I have designed a digital crossover and exported the impulse response. When looking at the hipass filter frequency response with REW it appears that the low frequency attenuation is quite poor, REW shows a positive slope. At 2Hz attenuation is only 20 to 30dB and continue to rise I guess up to 0dB at 0Hz.
Here is a plot of a simple LR36 at 8KHz:
I should have missed something but can't figure out what...
joël
Here is a plot of a simple LR36 at 8KHz:
I should have missed something but can't figure out what...
joël
Only from result is nearly impossible to find out what you had missed, describe your filter/impulse creation process with more details.
filter is very simple :
Impulses is created from the menu View - Impulse response :
Vituixcad file is attached, just remove the added .txt extension.
Thanks!
Impulses is created from the menu View - Impulse response :
Vituixcad file is attached, just remove the added .txt extension.
Thanks!
Is the measurements in post #1 form the real world? Does your physical equipment really support 64k taps?
When number of taps is not enough, the filter don't looks as anticipated. This can be seen in the program rePhase where when one generates a filter and the taps are not enough, the final curve don't match the specified one.
But this is somewhat of a long shot, so...
//
When number of taps is not enough, the filter don't looks as anticipated. This can be seen in the program rePhase where when one generates a filter and the taps are not enough, the final curve don't match the specified one.
But this is somewhat of a long shot, so...
//
not a real measurement, the wav file impulse response is imported in REW. I tried using less taps, it's the same. Yes the convolver support 64k taps.
This is is a real measurement in the digital domain of my crossover high channel:
This is is a real measurement in the digital domain of my crossover high channel:
Hi, haven't used VCad for a while...so I'm guessing here....
Check the "Center IR" box.
That will produce a linear phase FIR file, which I'm assuming is what you want anyway.
Because otherwise, why not just use IIR to begin with...
Of course, FIR can be used for IIR replication, and quite successfully by moving impulse towards start.
But I've found if impulse is immediately close to start, results can sometimes go unexpectedly wonky.
Hi, same 8KHz filter "Center IR" box checked :
I want minimum phase filters, in my case FIR is just a more convenient way to do it.
I want minimum phase filters, in my case FIR is just a more convenient way to do it.
No clue then...sorry. Like said, not much time with VCad.
Gotcha. What's the setup that make it more convenient? Just curious...
Every clue is welcome!
As i do digital room correction convolver is mandatory so let's use it for crossover filtering too.
As i do digital room correction convolver is mandatory so let's use it for crossover filtering too.
Totally OT, but looking back I see how readable and workable old Win98 style interface is. Everything is clearly deilenated, active fields clearly separated from the backgrounds and inactive fields, buttons have this simplistic yet clear "pressable" look.
(compare to Win10 flat look above)
Last edited:
In total? Btw: an 8k filter 1t order 88ksps should be OK with say 4k taps.
Did you try with an IIR filter in REW to see that you are able do a good filter in the processing chain at all??
Post your wav. impulse file?
//
Thanks! Looks like something is wrong with IR export. Tried even more simpler:
IR export shows a Dirac which is good:
But once exported and imported in REW :
Looking at the impulse with a dBFS scale, it looks like a DC component is added somewhere in the export process :