Modular active 3 way - work in progress

I also refactored my subwoofer to compare it to the BP6 using the same DCS305 driver.

The box volume (55L) lets me try it as either a BR or sealed (Q=0.66). The port pipe exits the front to make room placement easier (only one side radiates).

The sim uses the published T/S and the simple inductance model. The measurements are nearfield.
.

This is a brilliant idea to feed the port from the rear of the box! I have never seen that before.
 
Thanks. The port out the front is less common but there are other examples. I already prefer it over the BP6.

Last year I did some experiments on a basic BR and port location. I tried the side and back with the port tube flush inside, flush outside, and halfway. It made no significant difference in performance. It seemed reasonable as the wavelengths are ~10m so there was little effect changing the distance between the port and driver. May as well place it where it benefits me :)
 
Last edited:
Great work!

That horn is definitely a constant directivity device. A copy of the QSC horn that became popular with the diy crowd about a decade ago. One of the best measuring waveguides around if you don't count diagonals.

Not really good below 1khz as you have noticed. Modern 1" drivers aren't really designed to be used that low. Might have decent luck with a BMS 5530.

A 1.4" exit driver would be much more at home below 1khz in a home setting. Something like this would match up nicely.
Eighteen Sound - Professional loudspeakers

Thanks. The PHRN-1014 is a good neutral sounding horn and it had good reviews so I tried it out. The D290py was suppose to go low enough so it might be this horn (advertised to load to 600Hz). I had better low end on the larger SH-402 (JBL2370 clone). I could raise the XO point a little or as you suggested use a larger driver + horn, once I settle on what I prefer.
 
comparing polars 1

I'm satisfied with the sub and woofer arrangements so moving up to compare the polars.

The measurements are @1m @midrange height and gated 5ms. The raw measurement is provide then the "normalized to 0 deg" plot is provide to see the polar with axial variations removed. You can see exactly where the polar narrows from the midrange horn XO=725Hz.

The random ebay diffraction slot tweeter is OK from 0-40deg but at more extreme angles 40-60deg it looks like there is interference from the tweeter diagonal @XO=4550Hz. Both horns are relatively constant directivity. Some of this may be from shifting the horns to get proper axial time alignment. I might look into this to see if it can be improved.

The HE07 is rounded (elliptic) and does not show interference and I assume its because its diagonal pattern is better.

.
 

Attachments

  • DSCN8047crop.jpg
    DSCN8047crop.jpg
    103.7 KB · Views: 311
  • Polar PHRN1014  EbayDifrSlot  10 deg steps.jpg
    Polar PHRN1014 EbayDifrSlot 10 deg steps.jpg
    101.3 KB · Views: 314
  • Polar Normalized PHRN1014  EbayDifrSlot  10 deg steps.jpg
    Polar Normalized PHRN1014 EbayDifrSlot 10 deg steps.jpg
    102.2 KB · Views: 314
  • DSCN8064r.jpg
    DSCN8064r.jpg
    89.7 KB · Views: 309
  • Polar PHRN1014  H07E  10 deg steps.jpg
    Polar PHRN1014 H07E 10 deg steps.jpg
    101.6 KB · Views: 308
  • Polar Normalized PHRN1014  H07E.jpg
    Polar Normalized PHRN1014 H07E.jpg
    99.7 KB · Views: 100
comparing polars 2

The cone + dome solution has less directivity control than the previous horns. It also has baffle diffraction to deal with.

The on axis diffraction prediction (pic#4) was EQ'd flattish (on axis) by applying -3dB@1Khz and +2dB@2Khz corrections. However, looking at the normalized polars it shows that same correction came back in the off axis at 1Khz and 2Khz. The J.Bagby diffraction calculator is consistent with the ABEC BEM sim in post#186

The diffraction needs to be corrected physically by sculpting the baffle for both the midrange and tweeter. Something like the correction in post#186 Modular active 3 way - work in progress

.
 

Attachments

  • DiffractionCalc-RS52AN.jpg
    DiffractionCalc-RS52AN.jpg
    288.5 KB · Views: 88
  • Polar Normalized - RS52AN DC28F 10 deg steps.jpg
    Polar Normalized - RS52AN DC28F 10 deg steps.jpg
    95 KB · Views: 92
  • Polar - RS52AN DC28F 10 deg steps.jpg
    Polar - RS52AN DC28F 10 deg steps.jpg
    95.6 KB · Views: 139
  • DSCN7821crop.jpg
    DSCN7821crop.jpg
    165.3 KB · Views: 136
Just another Moderator
Joined 2003
Paid Member
That dip at ~4.5K is definitely odd. I assume if you measure just the tweeter it isn't there.... The H07E definitely looks a lot better.

Can you notice the off axis iregularities in listening in room?

The cone + dome definitely looks a bit amiss around 2K (is that where the crossover is?) It is making me appreciate even more how well my Mid's and tweeters work together. They are not without their problems but their polars are very consistent (not CD but uniform)

I've attached my polar it is 10deg off axis measurements of my MTM. it's using a semi horn loaded dome tweeter. The 2K dip is an issue with the mids/cabinet and I've never been able to fix it, but it doesn't seem to cause too much of an issue, not sure about the one a bit above 5K but I think it is a diffraction problem.
edit: note that in the graph below crossover is 4th order (acoustic) at 2.8Khz bessel.

Tony.
 

Attachments

  • mtm_polar.png
    mtm_polar.png
    46.1 KB · Views: 91
Last edited:
Hi Tony,

Nice graph. I've always wanted to try Bessel filters.

The ebay slotted diffraction horn tweeter has the 40-60deg polar dip, IMO is because my measurement position is at the midrange horn and I'm catching the tweeter's diagonal field. It's a square mouth so diagonals are usually not good. It's also why the other H07E tweeter horn is better as it has no corner diagonal, its more rounded. I don't notice it when listening to them as I'm usually within the 40deg. However the horn solution sounds very different from the cone/dome even when they're EQ'd flat.

The XO points are BW2@80Hz, LR4@725 and LR4@4550. I think the polar variation in the mid/high is from diffraction. So a few experiments are coming.
 
polar 3

This is the start of sculpting the baffle for mid/high improvements. It's only the first layer so it should improve most of the MF+HF. Measurements @1m at midrange height gated 5ms.

There is more variation in 700-1000Hz because IMO there is change in width from the woofer cabinet. There will be more baffle layers added to widen the back for LF.

There is an improvement seen in the normalized polars. The baffle got narrower and the diffraction ripple is reduced from -3dB@1Khz / +2dB@2Khz to -1dB@2.5Khz / +1dB@6Khz. That is good enough for me.

I also replaced the DC28F (1-1/8") with a TD20F (3/4") hoping to improve the HF off axis. It actually does but it's still -3dB @30deg @15Khz off axis. I think the only way to improve the tweeter's offaxis is a shallow waveguide.
.
 

Attachments

  • Polar Normalized QuadWoofer RS52AN TD20F.jpg
    Polar Normalized QuadWoofer RS52AN TD20F.jpg
    100.5 KB · Views: 93
  • Polar QuadWoofer Rs52AN TD20F.jpg
    Polar QuadWoofer Rs52AN TD20F.jpg
    105.3 KB · Views: 126
  • DSCN8065r.jpg
    DSCN8065r.jpg
    87.1 KB · Views: 126
Last edited:
polar 4

I was curious if adding a wider back, the same width as the woofer cabinet would help the low mid polar.

First pic is an additional layer with cove cut to transition to a flat surface. The second pic is the flat baffle surface, same width as the woofer cabinet.

Clearly more transitions are making the polar bumpier. I think I would be better off with a simple trapezoid with radius edge. :)
.
 

Attachments

  • Polar Normalized- Layer3 added - QuadW RS52 TD20.jpg
    Polar Normalized- Layer3 added - QuadW RS52 TD20.jpg
    92.8 KB · Views: 81
  • Polar - Layer3 added - QuadW RS52 TD20.jpg
    Polar - Layer3 added - QuadW RS52 TD20.jpg
    93.1 KB · Views: 102
  • DSCN8067r.jpg
    DSCN8067r.jpg
    87 KB · Views: 101
  • Polar Normalized - Layer2 added - QuadW RS52 TD20.jpg
    Polar Normalized - Layer2 added - QuadW RS52 TD20.jpg
    90.8 KB · Views: 88
  • Polar  - Layer2 added - QuadW RS52 TD20.jpg
    Polar - Layer2 added - QuadW RS52 TD20.jpg
    91.9 KB · Views: 107
  • DSCN8066r.jpg
    DSCN8066r.jpg
    101.2 KB · Views: 112
Just another Moderator
Joined 2003
Paid Member
Interesting! great for really clearly showing how the effects of diffraction show up in a polar response too! My comment about diffraction on my own polars is clearly wrong! :)

Can you explain what you are doing for the normalized response graphs? Is it the response if it is eq'ed flat on axis?

Tony.
 
Last edited:
Hi Tony,

I can EQ out the diffraction effects flat on axis but I can't escape them in the polars. Physical problems need physical solutions. :)

I EQ flattish (but not obsessively) on axis, take my measurements (gated 5ms + FD8 to smooth), then normalize by dividing all measurements (in REW) by the 0 deg on axis measurement. So the normalized 0 deg curve is now a flat line at 0dB. It's easier to see the polar variations in the normalized curves because it removes the driver's axial variations.
.
 
diyAudio Moderator
Joined 2008
Paid Member
I can EQ out the diffraction effects flat on axis but I can't escape them in the polars. Physical problems need physical solutions.
This is good advice.

Regarding what and not to take out of polars to make artefacts easier to see, in the first attachment I can differentiate resonances and diffraction effects on this non-normalised plot.

Once worked on and improved, the resultant adjustments would not necessarily look like (be representative of) any of the axes. It is most thorough when done by hand.

If anything, there will be the least discrepancy at some unique axis between 10-40 degrees typically. 0 degrees isn't necessarily a good choice. I've shown the same plot normalised to 0 degrees to show how unhelpful that was in this case.
 

Attachments

  • po2.gif
    po2.gif
    107.7 KB · Views: 419
  • po.gif
    po.gif
    68.1 KB · Views: 417
diyAudio Moderator
Joined 2008
Paid Member
I see your point. I think we mostly agree, but I think not all others would see the signs before making a normalisation decision. Earl Geddes does not normalise in preparation for inspection either, but goes through by hand the way I've learned to.
 
polars

Another kick at the can, trying for a smoother polar.

This trapezoidal design uses the RS52AN and DC28F. The sides slope in at 15deg (off vertical) and the edges are R20mm. Measurements were made at midrange height, @1m, and gated 5ms+FDW8.

There is still a midrange bump (now @2.5KHz) in the normalized polars, however it's moved higher because the avg baffle width has decreased. The diffraction peak has lower amplitude and has been spread out in frequency. I have a new 9Khz dip at >40deg off axis however the LF-MF XO region is similar to the box.

I think it would be better if I used the smaller TD20 because I could increase the slope from 15deg to 20deg (or more) to further spread out the bump. The TD20 also has better HF off axis performance.
.
 

Attachments

  • DSCN8068r.jpg
    DSCN8068r.jpg
    86 KB · Views: 371
  • Polar - QuadW RS52 DC28.jpg
    Polar - QuadW RS52 DC28.jpg
    106.1 KB · Views: 372
  • Polar Normalized - QuadW RS52 DC28.jpg
    Polar Normalized - QuadW RS52 DC28.jpg
    106.9 KB · Views: 377
Although I fully realize I am only giving an opinion here, but the normalized plots of the RS52/DC28 i.m.h.o. demonstrate why many people seem to like the "clarity" or "transparancy" of such small dia mid/tweet combinations. Not having crossing off axis curves may lead to more laid back presentation . It would be an interesting topic to research somewhat deeper and check whether this is just an assumption.
 
Hi Boden, opinions are welcome. The system polars are different, but the horn solutions don't have crossing polars either. I've also heard that statement, that people prefer cone/dome over horn. I'm still deciding so I'm still carrying multiple designs.

The small domes have less directivity until 10KHz+ as seen in the curves being tighter together. So they might be filling in the room more than the horn solution, or maybe less sensitive to a sweet spot. Still, it would be interesting to understand if there is a technical basis for that preference and if its consistent across different systems.
 
Last edited: