• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Active DSP DIY speakers, for testing a few ideas

MAB

Major Contributor
Joined
Nov 15, 2021
Messages
2,137
Likes
4,783
Location
Portland, OR, USA
FFLOTSAM (Fabricated From LeftOver Things Studio Active Monitor)
Active DSP speakers, for testing a few ideas
I had some stuff lying around; a few random Baltic Birch-ply panels, some countertop material, a few cans of partially used brown spray paint, and some not-too clumpy spar varnish (that caused more effort than savings…) I also had the driver elements from my old disassembled Seas Thor speakers. The Thors had evolved over nearly 20 years from the original DIY design through the full range of cabinet and crossover modifications that many Thor builders pursued. I later converted them to analog active using Linkwitz crossover boards including integrated subwoofers. Thor eventually ended up active-DSP with a MiniDSP 4x10HD, which worked great. I really enjoyed the flexibility of active, even more so active with DSP. I scrapped the Thors last year since they had been eclipsed by a pair of Genelec Ones. While they sounded great, they were gigantic with an umbilical running to a rack of amps, always a bit too much of a science project…
Science has struck again now that I joined ASR. And I thought I would build a pair of mini monitors to test out a few ideas. One involves dealing with resonance peaks in midbass drivers. The Seas W18E001 woofer from the Thors is ideal for this since it has an epic breakup mode at 4.3kHz. I also wanted to go through a basic active speaker build since there have been recent discussions here about how easy and superior DSP is. Lastly, I figured that it would be fun and I wanted an alternative to a pair of Rauna Freja that are my primary computer speakers.

Table of contents:
Build and initial testing of the W18E001 woofer
Reduction in 3rd order distortion observation
Part 2 of testing the woofer's resonance
DSP vs. Passive/Active Hybrid
Full speaker testing with both passive and DSP notch

The drivers:
1666400434940.png

The tweeter is the T25CF-002. It’s a soft-dome tweeter made out of synthetic fabric. It can be crossed over pretty low for a tweeter (500Hz resonance frequency), but I did overheat a pair experimenting with a low crossover point a few years ago. So despite the capability to play low, there is only so much thermal power that can be dissipated from the motor.
The woofer is the W18E001. It has a magnesium cone. I have both the first and second versions of this driver. Here are the published and measured frequency responses of the first version:
1666988737065.png


There is a lot going on centered at 4.3kHz. There is a breakup mode where the cone starts looking more like a potato chip with a 9dB fundamental peak, but there are other things going on as well, like a null at 3kHz. These have harmonics that can be seen at high frequencies. The measurement and the data from Seas are in agreement. Like driving a car on a washboard road, you will either need to avoid driving the cone near these resonances or additional damping needs to be added to mechanically control the resonance (like a shock-absorber!!!). You could use a low tweeter crossover and risk melting tweeters (like me) or build a three-way with a proper midrange. Or you could use a DSP notch to reduce the signal at resonance. You could also use a passive notch filter.

Construction:
I built the drivers into 11 liter sealed boxes with an 20cm wide front baffle.
1666400550695.png

The front panel is 38mm thick cheap countertop. I rounded the front edges, chamfered the cutouts, etc. I recently adjusted the saw and things came out pretty true.
1666400575730.png

I use PVA wood glue for the cabinet joints, but I like the polyurethane glue for braces since it expands, especially if you lightly wet the wood surface before gluing. The braces are kiln-dried Douglass Fir which should add warmth to the midrange;). I’m not sure how long the front face of the cabinet is going to last since it will expand and contract much differently than the plywood cabinet.
1666400594076.png

The Epifanes varnish I had was old and a bit clotted, I eventually made it work but it is the least appealing part of the cabinets. Even the mismatched brown spray paint turned out better. There is a reason that artisans put on 12-coat finishes, the first 11 didn’t work out and had to be sanded off!:mad: The cabinets came together well and the faces press-fit nicely. They might look at home on top of some cinder-block bookshelves, perfect for the man-cave but not quite fine furniture.

Setup:
The speakers went together easily. I stuffed them with jute and poly-fill. The cabinets are nice and dead. I don’t have stands built yet, but I was able to put the speakers in my basement office, plug them in, implement simple Rev0 DSP crossover at 2.4kHz and start playing music within an hour of final assembly. I even got the levels pretty much right.
1666400631768.png

I forgot to mention the sea of electronics. And the wire. Lots of cable-lifters needed too I guess!!!:eek: I also forgot to mention that FFLOTSAM will include subwoofers, called JETSAM of course, using Seas W26FX001 aluminum/magnesium alloy woofers in 28 liter boxes.;) I have four but will likely only use two! The Marantz monoblock amps drive the FFLOTSAM, the Bryston powers the JETSAM subs. A MiniDSP Flex Eight does the rest very nicely.
1666400648339.png


Right off the bat I notice a metallic leanness to the sound, presumably due to the woofer’s breakup mode (even with 24dB/octave crossover at 2.4kHz) and the lack of baffle compensation. Moving the crossover point to an unreasonably low 1.4kHz subjectively improved the shrieking, moving the crossover to 3kHz made it sound worse. This kind-of confirms the woofer breakup needs dealing with at any reasonable crossover frequency. All in all, with the first-pass levels and filters it sounded pretty OK and played so loud I had a hard time fighting my way back to the desk to grab the remote. Engaging the subwoofers addressed (or covered) some of the leanness, but something needs to be done about the woofers.

Measurements:
Going back to the W18E001 v1 frequency response, I applied EQ to reduce the signal at the 4.3kHz peak. A -9dB notch with Q=4 seemed best. Adding a notch at 3kHz for the null and a shelf for the baffle and we end up with pretty flat response I think. You can really hear the resonance at 4.3kHz during the sweeps when no notch is present, I am pretty sure the dip at 3kHz makes it sound even worse.
1666400818029.png


Re-implementing the Rev0 crossover slopes and levels with these DSP filters the speakers sound surprisingly good to me without further fiddling. Any additional Bass EQ would be perilous at anything but moderate volume so subwoofers are needed, at least to suit my taste. There are nits to pick, but I think I could live with this. And, it was easy for me to get up and running quickly and easily, and implement EQ based on measurements. Granted, I have experience with these drivers from the DIY Thor saga, but no way I could have had the system up and running and sounding like this with passive crossovers so soon. I felt the same when I went from passive crossovers on the Thors to analog active; implementing changes was just so much easier:cool:. Even more so with DSP:cool::cool:. The penalty with active is that sea of electronics and wires.

But before I build this out to the greatest sounding speakers in the world, my goal is to evaluate notches first.;) And I wanted to do a comparison of active vs. passive notch approaches. Also, I was made aware by @thewas of a PuriFi whitepaper on resonant breakup modes which indicated a parallel notch filter (implemented in series with the driver) potentially outperforms an active notch.

So, I made a fixture to allow easy swaps of components for a passive notch filter.
1666401239633.png


Below is the response of the W18E001 v1 driver with and without notch. DSP notch is at 4.3 kHz, Q=4, and Gain = -9dB. The passive notch was achieved with C=8.2uF, L=0.15mH, and R=100Ω. 12.5Ω, 25Ω, and 50Ω were tried as well. I think somewhere between 50Ω and 100Ω is where the attenuation needs to be. I tried moving the L and C around too, these values worked the best but I admit I was limited slightly to parts on hand. Suffice to say, both passive and active notch filters do a good job of taming the peak. Interesting observation, there is less oscillation in the higher order harmonics (8.6kHz) with the passive notch than the DSP notch. Perhaps because my DSP and passive filter shapes are not absolutely the same. Also, I did feel that there may be more than one thing going on at 4.3kHz which may require more than one notch. But both filter approaches work well I think. Sweeps sound smooth with either, with no more ear-ringing howl.
1666401491022.png

Looking at the range of notch filters, 12.5Ω provides insufficient attenuation. 50Ω and 100Ω look best, the truth may be in between. Note the distortion; 100Ω appears to be best, 50Ω close second, 12.5Ω third for the passive notch. The DSP EQ (-9dB, Q=4, f=4300) does not appear to perform as well as the 50Ω and 100Ω passive notch for distortion.
1666401547823.png



I can’t honestly hear the differences between the 50Ω and 100Ω passives and the DSP notches. Sweeps sound great with all three. Adding the 3kHz notch, the shelf, and the crossover filters back in, I swapped the passive for active notch a few times. Both sound good listening to music.

Putting it all together, I added the 3kHz notch and the 500Hz step for the baffle and ran sweeps with no notch, the passive notch with R=100Ω, and the DSP notch. Both notches provide useable response and sounded smooth on the sweeps. I still wonder if I have the right DSP notch when I look at the distortion graph
1666401648511.png

Another question, what is going on at 1.5kHz on the distortion plot??? The Basic DSP only and Basic DSP plus active notch have a spike in distortion, while the Basic DSP plus passive has a much lower spike? I can’t imagine why the passive notch is reducing distortion at 1.5kHz too..

All those details aside, I got the speakers back on my desktop with the crossovers in place and they sound good. I am going to swap back and forth between the active and passive notches and see if I can hear a difference.

Regarding notches, I feel that the passive notch is less flexible and convenient but does outperform the active here. Whether it is audible, I am not sure, especially if you cross the driver over low. I reject the notion that a passive element somehow reduces the dynamics, or creates some high frequency grain, or prevents your amp from controlling the driver, or blurs the image, etc. In fact, I think it is incrementally better in the case where you have a large cone breakup mode. And, if you quickly check out the Seas Excel woofer line, you will see they have lots of drivers with large breakup modes (not just the metal ones, graphene, and treated-paper as well and they will all have a metallic ring...:eek:) High-end drivers with large resonances aren't even unique to Seas.

Regarding ease... In a way, DSP reduces building speakers to building good cabinets. There were threads a few months ago wondering why everybody didn't just convert their passive speakers to active. And threads that passives sound worse than actives, we seemed to damn every capacitor and inductor to the hell of lo-fi in that one. Another that in the future no more capacitors will exist, or all the electricity shunted to ground through everybody's passive crossovers was enough to power all of the clothes driers in the world... OK, those two I made up.:) But seriously, there are just too many great sounding passive speakers (and perhaps too many bad sounding actives) for these to be reasonable. Not to say that active DSP speakers are not the bright future, but the picture is a bit more nuanced than some of the discussions. Hope this helps.

I will test out the v2 of the the W18 woofer next. It has a larger peak, at higher frequency than v1.

Edit (Oct 28): replaced graph of W18 spec sheet actual vs. published frequency response with a graph that has correctly aligned frequency axes. Added linked table of contents.
 
Last edited:

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,755
Likes
16,201
Really great work, analysis and presentation, highly appreciated! I would guess the 1.5 kHz HD peak is a third harmonic which maps to the reduced 4.5 kHz distortion, could you maybe check it?
 

napfkuchen

Active Member
Joined
Mar 9, 2022
Messages
295
Likes
397
Location
Germany
Cannot judge the technical aspects, but I like the compact dimensions and the front plate finish. :D
 
  • Like
Reactions: MAB

olbobcat

Member
Joined
Oct 1, 2022
Messages
75
Likes
38
FFLOTSAM (Fabricated From LeftOver Things Studio Active Monitor)
Active DSP speakers, for testing a few ideas
I had some stuff lying around; a few random Baltic Birch-ply panels, some countertop material, a few cans of partially used brown spray paint, and some not-too clumpy spar varnish (that caused more effort than savings…) I also had the driver elements from my old disassembled Seas Thor speakers. The Thors had evolved over nearly 20 years from the original DIY design through the full range of cabinet and crossover modifications that many Thor builders pursued. I later converted them to analog active using Linkwitz crossover boards including integrated subwoofers. Thor eventually ended up active-DSP with a MiniDSP 4x10HD, which worked great. I really enjoyed the flexibility of active, even more so active with DSP. I scrapped the Thors last year since they had been eclipsed by a pair of Genelec Ones. While they sounded great, they were gigantic with an umbilical running to a rack of amps, always a bit too much of a science project…
Science has struck again now that I joined ASR. And I thought I would build a pair of mini monitors to test out a few ideas. One involves dealing with resonance peaks in midbass drivers. The Seas W18E001 woofer from the Thors is ideal for this since it has an epic breakup mode at 4.3kHz. I also wanted to go through a basic active speaker build since there have been recent discussions here about how easy and superior DSP is. Lastly, I figured that it would be fun and I wanted an alternative to a pair of Rauna Freja that are my primary computer speakers.

The drivers:
View attachment 238746
The tweeter is the T25CF-002. It’s a soft-dome tweeter made out of synthetic fabric. It can be crossed over pretty low for a tweeter (500Hz resonance frequency), but I did overheat a pair experimenting with a low crossover point a few years ago. So despite the capability to play low, there is only so much thermal power that can be dissipated from the motor.
The woofer is the W18E001. It has a magnesium cone. I have both the first and second versions of this driver. Here are the published and measured frequency responses of the first version:
View attachment 238747
There is a lot going on centered at 4.3kHz. There is a breakup mode where the cone starts looking more like a potato chip with a 9dB fundamental peak, but there are other things going on as well, like a null at 3kHz. These have harmonics that can be seen at high frequencies. The measurement and the data from Seas are in agreement. Like driving a car on a washboard road, you will either need to avoid driving the cone near these resonances or additional damping needs to be added to mechanically control the resonance (like a shock-absorber!!!). You could use a low tweeter crossover and risk melting tweeters (like me) or build a three-way with a proper midrange. Or you could use a DSP notch to reduce the signal at resonance. You could also use a passive notch filter.

Construction:
I built the drivers into 11 liter sealed boxes with an 20cm wide front baffle.
View attachment 238748
The front panel is 38mm thick cheap countertop. I rounded the front edges, chamfered the cutouts, etc. I recently adjusted the saw and things came out pretty true.
View attachment 238749
I use PVA wood glue for the cabinet joints, but I like the polyurethane glue for braces since it expands, especially if you lightly wet the wood surface before gluing. The braces are kiln-dried Douglass Fir which should add warmth to the midrange;). I’m not sure how long the front face of the cabinet is going to last since it will expand and contract much differently than the plywood cabinet.
View attachment 238750
The Epifanes varnish I had was old and a bit clotted, I eventually made it work but it is the least appealing part of the cabinets. Even the mismatched brown spray paint turned out better. There is a reason that artisans put on 12-coat finishes, the first 11 didn’t work out and had to be sanded off!:mad: The cabinets came together well and the faces press-fit nicely. They might look at home on top of some cinder-block bookshelves, perfect for the man-cave but not quite fine furniture.

Setup:
The speakers went together easily. I stuffed them with jute and poly-fill. The cabinets are nice and dead. I don’t have stands built yet, but I was able to put the speakers in my basement office, plug them in, implement simple Rev0 DSP crossover at 2.4kHz and start playing music within an hour of final assembly. I even got the levels pretty much right.
View attachment 238751
I forgot to mention the sea of electronics. And the wire. Lots of cable-lifters needed too I guess!!!:eek: I also forgot to mention that FFLOTSAM will include subwoofers, called JETSAM of course, using Seas W26FX001 aluminum/magnesium alloy woofers in 28 liter boxes.;) I have four but will likely only use two! The Marantz monoblock amps drive the FFLOTSAM, the Bryston powers the JETSAM subs. A MiniDSP Flex Eight does the rest very nicely.
View attachment 238752

Right off the bat I notice a metallic leanness to the sound, presumably due to the woofer’s breakup mode (even with 24dB/octave crossover at 2.4kHz) and the lack of baffle compensation. Moving the crossover point to an unreasonably low 1.4kHz subjectively improved the shrieking, moving the crossover to 3kHz made it sound worse. This kind-of confirms the woofer breakup needs dealing with at any reasonable crossover frequency. All in all, with the first-pass levels and filters it sounded pretty OK and played so loud I had a hard time fighting my way back to the desk to grab the remote. Engaging the subwoofers addressed (or covered) some of the leanness, but something needs to be done about the woofers.

Measurements:
Going back to the W18E001 v1 frequency response, I applied EQ to reduce the signal at the 4.3kHz peak. A -9dB notch with Q=4 seemed best. Adding a notch at 3kHz for the null and a shelf for the baffle and we end up with pretty flat response I think. You can really hear the resonance at 4.3kHz during the sweeps when no notch is present, I am pretty sure the dip at 3kHz makes it sound even worse.
View attachment 238754

Re-implementing the Rev0 crossover slopes and levels with these DSP filters the speakers sound surprisingly good to me without further fiddling. Any additional Bass EQ would be perilous at anything but moderate volume so subwoofers are needed, at least to suit my taste. There are nits to pick, but I think I could live with this. And, it was easy for me to get up and running quickly and easily, and implement EQ based on measurements. Granted, I have experience with these drivers from the DIY Thor saga, but no way I could have had the system up and running and sounding like this with passive crossovers so soon. I felt the same when I went from passive crossovers on the Thors to analog active; implementing changes was just so much easier:cool:. Even more so with DSP:cool::cool:. The penalty with active is that sea of electronics and wires.

But before I build this out to the greatest sounding speakers in the world, my goal is to evaluate notches first.;) And I wanted to do a comparison of active vs. passive notch approaches. Also, I was made aware by @thewas of a PuriFi whitepaper on resonant breakup modes which indicated a parallel notch filter (implemented in series with the driver) potentially outperforms an active notch.

So, I made a fixture to allow easy swaps of components for a passive notch filter.
View attachment 238757

Below is the response of the W18E001 v1 driver with and without notch. DSP notch is at 4.3 kHz, Q=4, and Gain = -9dB. The passive notch was achieved with C=8.2uF, L=0.15mH, and R=100Ω. 12.5Ω, 25Ω, and 50Ω were tried as well. I think somewhere between 50Ω and 100Ω is where the attenuation needs to be. I tried moving the L and C around too, these values worked the best but I admit I was limited slightly to parts on hand. Suffice to say, both passive and active notch filters do a good job of taming the peak. Interesting observation, there is less oscillation in the higher order harmonics (8.6kHz) with the passive notch than the DSP notch. Perhaps because my DSP and passive filter shapes are not absolutely the same. Also, I did feel that there may be more than one thing going on at 4.3kHz which may require more than one notch. But both filter approaches work well I think. Sweeps sound smooth with either, with no more ear-ringing howl.
View attachment 238758
Looking at the range of notch filters, 12.5Ω provides insufficient attenuation. 50Ω and 100Ω look best, the truth may be in between. Note the distortion; 100Ω appears to be best, 50Ω close second, 12.5Ω third for the passive notch. The DSP EQ (-9dB, Q=4, f=4300) does not appear to perform as well as the 50Ω and 100Ω passive notch for distortion.
View attachment 238759


I can’t honestly hear the differences between the 50Ω and 100Ω passives and the DSP notches. Sweeps sound great with all three. Adding the 3kHz notch, the shelf, and the crossover filters back in, I swapped the passive for active notch a few times. Both sound good listening to music.

Putting it all together, I added the 3kHz notch and the 500Hz step for the baffle and ran sweeps with no notch, the passive notch with R=100Ω, and the DSP notch. Both notches provide useable response and sounded smooth on the sweeps. I still wonder if I have the right DSP notch when I look at the distortion graph
View attachment 238760
Another question, what is going on at 1.5kHz on the distortion plot??? The Basic DSP only and Basic DSP plus active notch have a spike in distortion, while the Basic DSP plus passive has a much lower spike? I can’t imagine why the passive notch is reducing distortion at 1.5kHz too..

All those details aside, I got the speakers back on my desktop with the crossovers in place and they sound good. I am going to swap back and forth between the active and passive notches and see if I can hear a difference.

Regarding notches, I feel that the passive notch is less flexible and convenient but does outperform the active here. Whether it is audible, I am not sure, especially if you cross the driver over low. I reject the notion that a passive element somehow reduces the dynamics, or creates some high frequency grain, or prevents your amp from controlling the driver, or blurs the image, etc. In fact, I think it is incrementally better in the case where you have a large cone breakup mode. And, if you quickly check out the Seas Excel woofer line, you will see they have lots of drivers with large breakup modes (not just the metal ones, graphene, and treated-paper as well and they will all have a metallic ring...:eek:) High-end drivers with large resonances aren't even unique to Seas.

Regarding ease... In a way, DSP reduces building speakers to building good cabinets. There were threads a few months ago wondering why everybody didn't just convert their passive speakers to active. And threads that passives sound worse than actives, we seemed to damn every capacitor and inductor to the hell of lo-fi in that one. Another that in the future no more capacitors will exist, or all the electricity shunted to ground through everybody's passive crossovers was enough to power all of the clothes driers in the world... OK, those two I made up.:) But seriously, there are just too many great sounding passive speakers (and perhaps too many bad sounding actives) for these to be reasonable. Not to say that active DSP speakers are not the bright future, but the picture is a bit more nuanced than some of the discussions. Hope this helps.

I will test out the v2 of the the W18 woofer next. It has a larger peak, at higher frequency than v1.
I think that there might be an advantage with a bi-amp active crossover with added room EQ compared to a passive crossover with room EQ?
 
  • Like
Reactions: MAB
OP
MAB

MAB

Major Contributor
Joined
Nov 15, 2021
Messages
2,137
Likes
4,783
Location
Portland, OR, USA
Really great work, analysis and presentation, highly appreciated! I would guess the 1.5 kHz HD peak is a third harmonic which maps to the reduced 4.5 kHz distortion, could you maybe check it?
Thanks. You are right, looks like third harmonic. Here the second and third harmonics for the W18E001 with no notch at 4.3kHz, and with both the active notch and the passive notch.
1666453067334.png

The third-order peak at ~1.45kHz is almost completely unattenuated with the active notch!
 
OP
MAB

MAB

Major Contributor
Joined
Nov 15, 2021
Messages
2,137
Likes
4,783
Location
Portland, OR, USA
Cannot judge the technical aspects, but I like the compact dimensions and the front plate finish. :D
Thanks!
I just realized that the FFLOTSAM look similar to the Verdant Audio Bambusa MG-1!
1666454884357.png

Same drivers, but the FFLOTSAM is sealed... I also saw in the product description that the Bambusa has 12-coats of polyurethane finish, so maybe we all struggle with getting even coats and need multiple chances!
 

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,755
Likes
16,201
Thanks. You are right, looks like third harmonic. Here the second and third harmonics for the W18E001 with no notch at 4.3kHz, and with both the active notch and the passive notch.
View attachment 238831
The third-order peak at ~1.45kHz is almost completely unattenuated with the active notch!
Thank you for checking, isn't it great when theory is confirmed by own measurements? :cool:
 
OP
MAB

MAB

Major Contributor
Joined
Nov 15, 2021
Messages
2,137
Likes
4,783
Location
Portland, OR, USA
In addition to the W18E001 v1 drivers from the Thors, I also have a pair of W18E001 v2 woofers lying around and wanted to include those measurements here. The reason is v2 and v1 are really quite different. Here are the published frequency response graphs, I have changed aspect ratios to make the frequency axes the same scale for relative comparison.
1666666349285.png

The second version of this driver has a 12dB peak at 5.1kHz instead of 9dB at 4.3kHz for the first version so that is interesting. I am doubly interested because of some recent discussions about the advantages and disadvantages of active vs. passive filters, and just a few of the claims needed to be further investigated. And, after it was pointed out to me to look more closely at the harmonic distortion components in REW, I got excited to compare the two versions. Plus, maybe the second version has some characteristics that bring this whole thing to the next level! Or maybe v2's 12dB peak is part of a train of resonant modes that makes it less useful.

I built a fixture to make it a bit easier to measure the drivers. One thing for sure about doing any speaker build: prototyping is hard on the finished product. This is one area that DSP is a sure winner since you can change the filters without messing with the finished box or dealing with external crossovers. Even my tank-circuit fixture requires some care when fiddling. To minimize the thrash on the FFLOTSAM cabinets, I routed a hole in a piece of MDF and installed threaded t-nuts to hold the drivers down.
1666661874318.png


On the upside, it seems I am getting better distortion measurements with the new fixture. On the downside, I run into a problem with one of the v2 drivers. It looks like an entire edge had some marginal adhesive from the factory. You can even see it if you feed enough power at low frequency.
1666662513427.png

I sent an email with a picture to Madisound who sold me the unit. They immediately responded and said they would contact Seas. The next morning, Madisound said the response from Seas was "that's not right, send the customer a new unit." Later that afternoon I received confirmation from Madisound that my replacement driver shipped. This is great customer service from Madisound and Seas, just the way it should be. This is in stark contrast to some of my other audio customer service experiences, but that's another story.

It actually took me some time to figure out what was going on:confused:. I am a bit embarrassed to admit, I had started a build with these v2 drivers and was unhappy with the way one of them sounded but I wasn't aggressive enough mechanically probing the speaker to see what was wrong (the surround lays perfectly flat and needs the cone slightly deformed to see the delamination. Or send 100 watts of bass to get it flapping in the breeze.) Looking at just the frequency response, nothing stands out. But you can see the dramatically rising distortion below 800Hz if you look. This is very audible to the ear, so I also missed this originally by not measuring the distortion!
1666665708376.png

The distortion is primarily 2nd order and sounds bad as the rubber surround flaps around.
1666666112000.png

Small detour, glad I stumbled though all this, will have a shiny new v2 driver arriving later this week. In the meantime, I will look at the good v2 driver with a number of passive and active filters next, see what can be done to tame that 12dB peak.
 
OP
MAB

MAB

Major Contributor
Joined
Nov 15, 2021
Messages
2,137
Likes
4,783
Location
Portland, OR, USA
I think that there might be an advantage with a bi-amp active crossover with added room EQ compared to a passive crossover with room EQ?
If you are saying that room-EQ may perform better on an active speaker compared to a passive speaker all other things being equal, I think no.
I would agree that while integrating a subwoofer, room EQ and an active crossover can be used together to deal with a number of room issues. But I am not so sure that the individual midrange or tweeter driver responses and crossover can be used to meaningfully solve room issues.
Here, I am trying to first to optimize each driver, then integrate into a system, then apply room EQ. And right now I am somewhat interested in the distortion products from the resonant peaks in these woofers. And, if the first post made sense, I am seeing some areas where active is actually worse performance than passive.
Yes, I can imagine an approach that uses DSP at the driver level to deal with some room issues at midrange frequencies, (like maybe you could do some DSP magic at the driver level to allow a speaker to be used either: midfield away from boundaries or: on a desktop in a nearfield application...) But that is outside of my scope!
 
OP
MAB

MAB

Major Contributor
Joined
Nov 15, 2021
Messages
2,137
Likes
4,783
Location
Portland, OR, USA
Before I get into the W18E001 v2 filters and responses, let's look at v1 vs. v2 as measured. I got the levels matched but v1 is measured in the FFLOTSAM box, and v2 is measured in the large baffle panel. I think it is OK at the frequencies we are interested in.
1666676489114.png

Gosh, v2 has an even larger distortion peak at resonance, and also a larger peak at 1/3 the of the resonance frequency. While the breakup mode's fundamental is perhaps easier to mitigate with the upper crossover slope, it does have this bigger peak at 1.7kHz. That distortion peak at 5.1kHz is HUGE!

DSP Notch Filter:
I tried a DSP notch with f=5.1kHz, G=-12dB, and Q=4. In moments, I had it implemented in the Flex Eight, and moments later I had the data in REW:
1666677183603.png

The 5.1kHz peak is flattened and the distortion at resonance tamed. But the 3rd order harmonic resonance peak and associated distortion at 1.7kHz is absolutely unchanged. I tried a bunch of other DSP notch filters of various depths and Q, this one has the flattest response, and yields equivalent reduction in distortion. None of these active notches had any impact on the distortion peak at 1.7kHz.

Passive Notch Filter:
I tried a bunch of different passive filters. I tried a range of L and C all with resonance at 5.1kHz and compared to the 5.1kHz DSP notch. A few things stand out. The peak can be flattened with more or less effectively with passives and analog. Using high value inductors tends creates a rolloff that may or may not be acceptable. Looking at the distortion graph, please note that the distortion of both the 5.1kHz fundamental and the 1.7kHz harmonic are decreased with increasing L. The reason for this is discussed in the Purifi Whitepaper I mentioned earlier (section 1.3) As L is increased the motor hysteresis distortion finds a higher series impedance, the larger the inductor the larger the effect for a given distribution of distortion components. The penalty is an increasing downward slope in frequency response with larger L. I will deal with that next.
1666681303585.png


Hybrid - Passive Notch and DSP shelf
The distortion results look good, but it occurred to me that the downward frequency response was in effect cheating the distortion measurement due to the simple reduction in energy at higher frequencies. I used a high frequency shelf to level the frequency response of filter (e) with the 1mH inductor. I also thought that additional attenuation at the 5.1kHz fundamental was good, so I tried increasing R to 900Ω. These are the (f) and (g) in the graphs below. I think it is fair to say that I didn't get the responses of the passive and the DSP matched. I might say that the response of the passive notch + DSP shelf is actually more useable. Particularly, (f) the passive notch with 900 Ohm resistor is flat through the resonance and has dramatically lower distortion at the resonance and retains all of the benefit at the 1.7kHz peak in 3rd order harmonic distortion.
1666686603857.png

Conclusion of passive vs. active notch investigation:
I found a passive/active hybrid approach is superior for dealing with a cone's breakup mode if distortion is taken into account: damp the peak and snub the current induced by the back-emf with the passive notch, and level the frequency response to target with a DSP shelf. Passive-only is superior for distortion performance, but has a potentially undesireable frequency response impact. Active-only (DSP or otherwise) can flatten the response, even more so if multiple filters are used, but has less impact on distortion at the fundamental, and zero impact at the lower harmonic frequencies that are not attenuated by the crossover network.

My comparison between the two approaches is still not apples to apples, I need to better align the frequency responses to see how much closer the active notch can get to the distortion performance of the passive notch. I think the W18E001 is great for demonstrating all of this because it does have one of the largest breakup modes of any driver in this class that I have seen data on. Maybe v2 is even better case-study than v1 I think...
 
Last edited:

voodooless

Grand Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
10,227
Likes
17,807
Location
Netherlands
Fascinating! Thanks for the time spent on this.

It’s very clear that passive notch filters are the way to go, even in an active system.

I see you used a series filter? Did you try a parallel filter? It will have a lower RDC, which may be an advantage in some cases.
 
OP
MAB

MAB

Major Contributor
Joined
Nov 15, 2021
Messages
2,137
Likes
4,783
Location
Portland, OR, USA
Fascinating! Thanks for the time spent on this.

It’s very clear that passive notch filters are the way to go, even in an active system.

I see you used a series filter? Did you try a parallel filter? It will have a lower RDC, which may be an advantage in some cases.
Yes! I was focusing on following the Purifi Whitepaper on the subject. They discuss series vs. parallel there, but my goal was to explore this reduction in THD they observed with the series filter. Agreed, the DC resistance of the series filter can be more or less of an issue. The 1mH air-core inductor I used has 250mΩ DCR. I think you can get to 1mH 150mΩ iron-core reasonably. I have a 1mH inductor that's 120mΩ but is has an unreasonable gravitational field!
1666692752606.png

Also, all of the series notches are outperforming the active notch (and I assume the parallel passive too)! So, you could back off on L to 0.3mH (for instance) and end up with ~1/3 DCR of the equivalent 1mH piece.
But, now that I found the THD tab on REW, and have a driver that sprays harmonics like a fire hydrant, and a way to somewhat control the spray, I'm having fun:)...
 

DanielT

Major Contributor
Joined
Oct 10, 2020
Messages
4,751
Likes
4,633
Location
Sweden - Слава Україні
Great job and very interesting!

Hm, I wonder if that's why so many find the sound in paper based speaker cones to be good? Hard cone materials can have nasty breakup tendencies in the higher frequencies, as you show, but have other advantages. Low distortion for example. If I generalize now.
Paper-based conmatrial may have higher distortion but "kinder" roll-off in the higher frequencies. Again generalizing BUT I think they are true in many ways.

Here an example, an 8 inch bass element:
monacor-kt-momo-selbstbauprojekt-15297.jpg



According to Monacor's own data, it has this FR:

G100710K (3).jpg



Some more information here:

Screenshot_2022-10-25_150707.jpg


 
OP
MAB

MAB

Major Contributor
Joined
Nov 15, 2021
Messages
2,137
Likes
4,783
Location
Portland, OR, USA
Great job and very interesting!

Hm, I wonder if that's why so many find the sound in paper based speaker cones to be good? Hard cone materials can have nasty breakup tendencies in the higher frequencies, as you show, but have other advantages. Low distortion for example. If I generalize now.
Paper-based conmatrial may have higher distortion but "kinder" roll-off in the higher frequencies. Again generalizing BUT I think they are true in many ways.

Here an example, an 8 inch bass element:
View attachment 239360


According to Monacor's own data, it has this FR:

View attachment 239359


Some more information here:

View attachment 239358

I've wondered about the sound of the driver vs. the material the cone is made out of.
One thing for sure, these Seas Excel woofers sure have large cone breakup modes around 4-5kHz, even the coated paper ones:
1666715304855.png

The graphene-treated magnesium is actually the worst, Seas couldn't even find enough area on the graph to fit the peak, it's got to be 15-20dB!;) And, while the peak at 5kHz is likely addressable, it looks like it also has a large hump at 1/3 the fundamental (~1.6kHz), and if that hump behaves like the W18E001, then it will have a large amount of third order distortion that an electronic filter will do nothing to reduce, and that distortion is in a pretty important region for your ear and your perception. It does make me wonder about the sound of a driver based on materials. But it also makes me wonder when I look at a speaker like the Scan Speak 18W/8545-01 which is one of those classic paper pulp cone drivers that just look warm and non resonant.

18w_8545-01.jpg

Looking at the published data from Scan Speak, it has lots of stuff going on at the edge of it's useful frequency range (~2.4kHz), and it also has something strange at 700 - 800 Hz - is this the third harmonic distortion hump similar to the W18E001's, just at a lower frequency???
18w-8545-01-curve.jpg


Looking at the Revelator, Illuminator, and Ellipticor lines from Scan Speak, do they all push that breakup mode higher up to make a woofer more suitable to be crossed over high, but suffer the larger resonant peak??? The Illuminator aluminum and paper cone woofers have really large breakup modes!

I gotta tell you, listening to hundreds sweeps in REW with the W18E001 I can really hear the peak at 5kHz if it isn't filtered. The sweep is dramatically revealing, much more so than if I play music full-range through the speaker. I can even tell between the magnitude of the filter, but I can't tell the difference by ear from experiment a (DSP-only notch) and experiment g (passive notch plus DSP shelf). I can hear that I didn't EQ the FR to match, but the shriek at 5kHz is absolutely gone either way. It doesn't mean I'm not suffering that distortion blast at 1.7kHz, but looking at these other drivers makes me wonder if that 3rd order distortion peak exists on every driver if one is to look, but it doesn't show up in the frequency response graphs.
1666717861734.png



I'm going to work on matching the DSP and hybrid filters to get equivalent peak reduction and equivalent frequency response, and see who the distortion winner is. Unfortunately I won't be able to answer the "how does it sound?" question.
 

Thomas_A

Major Contributor
Forum Donor
Joined
Jun 20, 2019
Messages
3,422
Likes
2,407
Location
Sweden
Thanks for doing this. There have been several threads about the ”superiority” of active/DSP solutions while few have understood the benefits of passive components to reduce distortion in drivers. Hope this will bring some more science into the discussion.
 
  • Like
Reactions: MAB
OP
MAB

MAB

Major Contributor
Joined
Nov 15, 2021
Messages
2,137
Likes
4,783
Location
Portland, OR, USA
Thanks for doing this. There have been several threads about the ”superiority” of active/DSP solutions while few have understood the benefits of passive components to reduce distortion in drivers. Hope this will bring some more science into the discussion.
Yes, this effort is specifically to test some of the ideas and statements that were made in those threads. I am a big believer in active and DSP, just not to the point of ignoring the merit of other approaches.
 

Thomas_A

Major Contributor
Forum Donor
Joined
Jun 20, 2019
Messages
3,422
Likes
2,407
Location
Sweden
Yes, this effort is specifically to test some of the ideas and statements that were made in those threads. I am a big believer in active and DSP, just not to the point of ignoring the merit of other approaches.
I have mentioned this in other threads as well, ”impedance-match” to reduce distortion but this has been ignored as BS. It is not anything new, but it is good that you and Purify refresh these clever solutions.
 
  • Like
Reactions: MAB
OP
MAB

MAB

Major Contributor
Joined
Nov 15, 2021
Messages
2,137
Likes
4,783
Location
Portland, OR, USA
I have mentioned this in other threads as well, ”impedance-match” to reduce distortion but this has been ignored as BS. It is not anything new, but it is good that you and Purify refresh these clever solutions.
Yes, I recall some of the conversations;). In fact, these are the old solutions that have been used for ages. Like, a parallel RLC network as found in one of my all-time favorite speakers:
1666730911939.png

I wonder if the JBL XPL-200A would sound as good without the passive 1200Hz filter on that glorious eye-of-the-beholder midrange dome for instance.

Or the KEF LS50 Meta has benefit from the same in the LS50 Meta:
1666731093570.png

Does KEF use a simple DSP notch for the active implementation or retain the passive? In this case, the notch is claimed to be broad so that is different than the narrow notches being used to explore these resonances. I don't know the actual component values for the LS50...
I'm actually trying to find an example of a series RLC implementation in a product like in the PuriFi Whitepaper.
 
Last edited:

Tangband

Major Contributor
Joined
Sep 3, 2019
Messages
2,994
Likes
2,789
Location
Sweden
Nice testing. :)

My take on notch filtering, when I worked a couple of years with my HYBRID loudspeaker project, ( this was an active dsp loudspeaker with some passive components between the tweeter-power amp a capacitor equalizing the waveguide + a resistor to lower harmonic tweeter distortion, and trying to notch out the Seas er18rnx resonances with a passive filter, all set in two boxes outside the speaker ) .
Measurement microphone was line audio om1 , Audient id14 and Audiotools with an ipad.

1. Unfortunaltely my two Seas er18rnx had different resonances at different frequencies - one had a peak at 4,3 kHz , and the other had its peak at 4,8 kHz . I had to use different components for the passive notch for L and R speaker .
A pain to do passive, but easy to do with the dsp crossover ( dbx pa2 ) .

2. Even then - I thought that the sound was better without the passive notch , and instead using a dsp notch .
With a dsp-notch its easy to get the exact Q value of the notch.

3. I tried many options, but the best sound was without any notch filtering at all, using steeper crossovers instead . A dsp 12 dB/oct with a passive or active notch filtering sounded worse than just using a dsp 24 dB/oct or 48 dB/oct crossover - without notch filtering.

4. My conclusion of this after some years of comparing, is that its probably much better to use drivers that dont need any notch filtering in the beginning, these will sound better in the midrange . This is the way Genelec does it, as you can see with the M040 and 8030 monitor .

5. For an even better, clearer sound - a dsp crossover without an A/D is probably the way to go, keeping the signal digital as long as you can.

.EEB6B6AF-472F-46AC-B180-0708C9B58FC9.jpeg7A1543EA-5598-4E56-BDB4-F8B91E741730.jpeg0A051317-C55A-409A-A257-8619E7647E89.jpeg2F15339B-8641-4F03-889B-C35CB6A4A922.jpegAEF5D118-509D-4018-A1BF-D3BCD5FBDB50.jpeg
 
Last edited:
Top Bottom