noise floor

hello...

how do you measure the noise floor for a resampler algorithm?

thank you

"John" <john@nospam.thanks> writes:

> hello...
>
> how do you measure the noise floor for a resampler algorithm?

Depends on the resample ratios, but one possibility would be this: Feed
in a theoretically perfect sine wave at Fs/4: 0, 1, 0, -1, notch the
result, and measure the residual.
-- 
Randy Yates                      % "Ticket to the moon, flight leaves here today 
Digital Signal Labs              %  from Satellite 2"
mailto://yates@ieee.org          % 'Ticket To The Moon' 
http://www.digitalsignallabs.com % *Time*, Electric Light Orchestra

perform FFT and find out.

ok...i get it...

so everything around the sine spike in the frequency spectrum is "noise"...

and i should just sum up the magnitudes in the power spectrum (after 
notching out the sine) to get the variance of the noise ??

John,

    you need to perform abs, then square each term and add it.

    FFT values will be complex and hence.

Bharat

On Nov 22, 1:56=A0pm, "bharat pathak" <bharat@n_o_s_p_a_m.arithos.com>
wrote:
> John,
>
> =A0 =A0 you need to perform abs, then square each term and add it.
>
> =A0 =A0 FFT values will be complex and hence.
>
> Bharat

He did say "magnitude".

>ok...i get it...
>
>so everything around the sine spike in the frequency spectrum is
"noise"...
>
>and i should just sum up the magnitudes in the power spectrum (after 
>notching out the sine) to get the variance of the noise ??
>
>
>

sum of magnitude squares will give you noise power.

noise power = variance + (mean*mean)

Rgds
Bharat

On Nov 22, 8:56=A0am, "bharat pathak" <bharat@n_o_s_p_a_m.arithos.com>
wrote:
> John,
>
> =A0 =A0 you need to perform abs, then square each term and add it.
>
> =A0 =A0 FFT values will be complex and hence.
>
> Bharat

You don't need to take abs before squaring...

Rick

John wrote:
> hello...
> 
> how do you measure the noise floor for a resampler algorithm?

In dBFS.

John wrote:

> ok...i get it...
> 
> so everything around the sine spike in the frequency spectrum is "noise"...

It is a nasty, nasty problem as resampling artifacts aren't noise.

> and i should just sum up the magnitudes in the power spectrum (after 
> notching out the sine) to get the variance of the noise ??

As the level of artifacts depends on the fundamental frequency, you've 
got to integrate it over the spectrum.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com