With apologies to readers for carrying on this debate too long, but.... as long as Boden argues that we gotta have feedback PERFECTION between 5 Hz and light speed, DIYers will despair ever trying it. In practice, we just need is loop stability and some spooky-force* operating up to the first harmonic or so (see footnote).
Agreed that you need "loop shaping" done through amplitude or phase or sticking bits of crazy-dough on the cone.
B.
footnote: as far as my ears are concerned, you get 80% of all the benefit of MFB by just wrangling the basic resonance cone motions, since the music sound quality in the bass is not too much impaired by HD in woofers but shot to hell by boominess.
* to borrow Einstein's famous quip
Agreed that you need "loop shaping" done through amplitude or phase or sticking bits of crazy-dough on the cone.
B.
footnote: as far as my ears are concerned, you get 80% of all the benefit of MFB by just wrangling the basic resonance cone motions, since the music sound quality in the bass is not too much impaired by HD in woofers but shot to hell by boominess.
* to borrow Einstein's famous quip
New carbon fiber accelerometer support is attached.
Very exited to see the response changes... tune in tomorrow Ben!
Very exited to see the response changes... tune in tomorrow Ben!
Your 4 measurements certainly provided the evidence needed to identify the accelerometer support as the root cause of the majority of the HF issues. Nicely done 
One other thing that is troubling is that there appears to be some interaction between the CFB amplifier and the accelerometer preamp circuitry. Note that at LF the mic response is essentially the same for Voltage amp and CFB amp. Looking at the acceleromter data though, it is clear something odd is happening at LF when the accelerometer is used with the CFB amp. Perhaps there is some sort of shared ground trace or wire?
Looking forward to seeing results from your improved mount. As Boden mentioned, the accelerometer and Mic NF data should overlay nicely up to the limit of the NF measurement validity. Here is an old example I had laying about to see what things hopefully will look like.
One other thing that is troubling is that there appears to be some interaction between the CFB amplifier and the accelerometer preamp circuitry. Note that at LF the mic response is essentially the same for Voltage amp and CFB amp. Looking at the acceleromter data though, it is clear something odd is happening at LF when the accelerometer is used with the CFB amp. Perhaps there is some sort of shared ground trace or wire?
Looking forward to seeing results from your improved mount. As Boden mentioned, the accelerometer and Mic NF data should overlay nicely up to the limit of the NF measurement validity. Here is an old example I had laying about to see what things hopefully will look like.
I had to make PCB for the new accelerometer interface, and some epoxy gluing..
Then comes side supports as well.
Then comes side supports as well.
Comparison between old aluminum dome support ORANGE and new carbon fiber support for the accelerometer RED.
There is a new OP-amp gain stage for the accelerometer with higher gain, so the curves has been moved to fit close to each other.
I marked in blue the difference between free field measurement RED and the old H baffle Orange measurement. The air mass in front/back of the driver lowers the sensitivity and lowers the fs.
The interesting part is the marking in RED... that's so close to the intended bandwidth of the feedback!
Progress i guess....y
And a comparison between acoustic measurement GREEN in a H baffle, and from accelerometer in free air RED (New carbon fiber support) . Still a breakup at 700Hz...
There is a new OP-amp gain stage for the accelerometer with higher gain, so the curves has been moved to fit close to each other.
I marked in blue the difference between free field measurement RED and the old H baffle Orange measurement. The air mass in front/back of the driver lowers the sensitivity and lowers the fs.
The interesting part is the marking in RED... that's so close to the intended bandwidth of the feedback!
Progress i guess....y
And a comparison between acoustic measurement GREEN in a H baffle, and from accelerometer in free air RED (New carbon fiber support) . Still a breakup at 700Hz...
Last edited:
This is a more stiff support, the accelerometer is resting on a carbon fiber plate same size as the sensor to decrease stress in the ceramic base substrate.
The LF response now looks reasonably fine. That's step 1!
Now for step 2: The resonant peak @720 Hz looks a bit fishy to me. You have to check whether that peak is also present in the SPL, but you have to do a farfield measurement for that: nearfield is no longer valid for a large woofer like yours at 700 Hz. Outdoors with a decent distance from reflecting surfaces is mandatory for a clean and reliable measurement.
If no peak at all is present in the SPL, further mounting experiments are required. Sometimes a solid PCB disk in the voicecoil-former will do the trick.
BTW: my compliments for this audacious journey! Very insightfull.
Now for step 2: The resonant peak @720 Hz looks a bit fishy to me. You have to check whether that peak is also present in the SPL, but you have to do a farfield measurement for that: nearfield is no longer valid for a large woofer like yours at 700 Hz. Outdoors with a decent distance from reflecting surfaces is mandatory for a clean and reliable measurement.
If no peak at all is present in the SPL, further mounting experiments are required. Sometimes a solid PCB disk in the voicecoil-former will do the trick.
BTW: my compliments for this audacious journey! Very insightfull.
Added a small carbonfiber support, some hf improvement.
Green is with the extra diagonal support, red is without, 5kHz resonance is reduced!
Green is with the extra diagonal support, red is without, 5kHz resonance is reduced!
Now when the accelerometer seems to work better i placed one of the double dustcaps. This dustcap ads some rigidity to the cone and here is the measurement difference. The woofer is placed in free air, but will be placed back in the baffle for further measurements. The blue trace is with the attached dustcap.
The added mass lowers the fs, and maybe you can see some reduction in breakup? Or is it only me who hope for improvement... ?
I had to make a hole in the center to make room for the accelerometer.
The added mass lowers the fs, and maybe you can see some reduction in breakup? Or is it only me who hope for improvement... ?
I had to make a hole in the center to make room for the accelerometer.
Since the pole piece is open in both ends right now I do not expect any air movement whatsoever. But when the second dust cap is in place there will absolutely be some air movement.
Voltage drive Blue, vs Current drive Green, with the new improved stiffer carbon fiber accelerometer support.
The driver is in freefield with no baffle. Maybe I can see an improvement in less resonances?
How come? Ok. good for me anyway since I will use current drive..
20dB resonance peaks from Himalaya is not what i embrace. So feel free to liberate me from those!
The driver is in freefield with no baffle. Maybe I can see an improvement in less resonances?
How come? Ok. good for me anyway since I will use current drive..
20dB resonance peaks from Himalaya is not what i embrace. So feel free to liberate me from those!
In post 68 I said that that 720 Hz peak looked fishy to me. Looking at your woofers, I think I was wrong: this looks like the first cone-surround break up. Apart form stiffening the surround at the sharp edge, that is where the flat cone attached section changes into the half roll, i.m.o. there is little you can do about that with such massive paper coned woofers.
All I can think of for now is to experiment in VituixCad with 2 or 3 cascaded 12 dB filter blocks in order to radically filter out all the garbage above 400 Hz. You may be able to achieve a relatively narrow band, but clean transfer function. After that you can experiment with further loopshaping around -say- 10-40 Hz and 200-400 Hz in order to create the necessary phase margin for closing the loop with a decent amount of feedback. You may find yourself playing with VituixCad for a couple of days before something useful comes out. It is all a matter of trial and error and the possibilities are enless.
BTW: I struggle with exactly the same issues with my 25cm Starbass equipped woofer in post 45: in the low end +180 degrees is approached inconveniently quickly, whereas the massive 1.9 kHz breakup is a beast to tame. But fortunately I have no issues between 100 Hz-1 kHz.
All I can think of for now is to experiment in VituixCad with 2 or 3 cascaded 12 dB filter blocks in order to radically filter out all the garbage above 400 Hz. You may be able to achieve a relatively narrow band, but clean transfer function. After that you can experiment with further loopshaping around -say- 10-40 Hz and 200-400 Hz in order to create the necessary phase margin for closing the loop with a decent amount of feedback. You may find yourself playing with VituixCad for a couple of days before something useful comes out. It is all a matter of trial and error and the possibilities are enless.
BTW: I struggle with exactly the same issues with my 25cm Starbass equipped woofer in post 45: in the low end +180 degrees is approached inconveniently quickly, whereas the massive 1.9 kHz breakup is a beast to tame. But fortunately I have no issues between 100 Hz-1 kHz.
Hi esl 63,
I am running ESL speakers with dual SVS subs. I did distortion measurements on them and they could blend quite well distortion wise.
These are the measurements:
10 times? I thought the ESL-63 was reaching 1% at 95dB at 40Hz?
Do you have THD curves for your ESL at the same levels that I measured the SVS?
Then we can combine them and see if there still is use for such sub at some xover point.
(And yes, they are indeed not Dipole 🙂
You can play 97 dB at 40 Hz with 1% distortion if you bypass the protection electronics.
But after that the membrane will hit the stator... Maybe not with a sinus test tone but with music signal the center segments that is going full range will hit the stator first...
That's why you can play approx. 10dB higher level when you have an active crossover >100Hz.
Dipoles makes most sense in the low frequencies. This is because the room resonances that is limited with Dipoles. Another benefit with dipole bass is the low coloration and the excellent step response! Properly phase adjusted you can play a square wave. That's seldom the case with a boomy bass reflex box.
Anyway, Dipoles can not play in the infrasonics, so you then you need to do something else. Below 27Hz I have an IB subwoofer with an "infinite" box size ( four JBL 15" 2226H with different spider and softer suspension, fs<20 Hz! and that with no added mass) custom woofers . Actually it is the listening room that becomes the box, almost 100m3. It is only pressurizing the room in the infrasonic region. Later I will get the MiniDSP Flex and do a DIRAC FIR correction to make it blend in better. By the way SVS seems to be a really good value woofers, and low distortion as well!
But after that the membrane will hit the stator... Maybe not with a sinus test tone but with music signal the center segments that is going full range will hit the stator first...
That's why you can play approx. 10dB higher level when you have an active crossover >100Hz.
Dipoles makes most sense in the low frequencies. This is because the room resonances that is limited with Dipoles. Another benefit with dipole bass is the low coloration and the excellent step response! Properly phase adjusted you can play a square wave. That's seldom the case with a boomy bass reflex box.
Anyway, Dipoles can not play in the infrasonics, so you then you need to do something else. Below 27Hz I have an IB subwoofer with an "infinite" box size ( four JBL 15" 2226H with different spider and softer suspension, fs<20 Hz! and that with no added mass) custom woofers . Actually it is the listening room that becomes the box, almost 100m3. It is only pressurizing the room in the infrasonic region. Later I will get the MiniDSP Flex and do a DIRAC FIR correction to make it blend in better. By the way SVS seems to be a really good value woofers, and low distortion as well!
Last edited:
IB is indeed ideal if you dont have neighbors 🙂
I also use a MiniDSP SHD and dirac to get things right ..