The title is the heading of one part of Scott Wurcer’s excellent article in #10 LA (*)

Linear Audio | your tech audio resource

I have taken the liberty of using it because it gives me a perfect handle for to discuss a few topics regarding FFTs.

The heretical (his words) side of Scott prevailed in this part of #10 LA article and discussed the merits of ‘non windowed’ FFT spectrums (see attached image).

This strung a chord in me.

I am one of the few that shamelessly use FFT, this powerful analysis tool without having an in-depth knowledge of the fundamentals

Around a month ago, while doing some tests on CD players, I thought of using no window function for my analysis, as a way of presenting measured data as naked as possible.

Well presenting a naked figure, is a difficult task and requires possessing or obtaining first some talents. I had to look a bit deeper into theory for to understand what I am doing, thus I realized a few things that are of some importance.

One of these important things is the FFT bin and it’s width. I am not very good in explaining, here is an informative and easy to read article which helped me understand a few basic things on FFT, like what is the effect of choosing an FFT length (N) in combination with the sampling frequency (Fs).

http://www.ni.com/white-paper/4278/en/pdf

I built a simple spreadsheet mostly to investigate the effect that the bin width (equal to Fs/N) will have on the result of the FFT and what happens when certain test frequencies are used.

Then I was then introduced to the concept of coherent sampling

https://en.wikipedia.org/wiki/Bin-centres

https://en.wikipedia.org/wiki/Coherent_sampling

and I added a few lines on the spreadsheet (**)

Δf is the Interval of FFT frequency lines or spacing of the FFT bins. Second attachment is a diagram showing how Δf changes with Fs and N

George

(*) In this article, Scott is taking the reader by the hand for a guided tour at the digital landscape around the main subject (“Record replay RIAA correction in the digital domain”). I want to thank Scott for his time preparing and writing this ‘complete’ article and Jan for publishing it. Reading it gave me a rare educative enjoyment (meaning -among other things- that I have much to learn).

(**) The spreadsheet is simple (primitive). Input data only into yellow heavy lined cells

https://www.dropbox.com/s/fm9q19qlxevkfia/FFT%20issues%20light.xls?dl=0

Linear Audio | your tech audio resource

I have taken the liberty of using it because it gives me a perfect handle for to discuss a few topics regarding FFTs.

The heretical (his words) side of Scott prevailed in this part of #10 LA article and discussed the merits of ‘non windowed’ FFT spectrums (see attached image).

This strung a chord in me.

I am one of the few that shamelessly use FFT, this powerful analysis tool without having an in-depth knowledge of the fundamentals

Around a month ago, while doing some tests on CD players, I thought of using no window function for my analysis, as a way of presenting measured data as naked as possible.

Well presenting a naked figure, is a difficult task and requires possessing or obtaining first some talents. I had to look a bit deeper into theory for to understand what I am doing, thus I realized a few things that are of some importance.

One of these important things is the FFT bin and it’s width. I am not very good in explaining, here is an informative and easy to read article which helped me understand a few basic things on FFT, like what is the effect of choosing an FFT length (N) in combination with the sampling frequency (Fs).

http://www.ni.com/white-paper/4278/en/pdf

I built a simple spreadsheet mostly to investigate the effect that the bin width (equal to Fs/N) will have on the result of the FFT and what happens when certain test frequencies are used.

Then I was then introduced to the concept of coherent sampling

https://en.wikipedia.org/wiki/Bin-centres

https://en.wikipedia.org/wiki/Coherent_sampling

and I added a few lines on the spreadsheet (**)

Δf is the Interval of FFT frequency lines or spacing of the FFT bins. Second attachment is a diagram showing how Δf changes with Fs and N

George

(*) In this article, Scott is taking the reader by the hand for a guided tour at the digital landscape around the main subject (“Record replay RIAA correction in the digital domain”). I want to thank Scott for his time preparing and writing this ‘complete’ article and Jan for publishing it. Reading it gave me a rare educative enjoyment (meaning -among other things- that I have much to learn).

(**) The spreadsheet is simple (primitive). Input data only into yellow heavy lined cells

https://www.dropbox.com/s/fm9q19qlxevkfia/FFT%20issues%20light.xls?dl=0

#### Attachments

Last edited: