Yes I agree. I am humbled that so many people have contributed so much thought to my simple project.
While I am waiting for crossover parts to arrive, I did some experimenting with the passive radiator.
I currently have 30 g attached to the DS270 PR. This is the added mass predicted by VituixCad to give the optimal bass response. This PR design has a convenient threaded connection on the rear of the cone (where the voice coil would be) and was supplied with a bolt. By adding washers of various sizes, it is easy to adjust the added mass in increments as small as 2 g.
One thing that is interesting is the effect of woofer impedance on the passive low-pass filter. In the plot below, we can see that there is a 1.6 dB gain at 80 Hz, and then a -0.9 dB dip at 38 Hz. The original Vcad woofer+enclosure sim did not take this filter transfer gain into account, and I was curious if the optimal tuning was no longer 30 g added mass.
I removed one of the PRs, and experimented with reducing the added mass from 30 g, to 15g, and then to 0 g. I measured the near field response of the woofer and PR, and also the impedance for each test case. For each case, I merged the NF responses with the woofer on-axis far field gated response, also accounting for the transition from 2-pi NF measurement to 4-pi.
Here are the results for the 30 g and 0 g cases, using the woofer circuit that will be employed. I was hoping that reducing the mass would give a nice boost in the 40-50 Hz range, but it is not as much as I had hoped. The 0 g case raises the response at 50 Hz by 0.8 dB. The two cases have equal output at 40 Hz, and below 40 Hz the 30g case rolls off more slowly. The test case of 15 g added mass was evenly between the 30g and 0g cases.
Solid line = 0 g added mass. Dash line = 30 g added mass
The change in filter gain is also quite similar. The 0 g case has a gain of 1.6 dB at 80 Hz and a dip of -0.7 dB at 43 Hz… not much difference.
All in all, it is such a small difference that I will have to make this decision subjectively.
j.
And just now a box of capacitors and inductors was delivered... time for the next phase.
I currently have 30 g attached to the DS270 PR. This is the added mass predicted by VituixCad to give the optimal bass response. This PR design has a convenient threaded connection on the rear of the cone (where the voice coil would be) and was supplied with a bolt. By adding washers of various sizes, it is easy to adjust the added mass in increments as small as 2 g.
One thing that is interesting is the effect of woofer impedance on the passive low-pass filter. In the plot below, we can see that there is a 1.6 dB gain at 80 Hz, and then a -0.9 dB dip at 38 Hz. The original Vcad woofer+enclosure sim did not take this filter transfer gain into account, and I was curious if the optimal tuning was no longer 30 g added mass.
I removed one of the PRs, and experimented with reducing the added mass from 30 g, to 15g, and then to 0 g. I measured the near field response of the woofer and PR, and also the impedance for each test case. For each case, I merged the NF responses with the woofer on-axis far field gated response, also accounting for the transition from 2-pi NF measurement to 4-pi.
Here are the results for the 30 g and 0 g cases, using the woofer circuit that will be employed. I was hoping that reducing the mass would give a nice boost in the 40-50 Hz range, but it is not as much as I had hoped. The 0 g case raises the response at 50 Hz by 0.8 dB. The two cases have equal output at 40 Hz, and below 40 Hz the 30g case rolls off more slowly. The test case of 15 g added mass was evenly between the 30g and 0g cases.
Solid line = 0 g added mass. Dash line = 30 g added mass
The change in filter gain is also quite similar. The 0 g case has a gain of 1.6 dB at 80 Hz and a dip of -0.7 dB at 43 Hz… not much difference.
All in all, it is such a small difference that I will have to make this decision subjectively.
j.
And just now a box of capacitors and inductors was delivered... time for the next phase.
How strong was the cabinet vibration caused by the PR?
In my PR vs port experiments I noticed quite strong vibration induced by the radiators. Of course it's mostly concentrated at the PR's resonance frequency and at higher levels.
I have been assembling the filter networks in sections.
Here are the results of the validation/verification testing of the woofer filter network. I validated the network using a near field response.
I measured the impedance of the woofer, with and without the filter. I then made NF scans of the woofer with and without the network. Then I assessed how close the measured response with network matches the simulation.
The light blue dashed line is the raw woofer NF response. The dark blue dashed line is woofer NF response with the filter in place. The solid line is the Vcad simulation using the measured impedance and the raw woofer NF response. I think the acoustic response and the impedance match the simulation quite nicely.
Here are the results of the validation/verification testing of the woofer filter network. I validated the network using a near field response.
I measured the impedance of the woofer, with and without the filter. I then made NF scans of the woofer with and without the network. Then I assessed how close the measured response with network matches the simulation.
The light blue dashed line is the raw woofer NF response. The dark blue dashed line is woofer NF response with the filter in place. The solid line is the Vcad simulation using the measured impedance and the raw woofer NF response. I think the acoustic response and the impedance match the simulation quite nicely.
Here are the same set of plots for the midrange, but I used a far field response instead of NF.
The light blue dashed line is the raw midrange FF response. The dark blue dashed line is midrange FF response with the filter in place. The solid line is the Vcad simulation using the measured impedance and the raw FF response.
The mid is performing about 0.5 dB higher than predicted by simulation, and the impedance is slightly higher as well. It is well within the "voicing range" tolerance, so there is no problem, it is just interesting.
The light blue dashed line is the raw midrange FF response. The dark blue dashed line is midrange FF response with the filter in place. The solid line is the Vcad simulation using the measured impedance and the raw FF response.
The mid is performing about 0.5 dB higher than predicted by simulation, and the impedance is slightly higher as well. It is well within the "voicing range" tolerance, so there is no problem, it is just interesting.
@tktran303 - Thanks for the lead up... I was going to discuss this situation this morning. 
In my opinion, I think this is a measurement accuracy issue.
I use DATS v3 to measure impedance. It has the capability to check individual components (L, C, and R) as well as individual drivers and entire systems.
When measuring the individual inductors and capacitors, I found the measured values of C were all low. All of my caps are 1% caps, but I was measuring about 8% low on all of them. For example, the two 3.3 uF caps measured as 3.05 and 3.07 uF. They were very closely matched, but too low. All of the caps I measured were about 8% low even though the matched pairs had close values.
I found the same situation when measuring 1% resistors... Excellent agreement between pairs, but always about 8% low. By the time I got to measuring the inductors I understood what was going on, although the inductance values had an error that was less than 8%.
It is extremely unlikely that all of these components would just so happen to be 8% low. So I made the decision to use the stated value of each component in the simulation, rather than the measured value. This includes the stated insertion loss (DCR) of each inductor.
If the DATS v3 has an error, then that also affects my measurements of the drivers and the systems.
The whole situation is rather academic, because I have the ability to adjust the level of midrange and tweeter attenuation by adjusting the values of just two resistors. The two circled in red provide me with a lot of control.
j.
In my opinion, I think this is a measurement accuracy issue.
I use DATS v3 to measure impedance. It has the capability to check individual components (L, C, and R) as well as individual drivers and entire systems.
When measuring the individual inductors and capacitors, I found the measured values of C were all low. All of my caps are 1% caps, but I was measuring about 8% low on all of them. For example, the two 3.3 uF caps measured as 3.05 and 3.07 uF. They were very closely matched, but too low. All of the caps I measured were about 8% low even though the matched pairs had close values.
I found the same situation when measuring 1% resistors... Excellent agreement between pairs, but always about 8% low. By the time I got to measuring the inductors I understood what was going on, although the inductance values had an error that was less than 8%.
It is extremely unlikely that all of these components would just so happen to be 8% low. So I made the decision to use the stated value of each component in the simulation, rather than the measured value. This includes the stated insertion loss (DCR) of each inductor.
If the DATS v3 has an error, then that also affects my measurements of the drivers and the systems.
The whole situation is rather academic, because I have the ability to adjust the level of midrange and tweeter attenuation by adjusting the values of just two resistors. The two circled in red provide me with a lot of control.
j.
Yesterday I did a full system test of the filter network. All the pieces were strung together with lever lock connectors.
The Rev 20.8 filter shown in post 409 https://www.diyaudio.com/community/...ivers-a-new-3-way-project.413182/post-7773253 turned out to be a bit too bright. I needed to reduce the mid a treble by about 1.5 dB
The current iteration is the filter shown in the post above.
I made a scan on the tweeter axis, with a time window of 4.3 ms. In the plot below I show the solid line is measured response, the dashed line is simulated (but with different resistor values).
The Rev 20.8 filter shown in post 409 https://www.diyaudio.com/community/...ivers-a-new-3-way-project.413182/post-7773253 turned out to be a bit too bright. I needed to reduce the mid a treble by about 1.5 dB
The current iteration is the filter shown in the post above.
I made a scan on the tweeter axis, with a time window of 4.3 ms. In the plot below I show the solid line is measured response, the dashed line is simulated (but with different resistor values).
All the caps I see in the pictures are +/- 5% was that a typo? Capacitors of this type are rare at 2% tolerance let alone 1.
If you are getting a bad reading on precision resistors it is time to get something new to measure with. You can use a precision resistor to make a basic impedance jig to use with REW or limp.
You have the eagle eye... The woofer circuit uses 5% caps, and this is what is most visible in the photo...
The caps in the midrange and tweeter circuit are Dayton PMPC series, which are advertised as +/- 1%.
https://www.parts-express.com/Dayto...-Precision-Audio-Capacitor-027-220?quantity=1
... except one Jantzen 1.8 uF cap which now that I look at it is 5% it seems. So I definitely exaggerated when I said "all 1%".
Yes I agree... I think I may have dropped the DATS one too many times... it may be time for a new one. I certainly like the DATS software and ease of use, and it is not expensive compared to a good Fluke DVM multimeter.
I find it interesting and have some practical ramifications. Suppose in a commercial design, where variations components can result in 0.5 - 0.8 dB of variation over a wide bandwidth (entire driver). This is probably not a deal breaker but the evidence suggests it is audible and voicing seems entirely appropriate.
What should they adjust? In the traditional audiophile landscape the general thrust is to change one's amplifier, source, cables. So the audiophilia fun never stops.
Suppose a builder wants to build your design, and it doesn't sound quite right. At least it can be fixed with a simple resistor change (if he had some way to take measurements)
Last edited:
I have quite a few routers and the dust collection of the dw621 is by far the best . was interested to see they went back to this type of extraction on there new 1/2 inch cordless model although it's well out of my price range .
I am not sure what goes on at most commercial speaker manufacturers, but here is how I would do it.
Each finished speaker (with filter) would be frequency response tested under identical conditions. A software program would compare the measured response and compare it to the gold standard for that model. The software would calculate the necessary trim resistors to install at key spots in the filter. The filter would be laid out in a way to make trim resistors easy to install. That is how I would do it....
j.