That's a good point I hadn't considered. Hopefully diffraction is not having a big influence! @Ugg10 Can you share a photo or two, it would be cool to see the speaker.
Since we know the adjustment needed to make the original measurements align with the more recent loop-back ones, simply add or subtract that amount in VituixCAD either as a physical offset or pure delay. I think this is better than baking it in to the data in REW.
TBH driver alignment is not an exact thing because it varies with frequency (and visual alignment of impulse / step is also tricky for this reason). I measured a mid dome recently and the lower freq. were further back than the highs - the highs come from the tip of the dome while the lows from the VC attachment at the outside. You will likely need to tweak the filter slopes slightly to phase align both drivers.
The plot thickens.....
I decided to load in the original measurement on axis (horzontal 0 degs), apply minimum phase and also loaded in the loopback measurements taken yesterday with the plan of adding in the derived delays form the impulse plots to the original minimum phase data.
However, when I did this the phase and impulse plots did not match therefore there seems to be some sort of delay/offset data in the original measurements. So, I adjusted the individual driver delays so the two sets of measurement matched in terms of phase and impulse. I have modified my active crossover and will try and measure later. (Note plots below are not volume matched and there is obviously a lot of noise in the loopback measurements, delay in uS shown in crossover box).
woofer (green = loopback, red = original)
Mid (yellow = loopback, Blue = original)
Tweeter (brown = Loopback, Blue = original)
I decided to load in the original measurement on axis (horzontal 0 degs), apply minimum phase and also loaded in the loopback measurements taken yesterday with the plan of adding in the derived delays form the impulse plots to the original minimum phase data.
However, when I did this the phase and impulse plots did not match therefore there seems to be some sort of delay/offset data in the original measurements. So, I adjusted the individual driver delays so the two sets of measurement matched in terms of phase and impulse. I have modified my active crossover and will try and measure later. (Note plots below are not volume matched and there is obviously a lot of noise in the loopback measurements, delay in uS shown in crossover box).
woofer (green = loopback, red = original)
Mid (yellow = loopback, Blue = original)
Tweeter (brown = Loopback, Blue = original)
@Tenson please see below. The box is a re-purposed Akai SW125 with the front baffle cut out and a new flush one added. Sealed at around 40L and stuffed with 1.25lbs of teddy bear stuffing. You can see why I went for sealed box given my placement limitations. (I normally have a heavy rug in front of them).
Last edited:
Since you have the dsp xover, how about simulating a dsp xover with a deep reverse null then see if real life matches? This could help tweak the data to match.
WhooHoo, small success !!!
As @Tenson suggested, I loaded up the two models above, one delays based on the Pairs measurement data and one based on the Loopback impulse offsets.
I did a full system sweep at 1m in my small room with the mid in normal polarity and inverse polarity plus individual driers. I imported the VituixCad modeled responses for comparison. Finally, for the inverse mid measurements I added an IR window, reducing it until I got the deepest nulls (in my case 3ms due to the small room and closeness of walls). Oh, I also level matched the model and measurements by adding 3dBto the measurements in REW, I obviously measured slightly quieter than in the anechoic chamber.
The results - the Loopback matched delays win!!! The woofer/mid null on the Pairs delay version (green) is no where as deep as it should be, whereas the Loopback matched is as expected (purple)..
Comparison of the Loopback Matched VituixCad model with measurements, IR Window applied, ligher colours are measurements.
Similar for individual drivers (No IR window on woofer), above 300hz there is good agreement, below that there are all sorts of room resonances.
So, I think I will listen to the loopback version for a while and then I need to do the loopback measurements again without the protection capacitors in place to set the delays for a passive design. But so far it looks like I have a pretty flat on axis frequency response with pretty deep nulls when the mid is inverted.
** As I have already said, many thanks to one and all with help figuring out this, it's been a great learning experience for me. *
As @Tenson suggested, I loaded up the two models above, one delays based on the Pairs measurement data and one based on the Loopback impulse offsets.
I did a full system sweep at 1m in my small room with the mid in normal polarity and inverse polarity plus individual driers. I imported the VituixCad modeled responses for comparison. Finally, for the inverse mid measurements I added an IR window, reducing it until I got the deepest nulls (in my case 3ms due to the small room and closeness of walls). Oh, I also level matched the model and measurements by adding 3dBto the measurements in REW, I obviously measured slightly quieter than in the anechoic chamber.
The results - the Loopback matched delays win!!! The woofer/mid null on the Pairs delay version (green) is no where as deep as it should be, whereas the Loopback matched is as expected (purple)..
Comparison of the Loopback Matched VituixCad model with measurements, IR Window applied, ligher colours are measurements.
Similar for individual drivers (No IR window on woofer), above 300hz there is good agreement, below that there are all sorts of room resonances.
So, I think I will listen to the loopback version for a while and then I need to do the loopback measurements again without the protection capacitors in place to set the delays for a passive design. But so far it looks like I have a pretty flat on axis frequency response with pretty deep nulls when the mid is inverted.
** As I have already said, many thanks to one and all with help figuring out this, it's been a great learning experience for me. *
Last edited:
Good results! Looks like the woofer to mid was where the biggest difference was.
Careful when sweeping without the protection cap, I guess you know this already. I use ARTA rather than REW and can bandwidth limit the test signal, not sure if REW can do this as well?
Also, those nulls are very deep, that is an excellent result for acoustic cancellation.
Careful when sweeping without the protection cap, I guess you know this already. I use ARTA rather than REW and can bandwidth limit the test signal, not sure if REW can do this as well?
Also, those nulls are very deep, that is an excellent result for acoustic cancellation.
@Tenson thanks for the comment. Pretty happy with the result so far, interesting that the woofer/mid null does not really appear until you add the IR Window and it changes quite a bit as you reduced the window to a point and then it goes the other way. Below is a 4.0, 3.5, 3.0, 2.9 (best), 2.5ms window. The Pairs version is in grey, max null depth was at 3.4ms.
Yep, got the "no capacitor" warning. In REW I only measure the Mid Dome from 200-20khz and the tweeter from 300-20khz as a precaution with or without the caps in place (woofer obviously form 20-20khz).
Yep, got the "no capacitor" warning. In REW I only measure the Mid Dome from 200-20khz and the tweeter from 300-20khz as a precaution with or without the caps in place (woofer obviously form 20-20khz).
I don't really follow what you mean about the window. A long window time will be necessary to have any resolution in the low frequencies so I would expect that to show the null best. Someone more familiar with REW might need to comment on that.
edit: All the traces are very smooth below 1KHz. If this is in-room at 1m then the data in that range seems un-trustworthy, as you should be seeing lots of modal bumps and dips from the room.
To give some more insight here is anechoic response of a speaker with different gating times.
Ungated phase response
5ms Gated Phase response
The phase also gets smoothed with the windowing.
I would not use a gate shorter than 8ms if you are looking at the response around 600Hz for your xover. Since this will not likly help for any room reflections you may as well go un-gated when measuring this lower xover.
edit: All the traces are very smooth below 1KHz. If this is in-room at 1m then the data in that range seems un-trustworthy, as you should be seeing lots of modal bumps and dips from the room.
To give some more insight here is anechoic response of a speaker with different gating times.
Ungated phase response
5ms Gated Phase response
The phase also gets smoothed with the windowing.
I would not use a gate shorter than 8ms if you are looking at the response around 600Hz for your xover. Since this will not likly help for any room reflections you may as well go un-gated when measuring this lower xover.
Last edited:
In REW the IR Window is effectively the gate, just called something else. The woofer unrated datatcan be seen in #46 last graph, light purple line) which is a rollercoaster below about 300hz due to room reflections (The dark purple is the anechoic data mergers with a nearfield measurement which is why there are no fluctuations). If I look at the impulse plots then my first reflection at around 3.5mS which is why I need a low gate / IR window to get rid of them, in REW this smooths of the lower octaves and also truncates the data whereas it looks like on your system below the gate frequency it elongates them.
@Tenson without the gating the hill looks almost non existent (both cases, pairs offset and loopback matched of
ffset), it only really appears when you gate but when you gate only the loop back one goes deep and matches the model in Vituix. This is also true for the mid/tweeter null which again is only a slight dip without any gating.
Green ungated, Blue Ungeted with mid inverted, Orange gates at 2.9ms, Red gates at 2.9ms mid inverted.
ffset), it only really appears when you gate but when you gate only the loop back one goes deep and matches the model in Vituix. This is also true for the mid/tweeter null which again is only a slight dip without any gating.
Green ungated, Blue Ungeted with mid inverted, Orange gates at 2.9ms, Red gates at 2.9ms mid inverted.
Last edited:
I don't particuarly trust that gated low frequency, but at 600hz you should be close enough with a physical offset measurement of the mid dome voice coil to the woofer dust cap.
P.s. turn off smoothing when looking for the null. Anyway I'll shut up now, I think you've got close enough!
P.s. turn off smoothing when looking for the null. Anyway I'll shut up now, I think you've got close enough!
Yes REW can. Unsure though how it "engages" the sound card. there's still the issue I think of a pop / DC outside REW's control (depending on hardware and drivers) so I always use a protection cap above a 0.5v drive level if doing full range