I don't think so: the tweeter output would be visible as more even and stronger output up to 20 kHz.
Also, phase is still all over the place. I must be doing something wrong. There seems to be no improvement from room size, so must be another cause.Will add full files tomorrow.
Indeed, I should have been checking more carefully during measurements. Such an opportunity lost! But I think with this experience I can manage a new set either in a similar large space, outdoors or even at home and it will be goud enough for final DSP setup. I am starting to have a good feel on behaviors of the three drivers. For example, I am now confident on polarity of all three channels. I can see where and how much EQ is needed and where the crossover points should be.
However, I am still completely clueless on phase. How to measure it, how to interpret the results, and what to do with it is a mystery.
I will add the files later today once I am back at home and would appreciate any help understanding the phase.
However, I am still completely clueless on phase. How to measure it, how to interpret the results, and what to do with it is a mystery.
I will add the files later today once I am back at home and would appreciate any help understanding the phase.
For better readability with phase plots you need to remove the IR delay. The delay the sound needs to reach the microphone.
My workflow wich is based on 'Use loopback as timing reference':
First measure the tweeter on axis. In the SPL&Phase window actions use 'Estimate IR delay' after that press 'Shift an update timing offset'
This will shift the IR and add a timing offset for subsequent measurements. Do not move the microphone any more or change that number. As you're using the Fusion Amp the timing offset set will not equal the actual distance between speaker and microphone due to the processing delay. This makes the speaker appear further away.
As you already did all the measurements you can note the estimated IR delay and manually add this to the other measurements. I would not use the estimation on mids or woofers as this relies on the highest frequencies to be accurate. It might be good enough though as you're not measuring with any XO in place. (There is an action in the ALL SPL window to 'Remove IR delays')
My workflow wich is based on 'Use loopback as timing reference':
First measure the tweeter on axis. In the SPL&Phase window actions use 'Estimate IR delay' after that press 'Shift an update timing offset'
This will shift the IR and add a timing offset for subsequent measurements. Do not move the microphone any more or change that number. As you're using the Fusion Amp the timing offset set will not equal the actual distance between speaker and microphone due to the processing delay. This makes the speaker appear further away.
As you already did all the measurements you can note the estimated IR delay and manually add this to the other measurements. I would not use the estimation on mids or woofers as this relies on the highest frequencies to be accurate. It might be good enough though as you're not measuring with any XO in place. (There is an action in the ALL SPL window to 'Remove IR delays')
Not sure if this has been already posted, this is the official guide how-to do meassurements in REW for usage in VituixCAD:
https://kimmosaunisto.net/Software/VituixCAD/VituixCAD_Measurement_REW.pdf
https://kimmosaunisto.net/Software/VituixCAD/VituixCAD_Measurement_REW.pdf
The problem is in making decisions which of the many ones is the most true to a real thing.
Your mid measurements are not of any use in the manor that you plan to use them. However, there's enough good information to get it done using "old school" techniques.
Don't take this too literally, but phase is mostly a result of the response changing.
I will share with you the step by step process I use. I work in a few stages measuring at home. The manual for Arta measurement software is also an excellent reference. My process isn't perfect but it's simple and gets you in the ball park.
1) I have a stick that is 18" long. I use it to position the microphone directly in front of one driver at a time at exactly the same distance. I use a single USB interface for both input and output so that they are synchronized in time. So this means I'm using the microphone input on the interface and the analog out of the same interface to the analog input on the Hypex. I plan to use the digital input on the Hypex in daily use, but that may not work well for these measurements. I happen to use a Motu USB interface, but that's not important. If I measure three impulse responses they all lay on top of each other with no time shift. That is the goal. There are two channel measurement techniques that can be used, but I don't need that with this setup.
2) I measure each driver individually at 18" and gate the measurement at around 7 ms as this is where I can clearly see the first room reflection show up when I gain up the impulse response. This does not give a good measurement of the bass, but it is low enough to work perfectly for the crossover frequencies of most speakers. In a larger room or outdoor you will see the first reflection happening later and can leverage that for larger distance from the speaker and longer gate times.
3) I work with the Hypex filter designer in real time to first equalize each driver flat in the desired range, adding filters and repeating measurements as I go. After the mid band is flat, I start on the crossover. I use the asymmetric second order shelf filter to raise the 2nd order natural low frequency cut off of the midrange and tweeter up to the desired crossover frequency with a Q of 0.707. This will enable you to produce a perfect second order high pass response without the transition to a fourth order response at the drivers natural low frequency cutoff with the associated extra phase shift at the crossover frequency that thwarts creation of an ideal crossover.
4) I cascade a second high pass filter with Q of 0.707 and frequency at the desired crossover frequency. I now have an ideal LR 4th order alignment for the high pass filter. Something that is nearly impossible to achieve with a passive crossover.
5) If it's a midrange, add in two more second order low pass filters with Q .707. Do this with each driver or amplifier section individually.
6) Match the gain of each driver. No gain on the least sensitive driver. Negative dB gain on all the others to match.
7) Set the delays. Move the microphone back about a meter or two. Running two drivers, with one having inverted polarity ( see Hypex menu diagram) , adjust the delay to get the deepest notch at the crossover frequency.
After doing this the speaker should produce a pretty good response with some tuning required for the bass due to room position. As you move the microphone farther back you get into room response due to floor, ceiling and wall bounce. The Hypex Filter design has a separate set of filters for room EQ in the amplifier setup area if you want to address the room without mucking up the crossover.
Of course you can load all of those .frd files from the drivers measured at 18" into Xsim4 or Vituix CAD and design all the filters in software. Adjusting the delay the exact same amount for all the drivers will flatten out the phase plots so they make sense. I did it with Xsim4 and then transferred the filters into Hypex filter design. The Hypex software also has measurement capability that works pretty well.
1) I have a stick that is 18" long. I use it to position the microphone directly in front of one driver at a time at exactly the same distance. I use a single USB interface for both input and output so that they are synchronized in time. So this means I'm using the microphone input on the interface and the analog out of the same interface to the analog input on the Hypex. I plan to use the digital input on the Hypex in daily use, but that may not work well for these measurements. I happen to use a Motu USB interface, but that's not important. If I measure three impulse responses they all lay on top of each other with no time shift. That is the goal. There are two channel measurement techniques that can be used, but I don't need that with this setup.
2) I measure each driver individually at 18" and gate the measurement at around 7 ms as this is where I can clearly see the first room reflection show up when I gain up the impulse response. This does not give a good measurement of the bass, but it is low enough to work perfectly for the crossover frequencies of most speakers. In a larger room or outdoor you will see the first reflection happening later and can leverage that for larger distance from the speaker and longer gate times.
3) I work with the Hypex filter designer in real time to first equalize each driver flat in the desired range, adding filters and repeating measurements as I go. After the mid band is flat, I start on the crossover. I use the asymmetric second order shelf filter to raise the 2nd order natural low frequency cut off of the midrange and tweeter up to the desired crossover frequency with a Q of 0.707. This will enable you to produce a perfect second order high pass response without the transition to a fourth order response at the drivers natural low frequency cutoff with the associated extra phase shift at the crossover frequency that thwarts creation of an ideal crossover.
4) I cascade a second high pass filter with Q of 0.707 and frequency at the desired crossover frequency. I now have an ideal LR 4th order alignment for the high pass filter. Something that is nearly impossible to achieve with a passive crossover.
5) If it's a midrange, add in two more second order low pass filters with Q .707. Do this with each driver or amplifier section individually.
6) Match the gain of each driver. No gain on the least sensitive driver. Negative dB gain on all the others to match.
7) Set the delays. Move the microphone back about a meter or two. Running two drivers, with one having inverted polarity ( see Hypex menu diagram) , adjust the delay to get the deepest notch at the crossover frequency.
After doing this the speaker should produce a pretty good response with some tuning required for the bass due to room position. As you move the microphone farther back you get into room response due to floor, ceiling and wall bounce. The Hypex Filter design has a separate set of filters for room EQ in the amplifier setup area if you want to address the room without mucking up the crossover.
Of course you can load all of those .frd files from the drivers measured at 18" into Xsim4 or Vituix CAD and design all the filters in software. Adjusting the delay the exact same amount for all the drivers will flatten out the phase plots so they make sense. I did it with Xsim4 and then transferred the filters into Hypex filter design. The Hypex software also has measurement capability that works pretty well.
Last edited:
You only delay as much as needed. So the driver that is the latest (farthest away from microphone) is not delayed.
I was describing how to form the crossover filters for each driver. Fourth order crossovers work very well. A tweeter requires just a high pass filter. A midrange requires a high pass and a low pass filter.
Step 4 is required to produce a 4th order Linkwitz - Riley crossover filter. In the Hypex filter designer a 4th order filter is created by using two 2nd Order filters. To do this two second order filters at the same frequency with Q = 0.707 are cascaded. When used to form a crossover the magnitude response is flat. This filters are frequently used in high performance professional monitors. They are not common in lower cost speakers with passive crossovers because that requires a lot of expensive parts. It's all free with digital crossovers.
https://en.wikipedia.org/wiki/Linkwitz–Riley_filter
https://www.ranecommercial.com/legacy/note160.html
Step 6. The driver that is farthest from the listener requires no delay. All other drivers get some delay to match the time of flight for sound from the woofer.