Looking better, why is there smoothing on a FDW view? No need to do both.
Remove the smoothing from that graph and use it for EQ.
The IR falls out of the picture on the left side, slide it so we get to see the peak and the first 15 ms. We now see 10 to 23 ms instead. -5 to 15 ms should work
Remove the smoothing from that graph and use it for EQ.
The IR falls out of the picture on the left side, slide it so we get to see the peak and the first 15 ms. We now see 10 to 23 ms instead. -5 to 15 ms should work
@Juhazi
@mkane77g
@wesayso
Sorry I'm late, day job cluttered my activity here.
@Juhazi
@wesayso
Can we discuss if there is hope we can calibrate that speaker up close and maybe get it pretty good optionally without using a ladder and tons of pillows stuff
Have observed some pattern when we filter IR, below is a textbook synthetic perfect impulse for a known passband plus all the 1/3 filters REW can offer us turned on, for me it looks there is a clearly balanced pattern to see plus the unfiltered response ride exactly always +12dB above the filtered ones, and BTW all responses is a FDW and setting is 1/6 width in octaves / same as 4,3 width in cycles:
Now to a live measurement, here is my two way on wall FAST speaker calibrated on wall within a meter to exactly same known pass band as above and same FDW settings, for me it looks pretty textbook even its a real acoustic domain live measurement and if we compare to above it looks lowest filtered response (blue) have error not as the others being exactly -12dB below unfiltered response and also we begin see resonances in pattern below 20Hz area and deviation scheme from the textbook one looks by eye to start about 30Hz and below:
Hope is we can use patterns from above and get mkane77g's OB calibrated, here is where he got woofer a couple of days ago, many patterns are not real symmetric and there is resonances plus non filtered response don't ride exactly +12 above filtered responses which think should give clues to where to set in:
What do you think about it ?
@mkane77g
@wesayso
Sorry I'm late, day job cluttered my activity here.
@Juhazi
@wesayso
Can we discuss if there is hope we can calibrate that speaker up close and maybe get it pretty good optionally without using a ladder and tons of pillows stuff
Have observed some pattern when we filter IR, below is a textbook synthetic perfect impulse for a known passband plus all the 1/3 filters REW can offer us turned on, for me it looks there is a clearly balanced pattern to see plus the unfiltered response ride exactly always +12dB above the filtered ones, and BTW all responses is a FDW and setting is 1/6 width in octaves / same as 4,3 width in cycles:
Now to a live measurement, here is my two way on wall FAST speaker calibrated on wall within a meter to exactly same known pass band as above and same FDW settings, for me it looks pretty textbook even its a real acoustic domain live measurement and if we compare to above it looks lowest filtered response (blue) have error not as the others being exactly -12dB below unfiltered response and also we begin see resonances in pattern below 20Hz area and deviation scheme from the textbook one looks by eye to start about 30Hz and below:
Hope is we can use patterns from above and get mkane77g's OB calibrated, here is where he got woofer a couple of days ago, many patterns are not real symmetric and there is resonances plus non filtered response don't ride exactly +12 above filtered responses which think should give clues to where to set in:
What do you think about it ?
Attachments
Last edited:
today i tweaked my car audio, slight changes to xo points which are, 60,100 and 3024hz.
is freaking hard to get it right in car, dual channel fft is only way to get it right, still have to use hard smoothing.
with s.harsch i allmost run out of delay, max is 15.6ms.
i wish i could change 60hz to 50hz to get bass more on front but delay hits limit and 6liter boxes under the front seats start to moan.
s.harsch is much better sounding than full LR4
Last edited:
You could use a higher Q value for that tweak, right now its tweaking 60 to 600Hz.
Is that smooth line the target we want to hit?
I have no idea how many PEQ's you have available, but you don't need to actually set a named crossover to hit the slope you need. You could do it entirely with PEQ.
Did you share the actual measurements? Maybe an example done by one of us would help you see what's possible.
One more question, do you have the mic setup with the speaker and can walk away and do another sweep at will under the exact same circumstances?
We'd need to be in the exact same spot each time.
How far from the speaker is the microphone?
Is that smooth line the target we want to hit?
I have no idea how many PEQ's you have available, but you don't need to actually set a named crossover to hit the slope you need. You could do it entirely with PEQ.
Did you share the actual measurements? Maybe an example done by one of us would help you see what's possible.
One more question, do you have the mic setup with the speaker and can walk away and do another sweep at will under the exact same circumstances?
We'd need to be in the exact same spot each time.
How far from the speaker is the microphone?
Back in post 515, I said
So use those filter dialogs to smooth and flatten your responses - hammer them like a blacksmith with gentle broad PEQ’s to force them to fit the textbook curve. I mean the woofer separately and the fullrange separately. Should not need more than maybe 6 individual filters to get close.
No point in EQ’ing the composite flat when the individual ones are off.
You might want to look at Charlie Laub’s website and check out his spreadsheet for designing miniDSP active XO’s - it will have a handy textbook overlay generator and you import your measurements FRD file. Then tweak and massage the response to for the profile. Then upload the resulting PEQ table to miniDSP.
the Active Crossover Designer web page
Jeff Bagby’s PCD (also found on same Charlie Laub’s website) can do this too.
Loudspeaker Design Software
I actually use PCD for making the PEQ adjustments. It works perfectly, that is, the simulated matches the resulting measurement perfectly.
Last edited:
Can't figure out how to get a FR loaded. I have them in a file but when asked to choose I don't have a file labeled FRF
You know this, right?
This post was a revelation and I didn't miss it back a few pages. I guess I didn't pay enough attention that when entering numbers manually the changes immediately show up. Too bad, it blocks half the page