Hi all,
After some great help I was able to measure my WAW open baffle 2 way speakers that I recently build so that I could generate my own FRD and ZMA from my drivers in the baffle to be able to then work on crossovers in simulation software with all the baffle and stuff included in the measurements. I've made some progress and figured I'd share where I am currently and see what else can be done from this point or if I'm going in the right direction.
I measured the speaker outside on a ladder with a heavy blanket so that it was over 6 feet (1.8 meters) off the ground and nothing near by and measured from 1 meter away. I measured each driver individually at 1 meter farfield at the same input signal level. I then measured each driver individually nearfield at the same input signal level. I used the distance to the ground to guide my impulse response time gating for reflections. I corrected for the near field half space to full space response with script I wrote in libreoffice with some great help from some very friendly folk recently. And I summed the farfield and nearfield responses together to make my FRD files for each driver. I used a DATS to measure the impedance of each driver on the baffle and made ZMA files.
Here are the resulting corrected measurements from the process with 1/24 smoothing.
BG20 full range:
8PT-8 woofer:
These measurements have all the issues from diffraction and step from the baffle, waveguides, etc, that I built already and mounted into. From here, it's just working with the metric data to build a passive crossover. I've already done an active crossover with DSP and listened to them and liked them. I liked how they were from 700~1000hz crossover. I figure I would start working on a passive crossover now in VituixCad with these measurements.
Here is my first current draft from these measurements in VituixCad.
I'm attaching my VituixCad file in case anyone would like to play with it or help in any way, just rename it from *.txt to without that.
When I invert a driver, to see the resulting null at the crossover:
At this point, I like the basic curve. However, I see that for some reason I don't see all the individual spikes I should see from the BG20. It's overly smoothed in VituixCad so I cannot get a good idea of where to place notch filters and see the response here. I can use my measurements to place these filters, but it would be nicer to see it in the CAD software here.
Any ideas on how to remove the over-smoothing in VituixCad? I'm looking for options to turn that off or reduce it.
I know I want to target notch filters around 750hz on the woofer, 2400hz on the full range, 3200hz on the full range and 8800hz on the full range. I've already designed notches for this to address those peaks. I just would like to see it in the simulation. Currently its so smooth it just looks like I'm dropping nulls into the response.
This may be too much really, I'm not sure. This is where I need some guidance I bet.
Notches to target those spikes in the measurements from the baffle and waveguides. It would be nice to see the peaks in Vituix and then apply notches to drop them. I'm not sure why its so smoothed out. Can't seem to find a setting in the preferences to unsmooth it.
Any help appreciated!
Very best,
After some great help I was able to measure my WAW open baffle 2 way speakers that I recently build so that I could generate my own FRD and ZMA from my drivers in the baffle to be able to then work on crossovers in simulation software with all the baffle and stuff included in the measurements. I've made some progress and figured I'd share where I am currently and see what else can be done from this point or if I'm going in the right direction.
I measured the speaker outside on a ladder with a heavy blanket so that it was over 6 feet (1.8 meters) off the ground and nothing near by and measured from 1 meter away. I measured each driver individually at 1 meter farfield at the same input signal level. I then measured each driver individually nearfield at the same input signal level. I used the distance to the ground to guide my impulse response time gating for reflections. I corrected for the near field half space to full space response with script I wrote in libreoffice with some great help from some very friendly folk recently. And I summed the farfield and nearfield responses together to make my FRD files for each driver. I used a DATS to measure the impedance of each driver on the baffle and made ZMA files.
Here are the resulting corrected measurements from the process with 1/24 smoothing.
BG20 full range:
8PT-8 woofer:
These measurements have all the issues from diffraction and step from the baffle, waveguides, etc, that I built already and mounted into. From here, it's just working with the metric data to build a passive crossover. I've already done an active crossover with DSP and listened to them and liked them. I liked how they were from 700~1000hz crossover. I figure I would start working on a passive crossover now in VituixCad with these measurements.
Here is my first current draft from these measurements in VituixCad.
I'm attaching my VituixCad file in case anyone would like to play with it or help in any way, just rename it from *.txt to without that.
When I invert a driver, to see the resulting null at the crossover:
At this point, I like the basic curve. However, I see that for some reason I don't see all the individual spikes I should see from the BG20. It's overly smoothed in VituixCad so I cannot get a good idea of where to place notch filters and see the response here. I can use my measurements to place these filters, but it would be nicer to see it in the CAD software here.
Any ideas on how to remove the over-smoothing in VituixCad? I'm looking for options to turn that off or reduce it.
I know I want to target notch filters around 750hz on the woofer, 2400hz on the full range, 3200hz on the full range and 8800hz on the full range. I've already designed notches for this to address those peaks. I just would like to see it in the simulation. Currently its so smooth it just looks like I'm dropping nulls into the response.
This may be too much really, I'm not sure. This is where I need some guidance I bet.
Notches to target those spikes in the measurements from the baffle and waveguides. It would be nice to see the peaks in Vituix and then apply notches to drop them. I'm not sure why its so smoothed out. Can't seem to find a setting in the preferences to unsmooth it.
Any help appreciated!
Very best,
You can't sum/combine the nearfield and farfield responses for a dipole like you can with a closed box. The nearfield does not include the dipole response, only the driver's own response. I did not see any correction for this. It's like how with a closed box system you must add the baffle step to a nearfield before merging/blending it with a farfield measurement. Without correction a nearfield measurement is not very useful for developing the crossover for a dipole or OB loudspeaker. Unfortunately it is difficult to calculate a frequency and phase correction (unlike for closed box loudspeakers) that accounts for the front-back interaction of a driver in an OB configuration at higher frequencies.
To avoid this problem I measure in the farfield for my own dipoles whenever possible. I use various techniques and then adjust for measurement distance and "space" in each:
If you want more details on how to do the SPL and phase corrections let me know.
To avoid this problem I measure in the farfield for my own dipoles whenever possible. I use various techniques and then adjust for measurement distance and "space" in each:
- Always made 2-channel measurements where one channel is a time reference (eg for ARTA) or with a measurement system/software that has a time reference built in. This ensures that all measurements will have the correct phase angle.
- For higher frequencies, the measurement distance should be at least three times the largest dimension of the baffle. This is where very small baffles or nude drivers have a big advantage over larger baffled OB systems! I typically use a distance of 20in/0.5m when measuring the mid and tweeter (that are nude) while they are elevated e.g. 2m or more above the ground to minimize the ground reflection. I make sure to measure and record the distance between source and mic.
- When measuring LF below about 300Hz it's often better to make a groundplane measurement and then subtract 6dB from the result to convert it to full-space (e.g. the "space" during the measurement of the tweeter or midrange). I typically measure at around 3m/10ft, making sure to record the distance from source to mic.
- To harmonize all the measurements, I choose some distance (typically the groundplane distance) and then correct the phase and SPL of the measurements for the mid and tweeter as if these drivers were magically moved horizontally from their measurement distance (e.g. 0.5m) and placed above the postion where the woofers were measured (e.g. 3m) just like they will be in the final speaker. E.g. I account for reduced SPL and greater phase rotation that would have been recorded if the mid and tweeter were moved by 3m-0.5m = 2.5m.
If you want more details on how to do the SPL and phase corrections let me know.
One (not very precise) way to get OB speaker response approximation is to measure driver near filed response and add to this response 180 degrees turned in phase same response what is delayed by distance from rear to front radiation side of driver (separation distance). This calculations can be done in REW. Mainly this works for bass and lower mirdange.
Thank you very much! I have read quite a few of your publications and software.
Hah, what a bone head move I did there. I completely forgot I was even dealing with a dipole there when I measured near field, I was so caught up in measuring 3 different speakers back to back (different designs) and when this one's turn came up I just repeated the process just like all the previous ones. Good lord! Thanks for the sobering reminder!
I did far field at 1 meter distance from speaker to microphone and the speaker was 1.8 meters off the ground.
It makes sense to just use groundplane. I will do this on the next go. Thanks!
I'll re-do some measurements and see if I can get a better overall measurement and then try to move into the next step as you mentioned for correcting the measurements.
Baffle width is 10" or 0.254m
Baffle height is 20" or 0.508m
Side wings next to woofer is 9.62" or 0.244m
Minimal wing to nothing next to tweeter (BG20 full range)
See below to help:
It has side wings though and a bottom. I'm not sure if this will change a few things in method. Sometimes pictures can say a lot more than words. I built this before designing, backwards to my usual, so now I'm going backwards and trying to sort it out after building it from measurements--a good learning exercise for me and difficult.
Ok, since the near field response and correction is incorrect for dipole, I took that away and went back to my basic 1 meter full sweep measurements to share as a restart.
This was my 1 meter measurement (1.8m from the ground) of the BG20 driver from 20hz to 20khz. I didn't gate it or modify it, just 1/24 smoothing. So under 300hz I should ground plane this at around 3 meters per your advice. I'll do this soon.
This was my 1 meter measurement of the 8PT-8 woofer, 1.8m from the ground, also 20hz to 20khz. No gating. No modifying. Just 1/24 smoothing. I'll re-do a groundplane for under 300hz around 3 meters soon.
Very best,
Last edited:
Thanks, very interesting. I'll look into this! If I cannot do a good measurement outside, I may try this one then.
Very best,
This is a very good technique to use when a groundplane measurement is not feasible. It's good up to a couple hundred Hertz. Because the front and rear will often be different (especially if a U-frame is used, but even with a planar baffle due to the magnet) the result will be better if nearfield measurements are taken at both front and rear (sometimes the rear is tricky), making sure to measure the phase correctly during each measurement. To get the far field (with dipole cancellation) you compute the distance between each "source" (the front and rear output) and the listening position or other position in space. Then you determine the SPL and phase modification due to propagation in air between the measurement location and the new location, make those modifications to the data, and then sum the two together. The assumption is that both sources are propagating like monopoles and wavelengths are much longer than the dimensions of the radiator and baffle. This is why there is an upper frequency limit above which this technique is not accurate.