"Mike Dennison" <[email protected]> wrote:
This really is a nice piece of software. It is easy to run, is very
sensitive and continues to run when you are working on another
program in a foreground window.
[snip]
Mike,
thanks for your nice words.
What you call 'frequency error', and which I would call 'frequency
uncertainty',
is caused by the relative values of the sampling rate and the FFT size.
In the full band view, the sampling rate is 11025 S/s, and the FFT size
is 4096 samples. This gives a frequency resolution of 11025/4096 = 2.69 Hz
(roughly). This is the uncertainty with which the frequency can be determined
when in that mode.
If you look at the peak display, you will see that it says :
Peak at nnn.nn Hz +/- 2.69 Hz The +/- value is the uncertainty.
For the other modes the values are different, but you get the idea.
When you scroll up and down the scale, it may happen that the peak falls
in one FFT bin or in the adjacent one. It's not a continous movement, but
rather a step one. So it maybe that, for example, if you measure the mains hum
picked up by a piece of wire, at the beginning of the scale
you find that the peak is, let's say, at 49.9 Hz, and scrolling you find
it at 50.2 Hz. This is not an error, it's just the consequence of what said
above. To measure with great precision, it would be necessary an FFT
length of several MegaSamples, with the inevitable long time of acquisition.
Then you could go in the milliHz region, as by example Andy does with his
EVM56002 (I have one, where can I download your software from, Andy ?),
but be prepared to long waits between screen refreshes.
To answer to John's question, to choose the 'magnitude squared' estimator,
go to the action bar, select Mode, then Spectral Estimator, then Magnitude
Squared. This is present only in build 113.
73 Alberto I2PHD
____________________________________________________________________
Get free email and a permanent address at http://www.netaddress.com/?N=1
|
