True resolution

[celef]

Lot of measurement graphs are so smooth that they almost looks like they are made up, like REW graphs looks amazingly fancy and smooth, but when looking at graphs made with hiend measurement systems like Klippel they look always a lot worse and i think more accurate

If a measurement due to setup has a low resolution of 200 Hz, is it not better to just show every true measurement point and nothing in between?

 

[AllenB]

A specific amount of smoothing can help crossover design by removing the high Q peaks that we probably won't hear. Another example of the idea when zooming a scope, you can focus on a single issue amongst many issues when you get this right.

 

[Mark Tillotson]

Its a case of not being able to see the wood for the trees sometimes. For EQ there's little point looking at details so fine you can't correct for them anyway - also there is noise in any measurement, smoothing removes a lot of noise very easily, although averaging multiple measurements is more accurate.

 

[celef]

oh, no, i did not talk about smoothing, i meant true resolution, if i measure something with gating then the resolution is dependent on the time window, a small time window will get poor resolution, i think it would be more truthfull and fair to show only frequency points that is actually measured and not something that is extrapolated

 

[multisync]

on holmimpulse you can increase the sweep time to more than 90 seconds, you can display 1/48 smothing or raw data. At max sweep time holmimpulse will take over 4million samples.

is that what you mean?

 

[celef]

not exactly, as far as i know, for fft measurements, basic resolution is: sampling frequency / fft size

but when using a gated measurement to get anechoic response, then true resolution is: 1 / the gated time window in seconds