Back to the question: was it (a) light-weight feedback that nicely smoothed-out the major system bumps common to all the drivers or was it (b) heavy feedback to correct the fine points of distortion and transient behaviour?
You would be very happy to play a system which had no other benefit than cutting the principle resonance in half. That's easy to do without risking instability. But that degree of feedback would not make the system tight... like a good MF system ought to be.
Ben
The Infinity servo I measured reduced distortion by about a factor of 10, so pretty close to the limit of what is possible with analog MFB. I'm not sure what you mean by trying to differentiate between feedback correction of response and correction of distortion. Whatever level of loop-gain you decide flattens and extends the LF response of a woofer to your liking will be reducing the distortion a considerable amount as well. The feedback system doesn't know what is causing the difference between input and output, it reduces the difference by the same factor no matter what the source.
Thanks for the interesting data point.
In the context of the earlier discussion about serve-your-one driver versus servo-a-whole-bunch of drivers, it is a different matter between improving by a few dB the large bumps versus fixing distortion. Further, the metric of distortion reduction isn't quite the same as feedback fraction since sometimes just a little feedback (as I suggested might be the case with the Infinitys) removes overloads and other extreme problems.
MF (at least when referenced to the research using VC bridge circuits) does not raise low end output. These systems linearize cone motion as seen by the bridge vis a vis signal and therefore, driver output being a kind of artful blend of factors including the turnover point for mass-control, it makes it softer. In such cases, you need to equalize to restore the low end (which ends up with the same power input as without MF). But you get a much better, tighter, nicer low end.
Ben
Last edited:
Apologies for the tardy reply, I somehow missed notification of your post.
I’m having a hard time understanding your statements. I think you are saying that it may take less feedback to flatten the response to some desired measure of flatness than to reduce distortion by some desired factor or to some desired percentage. If so, then I agree that could certainly be the case. But, distortion will still be reduced by an amount based on the loop gain you have chosen to use for flattening the response. For example, suppose you decide that a loop gain of 10dB is adequate to flatten the response the way you would like. The 10dB of loop gain will reduce distortion by a factor of 4; not insignificant, although only about a third what is possible with a loop gain of 20dB.
This is true because, as you may recall, the feedback signal generated by a VC or VC bridge circuit is proportional to velocity. So, once the feedback loop is closed the circuit tries to drive the woofer cone such that its velocity is constant for all frequencies.
But we don’t really care about cone velocity; we care about SPL which is proportional to cone acceleration not velocity. Constant cone acceleration results in flat SPL response, which is why using an accelerometer sensor in a MFB system is arguably a more straight forward approach.
Constant cone velocity on the other hand results in SPL falling 6db/oct as you go down in frequency.
So yes, if you use velocity feedback you will need to boost the low end by 6dB/oct to get a flat response. Alternatively, you can differentiate the velocity feedback signal before applying it to the feedback loop comparator and you will get flat response without the LF boost.
Both approaches have been used in published articles and patents.
How can you make a voice coil pick-up using the same magnet system as the driving coil without having interference? Putting two layers on top of each other will make a transformer, putting them at other heights of the vc former will introduce assymmetry, and in both cases they could be affected by eddy currents. Will copper rings do the trick like the copper bases of accelerometers in MFB systems?
I also experienced the mfb woofers (AD8067/mfb from Philips) to be very stiff and solid, not a very typical woofer. The cone is extra heavy and the rubber surrounds are thicker as well. This was mainly to reduce distortion of these parts as the boxes were small enough to suck in the rubber surround when the cone was moving outward.
This extra thickness and weight introduces colouring in the 150 hz area and above. This is why people would call 'em cheap sounding woofers I guess.
It almost is a miniature car subwoofer, stiff cone, stiff surround. Not great candidads for playing mid-range.
The 8" mfb woofers' passive, non-mfb related cousin AD8066/W4 sounds really nice in comparison, much better than the actual mfb woofer especially if you compare them free air without feedback. The same goes when you put them in 30l boxes running passively. Activating feedback will rule out a lot of the differences between the two, but the colouring in the low mids still are ugly in the 8067/MFB while the originally passive AD8066/W4 will sound a lot more real in the upper notes of it's band. It's lighter cone reacts better to transients, dampens break-up more pleasantly and has a lower fs of 27 hz vs the 43 hz of the original feedback woofer. The only things the MFB woofer does better are: not collapse in a 9l box, and handle more power. Those are not really of my interest, I just enjoy an already good woofer having a lot less distortion around system resonance and especially below that.
Personally I don't use MFB in the midrange. I think the added mass of an accelerometer and the phase-shifting filters just make the system too slow to really work transients at high frequencies. Add to that the distortion of the pick-up. Well.
Would be interesting to see at what frequency the system is a cycle behind.
Crossing at 70hz from activated AD8066/W4, to AD12100/M4 and then to some other units as well.
I also experienced the mfb woofers (AD8067/mfb from Philips) to be very stiff and solid, not a very typical woofer. The cone is extra heavy and the rubber surrounds are thicker as well. This was mainly to reduce distortion of these parts as the boxes were small enough to suck in the rubber surround when the cone was moving outward.
This extra thickness and weight introduces colouring in the 150 hz area and above. This is why people would call 'em cheap sounding woofers I guess.
It almost is a miniature car subwoofer, stiff cone, stiff surround. Not great candidads for playing mid-range.
The 8" mfb woofers' passive, non-mfb related cousin AD8066/W4 sounds really nice in comparison, much better than the actual mfb woofer especially if you compare them free air without feedback. The same goes when you put them in 30l boxes running passively. Activating feedback will rule out a lot of the differences between the two, but the colouring in the low mids still are ugly in the 8067/MFB while the originally passive AD8066/W4 will sound a lot more real in the upper notes of it's band. It's lighter cone reacts better to transients, dampens break-up more pleasantly and has a lower fs of 27 hz vs the 43 hz of the original feedback woofer. The only things the MFB woofer does better are: not collapse in a 9l box, and handle more power. Those are not really of my interest, I just enjoy an already good woofer having a lot less distortion around system resonance and especially below that.
Personally I don't use MFB in the midrange. I think the added mass of an accelerometer and the phase-shifting filters just make the system too slow to really work transients at high frequencies. Add to that the distortion of the pick-up. Well.
Would be interesting to see at what frequency the system is a cycle behind.
Crossing at 70hz from activated AD8066/W4, to AD12100/M4 and then to some other units as well.
Always gratifying to be agreed-with by bolserst.
Scary as bolserst's implication of the need for calculus in "integrating" velocity into acceleration (or do I have that backwards?), we are only talking about adding a wee single capacitor into the feedback loop. That's the only difference.
Although not a binary distinction, I think it is helpful to distinguish "a little" MF from "a lot." It is helpful, easy, and not likely to blow up an expensive driver when you use a little feedback to smooth the freq response curve and tame some distortion. But the opposite when you go for serious MF to truly tame impulse response and get the driver truly into the feedback loop.
Apropos Robintip's post, whatever the mutual inductance might be, voice coil MF works great.
As much as I like the idea of extending the feedback loop, there is so much "art" in driver construction that for the moment, MF only makes sense for the lower range where something resembling piston behaviour of cones can be relied on. Interesting question how far up the scale you can go.
The Philips system has, from Day One, seemed pretty ill-conceived to me. Taking mighty efforts to get superior sound from inferior drivers and boxes. Why? Using cheap accelerometers to try to control cone motion precisely?
I remain puzzled why there is so much advocacy of accelerometer MF and so little interest in VC feedback. Granted, VC is simple, anybody can experiment with it in 20 minutes... no need to read data sheets... Takes no time to stick a resistor in series with a woofer and compare the signal across the driver to the input signal. Eyeball that because that's the MF feedback.
BTW, if you have a Kenwood Basic amp with the Sigma wires, you are doing it already.
Ben
Scary as bolserst's implication of the need for calculus in "integrating" velocity into acceleration (or do I have that backwards?), we are only talking about adding a wee single capacitor into the feedback loop. That's the only difference.
Although not a binary distinction, I think it is helpful to distinguish "a little" MF from "a lot." It is helpful, easy, and not likely to blow up an expensive driver when you use a little feedback to smooth the freq response curve and tame some distortion. But the opposite when you go for serious MF to truly tame impulse response and get the driver truly into the feedback loop.
Apropos Robintip's post, whatever the mutual inductance might be, voice coil MF works great.
As much as I like the idea of extending the feedback loop, there is so much "art" in driver construction that for the moment, MF only makes sense for the lower range where something resembling piston behaviour of cones can be relied on. Interesting question how far up the scale you can go.
The Philips system has, from Day One, seemed pretty ill-conceived to me. Taking mighty efforts to get superior sound from inferior drivers and boxes. Why? Using cheap accelerometers to try to control cone motion precisely?
I remain puzzled why there is so much advocacy of accelerometer MF and so little interest in VC feedback. Granted, VC is simple, anybody can experiment with it in 20 minutes... no need to read data sheets... Takes no time to stick a resistor in series with a woofer and compare the signal across the driver to the input signal. Eyeball that because that's the MF feedback.
BTW, if you have a Kenwood Basic amp with the Sigma wires, you are doing it already.
Ben
Last edited:
Most of the advocacy for accelerometer feedback must come from the folks that are still amazed at deep bass from small boxes, because omg and wow.
Not at all what we're after.. It's much more interesting to see good speakers do better. Fortunately, a lot of the tech from the later MFB's happens to do just that when coupled to good drivers.
The fact that this kind of feedback was used in a heavily compromised woofer and box in the original MFB boxes, is not a reason to ONLY use it that way. The tech works for any woofer capable of housing an accelerometer. Of course, it is always an option to lower the amount of feedback to your liking.
I think, these are the points why more people are getting interested:
- Removing distortion below system resonance
- Active amplification and filtering
- Lowering resonance
- General reduction of distortion
- Actively reducing the low roll-off point frequency by using an inverted filter and a rumble filter
Ultimately, to my ears it just is a sub-woofer thing. These frequencies are low enough to not notice a sort of 'echo' effect. Maybe the cheaper voice-coil feedback would handle higher frequencies better because of its simplicity. But, I'm still not convinced because of numerous tests stating eddy current, voice coil heating, magnet assymilarity problems are still not to be overcome.
It is, however, not the first time that technology is not on par with idea's on this subject. And as long as laser position sensors with low latency, high speed and high accuracy are not available for the average Joe, I'll just stick with accelerometer feedback in sub-woofer duty.
Not at all what we're after.. It's much more interesting to see good speakers do better. Fortunately, a lot of the tech from the later MFB's happens to do just that when coupled to good drivers.
The fact that this kind of feedback was used in a heavily compromised woofer and box in the original MFB boxes, is not a reason to ONLY use it that way. The tech works for any woofer capable of housing an accelerometer. Of course, it is always an option to lower the amount of feedback to your liking.
I think, these are the points why more people are getting interested:
- Removing distortion below system resonance
- Active amplification and filtering
- Lowering resonance
- General reduction of distortion
- Actively reducing the low roll-off point frequency by using an inverted filter and a rumble filter
Ultimately, to my ears it just is a sub-woofer thing. These frequencies are low enough to not notice a sort of 'echo' effect. Maybe the cheaper voice-coil feedback would handle higher frequencies better because of its simplicity. But, I'm still not convinced because of numerous tests stating eddy current, voice coil heating, magnet assymilarity problems are still not to be overcome.
It is, however, not the first time that technology is not on par with idea's on this subject. And as long as laser position sensors with low latency, high speed and high accuracy are not available for the average Joe, I'll just stick with accelerometer feedback in sub-woofer duty.
If you want to use VC feedback with a single magnet system and two coils, there will be inductive coupling as your describe. It tends to only come into play(ie be of significant magnitude relative to the desired signal voltage induced by motion relative to magnetic field in gap) at frequencies > 100Hz. The options are to use a separate magnet system for the sensor VC, or put the driven VC in a bridge arrangement to get the velocity feedback signal.
The apparent simplicity of the bridge arrangement is appealing. It can provide useful reduction in distortion with minimal circuitry. However, I was never able to achieve the high levels of stable open loop gain and distortion reduction that I could with an accelerometer.
The issues included:
- bridge imbalance(VC inductance function of VC position and current, VC resistance function of temperature)
- sensor distortion(BL variance with VC position, flux modulation from VC current).
My comment wasn’t meant to be scary. As you state, passive differentiation circuits can be made from a single capacitor and load resistor the response of which is a first order HP network. The trade off in the passive design is setting the roll off point. There higher frequency you choose(smaller capacitor) the more reduction in signal level…so lower S/N ratio. The lower frequency you choose(bigger capacitor) the more phase shift in the feedback loop...lower stability margins. All things considered, a bass boost EQ prior to the feedback control loop is probably the best approach.
The Sigma circuit does not connect the speaker VC as one leg in a balanced AC bridge for distortion reduction. Rather, the voltage is sampled directly across the VC to place the speaker wire impedance, protection relay, etc. inside the feedback loop. The result is increased damping factor, but no motional feedback.
The Sigma circuit was discussed a bit earlier in this thread(starting post #115)
http://www.diyaudio.com/forums/subwoofers/239941-analog-servo-sub-12.html#post3653431
Attachments
Last edited:
Robintip - as far as I can tell, exactly the same functional benefits using a VC/bridge system as using an accelerometer. The bridge approach seems to me to produce a more perfect feedback replica of cone motion than a separate winding but I have no test experience with a separate winding. Yes, bridge feedback can be unstable and takes a while to sort things out - as is also the case with the sometimes elaborate circuits for an accelerometer. I suppose somebody could manufacture drivers suitable for MF.
bolsterst - yes, the Sigma circuit kind of looks like the hook-up for remote-sensing of a bench power supply delivery. But in the Sigma amp circuit, the speaker wires (but not the sensor wires) are part of the bridge if their resistance is non-trivial. Beyond my understanding to know the difference between current feedback (like in the Sigma) and MF feedback.
I am always amazed at how much bolsterst knows. He's the expert on ESLs too. Accurate knowledge and even says things with Canadian-level courtesy. What a brain!
Ben
bolsterst - yes, the Sigma circuit kind of looks like the hook-up for remote-sensing of a bench power supply delivery. But in the Sigma amp circuit, the speaker wires (but not the sensor wires) are part of the bridge if their resistance is non-trivial. Beyond my understanding to know the difference between current feedback (like in the Sigma) and MF feedback.
I am always amazed at how much bolsterst knows. He's the expert on ESLs too. Accurate knowledge and even says things with Canadian-level courtesy. What a brain!
Ben
Last edited:
There is no way a bridge arrangement could ever work properly as a way of accurately sensing the velocity of the diaphragm because it relies on constant Bl and no voice coil inductance which are the very thinbgs that cause distortion in the first place. Also the balance of the bridge changes as the voice coil resistance changes with temperature and so goes out of balance quite easily. The bridge arrangement is a crude way of implementing motional feedback and if it was so effective it would almost be universally used since it would not require the use of an accelerometer.
regards
Trev
You're right... but.
First, there are no non-crude ways to do MF. To this day, there seems to be no way to commercialize an MF system that has big enough feedback to improve a good driver but is also not on the verge of instability. And that's because Rice-Kellogg cone drivers are a lousy engineering concept, barely limping along after 100 years to produce good sound.
You can balance a bridge to any degree of precision you want. In any experimenting I did (long ago), I found it wasn't helpful to simulate the driver with endless niceties within the bridge. And that is because the imbalances you are concerned about have little or no effect on the basic operation of the feedback loop. Basically, you just want to grab the back-EMF feedback and the factors you mention just are slight trimmers on the amount of feedback. In any case, I like to focus on the big effects before sweating the small (and sometimes just theoretical) stuff.
If you are saying your VC isn't good enough to provide accurate feedback, you are saying your VC isn't good enough to provide good music without feedback either.
Ben
You are missing the point. The parameters you are trying to correct such as Bl non linearities are the very things you are relying on to provide accurate feedback of the diaphragm. There are no benefits to using a bridge arrangement because it is only linear at low power and cone excursion where the distortion is low anyway. At high power and high cone excursion is where you need it to be accurate and it fails dismally at those extremes.
The only real and small advantage of a bridge circuit is to approximate constant velocity vs frequency operation and with a suitable bandpass shaping filter you can then arbitrarily determine the low frequency cutoff point but this could equally be achieved using a linkwitz transform eq circuit so no cigar on this one.
Also the voice coil inductance will cause havoc and create a peaking in the response at the mass/inductance resonant point if it is not taken care of.
regards
If your driver has "BL non linearities" I'd think it would be a pretty stinky driver, MF or not.
There are various advantages to MF. The biggest one is transient behaviour with distortion and freq response being nice but secondary. Of course, at DIYaudio many people use TS software which aims at efficient Qs instead of best Qs.
Whatever the theoretical issues, in numerous articles, some in respected publications, the results of VC MF are dramatically wonderful.
Frankly, I am sorry to engage in all this theoretical back-and-forth. Let's see some data.
Ben
There are various advantages to MF. The biggest one is transient behaviour with distortion and freq response being nice but secondary. Of course, at DIYaudio many people use TS software which aims at efficient Qs instead of best Qs.
Whatever the theoretical issues, in numerous articles, some in respected publications, the results of VC MF are dramatically wonderful.
Frankly, I am sorry to engage in all this theoretical back-and-forth. Let's see some data.
Ben
Last edited:
Have you taken a look at a klippel diagram? There are hardly any woofers without BL non linearities when talking about overhung voicecoils.
Isn't one of the common definitions of Xmax, that the driver motor has been reduced to 70% of full BL?
I recently bought a woofer that had this definition in it's specs.
That has to come from the voice coil partially leaving the gap.
It sounds pretty non linear to me.
Dave
Are we talking about a world where our VCs leave the gap (simultaneous with our ear drums rupturing) and the sound is awful or are we just having big fun with extreme theoretical "what if" scenarios?
Whatever the non-linearities are, they just can't be much or the sound would be awful. You can't have it both ways: (a) my speaker sounds good and (b) your VC MF works terrible.
Back to data please.
Ben
Whatever the non-linearities are, they just can't be much or the sound would be awful. You can't have it both ways: (a) my speaker sounds good and (b) your VC MF works terrible.
Back to data please.
Ben
No, a world where BL is not constant over the usable range which ist given as xmax in the data sheets of the speakes and where the slope of BL decreasing when using backwards often is very different than the one when the cone is moving forwards.
Please do read up on klippel.
Compare a good underhung system with a bad overhung system - with the bad overhung system being a regular pro sound woofer being used for decades and nobody complaining about "awful sound". If you listen to both, comparing them directly to each other, you will hear a great difference caused by the different linearities in movement where one cause of this effect is BL Non Linearity.
It doesn´t always have to be "awfull" or "non existend", there are many shades of grey between
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.