Thank you for your kind words! A lot of work (and that's an understatement) went into this and I'm glad people enjoy this.
I apologize to everyone that I haven't been able to update this thread for so long even though so much has happened since. This year has just been absolutely crazy for me and I'm just so time strapped. I will place greater priority to do some writing on my progress. But right now I just have too much going on in life.
I've been using a PC as a DSP for the past year now. The miniDSP is just so lacking in features and capabilities. In fact, I've found all hardware DSPs to be very limiting on what I can do. The most powerful and flexible solution is to use the PC and have a VST host. This way the PC provides near unlimited processing power while the VST host can load any type of processing that is available through a VST plugin, and I can write my own VST to do custom processing as well. More will be written in the future about this.
Congrats with your project! Seems like all the hard work was well worth the effort.
I see very much forward reading how you did your crossover. Slopes, amount of tabs, etc
Very much interested in how you managed to get Controlled directivity.
Also- Did you do some special tricks in regards to the Linkwitz transform. IMO when I try it, the bass becomes more "sloppy". It may be because already my bass play plenty low.
With something like REW, JRiver and Rephase I gues something like your build would be possible for those who like to test it.
I see very much forward reading how you did your crossover. Slopes, amount of tabs, etc
Very much interested in how you managed to get Controlled directivity.
Also- Did you do some special tricks in regards to the Linkwitz transform. IMO when I try it, the bass becomes more "sloppy". It may be because already my bass play plenty low.
With something like REW, JRiver and Rephase I gues something like your build would be possible for those who like to test it.
The cardioid part hasn't been done yet. I just don't have time lately. However, it is constant wide directivity above the baffle step transition.
Measurements is really hard, much, much harder and more complex than people think. I've literally done thousands of measurements on my speaker and I still don't have measurements that I would consider to be accurate. I do think I have the methodology down now, but just don't have the time. To give an idea of what's required:
For each axial measurement, it takes 3 splices
- Ground plane measurement for <1000Hz
- Quasi anechoic measurement for 1000-10000Hz in a large space like gym or auditorium to achieve ~10ms gating. The typical 4-5ms reflection free time achieved in a typical room just doesn't have enough resolution
- Flush mounted microphone for >10000Hz measurements to avoid reflections from the mic clip and mic stand.
Then the ground plane measurement has to be compensated because it shows the result of a double height baffle since the ground is a reflector. A compensation factor of the difference between the double height baffle and the normal height baffle needs to be applied.
Then for polar measurements, things get real tricky. The standard way of rotating a loudspeaker on a turntable does not give the true angle and distance because the centre of rotation is at the centre of the speaker instead of at the centre of the baffle. Therefore, there needs to be angle and distance compensation applied. The angle compensation needs to be done physically by moving the mic. Distance compensation can be done after measurement.
So as you can see, for polar measurements, which I do in 10 degree intervals, requires 36 measurements for horizontal, 36 measurements for vertical, so 72 measurements. Multiply 72 by 3 for 3 different drivers, and you get 216 measurements. Then another 72 measurements for the final response for a total of 288 measurements. But wait, I have 3 measurement conditions, ground plane, quasi anechoic, and flush mounted mic! So that's 864 measurements!
Therefore, I have to do 864 measurements to get a complete polar response for this speaker, AND I have to do the baffle, angle, distance, and splicing compensations for all of those measurements! Yes, there are some shortcuts that'll cut the work by half, but that's still a crazy amount of work. One day I'll actually do all that, but not now.
And yes, every one of those complications is important to do in order to get a truly accurate measurement, and it makes a HUGE difference.
I appreciate your suggestion!
We will be satisfied with much less! Just the whole system at 0, 15, 30 etc. angles at 1m with 4-6ms gating depending of room size. We can guess the rest. Perhaps individual drivers at 0 degree.
I understand that you need more for sending and xo.
One detail - set the speaker on turntable so that rotation axis is same as acoustic center!
I understand that you need more for sending and xo.
One detail - set the speaker on turntable so that rotation axis is same as acoustic center!
Since you asked, here is an older polar response of the speaker in vertical orientation (yes, meaning the midrange and tweeter are side by side). There is no smoothing at all in this measurement. I want to repeat this again, because accurate measurements should never be smoothed.
The measurements are in 20 degree intervals, 0, 20, 40, 60, 80, 100 degrees. I have measurements in 10 degree intervals, but not for outdoor ground plane. I have limited battery power when measuring outdoors, so I did the outdoor measurements in 20 degree intervals. <900Hz it is ground plane measurements. Above 900Hz it is quasi anechoic with 10.5ms of gating.
The speaker is even better now, since the minor crossover dip is now moved to ~3100Hz (the most sensitive part of our hearing, and also the region where the ear will be least noticeable to dips), and steered so the deepest dip happens at around 40 degrees instead of 20. The midrange is brought up to be flat with the treble as well. The graph has less wiggles due to further removal of sources of measurement reflection. But I don't have a complete set of measurement yet.
P.S. Putting the speaker so that the rotation axis is at the baffle would cause the speaker to fall off the turntable! It is also a source of unwanted reflection.
Last edited:
Thank you and congratulations , that's extremely good! Low and smooth directivity is what I like and just right for nearfield/multipurpose!
Is bass response splitted with ground plane or nearfield? There are surprisingly little wiggles. REW has variable gating nowdays but this is Omnimic that I don't use much.
Equalization works well when basics are right, like here. I might like to set rising on-axis above 6kHz to compensate increasing directivity. But that is very much a matter of taste and preference, and individual audiometry curve...
Is bass response splitted with ground plane or nearfield? There are surprisingly little wiggles. REW has variable gating nowdays but this is Omnimic that I don't use much.
Equalization works well when basics are right, like here. I might like to set rising on-axis above 6kHz to compensate increasing directivity. But that is very much a matter of taste and preference, and individual audiometry curve...
Thank you! It took a LOT of work to get to this point.
The bass and low midrange response is a ground plane measurement, done in an outdoor parking lot with no boundaries for over 30m. That's why there are no wiggles. It is the only way to get accurate and high resolution measurements for bass and midrange frequencies.
And yes, I do have a rising on axis response for actual usage. However, it starts at 9KHz, and it only serves to match the power response roll off in the tweeter's beaming region because the high frequency rolls off faster than the roll off starting at 6KHz since the tweeter starts beaming.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.