Desktop Recorder

[agraham]

I'm afraid there is something wrong. If Ttransform = 2 then "3 Sines" transforms and inverses fine but "3 Cosines" does not retrurn the correct time series after the IFFT. I don't know where the problem lies. I thought that it worked fine on your previous version but unfortunately I can't check it as the latest version overwrote that previous version in my project folder and I can't download it again as you have removed it. Please can you have a look if you still have a previous version available?

[redbird]

Originally Posted by klaus

How do you like the attached sample program ?
Tachometer with moving needle.

Well, after a bit of playing around, I came to this user-interface with a moving needle, and without much effort, thanks to your code.
I should be ready in a few days to post the final program, once I did some more testing. Your name ("Klaus") is in the credit list, if you don't mind. I would even put your real/complete name with pleasure, but that would be rather indiscrete to ask, I think.

[agraham]

@Klaus - I found the error, it was a silly typo that was so obvious I couldn't see it at first

EDIT :- It wasn't that. I'm presently totally baffled

[klaus]

Hi Andrew,

Unfortunately I don't have the previous versions either.

I don't rember having seen this before.
The CalcInvFFT function has not been changed in the FFTdemo program in the last versions.

I had a closer look to undestand what happens.
I modified the FFTdemo program to:
- set the second half of the real and imaginary part in B4PPC
- and calculate the InvFFT with FFT.Inverse

When you run the test program it will show a single cosine.
Click on FFT to show the first half of the FFT signals
Click on CalcInvFFT we get what we see in the FFTdemo program
Click on CFFT and FFT to calculate a new FFT with the inverse time signals with Transform1
We see that
sample 0 has a value of 2 instead of 0
sample 256 has a value of -2 instead of 0.
I don't know where these values do come from.

Click on CSetFFT the program adds the second half of the FFT signal
and sets DataReal(0)=2 and DataReal(256)=-2
Click on C Inv FFT you see the same as what we get with the FFTdemo program.

Then change lines 940 -943
DataReal(256)=0
DataImag(256)=0
' DataReal(0)=2
' DataReal(256)=-2

Repeat the same as above, the inverse transform works.

Best regards.

[mjcoon]

Originally Posted by klaus

Unfortunately I don't have the previous versions either.

Not sure what versions you want but I downloaded and still have FFT zips 1.1, 1.6 and 1.7...

Mike.

[klaus]

If you still have versions 1.6 and 1.7 of the demo programs could you please post them.
If you mean the FFT library versions I have them all.

Thank you and Best regards.

[mjcoon]

Originally Posted by klaus

If you still have versions 1.6 and 1.7 of the demo programs could you please post them.

No, sorry, I only have FFTDemoKC1.zip.

Mike.

[agraham]

Well spotted Kalus. "DataReal(n/2)=0" was the problem.

When rebuilding the complex conjugate part I earlier said in post #56 "This leaves no value to be relected into index 256 so I have left it as zero." The implicit assumtion was that the data array was initialised to zero but as you are reusing the array there is an unwanted value in there. Fixed dll attached together with almost final help that incorporates your description of the demo which I have taken the liberty of slightly amending to more idiomatic English - hope you don't mind!

Originally Posted by klaus

If you still have versions 1.6 and 1.7 of the demo programs could you please post them.

I'm afraid that I only have the latest as the later ones overwrote the earlier ones.

If you mean the FFT library versions I have them all.

So do I, and all the sources. My version control system is pretty strict - at least for the stuff I generate myself!

We are nearly there! A couple of comments about the demo.

1) Why not make the "Calc Inv FFT" button visible at the beginning? Once it is visible it stays visible anyway.
2) "Click on one of the sine buttons to continue" implies you can't click the cosines button! How about "signal buttons"?
3) The "Inverse FFT Imaginary" graph legend is wrapped as the label is too short.

[klaus]

Hi Andrew,

When rebuilding the complex conjugate part I earlier said in post #56 "This leaves no value to be relected into index 256 so ...

I remembered this post, and as the CalcInvFFT routine in the FFTdemo program was not changed I was looking at the second half of the FFT samples.

I have taken the liberty of slightly amending to more idiomatic English - hope you don't mind!

No problem, I'm glad that you do it.

Your suggestions are included.

Unfortunately in the FFT1.9.zip file the FFT.cs file is missing.

Attached the last FFTdemo program.

Best regards.

[agraham]

Originally Posted by klaus

Unfortunately in the FFT1.9.zip file the FFT.cs file is missing.

I didn't bother including it as I'll now post the library and other usual items in a release quality archive in a new thread, probably tomorrow.