw = load('S6H1_Hann_40_2000feedforward_bins.txt');
% Our object is to average all the calibrated amplitude rows corresponding
% to a given SFT number.
-beforeL = unique(beforeRaw(:,1));
-afterL = unique(afterRaw(:,1));
+% The text file is zero-indexed
+beforeL = sort(unique(beforeRaw(:,1)));
+afterL = sort(unique(afterRaw(:,1)));
before = ones(size(beforeL));
after = ones(size(afterL));
+
+
% Average over all the bins corresponding to a given SFT number,
% then multiply by sqrt(8/3) because of the Hann window normalization factor.
+% Have to shift index on the right-hand side, to compensate for
+% Matlab's one-indexing vs the C-generated text file's zero-indexing
for ii = 1:length(before)
- before(ii) = sqrt(8/3) * mean(beforeRaw(beforeRaw(:,1) == ii, 4));
+ before(ii) = sqrt(8/3) * mean(beforeRaw(beforeRaw(:,1) == ii - 1, 4));
end
for ii = 1:length(after)
- after(ii) = sqrt(8/3) * mean(afterRaw(afterRaw(:,1) == ii, 4));
+ after(ii) = sqrt(8/3) * mean(afterRaw(afterRaw(:,1) == ii - 1, 4));
end
+% Strip any NaN -- but there should not be any
+before(isnan(before)) = [];
+after(isnan(after)) = [];
+clear beforeRaw
+clear afterRaw
figure(1)
semilogy(beforeL, before,afterL, after)
@@ -43,6 +53,7 @@ differenceClean = difference;
differenceClean(isnan(difference)) = [];
arithmean = mean(differenceClean)*ones(size(difference));
arithmeanString = horzcat('Arithmetic mean of difference: ', num2str(arithmean(1)));
+length(difference)
figure(2)
plot(beforeL, difference, beforeL, arithmean)
@@ -57,3 +68,30 @@ ylim([-1e-24 1e-24])
print('-dpdf', 'HannAmpDiff.pdf')
print('-dpng', 'HannAmpDiff.png')
close(2)
+
+% Try some statistics
+HoftHistVector = 1e-24*(50:100);
+beforeHist = hist(before, HoftHistVector);
+afterHist = hist(after, HoftHistVector);
+arithmeanBefore = mean(before);
+arithmeanAfter = mean(after);
+arithmeanDifference = arithmeanBefore - arithmeanAfter;
+harmmeanBefore = harmmean(before);
+harmmeanAfter = harmmean(after);
+harmmeanDifference = harmmeanBefore - harmmeanAfter;
+harmmeanString = 'Harmonic mean: before, after, difference';
+
+figure(3)
+plot(HoftHistVector, beforeHist, HoftHistVector, afterHist)
+xlabel('Hoft')
+ylabel('Histogram count')
+legend('Before feedforward', 'After feedforward')
+title({'Hann windowed calibrated amplitude histogram';...
+ harmmeanString;...
+ num2str(harmmeanBefore);
+ num2str(harmmeanAfter);
+ num2str(harmmeanDifference)})
+grid on
+print('-dpdf', 'HannHist.pdf')
+print('-dpng', 'HannHist.png')
+close(3)
diff --git a/fullspectBoth.m b/fullspectBoth.m
index 23b099f7c26d93548c687926d224735152d6e9c2..f7b5057b8a696b039b94316bb741ac9a2d6bbc14 100644
--- a/fullspectBoth.m
+++ b/fullspectBoth.m
@@ -5,7 +5,7 @@ hold off
runlist = ['S6'];
run = ['S6'];
%windowlist = ['Hann ';'Tukey'];
-windowlist = 'Tukey';
+windowlist = 'Hann';
%ifolist = ['H1'; 'L1'];
ifolist = ['H1'];
@@ -54,10 +54,10 @@ irun = 1;
color = strtrim(plotcolor(iifo,:));
fnameroot = sprintf('%s%s',run,ifo);
fname1 = sprintf('%s_%s_40_2000test.txt',fnameroot,strtrim(windowlist(irun,:)));
- fnameFull1 = strcat('~gmeadors/2012/10/24/AMPS/', fname1);
+ fnameFull1 = strcat('~gmeadors/2012/11/21/AMPS/', fname1);
data1 = load(fnameFull1);
fname2 = sprintf('%s_%s_40_2000feedforward.txt',fnameroot,strtrim(windowlist(irun,:)));
- fnameFull2 = strcat('~gmeadors/2012/10/24/AMPS/', fname2);
+ fnameFull2 = strcat('~gmeadors/2012/11/21/AMPS/', fname2);
data2 = load(fnameFull2);
% Try resampling to smooth random variation.
% This combines neighboring bins.
@@ -69,11 +69,12 @@ irun = 1;
resampleFilterSize = 225;
% Theoretically, the frequency array is linear so it could be resampled
% using a less sophisticated algorithm. Yet why not try out resample here?
+ % Multiply by sqrt(8/3) for the Hann window normalization
freq{irun,iifo} = resample(data1(:,1), 1,...
resampleFactor, resampleFilterSize);
- amppsd{irun,iifo} = resample(data1(:,5), 1,...
+ amppsd{irun,iifo} = sqrt(8/3)*resample(data1(:,5), 1,...
resampleFactor, resampleFilterSize);
- amppsdwt{irun,iifo} = resample(data2(:,5), 1,...
+ amppsdwt{irun,iifo} = sqrt(8/3)*resample(data2(:,5), 1,...
resampleFactor, resampleFilterSize);
sprintf('Looping over single-IFO bands...')
for iband = 1:length(bandlolist)
diff --git a/fullspectDiff.m b/fullspectDiff.m
index edf5d507c201b0b678469f2102202d6c52dc9ec0..aa51f916e3a24f1aa4eee4ef9271d0541e4e0304 100644
--- a/fullspectDiff.m
+++ b/fullspectDiff.m
@@ -5,7 +5,7 @@ hold off
runlist = ['S6'];
run = ['S6'];
%windowlist = ['Hann ';'Tukey'];
-windowlist = 'Tukey';
+windowlist = 'Hann';
%ifolist = ['H1'; 'L1'];
ifolist = ['H1'];
@@ -54,10 +54,10 @@ irun = 1;
color = strtrim(plotcolor(iifo,:));
fnameroot = sprintf('%s%s',run,ifo);
fname1 = sprintf('%s_%s_40_2000test.txt',fnameroot,strtrim(windowlist(irun,:)));
- fnameFull1 = strcat('~gmeadors/2012/06/29/AMPS/', fname1);
+ fnameFull1 = strcat('~gmeadors/2012/11/21/AMPS/', fname1);
data1 = load(fnameFull1);
fname2 = sprintf('%s_%s_40_2000feedforward.txt',fnameroot,strtrim(windowlist(irun,:)));
- fnameFull2 = strcat('~gmeadors/2012/06/29/AMPS/', fname2);
+ fnameFull2 = strcat('~gmeadors/2012/11/21/AMPS/', fname2);
data2 = load(fnameFull2);
% Try resampling to smooth random variation.
% This combines neighboring bins.
@@ -69,11 +69,12 @@ irun = 1;
resampleFilterSize = 900;
% Theoretically, the frequency array is linear so it could be resampled
% using a less sophisticated algorithm. Yet why not try out resample here?
+ % Because we are using Hann windows, multiply by sqrt(8/3)
freq{irun,iifo} = resample(data1(:,1), 1,...
resampleFactor, resampleFilterSize);
- amppsd{irun,iifo} = resample(data1(:,3) - data2(:,3), 1,...
+ amppsd{irun,iifo} = sqrt(8/3)*resample(data1(:,3) - data2(:,3), 1,...
resampleFactor, resampleFilterSize);
- amppsdwt{irun,iifo} = resample(data1(:,5) - data2(:,5), 1,...
+ amppsdwt{irun,iifo} = sqrt(8/3)*resample(data1(:,5) - data2(:,5), 1,...
resampleFactor, resampleFilterSize);
sprintf('Looping over single-IFO bands...')
for iband = 1:length(bandlolist)
@@ -99,7 +100,7 @@ irun = 1;
%plot(freqdesign,sensdesign,'color','blue');
grid
legend('Average amplitude PSD','Weighted average amplitude PSD','Location','North')
- titlestr = sprintf('%s %s Average Spectra (%d-%d Hz)',run,ifo,round(bandlo),round(bandhi));
+ titlestr = sprintf('%s %s Average Spectra, Before - After (%d-%d Hz)',run,ifo,round(bandlo),round(bandhi));
title(titlestr)
fnamepdf = sprintf('%s_%s.pdf',fnameroot,bandname)
print('-dpdf',fnamepdf)