Glad to hear you/we are finally starting to understand what was going on with your phase response discontinuity around 13Hz. And, I finally understand what exactly was being said in that patent! I could never figure out how moving weights around would adjust the phase angle of feedback signal...made no sense at all. Based on your measurements, it seems that carefully tuned amounts of anti-phase cone rocking has the potential to provide increased phase margin which in turn would allow higher loopgain and lower distortion. I am also now wondering if the stability problem I mentioned back in post #51 was related to this behavior.
I found it beneficial to include a limiter circuit in the MFB electronics to avoid hard clipping. I experimented with two different types...one controlled the gain of the input buffer which effectively just turns down the volume, the second controlled the loopgain, so as clipping was approached lower distortion was traded for increased output.
Notch filters will add in their own rather abrupt phase changes, as covered in the Waslo article. It may be useful to acquire the open loop response/phase as measured by the accelerometer for the different locations of the notch filter and compare with the original "no-notch" results. Like you, I would have expected the results to be similar for any location of the notch filter within the feedback loop. Perhaps there are some headroom issues?
Is there a particular plot you are asking about? Armand mentions if data is with or without feedback for most plots, but he may have missed some.
I'm really confused about what is going on and can't claim my understanding is any good. Based on my likely erroneous understanding of your tests (or the set-up isn't quite what I think it is), sometimes the ringing seems to be in the amp/MF (which ought to include this mysterious rocking). Something in the feedback is introducing the 13 Hz error (and the 3-4 Hz error too). Putting the twin-tee outside the loop removes this but doesn't address the problem of instability.
In an MF circuit you should be seeing the amp output the inverse of the erroneous 13 Hz component (even if due to rocking), not "zero negative signal" you show. Zero would mean the feedback isn't working right, although at the twin-tee frequency it is hard to say what we should see.
Putting a twin-tee filter into a feedback loop may be OK for diagnosis but it does weird things to phase too. Possibly you could use a sharp low-cut filter in the final design.... assuming, as I do, that the dream of DC isn't attainable.
It would be good to see a check of feedback criteria, kind of Nyquist-lite just to examine gain and angle for the low freq region. Or even just a freq sweep.
Might be rocking (which has obvious mechanical diagnostic tests) or might be electronics (in power supply??).
Ben
Last edited:
Take a look at the openloop measurements for the microphone (Attachment #1) and the accelerometer(Attachment #2). There is nothing in the signal chain besides the power amplifier, woofer, and mic or accelerometer. No feedback loop. The 13Hz issue is seen by the accelerometer but not the mic, so it has to be some type of cone motion that the accelerometer would pick up on, but that doesn't change the overall average motion of the cone so that the mic misses it. A rocking mode is the simplest explanation I could think of.
Attachments #3 & #4 are my attempt to overlay the response and phase between the two measurements to make the difference more apparent. A cone rocking mode would explain the cancellation at 13Hz and then the in-phase summation lower down in frequency for the accelerometer response compared to the expected -12dB/oct slope measured by the microphone.
I Agree. That is why I recommended the additional openloop measurements with the notch filters to see exactly what they are doing to the loop response in terms of phase margin. It is difficult to solve stability problems looking only at the closed loop behavior.
Attachments
Bolserst - I appreciate your explanations (and no doubt, Armand even more so).
Why is there a hole in the accelerometer response at 13 Hz? Is that with the twin-tee filter in place?
Something in the loop is going funny at 13 Hz. I wonder if Armand can simply insert a 13 Hz signal to replace the accelerometer's input and see the amp output? Or "play" 13 Hz into the driver using another speaker. Or just tap the cone to see what's ringing?
While I can't claim any first-hand experience with dissected cones and glued-on accelerometers, I just don't see a 13 Hz rocking motion intuitively. Patents have all kinds of junk inserted in them for legal purposes such as to make the patent theory unique, confuse the competition, etc. Including, as I have previously suggested, MF patents from RCA and who knows who else.
Based on my sometimes painful experience, the instabilities have to be addressed. Luckily, my original experiments were done long ago using fancy tube amps that usually wouldn't instantly destroy a driver DAMHIK.
Ben
Why is there a hole in the accelerometer response at 13 Hz? Is that with the twin-tee filter in place?
Something in the loop is going funny at 13 Hz. I wonder if Armand can simply insert a 13 Hz signal to replace the accelerometer's input and see the amp output? Or "play" 13 Hz into the driver using another speaker. Or just tap the cone to see what's ringing?
While I can't claim any first-hand experience with dissected cones and glued-on accelerometers, I just don't see a 13 Hz rocking motion intuitively. Patents have all kinds of junk inserted in them for legal purposes such as to make the patent theory unique, confuse the competition, etc. Including, as I have previously suggested, MF patents from RCA and who knows who else.
Based on my sometimes painful experience, the instabilities have to be addressed. Luckily, my original experiments were done long ago using fancy tube amps that usually wouldn't instantly destroy a driver DAMHIK.
Ben
Last edited:
No. The response has no notch filter, only the power amp and woofer as shown in the diagrams.
The "hole" in the response is due to the rocking mode...at least all indications are pointing that way.
The rocking mode can be thought of as the mechanical analog of a vented box.
At port resonance in a vented enclosure the port is 180 degrees out of phase with the cone and if you have ever seen response plots showing separate piston and port outputs you will notice the cone output going to a minimum right where the resonant output of the port is at its maximum.
The rocking mode is similar in nature, being driven by the motion of the cone. At resonance the cone rocking motion is at a maximum and the side of the cone where the accelerometer is mounted is moving 180 degrees out of phase to the motion of the voice coil, hence the notch in the accelerometer measured cone response as the two motions are tending to cancel out.
Last edited:
Ah.... interesting. A mysterious hole in the circuit right at 13 Hz. Just because the cone has a rocking inclination, why should the accelerometer be reflecting that in its output?
Bolserst (once again) provides an incisive analysis and that may well be correct. But I am sure it can be tested in ways much easier to interpret than the twin-tee tests. A mechanical pulse to the cone picked up by a different accelerometer or even the VC (albeit at a low level)?
The rumble at the end of the tone burst looks like typical electronic instability. Maybe through the power supply. Or some other piece tied to the feedback network.
If there are ooodles of feedback everywhere but 13 Hz, wouldn't that cause bubbles of output at 13 Hz?
There are all kinds of unfriendly shakes in a driver (and not identified in those swell models I hear about so much). But my intuition is that 13 Hz is way too low for the rocking motion by hundreds of Heinrich Hertz' (purely a guess). The whole darn cone assembly can barely resonate below 30 Hz on a good day.
But let's not lose sight that Armand has created a servo system that does AMAZING good things by controlling the cone, as his tone bursts, etc. have shown. MF - the final frontier.
Ben
Bolserst (once again) provides an incisive analysis and that may well be correct. But I am sure it can be tested in ways much easier to interpret than the twin-tee tests. A mechanical pulse to the cone picked up by a different accelerometer or even the VC (albeit at a low level)?
The rumble at the end of the tone burst looks like typical electronic instability. Maybe through the power supply. Or some other piece tied to the feedback network.
If there are ooodles of feedback everywhere but 13 Hz, wouldn't that cause bubbles of output at 13 Hz?
There are all kinds of unfriendly shakes in a driver (and not identified in those swell models I hear about so much). But my intuition is that 13 Hz is way too low for the rocking motion by hundreds of Heinrich Hertz' (purely a guess). The whole darn cone assembly can barely resonate below 30 Hz on a good day.
But let's not lose sight that Armand has created a servo system that does AMAZING good things by controlling the cone, as his tone bursts, etc. have shown. MF - the final frontier.
Ben
Last edited:
I think he was trying to fix the 13Hz with the notch filter, not determine its exact cause.
A simple test would be to add a bit more weight on the counterweight side of the cone(ie opposite the accelerometer) and measure the open loop response as before and see if frequency or phase is affected. My guess would be that as weight is added the notch depth would be reduced. If weight is continued to be added, it wouldn't be unreasonable to think the the notch would turn into a peak. Also as weight is added, the frequency of the rocking mode would be lowered.
Now that you mention it, I don't think Armand has posted a closed loop frequency response that extended below 20Hz. But yes, you would expect to see a bump in the response at 13Hz to go along with the poor transient response and ringing mode he has shown measurements for.
Have you compared the suspension stiffness for rocking versus linear motion in a long throw woofer like the Peerless XXLS?
Unless your woofer happens to have dual spiders, usually the resistance to rocking is rather small.
Indeed!
Freq curves not below 13 Hz, open or closed loop? It would be nice to see that. Given the wholesale instability of woofers, I'm all for cutting off the feedback network at 18 - and for sure, not going for DC coupling.
Can't say as I man-handle (or should I say "person-handle") spiders any more than by accident, but they seem pretty stiff to me sideways. But I certainly have no confidence on my intuitions here.
I run my REW test through a Mac laptop sound components. Seems to work pretty good. But my gut-feel is that a computer sound card isn't in the same quality league as HiFi gear. I have misgivings about all DSP stuff. Has Armand's soundcard been tested or tested implicitly through tests in which it was included? DSPs are "black boxes" and I avoid all the "black boxes" I can.
Ben
Can't say as I man-handle (or should I say "person-handle") spiders any more than by accident, but they seem pretty stiff to me sideways. But I certainly have no confidence on my intuitions here.
I run my REW test through a Mac laptop sound components. Seems to work pretty good. But my gut-feel is that a computer sound card isn't in the same quality league as HiFi gear. I have misgivings about all DSP stuff. Has Armand's soundcard been tested or tested implicitly through tests in which it was included? DSPs are "black boxes" and I avoid all the "black boxes" I can.
Ben
He has posted plenty of open loop curves going down below 10Hz(like the ones are showed in post #64).
But I can't seem the locate any closed loop curves that go that low.
I agree spiders are quite resistant to sideways motion.
Rocking would be tested by lifting up on one side of the cone while pushing down on the opposite side.
Armand used loopback tests with REW software to create calibration files for his sound card so response and phase are corrected down to 2Hz.
I've been away from the computer for a couple of days and the measurements come flying in! Great to see the excitement on the subject.
There were some questions;
. In my set-up I made an aluminium disc which sits in and on the voicecoil. On that disc, I have mounted/glued the accelerometer.
Keep up the good work. I think this is a lovely topic.
There were some questions;
The closer your accelerometer is a match to the propagated pressure-waves, the higher the improvement by the feedback. Only drawback; stiffening the cone usually means that at the shifted frequency of break-up, the effects are more violent and need to be damped more. High stiffness and high damping don't usually go along too well.
I have performed the linkwitz mod and have implemented the opamp buffer setup in order to try and get the most out of the mic. Fortunately, I expect to be measuring with a different MIC any day soon.
I have used a scanning laser in order to map the motion of the cone. I too did some research on using a non-contact laser single point laser as a real-time feedback sensor. However, in practice, they are big, slow, expensive and have a limited measuring range. The accelerometer is a great sensor for this type of application.
Rocking is monitored when the accelerometer is mounted off-axis. Guess what, it is in Armand's set-up
. In my set-up I made an aluminium disc which sits in and on the voicecoil. On that disc, I have mounted/glued the accelerometer.Keep up the good work. I think this is a lovely topic.
Rocking movies!!!!
Very interesting discussion. I'll see what I can do to do some more measurements/explanations regarding my system. Sadly my job as a teacher starts up again this week and I am afraid I won't have so much time in the time to come.
But now I want to show you something I was thinking about when going to sleep last night. I have a Gopro Black edition with 240fps capability, and I am so excited about these movies I just have to post them now.
I played one single wave 70Hz at 25V peak amplitude without servo and recorded a video of the event. Have a look.
View attachment Rocking movies.zip
Very interesting discussion. I'll see what I can do to do some more measurements/explanations regarding my system. Sadly my job as a teacher starts up again this week and I am afraid I won't have so much time in the time to come.
But now I want to show you something I was thinking about when going to sleep last night. I have a Gopro Black edition with 240fps capability, and I am so excited about these movies I just have to post them now.
I played one single wave 70Hz at 25V peak amplitude without servo and recorded a video of the event. Have a look.
View attachment Rocking movies.zip
Wooooo...that's awesome!!!
Driving in to work today I thought to suggest to you to use a strobe light to visualize the rocking.
I have used strobed lighting to visualize surround and lead wire behavior before.
But, your video grab makes my idea sound a bit archaic.
Another thought I had was that the rocking may be partially caused by the lead-out wire from the accelerometer if it doesn't have the proper amount of slack.
Fantastic. That looks like rocking all right.
For sure, rocking needs to be eliminated to produce good sound. Is it the weight of the accelerometer? If so, adding more weight seems a contentious fix.
Doesn't all sound get distorted if the cone moves that way with the accelerometer weight added?
Is it 13 Hz resonance or part of every freq and loudness cycle? Is the voice coil rubbing?
Great pictures. Sure beats watching with my StroboTac or locopodium powder!!!
Ben
For sure, rocking needs to be eliminated to produce good sound. Is it the weight of the accelerometer? If so, adding more weight seems a contentious fix.
Doesn't all sound get distorted if the cone moves that way with the accelerometer weight added?
Is it 13 Hz resonance or part of every freq and loudness cycle? Is the voice coil rubbing?
Great pictures. Sure beats watching with my StroboTac or locopodium powder!!!
Ben
It's not the weight per se but the distribution of weight that can lead to rocking. You want center of mass to line up with center of force. Differences in surround tension around the circumference due to imperfect mounting of the surround can also lead to rocking. This is especially troublesome when excursion starts to exceed 25mm or so. The most common solution is to replace the single spider with a spaced set of spiders to stabilize this motion as shown in attached pic back in post#57.
http://www.diyaudio.com/forums/subwoofers/239941-analog-servo-sub-6.html#post3590443
With the wavelengths subwoofers are asked to put out, there isn't much if any distortion added since the average motion(ie volume velocity, or amount of air moved back and forth) of the cone is still the same. Well...that is unless you have an off-center mounted accelerometer picking it up and and feeding it back to the input.
Just like any resonance mode it will respond most when driven at its natural frequency and taper off on either side of that frequency. The leading and trailing edges of square enveloped tone bursts contain a broad frequency range, so they excite it pretty strongly.
Wait just a minute....
There seems to be some artifacts with the gopro camera.
I have made some new videos today from another woofer and it had some rocking problems too.
First I turned the whole box upside down and suddenly the rocking was inverted. Did it have something to do with earths gravity??
I then turned the camera upside down and the rocking was inverted again!
It seems that the gopro is scanning its picture chip from bottom to top with a time delay that shows up as rocking in the movie.
I turned the camera 90 degrees and the rocking was gone. (at least in the up-down axis)
The old trusted stroboscope or locopodium powder may not be so archaic after all
I'll see what I can do to make some more accurate videos. I saw there was a 1.000.000 fps camera available. (100.000 $)
There seems to be some artifacts with the gopro camera.
I have made some new videos today from another woofer and it had some rocking problems too.
First I turned the whole box upside down and suddenly the rocking was inverted. Did it have something to do with earths gravity??
I then turned the camera upside down and the rocking was inverted again!
It seems that the gopro is scanning its picture chip from bottom to top with a time delay that shows up as rocking in the movie.
I turned the camera 90 degrees and the rocking was gone. (at least in the up-down axis)
The old trusted stroboscope or locopodium powder may not be so archaic after all
I'll see what I can do to make some more accurate videos. I saw there was a 1.000.000 fps camera available. (100.000 $)
I must admit to being surprised by the magnitude of the rocking shown in the video clip.
Much, much more than I had ever seen with a strobe light, but I didn't question it..
Sooooo....holding off on purchase of a gopro till I see if you can get it sorted out
bentoronto should give himself a pat on the back for questioning if the apparent rocking motion wouldn't be causing voice coil rubbing.
Always a good idea to question new technology and measuring techniques when it doesn't pass the smell, errr rub test.
Last edited:
OMG!!!! Right!!! Of course!!!Wait just a minute....
There seems to be some artifacts with the gopro camera.
I have made some new videos today from another woofer and it had some rocking problems too.
First I turned the whole box upside down and suddenly the rocking was inverted. Did it have something to do with earths gravity??
I then turned the camera upside down and the rocking was inverted again!
It seems that the gopro is scanning its picture chip from bottom to top with a time delay that shows up as rocking in the movie.
I turned the camera 90 degrees and the rocking was gone. (at least in the up-down axis)
The old trusted stroboscope or locopodium powder may not be so archaic after all
I'll see what I can do to make some more accurate videos. I saw there was a 1.000.000 fps camera available. (100.000 $)
You're brilliant to discover this flaw of measurement. And rare honest person for posting it too.
In my business, we are constantly struggling to critique our tools of measurement and our research methods.
It is possible some of the rocking is the cone and some the nature of the measurement. So uncertainty remains. I think a StroboTac would make rocking motion visible it it wasn't too tiny.
But as I earlier asked, if there is rocking that big, you'd ask about the VC rubbing and distortion at all frequencies. Not to mention Doppler distortion* (Actually, I was going to say "hey, cheap driver that can have such a loose voice coil gap...." but I realized that would be so discourteous a reaction to your beautiful high speed photography.)
Ben
*I wonder if those small cone Philips advocates believe that Doppler distortion matters?
Last edited:
Yes, you were quite right Ben. With that rocking I would expect the voice coil rubbing. But I can assure you it is not. I remember that sound from a MF project years ago.
Since I started this project it has indeed been a hunt for measuring errors. There are so many things that can go wrong, and I have been through them all. (Or maybe not..)
But I have now a very stable power supply, calibrated soundcard and with finding the flaw in the gopro camera yesterday I think things are working OK.
BTW, I made another video using the gopro camera in a 90 degree angle and then the bad rocking we saw in the previous video was gone. (but it had of course moved to the righ-left axis). It does not seem that the gopro is suited.
While going to sleep yesterday I must have been inspired by the meteor storm and designed a stroboscope in my head. Now it is up and running and working really good but I need some tip on how to use it.
The strobe lamp is a 30W LED controlled as in the schematic attached.
As input signal I can an external wave generator or use the woofer signal. I have a switch that switches between trigging on positive or negative slope. I can also adjust where on the sinus curve I want to trig the flash.
When using an external signal generator I can make the woofer look like it is standstill in any position, or make it look like it is slowly moving in or out by slightly ingreas or decrease the frequency on the signal gererator.
When driving the stroboscope from the woofer signal it always appear as if it isn't moving and I have the option to adjust the trigger so I can make it "stand still" is any position.
I'ts really fascinating to look at but I have not been able to indentify any rocking. Do you have any suggestions?
Armand
Since I started this project it has indeed been a hunt for measuring errors. There are so many things that can go wrong, and I have been through them all. (Or maybe not..)
But I have now a very stable power supply, calibrated soundcard and with finding the flaw in the gopro camera yesterday I think things are working OK.
BTW, I made another video using the gopro camera in a 90 degree angle and then the bad rocking we saw in the previous video was gone. (but it had of course moved to the righ-left axis). It does not seem that the gopro is suited.
While going to sleep yesterday I must have been inspired by the meteor storm and designed a stroboscope in my head. Now it is up and running and working really good but I need some tip on how to use it.
The strobe lamp is a 30W LED controlled as in the schematic attached.
As input signal I can an external wave generator or use the woofer signal. I have a switch that switches between trigging on positive or negative slope. I can also adjust where on the sinus curve I want to trig the flash.
When using an external signal generator I can make the woofer look like it is standstill in any position, or make it look like it is slowly moving in or out by slightly ingreas or decrease the frequency on the signal gererator.
When driving the stroboscope from the woofer signal it always appear as if it isn't moving and I have the option to adjust the trigger so I can make it "stand still" is any position.
I'ts really fascinating to look at but I have not been able to indentify any rocking. Do you have any suggestions?
Armand
You might try feeding the woofer a signal just above 13Hz(what we think is a rocking mode) and trigger the strobe light from the same signal you are feeding the woofer amplifier. So if you fed it say a 14Hz signal, the strobe would "freeze" the 14Hz up and down motion, but you would see a slow-motion version of the 13Hz rocking mode.
What is the length in mS of your strobe pulse?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.