| MarkMcK |
A Cheap and Super Simple Motional Feedback (or Servo) Woofer
Two caveats:
One, I reserve the right to publish any or all of the following in other venues.
Two, if you do not feel comfortable working inside the covers of an
amplifier or are unwilling to face the consequences if something goes
wrong or are unwilling to void a manufacturer’s warranty, then do not
begin the motional feedback woofer project described herein.
A Really Simple Motional Feedback Woofer Design
This is a really simple design that I hope will provide listening
pleasure to any “diy”er, regardless of ascribed theory of enclosure
design.
An Introduction to Transient Thinking
When you input a transient signal to a woofer, the motor of the woofer
drives the cone producing what we call sound. When the transient input
signal ends, the suspension of the woofer acts to return the cone
toward its origination position producing more sound and often
overshooting the original pre signal position of the cone. This
undriven movement of the cone then decays into the natural resonant
frequency of the driver and its enclosure alignment. Again, this signal
looks similar to a decaying sinusoidal waveform, but never achieves the
regularity or consistency over time of a sinusoidal waveform. This
decay signal is the sound of the woofer. This decay signal is not the
sound of the inputted signal. Since woofers in particular and
loudspeakers in general are sound reproducers and not producers of
sound, this sound production is a bad thing.
While controversial, when you examine the acoustic performance in a
transient defined world, you can separate the various modalities of
sound reproduction across time. When excited by a broadband impulse,
the low order stop band filters define the onset system output. At
later periods, the driver and system resonance modes dominate. When we
speak of driver output at system resonance, the late period response is
where the majority of the sound is located. Because of this, output at
resonance bears little resemblance to the input. Since we usually
define distortion as a difference between the shape of the input and
output signal, it seems logical that the sound at resonance is highly
distorted. It also seems reasonable to say that this is a bad thing.
Feedback
Feedback does wonderful things, but most of the time it does them by
simply changing the closed loop gain of the amplifier. When the signal
at the output of the amplifier varies from the input by being less than
the input signal plus the theoretical gain of the amplifier, feedback
decreases and closed loop gain increases. When the signal at the output
of the amplifier varies from the input by being more than the input
signal plus the gain of the amplifier, feedback increases and the
closed loop gain decreases.
When we use motional feedback from a woofer, we are doing nothing
different than the amplifier is already doing to itself. With the
woofer we must assume some range of linear frequency and some limits of
linear operational magnitude as a set point. When cone motion exceeds
what is consistent with the input to the driven coil, the sensing coil
generates a voltage greater than that on the drive coil (as long as the
motion is within our operational set points), and this signal fed back
to the amplifier reduces the closed loop gain. When the cone motion is
less, the opposite occurs.
Both feedback in electronic circuits and motional feedback in
loudspeakers have limits. You can do a lot with feedback, but try to
use too much feedback and things can go wrong. Instead of making signal
amplification or signal transformation better, too much feedback can
make it worse. “All things in moderation” is a wonderful proverb. In
this MFB woofer project, a moderate target of 6 db of feedback is a
good thing.
Parts for the Project
This is a sealed box design requiring an extra amplifier channel
utilizing global feedback, a box to use as an enclosure, acoustic
damping material to stuff the box, a dual voice coil woofer, one extra
lead from the speaker to the amplifier and one or two resistors (and
the various assorted screws, terminals, and gaskets needed to put
together any loudspeaker, and maybe a couple of potentiometers if you
want to be able to dial in the performance).
Preparing the Amplifier
The feedback circuit of a global feedback amplifier consists of two
resistors connected as a voltage divider. There may be other parts,
such as capacitors, but they are not important for the servo. Hopefully
the graphic of the simplified schematic will appear about here:
R1 and R2 form the existing feedback voltage divider of a
polarity-conserving amplifier. The amplified signal is fed through R1
from the amplifier output to the inverting input of our global feedback
amplifier at the junction of R1 and R2. We add R3, connecting one lead
of the resistor at the junction point of R1 and R2. We connect the
other lead of R3 to the wire coming from the second woofer voice coil.
R3 inputs the signal from our sensing coil to the amplifier and
determines the amount of change in the closed loop gain of the
amplifier. If amplifier is polarity conserving and we have wired the
powered voice coil of our woofer as polarity-conserving, then we
connect the positive lead of the second voice coil to R3 and the
negative lead to ground. The ground wire to the voice coils may be
separate wires or one shared wire. If we make the value of R3 equal the
value of R1, then we will provide about 6db of feedback from our sensor
coil. The actual amount of feedback at resonance will vary depending
upon the cabinet/woofer tuning and Qts of the driver used.
You can make the circuit tunable by adding to adjustable controls.
Testing
One of the advantages of motional feedback is the control it gives you
over the alignment of the driver in the box. With motional feedback it
is not as critical to choose just the right woofer and fit it into the
perfect box. The woofer I am using in these tests is the MCM Audio
Select woofer, model number 55-1460 from MCM Electronics. They list a
price of $26.75 each in small quantities of this 10-inch driver. They
claim a “Vas” of 4.5 cubic feet, an “fs” of 29 Hz and a “Qts” of .49.
Whether in a sealed box or in free air, the driver exhibits rising
output as you approach resonance that belies it “Qts” rating. The
driver also shows a huge bell mode resonance above 2 kHz.
The graphs here show the before and after MFB response of the driver in
a 3.5 cubic foot box with about 3 lbs of medium density fiber fill. As
described in the graph annotations, we are controlling the resonance
peak and we getting viable feedback signal off the second voice coil
until about 1.6 kHz. The higher frequency bell mode resonance is not
affected by the application of feedback.
For the next set of tests I mounted a 55-1460 driver in a 3 cubic foot
box, densely packed with poly fill and driven by an old Dick Smith Kit
ETI 480 amplifier module. I have found the ETI 480 kit to be poor
performers and be less than unconditionally stable. If this
modification is stable with this amp kit, then it should work with just
about anything else out there.
This 3 cubic foot (internal volume) box is too small for this driver,
yet with MFB, it works. As shown in the impulse graphs, the resonant
decay is quickly damped with MFB.
This combination of a high Q woofer in too small of a box and driven by
a low quality power amplifier should not sound nearly as good as it
does. Because the high Q of the woofer is being compensated for by
decreased closed loop gain, the amplifier is loafing and plays as if it
were a much more powerful amplifier. The bass is tight, it is clean, it
sounds powerful, and it is at times startling in its detailed
reproduction of bass sounds. This power-conserving feature would be a
perfect set-up for to use one of Nelson Pass’ Zen amplifiers.
Limitations
I see two limitations of this incredibly cheap and easy MFB woofer
system. Because sensor output depends upon velocity and below system
resonance velocity is falling, little feedback is available. Unlike the
accelerometer systems, the speaker does not “harden up” when you
attempt to manually move the cone. I know some people who are very
impressed (wowed) by this. The second limitation is that because of
this loss of sensor sensitivity, you cannot push the cut-off frequency
super low. This MFB system is not the answer for really big woofers in
tiny boxes.
Saving the Electronics Just in Case
Because circuit boards are stuffed with components and often placed in
difficult to reach places, it is possible you may not attach the MFB
feedback resistor or lead in the correct position. Depending on where
you put it, things could go very wrong. It is nice to be able to have a
fail safe just in case. If you have a high idle current draw amp, the
series light bulb will work, it just needs to be a high wattage bulb
and the visual indication that something is wrong is not as obvious.
 |
|
|
| SY |
I experimented with dual VC systems like this some
years back. On the one hand, it's a convenient way of deriving the
feedback signal. On the other hand, you're tossing out half of the
motor strength to get the signal. That's why I ended up biting the
bullet and using an accelerometer.
As you noted, the feedback is pretty limited as you go down in
frequency- I wonder if this explains your observation about the
behavior when you manually move the cone.
One other thought- at a low feedback level like 6 dB, you're trading
off 2nd HD for 3rd HD at levels that could potentially have a negative
impact on sound quality. |
|
|
| JoeBob |
I applied the exact same thing to simmilar woofers
(MCM 55-1465, 12" version of the 10" driver Mark spoke off). I emailed
Mark asking him about his results with this before his post. And I must
agree with most of what he says.
Except I did notice one thing, the cone does stiffin up when tapped. It
doesn't become completely hard, but there is much resistance compared
to when the amp is turned off. I have my woofers mounted on an open
baffle.
My results were this. When comparing this type of MFB to no motion
feedback at all, I noticed an improvement in sound with the MFB. The
bass sounded more "right" and dipole bass sounds mighty great to begin
with. Given, this was with a cheap driver, but it really cleaned things
up, the bass was already punchy, this just made it sounds more clean,
sorry for the horrible explanation. Long story short, it made things
better.
In my book a drop in efficiency is alright when weighing how much it sounded better.
Give it a try, you'll see, much cheaper and simpler way to experiment than using an accelerometer. |
|
|
| MBK |
I thought about motional feedback as one possible
way of ultimate total systems integration, but maybe not the most
elegant one.
Ideally an amplifier should get feedback from the actual output device.
Unfortunately I don't have a clear idea of the options. Here my
thoughts and state of information:
- microphone feedback. Problem, hard to calibrate and not exactly direct.
- dual voice coil system discussed here. Problem, must use dual voice
coil woofers. I would think the midrange should benefit as well from
this.
- accelerometer: SY, how do you do that?
- I could imagine an optical system that measures woofer position by
reflected light, say, from a white spot on the dustcap (as done in
camera autofocus). Does anybody know of implementations of such a thing?
- Ideally I could imagine using power as feedback, by sensing amp power
output (not voltage or current which both depend on device impedance).
This would have the advantage of staying in the purely electrical
domain and to avoid any power compression at the same time. Do any of
you know a way how to electrically implement power feedback ?
If you don't |
|
|
| MBK |
I thought about motional feedback as one possible
way of ultimate total systems integration, but maybe not the most
elegant one.
Ideally an amplifier should get feedback from the actual output device.
Unfortunately I don't have a clear idea of the options. Here my
thoughts and state of information:
- microphone feedback. Problem, hard to calibrate and not exactly direct.
- dual voice coil system discussed here. Problem, must use dual voice
coil woofers. I would think the midrange should benefit as well from
this.
- accelerometer: SY, how do you do that?
- I could imagine an optical system that measures woofer position by
reflected light, say, from a white spot on the dustcap (as done in
camera autofocus). Does anybody know of implementations of such a thing?
- Ideally I could imagine using power as feedback, by sensing amp power
output (not voltage or current which both depend on device impedance).
This would have the advantage of staying in the purely electrical
domain and to avoid any power compression at the same time. Do any of
you know a way how to electrically implement power feedback ? |
|
|
| phase_accurate |
| quote: | | -
I could imagine an optical system that measures woofer position by
reflected light, say, from a white spot on the dustcap (as done in
camera autofocus). Does anybody know of implementations of such a thing? |
Not the way you describe it. But the German manufacturer T&A once
made a thingie consisting of a long cone shaped piece that was mounted
to the back-side of the dustcap and protruding out of the back of the
magnet assembly through the pole-piece venting hole. On the back of the
magnet the motion (i.e. coil POSITION to be exact) was then detected by
y light barrier (which was more or less masked depending upon cone
position).
Regarding the "electrical MFB" there are a lot of resources including
the EW&WW article by Jeff Macaulay that was discussed here
recently. There is still some discussion whether you can call this MFB
or not.
Another related technique is the one called ACE (Amplifier Controlled
Euphonic) which influences the TSP electrically by the use of
frequency-dependant current feedback.
Regards
Charles |
|
|
| MBK |
Thank you Charles.
The motion sensing I thought of could be done with say, a laser
distance metering system for instance. That would have the advantage of
not being mechanically or electrically coupled to the driver.
Electrical power feedback, I haven't really found anything so far in
that respect, but I'll keep on searching... People talk about current
drive or current feedback, yes, but I don't see how this differs from
voltage feedback in that in both cases the amplifier stays unaware of
the nonlinearity of the driver's response. |
|
|
| SY |
I used an AD accelerometer, the 100g version. It's
attached with epoxy glue to the voice coil in my JBL 2245H (the dust
cap has to be removed and then replaced). A couple of very fine wires
swiped out of an old tone arm bring the signal out to a preamp module
glued to the driver frame. This conditions the signal and sends it down
a shielded cable back to the amplifier mounted a few feet away. Very
simple, and if I'd bought some plate amps when I put this together, it
would have been even simpler.
I haven't seen an optical method yet, but there's a really, really cool
way to do that. If I can hook up with someone who can do the
microcontroller end of the project, I'll go optical. The advantages are
that one can sense from any part of the cone, not just the voice coil,
and one can get position measurement, which allows derivation of both
velocity and position. With all these in hand, some rather nice signal
processing/feedback can be implemented. |
|
|
| MBK |
Wow, sounds pretty nifty already...
Yes the optical method should get superb accuracy and zero interference
with the acting driver. The signal processing would need some tuning I
guess but it looks very elegant to me. Sadly I have no experience with
microcontrollers.
This kind of approach could really cut speaker driver distortion down to 1/10 or 1/100 of typical (gross) levels... |
|
|
| phase_accurate |
| quote: | | The advantages are that one can sense from any part of the cone, not just the voice coil, and one can get position measurement |
I think one source of failure for MFB systems is the assumption that
the radiated sound is accurately described by the motion of ONE PART
ONLY of the driver's cone.
Regards
Charles |
|
|
| MarkMcK |
Design, particularly in the concept phase, is
always fascinating and enjoyable. Implementation is where things can
get a little frustrating. The maximum frustration is always found the
first time you implement a concept/design. For example, radar or
ultrasonic range detection would also conceptually work to identify
absolute cone position. Yet, off the shelf range detectors have
sampling rates too low to work with audio spectrum MF woofers.
Substantial work is needed for implementation. Still, it is a really
cool idea. Just wanted to add this to the others suggested.
If you will allow, I would like to be just a little philosophical for a moment or two.
One value of sharing is to reduce the frustration of implementation for
other first timers. When sharing, maybe more information is better.
I love all the alternative ideas presented here. I would just like to
see more details about their real world implementation. I would like to
know what you were attempting to achieve (concept) and how close did
your implementation came to the design?
When you "looped" the feedback, where did you insert it into the
amplification block? Did you sum at the inverting input? Did you
construct a summing network and add this to the input? Did you try to
use integrators or in any way process the feedback signal? And so on?
This is not meant as a criticism, just a request for details I would
value and appreciate. My own post was just an offer I thought might be
helpful to members just getting started and maybe a little intimidated
by designs requiring more parts, more modifications of the drivers (or
woofers costing over $800 each), and sometimes the assistance of
expertise few people have access to.
I have been building these simple servo woofer systems for 21 years. I
know a lot about what they can do and what they cannot do. I searched
the site and could not find other detailed project descriptions. I
thought having one might be beneficial to a couple of other newbies.
For others at a more intermediate level, maybe more intermediate design
projects would be valuable. For the really sophisticated, we could have
really sophisticated projects. I do not see any of these as being
exclusive of any of the others.
Thanks to all for the comments and please keep on thinking about new ways to tackle old problems,
Mark |
|
|
| SY |
Mark:
The accelerometer I used is the Analog Devices ADXL190
. I described the mounting a few posts back. The app info at analog.com
show exactly how to hook the device up. Its response is quite flat to
well beyond where I use it (my subs have a 4th order LP at 70 Hz).
Once you have a good, clean acceleration signal, it's just a couple
opamps in standard blocks to mix it with the input signal (analog
adder) and adjust the feedback ratio (i.e., the proportion of input
signal to feedback signal), adjust the delay (one stage all-pass), and
then output to the power amp. In my particular implementation, I
buffered the accelerometer output (using a plain old LM310) to send it
down a cable to the rest of the circuitry, which I built on a perfboard
and glued inside of an Adcom 555. A smarter guy would have bought a
plate amp, installed that and the opamp board into the subwoofer
cabinet, and eliminated the need for a buffer.
Each 2245H is mounted in an 8 cu ft sealed enclosure. The measured small-signal f3
is about 16 Hz with the feedback, about 40-45 Hz without (if memory
serves). I don't have the capability to easily do woofer distortion
measurements, so can't comment on any improvement. I note that the
accelerometer has a 0.2% linearity spec.
The key in getting it to work is not trying to push the sub too high in
frequency. I wouldn't use anything other than a sealed enclosure, and
of course, the driver max SPL limits will not be improved. |
|
|
| Timn8ter |
| Being relatively new to DIY audio I'm grateful and
a little surprised to see this type of information generously shared.
Thank you for spending the time to put all this into a forum posting. |
|
|
| Konrad |
Setting up mfb sub (again). This time with acc
01-04-05 ( it has better noise performance than the adxl series from
analog devices)
The opamp konfiguration is werry similar dual vice coil or
accelerometer. Slowly applying feedback, tapping the element until it
you hear the element beeing harder and harder, until it resonates or
start to resonate, then redusing feedback or open loop gain abt 9 dB
makes a verry stable system.
After all my strugling with servo's it has become more and more easy to
understand what is going on. I am realy supriced that there is so few
servo subs out there!!!
feedback node - ref 200k, innput-ref 33k, ref output 100k, ref output 33n, and ref connected to negative innput of opamp
I'l make a sketch and post it after x-mas. It's nice to have a nice pair of sub's. |
|
|
| tiroth |
I've yet to see a documented MFB project that
achieves better than 6dB distortion reduction. Check out the Analog
accelerometer implementation, they achieved only 38% average reduction,
primarily due to phase problems. This simplistic implementation looks
about the same, but it isn't clear exactly what the perpetrator is.
Now, the images accompanying the recent AX article
looked to be more dramatic, but no measurements were given, especially
of distortion reduction over a range. I fear that only the "best
working" images were submitted with the article.
MarkMcK, I noted that you tried to implement this and didn't find much
success as yet. This is on the back burner for me at least until early
next year, but I'm following any research with interest.
I'm going to throw out the notion that these second-voicecoil based
techniques may be the easiest to successfully implement due to the lack
of a phase problem. The AX article may or may not prove this; hopefully
someone can duplicate this and definitively measure the performance. |
|
|
| CeramicMan |
Mark,
Perhaps part of the reason your design didn't work to very low
frequencies is because it didn't use an integrator. The coil voltage
can be seen as "differentiated" because it's sensitive to the "rate of
change" of movement as opposed to the "amount of movement". It's along
the same principle that grammophones need to have a "phono" stage in
order not to sound extremely tinny.
IME amplifiers are able to "stiffen the cone" when they're switched on
anyway. Any tapping/pushing of the speaker's cone will attempt to
induce a voltage across the voicecoil. However, the amplifier is a
voltage "source" and large currents will flow in order to stop the cone
from vibrating in any way that it's not meant to. The effect is
practically identical to shortcircuiting the speaker's terminals (when
it's not plugged in, AKA output impedance). At very low frequencies
though, the acceleration of the coil is very small, therefore the
voltage induced across it is also small, so the amplifier can't do much
about it.
I suspect that the biggest effect of your design was actually
equivalent to reducing the reactive component of the system's
impedance. When the impedance is very flat and close to DC resistance,
it is usually because the box is large and heavily damped. This would
explain why the speaker sounded much "tighter", more expensive etc. Of
course, it can't really change the impedance, it just reduces
sensitivity where the impedance is highest (at resonance), which I
intuitively think is quite similar.
I wouldn't worry much about the physical movement of the cone versus
the voltage supplied by the amplifier, because it's possible for sounds
to be virtually identical despite being visually unrecognizeable. The
importance of phase coherence across a frequency range is often grossly
overstated, and in many cases the efforts to compensate for phase
differences are just a waste of time, no pun intended. ;)
About accelerometers, they too measure the rate of change of movement
(ie: "acceleration"), just like coils. The only real advantage they
have over the voicecoil design is that they don't rely on the magnet
system that powers the speaker. With magnet-and-coil design that you
described you wouldn't be able to reduce distortion. The difficulty in
getting an accelerometer design to work effectively would be
integrating the frequency response (to make it flat rather than
sloping), and making sure that phase differences don't adversely affect
the performance. I suspect that this is why existing designs aren't
that great at reducing distortion despite the promising concept.
CM |
|
|
| MarkMcK |
CM,
Thanks for taking the time to read, think about, and respond to my
simple little posting. I commend you for thinking about the idea of MFB
in loudspeaker woofers.
I believe that this forum and others like it are much better for
learning than they are for teaching. As a result I have tried to be
careful about my responses. Nothing I say here is meant as criticism;
it is just my thinking about your thinking.
First, for those interested in the basics of integrators and differentiators, this is a good Web site:
http://www.allaboutcircuits.com/vol_3/chpt_8/11.html
As I model the coherent transient world the use of integration in this
application does not make any sense. I have designed an experiment to
test my hypothesis, but it awaits the correspondence of free time and
energy. These are two things that I don't seem to have at the same time
lately.
Second, I chose to limit feedback to 6-db in this implementation. As a
function of decay control, the test results included in the posting
show a 6-db reduction continuing down to the low frequency limit of the
test. There is a driver anomaly at about 20 Hz that appears in box or
free air that cannot be controlled by MFB (from any source), but
otherwise there is no drop off in feedback or correction. It is just
that the output has fallen so far by these frequencies that it is not
relevant. The application of MFB, however, is still working as intended
even at very low frequencies.
Third, is the impedance rise at system resonance a cause or an effect?
Lastly, my comment about resistance to cone motion was meant as a
relative comparison to a very expensive commercially available MFB
woofer. With thousands of watts available and relatively large amounts
of feedback, this system becomes extremely resistant to manipulation by
human hands. Indeed, you don't move the cone by pushing on it; you have
to lean your body weight into it. As I stated then, some are very
impressed by this. Is it acoustically important? Not sure, but it seems
doubtful. And yes, the application of MFB increases the apparent
stiffness over what you have by turning on the amplifier without MFB.
Keeping learning and thinking, you seem to be doing well at it so far.
Mark |
|
|
| Konrad |
Hello
I think one of the main things to do first, even before trying to apply
feedback is to make the amplifier a current amplifier. The pahse is
therfore changed 90* in our favor at high freq, and higher bandwith is
possible. The problems getting feedback at resonanse is gone (if you
have available plenty voltage out swing). At high frequensies the
current drive does not suffer from coil inductance. As current is the
force, the applyed voltage is not. The system easily becoms stable at a
much higher freq using current drive. Making the amplifier current
feedback is essential. The possibility to increase the amplifier gain
below the upper corner freq also increases the amount of feedback
possible. But then again i'm using accelerometer, thou the possibility
to use the DVC MFB feedback is there ( dual voice coil elements ) . |
|
|
| Astro-Muse |
Mark, please forgive my ignorance but I don't know
much about electronics. The MFB design really intriques me. I have a
dual voice coil sub which I drive with an old stereo amp driven in
mono. Would this system work on any amplifier? How do you determine the
values of the resistors?
Finally, I can't quite follow the wiring schematics so perhaps I'm too
ignorant to attempt this, however, it seems quite simple! ;) |
|
|
| tiroth |
MarkMck,
I took a break from other work this weekend to check out your "super
simple" MFB concept--and I must say I was impressed. (I am
investigating solely for subwoofer use at the moment, so I can live
with a very restricted operational bandwidth)
Unfortunately phase issues were a bigger problem than I had expected.
Open-baffle, the phase differences between the input signal and the
voicecoil measurement were fairly negligable, but in a sealed enclosure
the phase shift was 90 degrees at the frequencies of interest. I chose
to correct this about 30Hz via a capacitor and found that the voicecoil
signal could be made to match the input EXACTLY. There is no reason
this couldn't be done at 20Hz, and CM, note that this is without
differentiation!
The only problem with this is that we introduce a low frequency pole
into the feedback network...but this is somewhat inevitable anyway due
to phase shift. Possibly it is even beneficial from a stability
standpoint?
I'll post some scope traces in the next couple of days. I haven't taken
distortion measurements yet but the servo responds as expected when I
touch the cone...increases the input to sustain the voicecoil
measurement.
I am now inclined to believe that a "simple" differentiator-free solution might be quite reasonable. |
|
|
| MarkMcK |
Astro-Muse,
| quote: | | I
have a dual voice coil sub which I drive with an old stereo amp driven
in mono. Would this system work on any amplifier? How do you determine
the values of the resistors? |
Yes, it will work. By mono, do you mean bridged? Is this a feature
built in or did you modify to make it bridged? If you modified it, then
you know enough to complete the servo conversion. The conversion I
describe will work with a bridged amplifier. I have numerous friends
using this simple design with bridged amplifiers.
For a variety of design considerations (including stablity with unknown
amplifiers) I chose 6 db of MFB global feedback. To achieve 6 db, one
must find out the value of the series feedback resistor in the stock
amplifier. The value of the series resistor in the amplifier feedback
voltage divider network is the value of the resistor you add in series
to the sensor voice coil lead (hot side). You do need to know your
amplifier well enough to determine the values of its feedback voltage
divider network. This does take some electronic knowledge.
tiroth,
| quote: | | Unfortunately
phase issues were a bigger problem than I had expected. Open-baffle,
the phase differences between the input signal and the voicecoil
measurement were fairly negligable, but in a sealed enclosure the phase
shift was 90 degrees at the frequencies of interest. I chose to correct
this about 30Hz via a capacitor and found that the voicecoil signal
could be made to match the input EXACTLY. |
The "phase" difference you see in box is because fo the box/system
resonance. You want to correct the frequency response and "phase"
errors with the feedback.
There is a problem with the idea of phase shifting in the transient
coherent world. The sounds we listen to are never sine waves. The
sounds we listen to are always starting and stopping and even when they
cycle, one iteration is never exactly the same as the one before.
The "phase" problem one sees in the sensing coil output is that of the
signal lagging the input. When you introduce another "pole" or use an
integrator/differentiator you can only add more lag. While it may look
like it matches when testing with a sine wave, with a coherent
transient you have made the error worse. The feedback signal you are
trying to use to correct the response of the driver is farther behind
the signal you need to correct. This is a problem with a sensing voice
coil or with an accelerometer.
Yet, as tiroth found, unless there is a problem (such as system
resonance), the lag off of MFB sensing is not critically. The function
of the MFB is to correct this driver lag of the input signal (and to
suppress the decay overhang of the driver).
Mark |
|
|
| Astro-Muse |
I didn't bridge the amp Mark. It hasn't been
equiped with that feature and I'm not certain there would be an
advantage to doing it when I have a dual VC sub. I have just used the
mono switch. I'm not certain where to find the series feedback resister
in my amp, so maybe I'd better steer clear of this project, as tempting
as it sounds.
Let us know if you ever publish a detailed article on the web. Maybe I
could follow it if there are lots of detailed diagrams (lol). ;) |
|
|
| tiroth |
MarkMck,
Thanks for your comment. Ironically I came to this realization as well
on the bus home when I stopped to really think about the impossibility
of it all (Insert foot in mouth)
I hooked up my original filter-free circuit again and although the
feedback is certainly operating, I don't get any measurable distortion
reduction.
I then took a look at a FFT of the VC output, and found the distortion was about 10dB down from the acoustic distortion. Bingo.
For reference Shiva, sealed Fb=38Hz, signal 31.5Hz @ 25W, acoustic
-32dB 2nd harmonic, VC -42dB 2nd harmonic. Feedback 6dB. (Circuit is a
bit of a lashup and isn't stable above that)
I'll try to take another look at this soon with measurements above resonance, although that is purely academic for me.
:mad: |
|
|
| Circlotron |
| quote: | Originally posted by Konrad
Hello
I think one of the main things to do first, even before trying to apply
feedback is to make the amplifier a current amplifier. The pahse is
therfore changed 90* in our favor at high freq, and higher bandwith is
possible. |
Feels right to me too. With a sinewave signal, the peak of the waveform
(maximum voltage) corresponds to the zero-crossing of the voice coil
(maximum velocity). So it would seem the voice coil movement lags the
drive voltage by 90 degrees. Over-simplification? |
|
|
| MarkMcK |
Konrad,
Thank you for the article link. I had not read this article before. The
authors, however, use the second voice coil as a sensor. As I quickly
scan the article it appears they found no significant problems with the
sensor signal below 70 Hz. They did find problems above 70 Hz and they
used a second order filter for frequencies above 70 Hz.
My research has not duplicated the problem they describe above 70 Hz.
Plus, within my design concept, and since I am only using the woofer up
to 100 Hz my focus was on the performance near resonance. This is where
the distortion and decay overhang are most problematic.
The intent of my posting was to provide a starting point, both in
concept and implementation. I wanted to provide documentation to
demonstrate what the technique could provide.
There is nothing wrong with using the technique as described. It works
and it will improve the performance of a woofer at low frequencies.
Yet, as a progressive society, we are always trying to do better. As
stated in the article there are limitations to the technique.
Improvements are possible. Complexity, however, must increase. My
little project requires only one resistor, one wire and a couple of
soldier connections. Sometimes the increase in complexity is a small
increase and sometimes it is not. Any change, however, must be tested
to see if the design criteria are being achieved.
Within those considerations, I encourage DIY folk to think, to
experiment, and to implement other designs. And if you are motivated to
do so, please share as much detail as you can.
I wish all of you great fun as you think, experiment, and implement.
Mark |
|
|
| tiroth |
| Very good article. My first thought is that they
are using a lot more feedback in order to get the performance they
posted. There is a big difference between k=10 and k=50. |
|
|
| Konrad |
:) As K increase "it" will come closer and closer
to current feedback. The main thing abt this that it is possible to
easily increase the sub performance, must say a lot! say you have 6dB
feedback only, then som improvement for the fase, and som dB' at
corecting frequency response. Togheter this is a lot! It definetly
sounds like a lot. I've tried dvc feedback on my four 12" underhung
tc-sound With a K of 150 (without the correct correction sircuit, but a
aproximation). There is nothing even close, exept real life.
Currently adjusting my new circuit with accelerometers. There the
possible feedback increased as the amplifier driving the speaker was
made current feedback. There is limmits for the current fedback, open
loop etc. but more of it seem to improve performance for the total
circuit.
A real element is 3.rd order, makes the PID (Proposjonal Integrator
Derivator) circuit more difficult to make and adj. But after all
current feedback has near the same freq response above elements
resonance. Then i think integration from the resonance and down. Then a
small amount of D but high i mean 100-300 hz region as D also improves
phase you dont want to much at low or resonance as fase is .......
Therfore i vant to know: How abt a BIQUAD filter in front of the
amplifier, here the adjustments have better range ? compared to PID? (
P= K proposjonal gain) |
|
|
| tiroth |
Konrad,
Could you possibly post a sketch of your implementation? Stability at k=150 sounds good to me.
Thanks for your input. |
|
|
| Konrad |
| Simplified modell of speaker: |
|
|
| Konrad |
The "stepp" phaseshift That neads to bee compensated for.
The phaseshift at bass resonance: |
|
|
| Konrad |
Making The amplifier a current gain amp :In
this example the element's "Q" at resonance is high, thus high voltage
output is needed. So there is still a kind of requierements for high
power (Voltage swing) amplifiers. To lower this "Q" the volume of the
kabinett neads to bee increased. but this will lower q only to a
certain ekstent.
Current drive removes phase problems : |
|
|
| Konrad |
The "conjungate nettwork" C10 R17 is there to increase stability.
C9 R16 is there to increase fase margin near elements resonance.
In my case im using buffered accelerometer inserted at R14, and the R14 repleased with 3k and 10u series cap.
As the current is the force needed to corect for distorted acceleration
the phase of the current is important as this also gives/tell us the
direction of power
The fase in the current drive circ: |
|
|
| Konrad |
As for the voltage feedback at the secondary coil
of dvc elements the phase shift at resonance needs to bee compensated
for. It will then bee somthing near the nettwork attached to R14, L4
beeing the secondary voice coil.
Alternative is replacing the nettwork with a "doble T filter" as this
can bee made a good aproximation to the "stupid" values in the
compensating nettwork. |
|
|
| Konrad |
Splitting R10 so to apply current drive from
resonance, and voltage drive below vil make the possibility to increase
gain at lower freq and thus more feedback
(R10 series cap redused to below 10u and from amp negative input to gnd
same value as R10 and a "larger" cap ) This wil then become a sort of
"power drive" Bee carefull use fuses in the circuit wile testing!!! as
gain is gain is high.
So i realy think it is possible to make good aproximations by some meashurements and a litle simulation.
fase of T filter : |
|
|
| cm961 |
Does anyone know of a good way to remove the dust cap from a woofer without damaging it? Maybe a heat gun?
Pete |
|
|
| johnnyx |
I did a fair bit of work on this subject, using
dual voice-coil speakers, and I would like to share it with you. I will
need to do a few posts, so please be patient. Tell me what you think,
it aint perfect, but it is simple, suitable for DIY.:D
The first thing with a feedback system is to ensure that it's stable.
At low the low frequencyend I direct-coupled the amp. The only phase
shifting device at low frequencies is then the loudspeaker itself. At
high frequencies I had to start the roll-off at 1kHz. With these
precautions, the d.c. open-loop gain of my system is about 150dB, and
it’s perfectly stable.
The analysis of a speaker system can get very complicated. At the time
I had neither a computer nor sophisticated test gear, so I adopted a
different approach.
In the attached image, the gain equation is the same as that of an
op-amp. The gain is set by the feedback transducer, whatever that may
be, rather than the pair of resistors usual in an op-amp circuit. As in
that case, when the open loop gain, A, is very large,
The equation becomes;-
Gain = 1/
End of part 1:)
Gain = 1/B |
|
|
| johnnyx |
Computers can do anything - Except what you want.
I measured Vout and Vin at various frequencies using my sig. gen. and
‘scope, and calculated â for each frequency. The results are given in
the table below.
The maximum feedback voltage occurs at 61Hz, the bass resonance of the
speaker mounted in its box. This gives maximum damping at resonance,
which is what we want. The problem is what happens when the feedback
voltage is small? The system gain becomes large, and the response peaks
at very low frequencies and at the minimum impedance frequency, 220Hz
in my case. The size of these peaks is proportional to the open loop
gain.
If current feedback is added to the voltage feedback, then when the
feedback voltage is small, the current will increase, and the voltage
developed across the sensing resistor will prevent the peaks. Previous
posts have suggested using current feedback,
but how much?
Sorry guys, but you’ll have to wait for the next instalment. I have to convert the next diagram.
I'm new to this posting lark, but it's fun, I think :( |
|
|
| johnnyx |
I have attached a freehand sketch of the usual speaker impedence curve to illustrate the following points;-
Fs is the usual mass- compliance resonance of the speaker in a box. It
is a parallel resonant circuit with a maximum impedence at resonance.
Fm is the mass-inductance resonance. It is a series resonant circuit with a minimum impedence at resonance.
The phase shift at both Fm and Fs is therefore zero.
The e.m.f induced in both driven and feedback coils due to motion must
be the same, the coils are identical and they move together in the same
magnetic field. The impedence curve reflects the induced emf. If we
assume that the errors are small, we can write;-
B = (Z-Re)/Z
There are problems because of the phase shift in Z, it should be a
vector subtraction, but if only two points are chosen, Fs and Fm, then
phase shift is not a problem. This is good, only two measurements:)
The expression above can then be used to combine the gain due tocurrent
-feedbakk with the original expression for gain due to feedback from
the second voice-coil.
An expression for the complete system can then be derived. |
|
|
| johnnyx |
The expression for the complete system is;-
Gain = 1/(Rs/Re - B Rs/Re + B )
If Rs = 0, then Gain = 1/B as before
If Rs = Re, then gain = 1,
The current feedbackand voltage feedback cancel because their effects
are opposing. Current feedback damps the series resonant circuit by
effectively increasing the series resistance, while voltage feedback
damps the parallel resonant circuit by effectively reducing the
parallel resistance. So the response is as though there is no feedback
at all.
These results show that the expression is correct.
A flat response is obtained if the gain at Fs is the same as the gain
at Fm. So the question is, what value of Rs in the above expression
fould give the same gain for the measured values ofB at Fs and Fm?
In my system it came out at 0.43 ohms, I used 0.47.
Having measured it with what little equipment I had at the time, I
think that the errors and assumptions err on the side of too much
current feedback. but it's not far off, considering. My system has been
in use for 5 years.
BTW, this is only part of the system, you need an integrator to give an
output proportional to acceleration. If the input is double integrated
you get an output proportional to excursion, I thought of using that to
trigger the protection relay but never got round to it.
The idea was to get a computer and CAD stuff to make more accurate
measurements, but in the past few years I've got bogged down with
computers and neglected the audio. So thanks for the inspiration to
work at a subject I love.
Hope all this makes sense, any questions?
Re is voice coil resistance
Rs is R8 in the diagram |
|
|
| Svante |
May I jump in here. I might have missed something
earlier in the thread, but what are the general thoughts when using a
second voice coil for feedback? Is it that it somehow measures the cone
velocity, and that the feedback loop would correct the input to the
speaker such that the output (presumably proportional in some way to
the cone velocity) closer to the audio signal?
If so, how is the mutual inductance between the two windings accounted
for? I mean, even if the cone is forced to stand still, there will
still come a signal back from the second winding, the system will act
as a transformer.
Or did I miss something? |
|
|
| johnnyx |
Yes the mutual inductance between both coils is an error term.
I would like to find a way to either cancel it out, or to include a
term in the gain expression to account for it. When I get my new
measuring system maybe I could find a way to do both, but I couldn't do
it when I was designing the system.
The questions are, how big is the error? how does it affect the results?
To find out, I built it. The error is frequency dependant, as you would
expect, and affects the feedback at Fm much more than at Fs.
Even so, the system works. As I said previously, the errors result in a
bit too much current feedback, when calculated as shown previously. |
|
|
| johnnyx |
Maybe I should add that the frequency response of the system behaves like a low-pass filter, the corner frequency being Fm.
It's second order, 12dB/octave.
Fm is higher for smaller woofers.
This could be an advantage, no crossover required.:) |
|
|
| Svante |
Johnnyx:
Have you measured your system acoustically? If so, it would be
interesting to see the graphs. If not, may I suggest a measurement with
the microphone *inside* the box. The resulting curve will be tilted by
-12dB/octave, but a double differentiator will fix this. I could point
you to some software that does this too if you want. |
|
|
| Svante |
And, yes, forgot:
Have you had a look at Audio Pro's ACE-bass. It does feedback on the
current of the voice coil and manages to extend the lower cutoff of the
loudspeaker. It is however not best understood as a feedback system in
the traditional sense.
I found this link, but there might be better ones.
http://www.1388.com/articles/tech_Ace-Bass/ |
|
|
| johnnyx |
I have a 3rd octave RTA graph of the sub together
with my cheap satellite speakers. I have a LAUD measuring system, but
had problems with the computer. It uses an ISA sound-card by Turtle
Beach, and is good, but I've had no luck getting the computer working.
I had so many audio projects on at the time, I didn't take many of the
MFB sub. I've just ordered Praxis by the same company, Liberty
Instruments, which should be less dependant on the computer. So watch
this space!:D
Regarding the ace bass, it is just the same as Yamaha's Active Servo
Technology system. I think, if I remember correctly, Jeff Macauley uses
the same principle.
Negative current feedback increases output impedence, and positive current feedback reduces it.
I accidentally used positive current feedback in my earlier
experiments. These first attempts were published in Hi-Fi World DIY
letters, August 1996. I used a tapped , dual impedence woofer for my
first attempts, because they were cheap. I used a differential amp, -ve
input to one half of the winding, +ve input to ground, and I used
remote sensing; ie the inputs connected at the speaker. The centre tap
of the voice coil was grounded, and the amp drove the other half. The
+ve input then had +ve current feedback from the voltage drop along the
speaker wire. I couldn't understand why the offset voltage was so high,
and drifted so much.
There lies the problem; the sensing resistor has to have the same thermal characteristics as the voice-coil resistance.
Yamaha used to sell thermally matched resistors for some woofers, to promote their idea.
My system seems relatively immune to these problems, because it is not
dependant on just one parameter, but ideally Rs should have the same
thermal characteristics as Re. |
|
|
| Svante |
| quote: | Originally posted by johnnyx
So watch this space!:D
|
I will... :D
[/B]| quote: |
Regarding the ace bass, it is just the same as Yamaha's Active Servo
Technology system. I think, if I remember correctly, Jeff Macauley uses
the same principle.
Negative current feedback increases output impedence, and positive current feedback reduces it.
[/B] |
Yes, Yamaha bought the rights to use ACE-bass, because it was so
similar, and used it in their products. There is however more to it
than positive current feedback, as you may know. The extra loop
generates an output impedance that in effect modifies the mechanical
parameters by electrical means. Pretty neat to be able to choose fs,
Vas, Qts, Mms etc by just changing capacitors and resistors. :bigeyes: |
|
|
| johnnyx |
I did many experiments including components within
and without the feedback loop, like conjugate filters, capacitors and
resistors, coils you name it, and I've seen the parameters change. Its
strange to alter the resonant frequency or damping factor in such a
way. I wanted to increase the bandwidth of the system to higher
frequencies. In the end I decided to leave it as it is.
I feel frustrated when I see some design, but no method for calculating
the effects of component values. This is the reason I've given the
equations above; so you can calculate your own design, and verify (or
maybe not) the results. Putting numbers and equations to something says
so much more about it. |
|
|
| Al3x |
| quote: | | BTW,
this is only part of the system, you need an integrator to give an
output proportional to acceleration. If the input is double integrated
you get an output proportional to excursion, I thought of using that to
trigger the protection relay but never got round to it. |
It is my understanding that the signal from the voice coil is
proportional to velocity, hence it needs a DIFFERENTIATOR not an
INTEGRATOR to get an output proportional to acceleration?? |
|
|
| johnnyx |
| With my first attempts at MFB, I had no idea what
to expect. Once I had a stable system, I measured the frequency
response with a sig. gen., mic, and 'scope. Crude but effective, and
requiring a great deal of patience. I noticed an overlying +6dB/octave
rise in the frequency response, despite the presence of the peaks I
described earlier. (I had not applied current feedback at the time). I
had no idea why, but thought that a preceding integrator would combine
with this to give a flat overall frequency responce. Later I learned
that acoustic output is proportional to the acceleration of the
diaphragm. |
|
|
| johnnyx |
Tannoy had a commercial model using the principles
we are discussing here, the ALF625. It was while reading a review of
this in Hi - Fi World sometime in 1994 that I was inspired to try doing
something similar. Later that year, the ALFie 625 was reviewed in What
Hi - Fi? who described it as "slow" :confused:
They also ridiculed the acronym, "Advanced Low Frequency improved
electronics", saying; "More Arnie and less Alfie would be better".
Maybe such a review in the popular hi - fi press can stifle a product
in the marketplace, I think it ceased production in 1996.
In their system they differentiate the voltage feedback which is, I
think, what you are suggesting, this would give a flat response by
itself. They too use current feedback to raise the output impedence of
the amplifier, so maybe the current feedback provides sufficient
stability to allow such a thing.
Here's a block diagram of their system, hope it's readable:) |
|
|
| johnnyx |
| Here's a 3rd octave RTA response of the system.
The bass is my MFB system, the rest is some small speakers I made. I
was trying out an active crossover I bought, intended for in - car use,
but it was after I was burgled and lost my 4 channel amp with active
crossovers, so I just needed something quick. The LAUD system allowed
me to average several mic positions, but I don't think it eliminated
room nodes. I remember setting the mic in position for a near - field
measurement of the subs but I cann't remember what happened.:xeye: |
|
|
| Chris Lockwood |
ACE-Bass is a really effective alternative to MFB.
I've built a few sub boxes over the past 10 years using ACE-Bass, from twin 6.5" to quad 10", with 3db points as low as 20Hz.
The degree of control the electronics has over the cone is quite
staggering. If you try to manipulate the cone with the system on, it
actually feels heavier, stiffer, more damped, with a lower Fs (assuming
you choose to increase all of these terms).
You just have to listen to one of those little Yamaha subs to realise
that for such a small driver, being driven hard in a small box, the LF
extension is quite impressive. Applying the technology to less
compromised drivers running a much larger headroom results in very
impressive performance, especially if dual subs are arranged in
push-pull fomat.
If you look at Stahl's original AES paper on the technology, he prints
some fine examples of measured improvement in performance and range
extension, comparable to the best commercial MFB systems.
No messing around with accellerometers, and the system works great with
a vented box, giving plenty of efficiency down low (if you choose that
avenue).
For the DIYer, I think it would be much harder to get better
performance from a MFB system compared to an ACE-Bass system design
with the same effort. |
|
|
| johnnyx |
| quote: | Originally posted by Chris Lockwood
ACE-Bass is a really effective alternative to MFB.
The degree of control the electronics has over the cone is quite
staggering. If you try to manipulate the cone with the system on, it
actually feels heavier, stiffer, more damped, with a lower Fs (assuming
you choose to increase all of these terms).
|
Hi Chris
My first experiment with MFB had both voltage feedback and an
accidental negative output impedence, as I described above. So my first
attempt combined both techniques:bigeyes:
When I pressed the cone, it hardly moved, very slowly it would move inwards. It felt very stiff. I was impressed:)
With this set-up I was concerned by the large offset voltage that was
present across the speaker, and how the value drifted over time. How
did you solve the dc stability issue and compensate for the temperature
coefficient of the voice - coil resistance in your systems? |
|
|
| Chris Lockwood |
I never bothered with VC temp stabilisation. Since
the VC resistance has a +ve temp coefficient, the worst thing that
would happen was having slightly less than optimal -ve resistance in
the amplifier, which meant a less tightly controlled system.
The biggest problem I saw was the non-linear voice-coil inductance.
This alone required trimming of the negative resistance to be less (in
magnitude) than optimal, or else at high excursions the thing would go
into +ve feedback mode and clip hard. Its a pretty horrible sound when
that happens :)
As far as DC offsets go, it never became a serious issue at the levels
I was listening. I did notice some rather large LF displacements going
on, so I assume this is similar to your observations. Unfortunately I
sold or scrapped all of the ACE-Bass systems that I built and have lost
track of them.
The DC offset problem only happens when you allow the current feedback
to dominate the voltage feedback. The loop should be tailored for
increasing voltage feedback below the cutoff frequency of the system.
AC coupling the current feedback can help here.
I'm all keen to make another system now :) Its way easier to measure
the driver parameters these days, using soundcard-based test software.
Chris |
|
|
| johnnyx |
| quote: | Originally posted by Chris Lockwood
I'm all keen to make another system now :) Its way easier to measure
the driver parameters these days, using soundcard-based test software.
Chris |
Hi Chris
Yes I have ordered a Praxis system and hope to do more measurements. I
had problems with the computer that had the Laud measuring system on
it, but the new system will be so much more flexible. Even the free
version can measure driver parameters accurately, I'm waiting so that
all my test leads can be calibrated together. With it, I will be able
to see the nature and magnitude of the errors in my design method as
described above, I'm really excited about it:D
None of my systems is set in stone.
The speakers are Seas CA21 RE4X/DC. I have 4, 2 per enclosure. each
driver has about 20 litres of box, with a shelf brace between them.
There is a partition at the base of the box containing the amplifier
and other electronics. I could try different designs by changing the
electronics in the box. I will have a go at Ace bass and LT
equalization, and any other technique that is suggested, then compare
the results. I have a coupled cavity design using these speakers, which
I find interesting, but is not so easy to implement. |
|
|
| phase_accurate |
| quote: | | This
alone required trimming of the negative resistance to be less (in
magnitude) than optimal, or else at high excursions the thing would go
into +ve feedback mode and clip hard. Its a pretty horrible sound when
that happens |
More than 10 years ago I built an ACE sub as well. It was designed for
30 Hz (-3dB) cutoff only but it was quite impressive. Even at very low
volumes almost everything within the appartement shook ! The downside
however was this strange clipping. I didn't have the measurement
equipment to see where the effect came from back then, so I gave up on
that one.
The DC-offset problem was eliminated by a single-pole highpass, with
adequately low cutoff frequency, in the forward-path of the circuit.
It was also quite impressive to hear the differenc in sound, generated
by tapping the cones, with the circuit activated or deactivated !
The efficiency of the principle will be in the same ballpark as other
electronically assisted ones. It's advantage is that it can be used
with reflex enclosures, needing reasonable amounts of Vd only.
Regards
Charles |
|
|
| johnnyx |
Okay okay, I'll try it
I have a 12" Peerless unit that I used in a 125l enclosure. That was a
reflex design, but it was just too big. Enormous even. I wanted to
build a smaller enclosure for it, and maybe it would be an ideal
candidate to try the ace bass technique.
| quote: | Originally posted by phase_accurate
The efficiency of the principle will be in the same ballpark as other
electronically assisted ones. It's advantage is that it can be used
with reflex enclosures, needing reasonable amounts of Vd only.
|
I was also planning a reflex enclosure for it.
Do you have any links to the design procedures? |
|
|
| phase_accurate |
You can mail me your address and I can send you the according JAES article by snail-mail.
Regards
Charles |
|
|
| johnnyx |
| quote: | Originally posted by phase_accurate
You can mail me your address and I can send you the according JAES article by snail-mail.
Regards
Charles |
You've got mail!:)
Now that I am going to try this ace bass technique, despite having
unfavourable past experience of this kind of thing, is there anyone who
will try my design?
Here's the integrator and input sections. |
|
|
| Al3x |
| look forward to seeing more of your system jonnyx! |
|
|
| johnnyx |
Er....That's it.
I will post some measurements when I get the measurement system working properly.:bawling:
Got a few problems with the soundcard.:bawling:
140dB/W sensitivity anyone? Yes I'll have two of those, please.:D |
|
|
| johnnyx |
My Praxis system is working now, except distortion for some strange reason, so here are some measurements of the system.
I was surprised at how accurate my original measurements were, considering the equipment I had at the time.
I built a stereo pair, the resonant frequency of one is 61Hz, and the other is 64Hz.
I modelled the drive units is WinISD to see what they would be like
un-aided. The resonant frequency in the 45l box (each unit has 22.5l)
would be 64Hz, with a Q of 0.617. WinISD recommends a reflex enclosure
for them.
Here's the TS parameters, only one coil on each unit used. |
|
|
| johnnyx |
| Here's the impedence plot... |
|
|
| johnnyx |
| This is the frequency response but without the high-pass-filter, so you can see the raw response. |
|
|
| johnnyx |
I want to show the effect of changing the ammount
of current feedback. Rather than change the value of the current
sensing resistor, R8 in the diagram of post#41, I changed the value of
R2 to 11K, and inserted a 47K preset pot so that the value could be
"tuned".
This is with R2=11k, too much current feedback. |
|
|
| johnnyx |
| This is with R2=39k, too little current feedback.. |
|
|
| johnnyx |
The shape of the curves is the important thing.
The gain changes as you would expect. Praxis thinks it's a normal
speaker, and calculates it's sensitivity from sound out/volts in. In
this case, I'm using the input to the system as the reference for the
volts in.
The following is a bit of a surprise. In the equations, if Rs=Re then the B
terms cancel, so I thought it would be the response without feedback,
But when I replaced R8 (current sense or Rs) with a 3R3 resistor, I got
this..... |
|
|
| johnnyx |
The peak in the last one was at Fs. Curious. Any ideas?
When I "tuned" the pot to give the flattest response, guess what? I got 22k!:nod:
So I think that maybe the errors due to inductances cancel or something. It shows that this technique works:)
You will have noticed a slow downward trend towards the lower
frequencies of about -3dB/octave. Dunno what that is. It isn't a
measurement artefact, because I noticed it with the first plots made
with sig gen and scope and electret insert mic. It's not the room
either (different rooms) Leaky integrator? Dunno. Any ideas?
Ok let me know what you think.
I'll post distortion measurements when I sort the problem out. The
residual distortion is <0.003% above 45Hz, but rises below that to
28% @ 10Hz, this would cloud any subwoofer measurements, so they'll
have to wait for now. |
|
|
| johnnyx |
Of course, when the feedback cancels in the case
when Rs=Re, the output is no longer proportional to velocity. The
integrator skews he shape of the curve.
Below is a plot of the system without the integrator, with R8=3R3. A much more familiar result.:) |
|
|
| 454Casull |
| Can't have this thread going to waste... |
|
|
| johnnyx |
Well I think it's interesting..... but it looks like I am alone in this.
I was inspired by a review of Tannoy's 625ALF subwoofer, which in turn
was the product of Hawksford & Mills current driven loudspeaker
paper.
Of course, I got the whole thing backwards. It was only recently that I
saw their paper, I was trying to work backwards from their finished
product. It does work though, and my analysis is so much simpler than
theirs, but no less valid.
I have some more stuff to add, but I thought interest had dried up. Maybe I'll post it then...:) |
|
|
| tschrama |
Hi johnnyx
This thread is never going to waste.... Thanks for all the info!!! |
|
|
| Konrad |
Have a closer look at this abt speaker eqations:
Eq (D6) in this link.
Then compare to EQ (1) (2) (3) in UAF 42 datasheet
Finaly use inverted konfiguration. And sum Hp Bp Lp with gains as to fit driver eq.
using accelerometer the input is a, if a is like a funktion of sinus
integrating gives cos, integrating again gives -sin. and summing sin
and - sin is 0. or to say rather perfekt. So summing with corect phase
and gain is important (obvius). |
|
|
|
