Semi classic 3-way with AMT (off-axis measurements included)

[XMechanik]

Trying a 3-way design. The goals as always: balanced SPL and directivity related characteristics (acoustic power response, in-room response etc).

Drivers used:
tweeter : AMT2-4 link
midrange: RS100-4 link
woofer: DC160-8 link

I started from what I had of the shelf and these were RS100-4 and DC160-8. I had also a dome tweeter but I thought it was a good opportunity to play around with ribbon or AMT driver as the midrange (RS100-4) is a fullrange driver and can be filtered high.

The box is made of 18mm birch plywood with external dimensions 185x395x293mm. The volume for the woofer is approximately 12.5 liters, for the midrange there is a closed chamber with a volume of 1 liter. Bass reflex for the woofer tuned to ~40Hz. The box is filled with polyester fiber.

Front panel diagram above, dimensions in mm.

Due to the asymmetrical arrangement of the drivers, the measurements were taken for angles 0-350 on both orbits (H and V). Upgrading Clio Pocket to version 3 was very helpful for this task. It has an auto saving functionality, which makes the collection of off-axis characteristics few times faster, also with using a manual rotary table.

During the measurement session with manual angle adjustments you have to walk a bit. Out of curiosity, I estimated that a distance of about 1.3 km is covered (36 measurements per orbit x 2 orbits x 3 transducers x 3 m distance from the laptop to the rotary table + return).

After processing and merging I entered all characteristics into XMachina software, which generated a number of solutions. I transferred one of them to VituixCad.

Attached VituixCad *.VXP file, FRD and ZMA measurements (for what it's worth): LINK

The circuits are not complicated and could be further simplified. It seems that applying R7-C10 branch was not nesesery, it creates a trap that removes a peak in the woofer response at 3kHz. The peak untrapped is over 20dB lower than the system spl so I think the branch could be safely omitted.

The filters are assembled on an universal PCB with copper strips along the entire board. Filters are mounted with a screw on an anti-vibration pad.


In the midrange THD level is lower than the fundamental component by more than 50dB (1m, 86dB)


This is great improvement comparing to my previous design (link). For that design (with SB and Scan Speak) it was only 37dB. I suspect it was caused by the midwoofer cone breakup resonance which was exited by third harmonic leakage from the midwoofer's pass band. The midrange driver in this 3-way system has small diameter comparing to the midwoofer in the 2-way (10cm vs 15cm), it's breakup frequency is higher and not so easy to reach by harmonics from the pass band. From the distortion perspective the system is limited by the woofer as it's is not a Xmax champion

 

[eddantes]

looks very nice!

 

[alex-z]

Great work!

Preference score comes out to 5.7, or 7.9 with sub. Although those numbers may be artificially high, as the smoothing introduced by gating boosts them.

For anyone wondering, here is what the speaker looks like with the 50dB scaling, and polar maps that Amir normally uses. Normalized to SPL with 3dB contour lines.

 

[XMechanik]

Preference score comes out to 5.7, or 7.9 with sub. Although those numbers may be artificially high, as the smoothing introduced by gating boosts them.

I think you're right. Here is an example from Toole's book and it may be similar here. the impedance graph shows a trace of resonance at 370 Hz. It is almost invisible on the SPL characteristics, but the measurement conditions allowed only for 5 ms gate, for the example from the book it was much better, i.e. 10 ms and the resonance was masked.

 

[sweetchaos]

Can you generate the typical CEA-2034 graph please?

 

[XMechanik]

Can you generate the typical CEA-2034 graph please?

Is that it?

 

[sweetchaos]

It's not a complete graph.
Here's one:

from manual:

VituixCAD help 2.0

 

[XMechanik]

It's not a complete graph.

Early Reflections DI added. Should be complete now.

 

[Matt_Holland]

I started from what I had of the shelf and these were RS100-4 and DC160-8. I had also a dome tweeter but I thought it was a good opportunity to play around with ribbon Bass reflex for the woofer tuned to ~40Hz. The box is filled with polyester fiber.

Nice! Very handsome cabinets.
From the impedance graph it looks like the reflex tuning is around 35-37Hz and from the SPL graph the bass alignment looks de-tuned. Is this the case? If so was it deliberate and did you consider a higher tuning frequency to get more SPL between 50-100Hz?

 

[XMechanik]

Nice! Very handsome cabinets.
From the impedance graph it looks like the reflex tuning is around 35-37Hz and from the SPL graph the bass alignment looks de-tuned. Is this the case? If so was it deliberate and did you consider a higher tuning frequency to get more SPL between 50-100Hz?

Thank you. BR tuning was aiming 40Hz (but it turned out to be maybe 1-1.5Hz less) because it resulted in the highest level at 40Hz in the enclosure simulations. Even though the simulations are not fully confirmed in reality I don't feel there is lack of 50-100Hz representation.

 

[BenB]

Nice project. Thanks for sharing. Any reason the mid is more offset horizontally than the tweeter is?

 

[Trouble Maker]

I'm a novice, so I'm curious about a few things.
I'm trying to understand how the driver alignment affects the results.
Is on axis in alignment with the tweeter?
Why not align the tweeter and mid? <-Sorry, just saw the same question above.
What positives and negatives do you see in offsetting the mid and tweeter from the woofer?

Very nice results overall!

 

[abdo123]

Does this sound exciting in real life?

The spinorama has a very steep roll-off. Usually i aim for 0.6 to 0.8 db/oct

 

[XMechanik]

Nice project. Thanks for sharing. Any reason the mid is more offset horizontally than the tweeter is?

I'm trying to understand how the driver alignment affects the results.
Is on axis in alignment with the tweeter?

I simply tried to reduce diffraction effects by changing locations of the drivers. As a result of the horizontal offset the main lobe may have a slight side tilt near the crossover frequency, but I can't imagine whether this could have any audible effects.
Attached results from VituixCad diffraction tool with pass bands marked.

 

[XMechanik]

Does this sound exciting in real life?

The spinorama has a very steep roll-off. Usually i aim for 0.6 to 0.8 db/oct

I'd rather not bright the sound any further. But maybe that's because my listening environment is quite reflective.

 

[DearSX]

View attachment 328763
Trying a 3-way design. The goals as always: balanced SPL and directivity related characteristics (acoustic power response, in-room response etc).

Drivers used:
tweeter : AMT2-4 link
midrange: RS100-4 link
woofer: DC160-8 link

I started from what I had of the shelf and these were RS100-4 and DC160-8. I had also a dome tweeter but I thought it was a good opportunity to play around with ribbon or AMT driver as the midrange (RS100-4) is a fullrange driver and can be filtered high.

The box is made of 18mm birch plywood with external dimensions 185x395x293mm. The volume for the woofer is approximately 12.5 liters, for the midrange there is a closed chamber with a volume of 1 liter. Bass reflex for the woofer tuned to ~40Hz. The box is filled with polyester fiber.

View attachment 349276
Front panel diagram above, dimensions in mm.

Due to the asymmetrical arrangement of the drivers, the measurements were taken for angles 0-350 on both orbits (H and V). Upgrading Clio Pocket to version 3 was very helpful for this task. It has an auto saving functionality, which makes the collection of off-axis characteristics few times faster, also with using a manual rotary table.

During the measurement session with manual angle adjustments you have to walk a bit. Out of curiosity, I estimated that a distance of about 1.3 km is covered (36 measurements per orbit x 2 orbits x 3 transducers x 3 m distance from the laptop to the rotary table + return).

After processing and merging I entered all characteristics into XMachina software, which generated a number of solutions. I transferred one of them to VituixCad.
View attachment 328764
Attached VituixCad *.VXP file, FRD and ZMA measurements (for what it's worth): LINK

The circuits are not complicated and could be further simplified. It seems that applying R7-C10 branch was not nesesery, it creates a trap that removes a peak in the woofer response at 3kHz. The peak untrapped is over 20dB lower than the system spl so I think the branch could be safely omitted.

The filters are assembled on an universal PCB with copper strips along the entire board. Filters are mounted with a screw on an anti-vibration pad.
View attachment 328765 View attachment 328768

In the midrange THD level is lower than the fundamental component by more than 50dB (1m, 86dB)
View attachment 328769

This is great improvement comparing to my previous design (link). For that design (with SB and Scan Speak) it was only 37dB. I suspect it was caused by the midwoofer cone breakup resonance which was exited by third harmonic leakage from the midwoofer's pass band. The midrange driver in this 3-way system has small diameter comparing to the midwoofer in the 2-way (10cm vs 15cm), it's breakup frequency is higher and not so easy to reach by harmonics from the pass band. From the distortion perspective the system is limited by the woofer as it's is not a Xmax champion

Hi Xmechanic,

Would this design work with the woofer on the side like the ATC SCM 25A or KH310A?

 

[XMechanik]

I don't think it's a good idea. The woofer operates up to ~0.65kHz and it's not a frequency where 90deg driver axis change could be omitted because there is noticeable directivity there.

 

[DearSX]

I don't think it's a good idea. The woofer operates up to ~0.65kHz and it's not a frequency where 90deg driver axis change could be omitted because there is noticeable directivity there.
View attachment 380407

Thank you for response and explanation!