Window Correction FFT

[AndresID]

Hello,

In the FFT function in DIADEM there are two possible window corrections: Periodic and Random. The help doesn't give much information about the difference between the two.

I have read in the forum that the general idea is that if you are looking for a periodic wave then you should use "Periodic" and if you aren't sure about the signal you should use "Random".

Also there is a document called FFT Use in NI DIAdem that shortly explains the window correction, but only says that for the Hanning window the factors are: Periodic 2, and Random 1.63299.

My questions are:

Why exactly are there two different correction factors?

Is there a correction factor table for the two corrections for each window?

How is the Random correction factor calculated?

Thank you and best regards,

[AndresID]

Any help or comments will be much appreciated.

[Artemis125]

It sounds like the two correction types refer to bias or random error. The bias or periodic error is an error consistant throughout the entire dataset, so the correction would be the same throughout.  The random error is fluctuations in reading that changes value to value, all values average to around the same but each has its own error. Random error correction may try to account for random error in your data. These are both attempts to make your data closest to the true value as possible.

[Walter_Rick]

Hi AndresID,

I attached a document which covers a lot questions about FFT. Hopefully it can help.

Greetings

Walter

[AndresID]

Hello,

Thanks for the replies.

The document Walter attached is the one I mentioned in my original comment. Unfortunately it doesn't fully explain how both corrections are defined. The only correction factors given are for the Hanning window.

From what I understand, both corrections are constant factor that are independent from the signal you are transforming. One multplies the amplitudes by 2 and the other by 1.633.

My main concern is that I can't find any information about what correction factor are used for each window and how is the random correction calculated.

Best regards,

Andrés

[DJacques]

Hello,

Digging up an old topic!

I would like to analyze vibrations measured on a rotating machine. The rotating speed is constant, but amplitudes can vary quite significantly, depending on how the machine is used. In other worlds, the period of the signal never change (1/rev frequency around 6 Hz), only amplitudes do.

Should I use a periodic attenuation or a random one?

Thanks!

I attached an example of signal: