Many thanks.
1. There are non-linearities of the suspension and non-linearities of the magnet system. How do they measure-up in a quality woofer at moderate excursions?
2. I lack the math to know if the accuracy of the sensing mechanism can fully compensate for the inaccuracy of the motor, at reasonable feedback fractions?
3. I've always wondered if the feedback from a non-linear magnet system is happily complementary to the initial error?
4. Ummm, I guess the "air spring" isn't as linear as Vilchur thought*?
Ben
*a pair of 1955?? AR-1's with Altec pancake "tweeters" just sold on eBay for $4700. I wonder what my derelict one is worth.
@bentoronto
1. My experience has been that at moderate excursions(<30% Xmax) many high-quality woofers have motor non-linearities that are less than suspension non-linearities. Once you start pushing beyond 50% Xmax, the non-linearities are of similar magnitudes.
2. The inaccuracy of the motor can never be fully compensated by MFB. At best, the inaccuracy is reduced by a factor = (1 + loop gain). For typical analog MFB systems this tends to be limited to a maximum factor of about 11(ie 20dB loop gain).
3. Unfortunately not. The output of the sensor is considered "truth" as far as the MFB system is concerned. In the VC feedback system this is the EMF generated from VC velocity. The assumption is that this "truth" signal is linearly related to the desired sound output. The feedback loop is trying to force the output of the sensor to equal the desired signal input. So, when it is working properly, "truth" = input voltage. If "truth" is in fact non-linearly related to the VC motion, the measured sound pressure will be distorted. It doesn't matter if the force on the VC due to current from the amplifier has a similar non-linear behavior or not.
4. Unless you are using a really small box, "air springs" do tend to be more linear than even the best woofer suspensions.
1. My experience has been that at moderate excursions(<30% Xmax) many high-quality woofers have motor non-linearities that are less than suspension non-linearities. Once you start pushing beyond 50% Xmax, the non-linearities are of similar magnitudes.
2. The inaccuracy of the motor can never be fully compensated by MFB. At best, the inaccuracy is reduced by a factor = (1 + loop gain). For typical analog MFB systems this tends to be limited to a maximum factor of about 11(ie 20dB loop gain).
3. Unfortunately not. The output of the sensor is considered "truth" as far as the MFB system is concerned. In the VC feedback system this is the EMF generated from VC velocity. The assumption is that this "truth" signal is linearly related to the desired sound output. The feedback loop is trying to force the output of the sensor to equal the desired signal input. So, when it is working properly, "truth" = input voltage. If "truth" is in fact non-linearly related to the VC motion, the measured sound pressure will be distorted. It doesn't matter if the force on the VC due to current from the amplifier has a similar non-linear behavior or not.
4. Unless you are using a really small box, "air springs" do tend to be more linear than even the best woofer suspensions.
I don't mind you saying so, it just shows you don't really know the motive of MFB as used by Philips, if you don't mind me saying so.
Their motive was to get low bass from a tiny cabinet, thus reducing the need for huge 'monkey coffin' size speakers to get bass.
In short: they wanted 25 hz f3 in a box normally suited for like a >100hz f3.
Simply using a Linkwitz Transform like circuit to compensate for high roll-off on a woofer in a really small cabinet proved that indeed the bass level was up, but distortion was up with it too.
This is because the higher frequency of the air spring in the box, was intermodulating with the low frequencies that the woofer was trying to produce.
They then wanted the woofer cone to 'ignore' the forces that were acting on it from the air spring. However, just ignoring these forces is impossible. So they made a system to cancel out these forces by means of negative signal feedback of just the erroneous signal that was composed of the accelerometer feedback compared to the input signal.
The result was a woofer that followed the input signal instead of getting influenced by the air spring.
Reducing distortion this way was just a side product in the process of realising this vision of small boxes with deep bass, but never the intention at first.
The real accomplishment of Philips was that they could get the same woofer that was designed for a 80 litres enclosure to play much lower in just a 9 litres enclosure AND reducing distortion, just trading in max SPL.
The fact that the drivers used weren't the best (compared to other brands at least), doesn't mean feedback is useless in systems with better drivers:
Regarding your remark:
I share your interest. I have the following vision on this:
In normal speaker systems, there is only 1 ideal situation where the woofer and the 'air spring' are in harmony and produce close to accurate sound.
This is at the system resonance frequency, at reference level (the level used to measure the woofer's spec which is then used to calculate the box) and with this exact frequency used in a sine wave as an input signal.
At all other levels, at all other frequencies and with any other waveform put in, it isn't in harmony: the two elements are trying to interact, causing some level of distortion. Thats feels like saying: when you play music, it distorts to some level.
So when you enable feedback on a woofer, it isn't going to improve the woofer itself. It just makes one constant factor of a system: enclosure volume, irrelevant to the driver's performance. Thus improving the performance of the system. And that can be any system with a woofer.
I enjoyed the low end on the original MFB speakers, but I am moving on. Currently prototyping a personal system using MFB in high quality bass drivers in the bottom part of a 4-way, because I don't mind monkey coffins. I enjoy the improved sound on them a lot.
Side note: Philips ditched VC feedback because of non-lineairity and other problems such as eddy currents.
I think the VC still can provide a function in which current sensing of the VC is used as a limiter, just not for feedback.
Accelerometers are still measuring something relative to time, which makes them slow. So that's why I use them in low range only.
Last edited:
An interesting post. But the basic facts remain that Phillips took a cheap driver, put it in small cheaply constructed box, and used MF to improve it. I admire their accomplishment and I truly wish they (and Sony and few other manufacturers who tried to commercialize MF) has sold millions of speakers.
I wish I knew more research on "air suspensions" because like MF it is a kind of negative feedback which operates ONLY in sealed boxes, including the sealed boxes on Klipschorns. All I hear are the ancient adiabatic versus isothermal rumours. My impression is that an "air spring" does its linear action at all frequencies and has no special behaviour at the system resonance.... except that it becomes the main restoring force from resonance on down.
Interestingly, MF likewise makes no sense except for sealed boxes and infinite baffles (maybe horns too).
Ben
I wish I knew more research on "air suspensions" because like MF it is a kind of negative feedback which operates ONLY in sealed boxes, including the sealed boxes on Klipschorns. All I hear are the ancient adiabatic versus isothermal rumours. My impression is that an "air spring" does its linear action at all frequencies and has no special behaviour at the system resonance.... except that it becomes the main restoring force from resonance on down.
Interestingly, MF likewise makes no sense except for sealed boxes and infinite baffles (maybe horns too).
Ben
The latest models might be low-cost cabinets, parts and finish. The drivers used however, were the highest quality drivers philips had when the corresponding series of MFB speakers came out. However I still think the woofers were a compromise between sensitivity, output, stiffness and high end extension, and they should have been bigger 4-way from early on, not having as much compromises. (but then it would be a monkey coffin again)
The earlier models were nothing like cheap. They were built like tanks.
They would have been really expensive at the time, but not why not many of them were sold. Philips international was not a star at marketing with a company spread all over the world after the war. The japanese did, however.
The earlier, most expensive 545 studio monitor model features a 12" woofer and expensive amplifier built in. It is one of the more expensive speakers at around that time. Low end was the best, but the mids and highs were surpassed by the competition. Also, studio engineers wanted loud speakers, and did not care for 20Hz frequency reproduction apparently.
In early documents of MF by Philips, they indeed say that the system only seemed to work correctly in a sealed enclosure.
I don't have a lot of knowledge on air springs, the main thing I think is important, is that MF deals quite well with the distortion below system resonance of any closed woofer system.
It is remarkable that a 7" in a 4 litre box does sound similar to a 12" in a 40 litre box using MF.
If both have the rumble filters set up to f3 at 27 Hz, you would have great difficulty telling them apart, blindfolded A-B comparison when listening to 27Hz at the same level.
The benefit of MF (said on other forums) lies in aplications lower than 500Hz.
Personally, I don't cross them so high. I think the power lies in the even lower regions, because of the slow accelerometer and not being able to follow transients at higher frequencies, but again, others disagree.
Finding a way to actually measure the cone's position (in stead of acceleration), comparing it to the height of a signal and real-time correction at hundreds of khz might be the solution to have it working at higher frequencies. For the time being, I just use it below 200Hz with great results.
Power compression because of heat is not an issue, it is simply being compensated for by the accelerometer system. To limit damage because of this, in original MFB boxes, additional automatic limiters limit the power when pushed to hard, then automatically gives it back after the correspondent unit is supposed to have cooled down. Now this is also not a very cheap feature.
Good drivers still have problems below system resonance, this is solved by MFB, at least to my ears.
And if you've got the xmax to spare, try extending the low end too.
The earlier models were nothing like cheap. They were built like tanks.
They would have been really expensive at the time, but not why not many of them were sold. Philips international was not a star at marketing with a company spread all over the world after the war. The japanese did, however.
The earlier, most expensive 545 studio monitor model features a 12" woofer and expensive amplifier built in. It is one of the more expensive speakers at around that time. Low end was the best, but the mids and highs were surpassed by the competition. Also, studio engineers wanted loud speakers, and did not care for 20Hz frequency reproduction apparently.
In early documents of MF by Philips, they indeed say that the system only seemed to work correctly in a sealed enclosure.
I don't have a lot of knowledge on air springs, the main thing I think is important, is that MF deals quite well with the distortion below system resonance of any closed woofer system.
It is remarkable that a 7" in a 4 litre box does sound similar to a 12" in a 40 litre box using MF.
If both have the rumble filters set up to f3 at 27 Hz, you would have great difficulty telling them apart, blindfolded A-B comparison when listening to 27Hz at the same level.
The benefit of MF (said on other forums) lies in aplications lower than 500Hz.
Personally, I don't cross them so high. I think the power lies in the even lower regions, because of the slow accelerometer and not being able to follow transients at higher frequencies, but again, others disagree.
Finding a way to actually measure the cone's position (in stead of acceleration), comparing it to the height of a signal and real-time correction at hundreds of khz might be the solution to have it working at higher frequencies. For the time being, I just use it below 200Hz with great results.
Power compression because of heat is not an issue, it is simply being compensated for by the accelerometer system. To limit damage because of this, in original MFB boxes, additional automatic limiters limit the power when pushed to hard, then automatically gives it back after the correspondent unit is supposed to have cooled down. Now this is also not a very cheap feature.
Good drivers still have problems below system resonance, this is solved by MFB, at least to my ears.
And if you've got the xmax to spare, try extending the low end too.
Nicely done!No 13hz (rocking/cantilever/????) mode in sight
He, he. No rocking this time. Just me rocking a little..
I am going to experiment a little more and also have to make a PCB and some noise shielding. The circuit is VERY sensitive for 50 Hz hum. I had to run everything on battery power and remove the mains fuses for my lab (!) to get rid of the 50Hz during testing.
Thanks for your help and critical feedback all of you
Armand
Armand,
Are you using shielded twisted pair wire from the sensor to the circuit, shield lifted at the sensor, circuit shield terminating to a low impedance earth ground?
Art
??????
This brings up an interesting question. Could you somehow slave additional woofers to a single feedback woofer and spread some of the benefit of feedback to a multiple cheap woofer system? I am imagining the woofers would have to be pretty well matched. You might need to measure the input voltage of the feedback woofer, isolate that signal, scale it, and use a separate amp to drive the non-feedback woofers.
I'm guessing the result would be better than no feedback at all, but not as good as a pure feedback system. It's probably not worth the effort, but it is an interesting question.
Yes, I would get the same SPL from a 10 (or 9mm) x-max 10" woofer. But I thought that with only 3.3mm x-max the woofer would have less distortion than a 10" with 9mm. With 8 drivers in front I would have enough SPL anyway. But I was wrong. Distortion is all about magnet motor quality
It is a little wast of time doing all this work on cheap drivers. On the other hand I would not feel very comfortable cutting the dust cap off a $800 driver
Anyway, if my DBA project is succesful I can always upgrade later.
This brings up an interesting question. Could you somehow slave additional woofers to a single feedback woofer and spread some of the benefit of feedback to a multiple cheap woofer system? I am imagining the woofers would have to be pretty well matched. You might need to measure the input voltage of the feedback woofer, isolate that signal, scale it, and use a separate amp to drive the non-feedback woofers.
I'm guessing the result would be better than no feedback at all, but not as good as a pure feedback system. It's probably not worth the effort, but it is an interesting question.
@Armand:
@lehmanhill: experiments with multiple woofers using 1 feedback signals have been carried out, but not in the way you suggest I guess (I think you mean that each woofer has it's own cabinet).
See this document (english) for more info: http://docs.mfbfreaks.com/div/SubServoAccelDeGreef.pdf
I'm sure multiple woofers using 1 airspace is not the best way to reduce distortion (and thats a conservative statement). More accaptable results can be produced with multple woofers of the same brand/type in exact copies of cabinets.
Remember that if you alter the weight of the backfeeding woofer by placing the accelerometer, you change the resonant frequency of a woofer. If you do not add the exact same weight to the other woofers, you create a difference between the woofers that will add distortion to the 'extra' woofers.
Adding weight in the light of feedback is discussed in this (english) document, page 7: http://docs.mfbfreaks.com/div/speakbuil90mfb.pdf
Difference between individual examples (manufacturing tolerance) is also a factor that would be playing a big role in the actual reduction of distortion.
Measuring the woofers, and using the ones that qualify as 'average' as accelerometer feedback sources would be my best guess. That the system is somehow compromised does not mean that it would sound worse than without the feedback. I am positive that it will be an improvement anyway, because a non-feedback system containing multiple woofers will ignore driver-driver irregularities.
Using one feedback signal to correct multiple woofers is still an interesting experiment so I'm planning to try this in the near future. I will be using 4x Philips AD8066 (resonance of 27 Hz) per side, in a series-parallel configuration powered by a Sony ES530 amp that I still need to fix.
This experiment is purely for fun/entertainment, I am not aiming at lowest distortion factors (the woofers are not suitable for that target to begin with).
It is not a quest in pursuit of lowest distortion possible, so it's basically off-topic here.
Look at the resistor. This is part of a solution to reduce humm.
@lehmanhill: experiments with multiple woofers using 1 feedback signals have been carried out, but not in the way you suggest I guess (I think you mean that each woofer has it's own cabinet).
See this document (english) for more info: http://docs.mfbfreaks.com/div/SubServoAccelDeGreef.pdf
I'm sure multiple woofers using 1 airspace is not the best way to reduce distortion (and thats a conservative statement). More accaptable results can be produced with multple woofers of the same brand/type in exact copies of cabinets.
Remember that if you alter the weight of the backfeeding woofer by placing the accelerometer, you change the resonant frequency of a woofer. If you do not add the exact same weight to the other woofers, you create a difference between the woofers that will add distortion to the 'extra' woofers.
Adding weight in the light of feedback is discussed in this (english) document, page 7: http://docs.mfbfreaks.com/div/speakbuil90mfb.pdf
Difference between individual examples (manufacturing tolerance) is also a factor that would be playing a big role in the actual reduction of distortion.
Measuring the woofers, and using the ones that qualify as 'average' as accelerometer feedback sources would be my best guess. That the system is somehow compromised does not mean that it would sound worse than without the feedback. I am positive that it will be an improvement anyway, because a non-feedback system containing multiple woofers will ignore driver-driver irregularities.
Using one feedback signal to correct multiple woofers is still an interesting experiment so I'm planning to try this in the near future. I will be using 4x Philips AD8066 (resonance of 27 Hz) per side, in a series-parallel configuration powered by a Sony ES530 amp that I still need to fix.
This experiment is purely for fun/entertainment, I am not aiming at lowest distortion factors (the woofers are not suitable for that target to begin with).
It is not a quest in pursuit of lowest distortion possible, so it's basically off-topic here.
In January 2012, I had two Philips MFB 22RH541 for a while.
Something like 35 years old.
No "special" components inside but the quality and cleverness of construction are impressive. The boxes are robust and easy to open.
I checked the performances. Just as day 1. Voltage at the check points were strictly in accordance with what is indicated in the service manual, no need to adjust the trim-pots.
I currently would like to see products built like that.
Something like 35 years old.
No "special" components inside but the quality and cleverness of construction are impressive. The boxes are robust and easy to open.
I checked the performances. Just as day 1. Voltage at the check points were strictly in accordance with what is indicated in the service manual, no need to adjust the trim-pots.
I currently would like to see products built like that.
Well, if you'd want something simular, that is possible, but not in a ready-to-use way. Order a set of DHCP-2 modules, accelerometers and good PSUs, take a set of Exodus Anarchy 7" midbass (12.5mm Xmax one-way, usable midrange, low Vas) and a SB29RDCN tweeter (low distortion, usable low-end)
Exodus Anarchy:
SB29RDCN:
Thanks Zaph for measurements.
I will be using this configuration in a 541 box probably, if I can get a set for good value.
You might have to build a box yourself, but in the end you've got yourself a set of revamped 541.
Because of the high xmax of the Exodus, you can get f3 at 27Hz at very reasonable listening levels.
To hear this small thing digg that low and with such a small amount of distortion really must be counter-intuitive. Looking forward to it
Exodus Anarchy:
SB29RDCN:
Thanks Zaph for measurements.
I will be using this configuration in a 541 box probably, if I can get a set for good value.
You might have to build a box yourself, but in the end you've got yourself a set of revamped 541.
Because of the high xmax of the Exodus, you can get f3 at 27Hz at very reasonable listening levels.
An externally hosted image should be here but it was not working when we last tested it.
To hear this small thing digg that low and with such a small amount of distortion really must be counter-intuitive. Looking forward to it
I am using a sensor with a fixed cable. Shield is grounded but not lifted at the sensor.
I can try to cut the shield at the sensor if I continue to have problems. But first I have to make a proper circuit board etc.
@ Lehmanhill
I am going to try to correct more more woofers from the one I have when time allows.
@Robbintip
Thanks for the fine links. A perfect read during christmas holiday.
BTW. Nothing is off-topc here if it involves MFB
Let us suppose that errors are of two types: (1) structural (and shared by all the woofers) and (2) random or adventitious.
So it all depends on how much of the error is structural and how uniform are the woofers. There's also a paradox for better quality drivers that have more uniformity but less structural error.
For sure, the notion of sensing one system to control other systems seems vaguely abhorrent to me! For "V" engine cars that have dual exhaust paths, I wonder if auto makers now use a separate exhaust sensors ("O2") in each exhaust path?
Ben
The main reason for using dual O2 sensors in a "V" engine is that the O2 sensor must be installed about 4 inches from the exhaust port for accuracy due to a phenomenon called the "after-burn" effect.
As long as you have identical cylinders there is no problem measuring just one and adjusting the others accordingly. You don't see one O2 sensor per cylinder on cars either?
I will build the boxes, do the measurements and give you all the test results sometime next year.
Armand
EDIT: You can buy systems that measure O2 per cylinder if you want to fine tune each cylinder. http://www.dragzine.com/tech-storie...tios-from-every-cylinder-with-aem-electonics/
Off-topic??
Last edited:
Very good point. You would have to find cheap woofers with good uniformity in all characteristics and I agree, cheap woofers are less likely to be uniform.
As for controlling one system with another, it is probably more risky than any of us know. It is also something that I don't really need, as my single feedback sub can play deep bass as loud as I need. Its just that I can't stop my curiosity about the possibility and I applaud Armand and others for actually experimenting.
As for car engines, Armand is correct. Modern car engines can tell from crank position and O2 signal what each cylinder is doing and adjust, at least in the best systems. No point in controlling one system from another if you want accurate results.
Five minutes into this video Paul says that "one" driver has an accelerometer..
Now whether he is speculating I don't know.
https://www.youtube.com/watch?v=NPazywXrqJo
EDIT:
~5:38 into it...
.
Now whether he is speculating I don't know.
https://www.youtube.com/watch?v=NPazywXrqJo
EDIT:
~5:38 into it...
.
Thanks for video link.
I'm keen to play Mahler's 3rd Symphony in a way true to his intentions (which is not the same as a mic in Row H). But quite a different esthetic from trying to "reproduce" the cooked-up studio sounds of Pink Floyd. Both worthy endeavours but easier for me to recognize when the flute is carrying the second theme like it should be clearly doing at the end of the second movement as opposed to my personal-fantasy of what Pink Floyd ought to sound like.
Back to feedback. Here's the plain poop: a single-sensor strategy using small unrisky amounts of feedback can readily address major errors like major system resonances. I suppose the effect would be a dramatic improvement in cutting boom.
But I'd guess it would be risky to push the feedback loop to the point where distortion and transient response would begin to be addressed.
Let me characterize the MF systems I worked with. When you push on the dust cap with your hand, it feels quite solid (OK... maybe solid cardboard), not much like a driver usually feels. So I wonder if the Infinity subs (at least the one with the sensor) feels that way?
My 2-cents.
Ben
I'm keen to play Mahler's 3rd Symphony in a way true to his intentions (which is not the same as a mic in Row H). But quite a different esthetic from trying to "reproduce" the cooked-up studio sounds of Pink Floyd. Both worthy endeavours but easier for me to recognize when the flute is carrying the second theme like it should be clearly doing at the end of the second movement as opposed to my personal-fantasy of what Pink Floyd ought to sound like.
Back to feedback. Here's the plain poop: a single-sensor strategy using small unrisky amounts of feedback can readily address major errors like major system resonances. I suppose the effect would be a dramatic improvement in cutting boom.
But I'd guess it would be risky to push the feedback loop to the point where distortion and transient response would begin to be addressed.
Let me characterize the MF systems I worked with. When you push on the dust cap with your hand, it feels quite solid (OK... maybe solid cardboard), not much like a driver usually feels. So I wonder if the Infinity subs (at least the one with the sensor) feels that way?
My 2-cents.
Ben
Last edited:
You actually had it right the first time.
Infinity did make several speakers systems that used an accelerometer on one woofer to control 2, 4, or 6 woofers depending on the size of the woofer tower. There really isn't any other way to do it without using a separate amplifier for each woofer.
The IRS Vs shown in the video link had 6 woofers controlled by one.
The IRS Beta's had 4 woofers controlled by one(servo circuit for the Beta's on last page of attached tech sheet)
Stereophile review of the Beta's also mentions the single accelerometer.
Infinity IRS Beta loudspeaker | Stereophile.com
"...The Beta uses a much more sophisticated form of servo control. One of the four woofers in each column has a high-quality accelerometer attached to its cone which measures the cone's changes in motional velocity. A second pair of (lightweight) wires carries the accelerometer output back to the crossover unit, and the signal is compared with the input signal, the difference being applied out-of-phase to the signal to correct for the discrepancy. (The woofers are said to be closely enough matched that the behavior of one will be typical of all.)"
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.