10% distortion and down to 5% is easy! As I mentioned earlier I went from 0.5% to 0.18%
But in an earlier attemt i got more than 10 times, from 10% to less than 1% equals more than 20 dB.
Robert Munnig Schmidt RMS acoustics, has achieved 30dB in a paper that is linked in the first post #1. And it was done with the same accelerometer and current feedback as I use as well.
Big difference is that he knows what he is doing!
He knows the drill, the math, he uses the right speaker unit as well with much stiffer cone.
But in an earlier attemt i got more than 10 times, from 10% to less than 1% equals more than 20 dB.
Robert Munnig Schmidt RMS acoustics, has achieved 30dB in a paper that is linked in the first post #1. And it was done with the same accelerometer and current feedback as I use as well.
Big difference is that he knows what he is doing!
He knows the drill, the math, he uses the right speaker unit as well with much stiffer cone.
RMS Acoustics can tailor the circuitry (and the report write-up) finely. The theory is simple (as I modestly understand it) and just a matter of having suitable phases and amplitudes as needed. But as you point out, some drivers are more cooperative than others.
The problem starts by mixing "apples and oranges" in the feedback loop. Taking the signal coming from an accelerometer and related bits and mixing it into the audio signal is challenging. Of course as you suggest, even the direct electrical MFB (such as current feedback) reveals how disrespectful the speaker's phases and amplitudes can be to your goals.
The problem starts by mixing "apples and oranges" in the feedback loop. Taking the signal coming from an accelerometer and related bits and mixing it into the audio signal is challenging. Of course as you suggest, even the direct electrical MFB (such as current feedback) reveals how disrespectful the speaker's phases and amplitudes can be to your goals.
Als Bolsert already pointed out, the issue you are facing is the non-monotonous curve of the highpass. This is one of the biggest problems of any particular woofer-accelerator-combo. That should be fixed in the first place before even thinking of loopshaping, because any peak instead of a solid monotonous 12 dB curve mucks up phase just where you do not want that to happen. RMS had his student Robert Valk while writing his Master Thesis, (Valk was once active here as SuperR) to find out the required mounting of the AC01: unless one is very lucky, it is not trivial. Unless you remove the AC01 and try different ways of mounting, no loop shape is ever going to yield 30dB. So the issue is not the feedback loop.
This is an accelerometer output free-air measurement of a 10"/25cm woofer equipped with a ChrisCam Starbass. For the sake of clarity the 2 channel ARTA measurement is displayed in VCad. In pink a textbook 12dB curve. Still, phase i.m.o is moving a bit too swift to +180 degrees.
Also, the next real world problems shows its ugly face: cone break up is greatly amplified by the accelerometer. At 20 kHz we see the accelerometer resonance.
Right. Like the old joke, "You can't get there from here". With a 1925 Rice-Kellogg driver you can't play the whole music range and you can't get a well-behaved feedback signal. Why try? Not feasible.
But as an enthusiast for audio, I'd welcome the passing of the R-K driver or at least something to replace rubber as a bearing and spring.
As an enthusiast for MFB, I'd like to see folks work on mid-range drivers. The Klipsch folks are boasting about a new tweeter design, as are others. But unlikely you could glue an accelerometer to a light-weight cone. So other sensors are needed (light? sound? current?)
B.
But as an enthusiast for audio, I'd welcome the passing of the R-K driver or at least something to replace rubber as a bearing and spring.
As an enthusiast for MFB, I'd like to see folks work on mid-range drivers. The Klipsch folks are boasting about a new tweeter design, as are others. But unlikely you could glue an accelerometer to a light-weight cone. So other sensors are needed (light? sound? current?)
B.
Hello Ben,
Unfortunately I do not get your point.
We only need MFB in a very limited range, that is from 20 to 150-250 Hz or so. Why? This is the range where distortion skyrockets without MFB because of the large cone excursion. And yes, a Linkwitz Tranfser could be used as brute force, but surely THD will then quickly rise to 30-40% in case of 10 dB boost.
North of 150-250 Hz there is no issue at all with modern drivers at moderate SPL levels, no matter what type of radiator is being used.
Unfortunately I do not get your point.
We only need MFB in a very limited range, that is from 20 to 150-250 Hz or so. Why? This is the range where distortion skyrockets without MFB because of the large cone excursion. And yes, a Linkwitz Tranfser could be used as brute force, but surely THD will then quickly rise to 30-40% in case of 10 dB boost.
North of 150-250 Hz there is no issue at all with modern drivers at moderate SPL levels, no matter what type of radiator is being used.
Yes, that is my point. I am asking why we are stressed about the range north of 150 Hz where a woofer's phase in relation to sound output or even cone motion go weird.
The only interest we have is trying to keep those upper frequencies out of the loop or not of any amplitude inside the loop.
But unlike your unqualified optimism, I do wonder about the value of feedback in the range about 150 Hz and do not believe drivers are already perfect up there (OK, film drivers are not perceptibly lacking). Or perhaps there are systems with other compromises (such as being very small) that might benefit from feedback.
B.
The only interest we have is trying to keep those upper frequencies out of the loop or not of any amplitude inside the loop.
But unlike your unqualified optimism, I do wonder about the value of feedback in the range about 150 Hz and do not believe drivers are already perfect up there (OK, film drivers are not perceptibly lacking). Or perhaps there are systems with other compromises (such as being very small) that might benefit from feedback.
B.
If you desire to use MFB to reduce distortion in the typical subwoofer range (20-80Hz), you need feedback active well above 150Hz.
For example, if you wanted to reduce 3rd harmonic for 50Hz by a factor of 3, you would need at least 10dB of OLTF gain at 150Hz.
For example, if you wanted to reduce 3rd harmonic for 50Hz by a factor of 3, you would need at least 10dB of OLTF gain at 150Hz.
Your are (of course) right. Never occurred to me.
But viewed more broadly, the attention needed from say, 150 Hz and up is a fairly small part of the total effort. The fundamental range from say, 20-35 Hz and the fist harmonics are the major parts of interest being near (1) system resonance and (2) where you get driver bad behaviour. At 50 Hz and above, the distortion is relatively less demanding correction. But, there does remain the "fast bass" frequencies to sharpen up.
B.
But viewed more broadly, the attention needed from say, 150 Hz and up is a fairly small part of the total effort. The fundamental range from say, 20-35 Hz and the fist harmonics are the major parts of interest being near (1) system resonance and (2) where you get driver bad behaviour. At 50 Hz and above, the distortion is relatively less demanding correction. But, there does remain the "fast bass" frequencies to sharpen up.
B.
Yepp, thats why I stated early on that my goal was 20dB feedback up to 1kHz... (9th harmonic at 120Hz)
But I realize that cone breakup from this unit is is way too high. So, my new goal is to use ESL63 down to 100Hz and add some extra panels and fiddling with the the High voltage resistors in ESL63, I have done tons of measurements and listening over the years and I think I ended up in something that I like a lot. This is a topic for a separate thread...
Anyway 100Hz and third harmonic (300Hz) and 15dB of feedback is my plan B. Call me a chicken... ;-)
If I knew how hard it is to get the pure acceleration signal from voice coil movement to accelerometer "useful" signal, then i would have chosen different drivers such as
6 x RSS390HF-4 15" al cone or 8 x RSS315HFA-8 12" al cone, or maybe 8 x UM12 12" Nomex cone
and Piratelogic small accelerometers.Here is four measurements From Dayton IB 15" Elements in a H baffle.
The two first is acoustic amplitude and phase response with the current feedback amplifier, Followed by the voltage amplifier.
The latter two is the accelerometer signal from CFB amp and ordinary voltage amplifier. An then some amplitude comparisons.
above: Acoustic from Current Feedback amplifier.
No filters (No C12, C15, C16) C8 is 5,7 uF
above: Acoustic nearfield Voltage amplifier.
above: Accelerometer signal from Current Feedback amplifier.
above: Accelerometer from voltage amplifier.
Comparison between CFB amp red and Voltage amplifier in orange.
above: Finally the accelerometer signal from CFB amplifier in green and voltage- amp in blue...
Nasty breakups is capured by the accelerometer... I´m redesigning the mechanical interface between voice coil and ACH-01.
The two first is acoustic amplitude and phase response with the current feedback amplifier, Followed by the voltage amplifier.
The latter two is the accelerometer signal from CFB amp and ordinary voltage amplifier. An then some amplitude comparisons.
above: Acoustic from Current Feedback amplifier.
No filters (No C12, C15, C16) C8 is 5,7 uF
above: Acoustic nearfield Voltage amplifier.
above: Accelerometer signal from Current Feedback amplifier.
above: Accelerometer from voltage amplifier.
Comparison between CFB amp red and Voltage amplifier in orange.
above: Finally the accelerometer signal from CFB amplifier in green and voltage- amp in blue...
Nasty breakups is capured by the accelerometer... I´m redesigning the mechanical interface between voice coil and ACH-01.
Thanks. Very wonderful to see those comparisons.
But the next comparison I'd like to see is a comparison of the acoustic output voltage amp and the signal sent by the ACH-01. Ditto for the current amp. That would show how the driver is converting drive signal into motion of the dust cap (and how motion of the dust cap picked up by the ACH-01 compares to acoustic output).
Other interesting plots might be THD on the ACH-01 plots because they reveals the actual THD coming from the cone (or at least the dust cap). In turn, that should correspond to acoustic THD.
Interesting how well behaved the phase is for both drive plot (of course) but the ACH-01 on the current drive isn't orderly. ("Orderly" is common parlance for "easy to correct" inside the loop.) I think that's because the cone resonance is well-managed on the voltage amp (with high damping factor) but out of control on the current (with low damping factor, which is evident from the acoustic plots).
B.
But the next comparison I'd like to see is a comparison of the acoustic output voltage amp and the signal sent by the ACH-01. Ditto for the current amp. That would show how the driver is converting drive signal into motion of the dust cap (and how motion of the dust cap picked up by the ACH-01 compares to acoustic output).
Other interesting plots might be THD on the ACH-01 plots because they reveals the actual THD coming from the cone (or at least the dust cap). In turn, that should correspond to acoustic THD.
Interesting how well behaved the phase is for both drive plot (of course) but the ACH-01 on the current drive isn't orderly. ("Orderly" is common parlance for "easy to correct" inside the loop.) I think that's because the cone resonance is well-managed on the voltage amp (with high damping factor) but out of control on the current (with low damping factor, which is evident from the acoustic plots).
B.
i.m.o the a above clearly shows there is mounting issue with the ACH-01: between 5 Hz and 200 Hz NF mic measurements and accell. output should fully track. Mounting issues are the biggest elefant in the MFB room. @esl: in the Herbert Rutgers docs that you already possess, you can see a a solid plastic mounting ring is used that fits the top of the VC former. Robert Valk used a similar device.
Near field is approx 3" or 8cm. I guess that is enough if you have a 15" driver?
My carbon fiber "bridges" will be used again. (when you have something that works, stick with it...)
My carbon fiber "bridges" will be used again. (when you have something that works, stick with it...)
Biden - Motional feedback by attaching the ACH-01 to the dust cap, senses and corrects the motion of the dust cap, not the whole driver. So fair only to ask if it is doing a good correlation with the dust cap behaviour not a microphone. Which means the choice of driver matters so that the speaker output resembles the dust cap output at 3 inches (is that a good place?) or any reference location.
There have been past reports, as you say, that the way of mounting the accelerometer matters. I guess the scale of motion and off error is very tiny and any slop in the mounting is detrimental.
B.
There have been past reports, as you say, that the way of mounting the accelerometer matters. I guess the scale of motion and off error is very tiny and any slop in the mounting is detrimental.
B.
Last edited:
Ben, we are not in disagreement, but the point here is that ESL63 can only achieve his target of 30 dB MFB in case the ACH-01 perfectly tracks the woofer SPL between 5 and 500 Hz, otherwise you never get the phase lead between 5 and 200 Hz right and manageable. Nearfield is ideally less than 1 inch/2.5 cm mic distance.
No disagreement in principle, but like it or not, shooting for 500 Hz with Rice-Kellogg devices is historically beyond reach. Leaving the upper harmonics uncorrected, would be a trivial loss in performance. Perfection is the enemy of the good.
I'd taper off the loop gain above 120 Hz and that would save a lot of trouble for an insignificant hit to performance.
B.
I'd taper off the loop gain above 120 Hz and that would save a lot of trouble for an insignificant hit to performance.
B.
Yes, but for sufficient feedback at 120 Hz a more a less flat response to say 500-700 Hz is desirable: it makes the loopshaping a lot easier. That said: unless you are extremely lucky, a simple integrator is never going to yield more than 10-12 dB feedback. The Rob Munnig Schmidt trick, and that is why that technique is referred to as "loopshaping", consists of manipulating the frequency response of the loop in such a way you create phase headroom (less than 135 degrees lead and lag) just where you need it. Post 37 tells that story.
That is easier said than done, because first of all you need clean accelerometer output that perfectly tracks SPL between 5 and 500 Hz. This starting point requires a woofer specific experimenting approach.
That is easier said than done, because first of all you need clean accelerometer output that perfectly tracks SPL between 5 and 500 Hz. This starting point requires a woofer specific experimenting approach.