A 3 way design study

@fluid:Thanks a lot.. 🙂 I will read up the posts from mtg90 and others to see the deltas w. r. t expected curves.. Just by looking at above post, it seems like 2dB sort of deltas are to be expected above 3-4kHz without CSL calibration.

In other updates.. my two way speakers are home with their big brothers.. 😀
IMG_20221006_153641.jpg

IMG_20221006_153607.jpg


Next step is to take measurements for the new speakers.

I was wondering if i should be keeping the speakers like this
IMG_20221006_153451.jpg


or like this
IMG_20221006_153142.jpg

In this configuration, the tweeter is exactly at my ear height.. in the ome before, tweeter may be 5-10cm up from my ear height.. 😀
 
So I took some measurements with the Sonarworks sound ID mic for the new 2 way top module and played around with a 3 way configuration with existing woofer box for this configuration:
1665246794408.png


First, I tried to take some measurements to study repeatability. I took a measurement. Went and took the mic along with stand and moved the entire thing away to some random location. Came back after a minute put everything back and tried to take the same measurement again. I did this for the mid driver measurement for 0 degree, 20 degree and 40 degree or so. Here is how the 0 degree measurements looked like (green is first try. red is the second try):
compare_measurements_repeatability.jpg

I got similar results for other angles too.. They look pretty close: 🙂

Mid driver nearfield and impedance measurement
SB15CAC30_in_cab_nearfield.jpg


Here are mid driver horizontal polars up to 180 degrees
SB15CAC30_in_cab_hor_polar.jpg


Tweeter in waveguide: horizontal polars up to 180 degrees
SB26CDC004_in_cab_hor_polar.jpg


Quick and dirty 3-way crossover with the Satori woofer box:
3way_SB_WO24P_15CAC_26CDC_wavguide_v1 Six-pack.png


Even though I don't have vertical polars the ditches in vertical directivity (between mid and tweeter) seem to occur even at small angles with this one...
Now I want some help to do a better crossover 😀
Requesting everyone interested to please take a stab at crossover using the attached Vituixcad project files.. 🙂
 

Attachments

  • Like
Reactions: hifijim
Very excellent work so far, Vineeth... 🙂

Attached is the horizontal polar response of your midrange, no filtering.

1665283632787.png


Are you certain that you have this level of directivity at 50 - 100 Hz? I think that this 5" driver should be nearly omnidirectional at 100 Hz. When I was doing my nearfield merging, I made an error, and I was getting a level of directivity similar to what you show. After I corrected the merging process, the polar curves all blended together below 100 Hz, which I think is closer to reality.

Similarly, here is the polar response of the two woofers
1665283962707.png


It looks suspicious to me. Once again, I think the response should be nearly omnidirectional below 100 Hz.

My mistake was in using the VituixCad merger tool... I failed to check the box "force to gradient". I normally use the default value of "40 Hz" and "100% monopole".

One other question: does your WO24P nearfield response show large peaks at 275Hz and 630 Hz? I would normally expect a woofer to be much smoother between 100 Hz and 1000 Hz.

Once again, great work !

j.
 

Attachments

  • merger.png
    merger.png
    33.7 KB · Views: 76
  • Thank You
Reactions: vineethkumar01
As an example, here is my SB15CAC polar response, measured in my cabinet. My nearfield merge happens around 500 Hz. Notice that at 100 Hz, I have just 1 dB of variation between the 0 degree and 180 degree measurements... by 60 Hz, there is virtually no variation, it is pure omnidirectional.
1665285252697.png


It was Fluid who noticed the error with my measurements, or more correctly, with my merging procedure.

j.
 
Vineeth> can you post the 6-pack with the vertical Power & DI plot?
Hi @augerpro

Based on the artefacts in the measurement merging process pointed out by hifijim and fluid above, some of these graphs might be wrong, especially at low frequencies...
But here are the current plots (I will update these and upload them once we have a proper crossover compared to the one now).
3way_SB_WO24P_15CAC_26CDC_wavguide_v1 Power+DI.png


Six pack:
3way_SB_WO24P_15CAC_26CDC_wavguide_v1 Six-pack_vertical.png



Please let me know if these were the kind of plots you were looking for.
 
Requesting everyone interested to please take a stab at crossover using the attached Vituixcad project files..
Assuming all the driver responses are correct, I came up with a nice reasonable filter. A little less complexity than your initial first draft.

1665288723240.png


The shelf filter right after the amp is a global EQ which tilts down the response from 100 to 10k by -3 db. I find it easier to do development and measurement work with a flat on-axis speaker, and then apply global EQ tilt later.

1665288898486.png

Between 1k and 3k there is a -1 dB dip in the power response... considering that the current simulation assumes horizontal is equal to vertical, 1 dB is well within the measurement uncertainty.

The woofer response from 100 Hz to 1000 Hz still bothers me a bit. Why the big peaks and dips... ? Maybe it is real, but Vineeth - I recommend confirming the validity. If those peaks and dips are not real, and you apply EQ to flatten out the mythical response, the result will sound bad.

Your turn to be the killjoy this time, but the point is valid,
Engineering and science can be very humbling. When I find the truth, it is rarely a soft gentle landing... it is like flying high speed in the fog and crashing into a stone wall.
 
I think the SB15 is a bit tricky to get right in this respect. In mainframe99/McFly's thread on his very similar speakers the same sort of midrange off axis rise is there in the first measurements. When he swapped it to a Scanspeak Revelator that sounded better without the midrange glare guess what isn't there anymore.

Original SB15
https://www.audiosciencereview.com/.../revel-m105-copy-diy-build.29465/post-1029858

Scanspeak
https://www.audiosciencereview.com/.../revel-m105-copy-diy-build.29465/post-1158083
 
  • Thank You
Reactions: vineethkumar01
I think that one problem with this speaker is the midwoofer as midrange. SB15 radiating diameter is actually 4" and the surround rubber is very "heavy", making response and directivity in upper band difficult. However, the ceramic SB15 is pretty good until way up.

https://www.madisoundspeakerstore.c...coustics-sb15cac30-4-5-ceramic-woofer-4-ohms/

A real 4-5" radiating diameter and frame ¤150mm midrange and perhaps with inverted roll surrond is easier to lowpass 2-3(4) kHz even with LR2.

https://www.madisoundspeakerstore.com/approx-5-midrange/scanspeak-15m/4624g-discovery-5.25-midrange/
https://www.madisoundspeakerstore.c...ige-mca15rcy-h1262-5.5-coated-paper-midrange/

Another unavoidable thing making DR rollercoaster is vertical response around MT xo is interference because of vertical separation of mid and tweeter. Letting on-axis response to have a bump there compensates for this, and can be heard as bump or "harshness" only in nearfield, IMO!
 
  • Thank You
Reactions: vineethkumar01
^This lead me to take quickie handheld vertical polars of different type of speakers that I have. In-room at roughly 80cm distance, speakers only a bit drawn out from their natural locations... so I used 1/3 smoothing and 6ms gating.

Adam is Adam Audio A7V 2-way with approx. LR4 xo around 2 kHz
Ainogradient is my 4-way dipole with LR2 xo at 300/1kHz and 4kHz
Acalanche is my modified Avalanche AS1 with LR2 xo at 300/3500 Hz
AW-7 is a 2-way coaxial SEAS 7" with asymmetric xo ~2,5kHz

Coaxial is superior here, very smooth behaviour around xo. Typical coax wiggles above 10kHz
2-way has largest dip in Averaged response, becaus interference nulling stays at same frequency
3-way looks very uneven, but Averaged is very smooth, because interferences avy a lot with angle and freq
4-way dipole shows similar behaviour as 3-way, but off-axis response spl drops much more off-axis

Choose our poison! One trick to get smooth vertical off-axis is to use asymmetric xo, like kimmosto used to do in his old designs.

Verticals tile.jpg
 
Last edited:
  • Thank You
Reactions: vineethkumar01
Vineeth,

With DSP; can you try and brute force an acoustic LR2 around 2.5KHz?

I haven’t used the SB15, only the SB17, but haven’t spent much time with it for a crossover. I’m not sure can do a LR2.

If it can, you’ll be able to avoid that off axis rise @3KHz that fluid noted. And you will get a smoother and gradually falling power and in-room response.

In general you need a driver that is flat and smooth out past 2-3 times Fc. Eg. 8-10KHz, on axis. And off axis it doesn’t do anything wildly incongruous with the on axis. Some paper and poly cones can do this, and hard cones have difficulty. I haven’t proven this however….
 
Last edited:
  • Thank You
Reactions: vineethkumar01