Hi all!
I’m tuning the active crossover (CamillaDSP running on a Pi) in my 4-way system with side-firing woofers, and I’m having a hard time understanding one issue.
My listening position is about 2.7 meters away from the speakers, and my ears are at the same level as the mid-dome. This puts my ears approximately 2.975 meters from the side-firing woofers.
If there is no delay configured in CDSP between the woofers and the other three drivers, I observe a significant dip at the crossover frequency (around 125 Hz), which makes sense to me (blue line). Based on my calculations, assuming the speed of sound is 343 mm/ms, the sound from the woofers will reach my ears about 0.8 ms later than the sound from the other drivers. So, I added a delay of 0.8 ms to the other drivers (green line), and while there was some improvement, it didn’t fully achieve the expected summation of the individual drivers' measurements.
I then experimented with several different delay values, taking measurements after each adjustment, and found that a delay of 2.5 ms gives me a response (yellow line) that accurately reflects the summation of the woofer and midbass response.
What I’m trying to understand is: Why is there a difference between the measured results and the math?
I’m tuning the active crossover (CamillaDSP running on a Pi) in my 4-way system with side-firing woofers, and I’m having a hard time understanding one issue.
My listening position is about 2.7 meters away from the speakers, and my ears are at the same level as the mid-dome. This puts my ears approximately 2.975 meters from the side-firing woofers.
If there is no delay configured in CDSP between the woofers and the other three drivers, I observe a significant dip at the crossover frequency (around 125 Hz), which makes sense to me (blue line). Based on my calculations, assuming the speed of sound is 343 mm/ms, the sound from the woofers will reach my ears about 0.8 ms later than the sound from the other drivers. So, I added a delay of 0.8 ms to the other drivers (green line), and while there was some improvement, it didn’t fully achieve the expected summation of the individual drivers' measurements.
I then experimented with several different delay values, taking measurements after each adjustment, and found that a delay of 2.5 ms gives me a response (yellow line) that accurately reflects the summation of the woofer and midbass response.
What I’m trying to understand is: Why is there a difference between the measured results and the math?
Are you using standard filters otherwise? Delay is affected by more than just distance. You may need other non-standard filters to reach the goal.
This is no different to any other crossover in that respect. An active configuration doesn't relieve you of regular crossover concerns, such as the work often done in a simulator. This can include spatial design as well as geometric and electrical.
This is no different to any other crossover in that respect. An active configuration doesn't relieve you of regular crossover concerns, such as the work often done in a simulator. This can include spatial design as well as geometric and electrical.
There are other filters in this frequency range: an LPF (LR24 at 180Hz) on the woofer, a low shelf to slightly boost the bass, and a few PEQs to control room modes.
If I understand you correctly, are you suggesting that these other filters might be contributing to the issue?
If I understand you correctly, are you suggesting that these other filters might be contributing to the issue?
No, not if they're there for the right reason.
You can't apply LR24 to each driver and expect it to work. They have differences in response and phase which you should consider and manage like always.
You can't apply LR24 to each driver and expect it to work. They have differences in response and phase which you should consider and manage like always.
I think it is just phase vector math that threw you off. Blue down whopping ~12dB vs Yellow meant XO'ed drivers nearly antiphase (off 180deg) at 125hz (one cycle = 8ms). Green = Blue adjusted 0.8ms (1/10 cycle or 36deg) is still 1.7ms (76.5deg) from Yellow and down ~4.5dB. If you drew a circle with three radii Yellow Green Blue their vector-sum with the other driver (nearly antiphase from Blue) will work out lengthwise in the measured SPL ratios (approximately).I then experimented with several different delay values, taking measurements after each adjustment, and found that a delay of 2.5 ms gives me a response (yellow line) that accurately reflects the summation of the woofer and midbass response.
What I’m trying to understand is: Why is there a difference between the measured results and the math?
Disclaimer: I've never studied this, only worked out the geometry from first principles.
Could you define your measurement setup please? Frequency wise it could be a floor bounce issue you 'compensated' by using delay... frequency domain mod should be accompagnied by time domain check to be sure you really treated the issue at those freq range imho.
There could be latency issue within your dsp too...
Well, I’ll stop being lazy and take proper measurements this time, doing the best I can. I used some scrap metal I had lying around to build a minimalist mic stand, and I’ll take measurements for each driver. Will take measurements at 1m, 2m and at 2.7m, which is the distance of the speaker to my listening position. WIll try to take off-axis measurements too.
As you can see, the woofers are mounted on the sides (two, positioned opposite each other). I’m not exactly sure how to measure and model them in VituixCAD. I’ll do some research, but if any of you can point me in the right direction, I’d really appreciate it.
As you can see, the woofers are mounted on the sides (two, positioned opposite each other). I’m not exactly sure how to measure and model them in VituixCAD. I’ll do some research, but if any of you can point me in the right direction, I’d really appreciate it.
What is the side firing woofer upper bandwidth limit? Something in the 100hz/150hz area?
If so then don't bother: closemicing will do as they will behave as an omni source anyway. If xover an octave higher then there could be start of issue but otherwise...
Multiple distance measurements will tell if the issue you seen was linked to floor bounce or not. For me and with basic guessing this is what happen.
But without actual dimension of your loudspeakers and the xover freq....
If so then don't bother: closemicing will do as they will behave as an omni source anyway. If xover an octave higher then there could be start of issue but otherwise...
Multiple distance measurements will tell if the issue you seen was linked to floor bounce or not. For me and with basic guessing this is what happen.
But without actual dimension of your loudspeakers and the xover freq....
Last edited:
I use rew and a Dayton audio calibrated USB microphone (UMM-6).
But you are probably right, the measurements that I was using were all in-room, not gated (i.e., all reflections and everything else were interfering).
The midbass can handle properly the 90-100Hz range. Below that it starts to loose SPL, so the minimum XO frequency for the woofers would be 100-120Hz. At this moment, in my currenct CamillaDSP config, there is A LPF set at 180Hz, but it has nothing to do with measurements and math. I just set this up, liked the sound and there it is... 🙃
I was able to take the measurements (all at mid high level, as it is my listening position):
- Each driver at 1m, on axis
- Each driver at 2m, on axis
- Each driver at 2.7m, on axis
- Each driver at 2.7m, 15 degrees off axis
- Each driver at 2.7m, 30 degrees off axis
Then it started to rain....... 😂
I will work with what I was able to measure. But I already can say that woofers measurements are a mess. Maybe because of reflections on the table where the speakers were.
Ok, if the side firing woofer goes to 180 then you are seeing a floor bounce null imho:
Try simulate it with this ( use only one of the side woofer coordinate or an average at box width center) and from there see your measurement and if it tracks... usually it does, we can't break laws of physic and acoustic. 😉 this kind of reflection cannot be gated and will be present whatever you do as long you try to simulate your listening position. That's why you should use close micing technique ( or on a pole at 6m height and away from any reflective plan...)to be sure it's not an issue within the box/driver.
https://tripp.com.au/sbir.htm
Totally fine you seted up your filter to taste as if i understood corectly you wanted to take a listen at the build in a hurry. As long as it please you it's perfectly fine, that's the point.
Now if you want to perfect things then yes measurements will bring a step ( many steps) further.
Try simulate it with this ( use only one of the side woofer coordinate or an average at box width center) and from there see your measurement and if it tracks... usually it does, we can't break laws of physic and acoustic. 😉 this kind of reflection cannot be gated and will be present whatever you do as long you try to simulate your listening position. That's why you should use close micing technique ( or on a pole at 6m height and away from any reflective plan...)to be sure it's not an issue within the box/driver.
https://tripp.com.au/sbir.htm
Totally fine you seted up your filter to taste as if i understood corectly you wanted to take a listen at the build in a hurry. As long as it please you it's perfectly fine, that's the point.
Now if you want to perfect things then yes measurements will bring a step ( many steps) further.
Last edited:
@wchang
And this is the reponse of the MidBass:
The woofer is at -90 degrees and the midBass at +15 degrees aprox. If you combine the phase difference with the time arrival difference, that would explain why a 0.8ms delay was not enough. That would also match the explanation in your amazing drawing, @wchang 😂
But again, my measurements where very primitive. Now that I have better ones, I am already playing with VituixCAD. I'll let you know how it goes.
For the drivers to be almost 180 degrees out of phase, other filters have to be playing a part in this. It makes a lot of sense. Look at the response plot of the woofers:
And this is the reponse of the MidBass:
The woofer is at -90 degrees and the midBass at +15 degrees aprox. If you combine the phase difference with the time arrival difference, that would explain why a 0.8ms delay was not enough. That would also match the explanation in your amazing drawing, @wchang 😂
But again, my measurements where very primitive. Now that I have better ones, I am already playing with VituixCAD. I'll let you know how it goes.
Last edited: