Calculating FDW window width in Room Macro4 with different cycle values

[Mitch Barnett]

Hi Uli,

Congrats on the new forum!

My question is how do I calculate the the FDW windows times when the low and high frequency cycles are different?

I know how to do that when the cycles are the same, like 15/15 using http://www.sengpielaudio.com/calculator-period.htm At 20 Hz has a time period of 50ms, so using 15 cycles, a window width of 15 * 50 ms = 750 ms at 20
Hz. At 1 khz the window width is 15ms and at 10 kHz it is 1.5ms.

But what if I have a setting like 6/1 in Room Macro 4 excess phase window, how would I calculate the different window widths at different frequencies like above?

Kind regards,
Mitch

[UliBru]

Hi Mitch,

just a short response right now. Acourate is applying some interpolation between the LF FDW and the HF FDW. I need to re-check the applied interpolation method, memory is fading So I hope I can answer tomorrow, thanks for your patience

BR
Uli

[Uli]

Hello Mitch,

FDW calculates with changing window width from low frequency (about 4 Hz) up to about fs/2 . The width is frequency dependent. You may consider the width curve as a straight line in a logarithmic frequency plot.
So if e.g. the low parameter is 10 and the high parameter is 2 you will get a window width at 1 kHz of

Code: Select all

width1kHz = 10 + (2 - 10)/(Log(24000)-Log(4))*(Log(1000)-Log(4)) = 4.9225

The samplerate here is 48000.

I hope this helps

Greetings
Uli

[Mitch Barnett]

Hi Uli,

Thanks! I totally get the memory fading bit Perhaps you can check my math and if I got the right spots to pass in the parameters.

For 500Hz:

Code: Select all

width500Hz = 10 + (2 - 10)/(Log(24000)-Log(4))*(Log(500)-Log(4)) = 5.5599

For 500Hz and window high parameter is 6 and low parameter is 1:

Code: Select all

width500Hz = 6 + (1 - 6)/(Log(24000)-Log(4))*(Log(500)-Log(4)) = 3.2249

And to complete the example for window width in ms at 500Hz. 500Hz time period = 2ms. So,

Code: Select all

width500ms = 3.2249 * 2ms = 6.45ms

Did I get that right? It has been awhile

Kind regards,
Mitch

[UliBru]

That's perfect, Mitch

All the best
Uli

[Valentin Nicolaie]

This is very good to have as reference to get back to (suggest to sticky it), but the big question still remains: how do we determine in one's acoustic space what would be the right FDW? Few years ago I have ventured to propose a correlation with the reverberation time, RT30 if my memory serves me (as RT60 is not fully suitable for small spaces), but I am not convinced of this direction anymore. In all honesty I am using 15/15 these days.

Has there been any development, new ideas on the topic lately? Perhaps, Uli, from your vast experience gathered over the years, will you be able to indicate a base direction?

Thanks,
Valentin

[UliBru]

how do we determine in one's acoustic space what would be the right FDW?

Valentin,

at the very end only your own ears count.
From the past I know that someone (Josef W. Manger) has told that the ear needs 7 cycles at least to detect a tone.
If you follow discussions about room correction people like Floyd Toole propagate that for neutral speakers a correction is only necessary for frequencies below the Schröder frequency (around 300 Hz). From my experience neutral speakers are quite rare, at least most measurements in rooms look different. Any applying a very short FDW, e.g. 3 to 5 here simply corrects the speaker and takes out reflections.
There are Acourate users who apply a mix by combining FDW calculations, e.g. 15, with FWD calculations, e.g. 3. You can do this by the SplitNJoin function. But you can also apply Amplitude Max. Envelope to both curves. This results in a soft correction of the overall behaviour including peak corrections whereas dips are not corrected.
There are may roads to Rome. You can follow the basic path but you can also detect and try different paths.
Your ears count.

BR
Uli

[Valentin Nicolaie]

Uli,

Interesting take about mixing corrections. I would love to understand more about how to achieve this mixing of combining FDW calculations without making a mistake. If you could share the steps to take, I would appreciate (I am not sure I understood your initial guidance).

many thanks,
Valentin

[UliBru]

Valentin,

see a given pulse response where I have calculated macro1 with FDW3 and FDW15 (you can "save" curves by copying/moving them, Ctrl+C / Ctrl+M)

AmplitudePV1.png (57.76 KiB) Viewed 10227 times

The same display without the original pulse shows the difference between the two FDW settings. FDW15 has more peaks and dips than FDW3.

AmplitudePV2.png (23.56 KiB) Viewed 10227 times

Now we can use FD-Functions - Amplitude max. Envelope and we get the result

AmplitudePV3.png (27.91 KiB) Viewed 10227 times

So the FDW3 curve is limiting the correction, now less dips will get corrected. We accept them as they are given.
As the envelope function applies a hard "cut" the result has some sharp edges in the frequency response. Edges in frequency responses are never welcome. they cause some ringing. Thus we can apply a further FDW, here FDW25 on the result. This gives:

AmplitudePV4.png (19.5 KiB) Viewed 10226 times

Now we can save the final result as Pulse48L/Rmp and proceed with Macro2 ...

As you see now the correction applies a soft correction for the general shape of the speaker behaviour, then it corrects for peaks and it avoids to boost dips.
Be aware: this is another one of x possibilities. At the very end your ears count.

Haven fun
Uli

[Valentin Nicolaie]

Great guidance, thank you very much Uli. Is the SplitNJoin approach achieving the same outcome as the Amplitude Max Envelope, i.e. limiting the boosts while correcting properly the peaks?

Let the fun begin.