Armand, whatever you or David do I'm sure nothing is ever going to satisfy Ben.
Good work by both of you!
Anyone can PLAINLY see my post started with an apology ("sorry"... apologies are very rare on these anonymous forums).
Then I plainly stated that I thought his results were meaningful and impressive ("your charts show effective MF"). That's my supportive praise and I'll stick with it... unless there are later critical attacks which reveal flaws in Armand's work based on reasons I can't see now (and don't expect... which is why I apologized and stuck my neck out in praise of his work).
Too bad so many people read a PLAIN text as if it were some kind of sarcasm. Hard to say plain nice things without somebody accusing you of being sarcastic.
I then added, that now that he has established his tech cred (further praise and that's pretty plain to read and does not mean the opposite of what it seems to say) he kind of earned the right here also to talk subjectively (unlike many "golden eared" types who have not done measurements and only talk subjectively).
I also suggested a "next step" would be to show results for cases higher in frequency and lower in loudness where the distortion isn't as excessive. I'd like to see that.
Jonasz, I think it is time for you to make an apology to the forum and to me for accusing me of sarcasm. Will I have to wait till hell freezes over? And I don't know what Calvin's comment apparently agreeing with Jonasz means.
Ben
Last edited:
I didnt mean that quote alone but you negative tone through the whole thread. Sorry if you find me offending but a little encouragement to the ts wouldnt hurt I think.
Thanks for the apology, I think. But the right way to apologize is not that you are sorry I misunderstood but that you are sorry you misunderstood. Now THAT would be rare and brave.I didnt mean that quote alone but you negative tone through the whole thread. Sorry if you find me offending but a little encouragement to the ts wouldnt hurt I think.
I think if you read this thread, you will find that I am a very exuberant cheerleader for MF, not a critic of MF.
And specifically apropos David, in the post you mistakenly thought was sarcastic I said "we'd all like to see if David's DSP can help advance the application of MF." No need to apologize for your false accusation.
What I definitely do not like is guys posting teaser amazing results only to discover later that they plan to sell it to us. Bad smell.
But bad smell or not, Armand's work looks like affordable accelerometer MF* has arrived and maybe David's DSP developments can bring us even better MF. I hope so.
Again, thanks for the apology, I think.
Ben
*affordable for DIYers, trying to make good woofers better, not like Philips old design trying to make cheap speakers tolerable.
Last edited:
Mic is close. Somewhere between 5 and 10 cm. Note that the SPL levels are not calibrated. Everything is just relative.
I have btw come further in my project and have sub 1% THD at all frequencies up to 100Hz except 20Hz which is at 1.3%. This is tested with 18.5V RMS at speaker terminals.
The system is stable with loopgain of 42dB at 39Hz and about 30dB at 100Hz.
I have run Linkwitz's bursts at 20,50 and 100 Hz and it looks really good.
The thing I am working on now is a slight fast decaying 13Hz oscillation that occur after playing one cycle at 18.5V RMS.
I am also considering to implement a DC servo since just 1mV offset on the opamp input creates a lot of DC error on the speaker. Now I am dependent on expensive opamps with low drift and offset adjustments.
If anyone should wonder why so many tests are run at 18.5V RMS is because I use a DC coupled LM3886 with 29V supply and 18.5V RMS is just below clipping. If (when) things go wrong it will not tear up my speaker.
I have still not listened to music on it...
I have btw come further in my project and have sub 1% THD at all frequencies up to 100Hz except 20Hz which is at 1.3%. This is tested with 18.5V RMS at speaker terminals.
The system is stable with loopgain of 42dB at 39Hz and about 30dB at 100Hz.
I have run Linkwitz's bursts at 20,50 and 100 Hz and it looks really good.
The thing I am working on now is a slight fast decaying 13Hz oscillation that occur after playing one cycle at 18.5V RMS.
I am also considering to implement a DC servo since just 1mV offset on the opamp input creates a lot of DC error on the speaker. Now I am dependent on expensive opamps with low drift and offset adjustments.
If anyone should wonder why so many tests are run at 18.5V RMS is because I use a DC coupled LM3886 with 29V supply and 18.5V RMS is just below clipping. If (when) things go wrong it will not tear up my speaker.
I have still not listened to music on it...
How are you measuring loop gain? 42dB sounds pretty high for a stable MFB woofer system.
You may also want to check stability at higher frequencies where you were showing a resonance and phase spike around 700Hz in the plots from your post#247. Being a subwoofer, you won't be sending it a signal this high in frequency, but oscillation might be provoked by a bump or dropped book. You may need to add a notch filter in the loop to address this.
http://www.diyaudio.com/forums/subw...trolled-woofer-controller-25.html#post3566426
Concerning the 13Hz oscillation:
I'm assuming reducing the loop gain reduces the oscillation tendency?
You should be able to address this issue by adjusting your phase compensation to start at a lower frequency. Increasing C2 from 0.15uF to 1.0uF with RV3 set to maximum(100K) should help. You will need to readjust gain levels to get loop gain back where you want it.
Today I did my first acoustic measurements with a Panasonic WM61A cap.
I have not reached your level of THD. 4.5% is all I could get within reasonable excursion at 40 Hz (no feedback). For the acoustics part, I want to implement software like REW or Arta, since that leads to results easily interpreted in the audio-community. I too measured at around 10 cm distance from the dome.
Concerning LF oscillation and DC problems, bring out the term; compromise!
OpenLoop-shaping is all about gaining something at point x, and loosing performance elsewhere. I have had LF problems too, but I decided to go for "safety" over performance by setting openloop-gain to be sub-unity below 25 Hz. I can have 'all' the DC offset from opamps and the accelerometer I want
The reason I did this, was the fact that in the time domain I saw that below say 25 Hz, the measured output of the sensor was not even close to linearity. This therefore leads to the fact that stability by analyzing a linearized system is like fooling yourself. You implement rules that hold for linear systems and implement them on a non-linear system
By doing this, I had to lower my open loop gain, but at a value of over 15db from 60 to 300 Hz, I settled for that.
I asked you whether the woofer is in an enclosure. I did not seem to read your answer, but I must confess I have not read all 27 pages of this topic.
Next to that, I always, at every type of controller tapped the cone and/or surround 'violently' in order to check for LF stability. I can imagine that with your implementation, if in an enclosure, it would freak out at such high gains and such low margins of stability.
BTW, this is only sharing info. Not criticizing!
I have not reached your level of THD. 4.5% is all I could get within reasonable excursion at 40 Hz (no feedback). For the acoustics part, I want to implement software like REW or Arta, since that leads to results easily interpreted in the audio-community. I too measured at around 10 cm distance from the dome.
Concerning LF oscillation and DC problems, bring out the term; compromise!
OpenLoop-shaping is all about gaining something at point x, and loosing performance elsewhere. I have had LF problems too, but I decided to go for "safety" over performance by setting openloop-gain to be sub-unity below 25 Hz. I can have 'all' the DC offset from opamps and the accelerometer I want
The reason I did this, was the fact that in the time domain I saw that below say 25 Hz, the measured output of the sensor was not even close to linearity. This therefore leads to the fact that stability by analyzing a linearized system is like fooling yourself. You implement rules that hold for linear systems and implement them on a non-linear system
By doing this, I had to lower my open loop gain, but at a value of over 15db from 60 to 300 Hz, I settled for that.
I asked you whether the woofer is in an enclosure. I did not seem to read your answer, but I must confess I have not read all 27 pages of this topic.
Next to that, I always, at every type of controller tapped the cone and/or surround 'violently' in order to check for LF stability. I can imagine that with your implementation, if in an enclosure, it would freak out at such high gains and such low margins of stability.
BTW, this is only sharing info. Not criticizing!
We've all seen the oscilloscope traces when clipping occurs in an amplifier. Squaring off or flat-topping of a sine wave is a common occurance when exceeding the limits of a power amplifier or perhaps pre/processing circuit. However, when a woofer is overdriven it does not produce a flat-top or squarish looking waveform...instead it takes on a triangular shape. I remember this surprising me the first time I saw it.
Square waves have 3rd(and higher order odd) harmonic in phase with the fundamental.
Triangular wave have 3rd harmonic 180 degrees out of phase with fundamental.
Woofer cone acceleration is 180 degrees out phase with cone position.
Since radiation SPL corresponds with cone acceleration not position, the triangular shape makes sense.
If there is substantial 2nd harmonic you will see some frequency doubling as well.
Last edited:
Armand - sounds like you are truly succeeding with your R&D. Very wise ensuring the power rails poop out before the driver destructs.... DAMHIK.....
The problem of stability when working with drivers is inherent: they are working outside what would be an intelligently limited passband for other kinds of devices, say 60-300 Hz. Just maybe folks like us can make a system for our own use that has certain gremlin oscillations. But on a rainy day or if we want others to use it too, the instabilities have to be excised.
The Nyquist snail plot brings together the two critical issues of gain and phase and shows where they go bonkers - but you know that already. Very brave of you to work down to DC, but it may be quite hopeless to aim that low. As they say about DC-coupled systems, "pride goeth before a fall." Compared to test signals, you may find music (and your whole train of gear turning on and off and changing channels) triggers many undesirable burps.
http://en.wikipedia.org/wiki/Nyquist_plot
As Bolserst says, you gotta cut the gain* to cut the pain... but then you are cutting the benefit.
Ben
*or straighten the angle
The problem of stability when working with drivers is inherent: they are working outside what would be an intelligently limited passband for other kinds of devices, say 60-300 Hz. Just maybe folks like us can make a system for our own use that has certain gremlin oscillations. But on a rainy day or if we want others to use it too, the instabilities have to be excised.
The Nyquist snail plot brings together the two critical issues of gain and phase and shows where they go bonkers - but you know that already. Very brave of you to work down to DC, but it may be quite hopeless to aim that low. As they say about DC-coupled systems, "pride goeth before a fall." Compared to test signals, you may find music (and your whole train of gear turning on and off and changing channels) triggers many undesirable burps.
http://en.wikipedia.org/wiki/Nyquist_plot
As Bolserst says, you gotta cut the gain* to cut the pain... but then you are cutting the benefit.
Ben
*or straighten the angle
Last edited:
The loopgain is 38,5dB just before oscillation if I have understood things correct. (see my new thread for detail about measuring)
After lowering 8dB from there everything is stable, kicking, knocking and and eaven screaming at it does't affect it. It's calm
Yes, lowering the gain lowers the tendency.
I had already increased C2 0.15uF to 0.33uF and set RV3 to 112k before you posted your suggestion. I tried to increase the C2 to 1uF as you proposed, but it did not have any effect.
Did you do the Linkwitz mod?
At 40Hz my distortion is also a *lot* less than at 20Hz. About 3% with 18V RMS in the terminals. You should try 20Hz.
I use REW too. Really good tool!
I know. but I don't want to
OK, but this will not kill distortion in the worst frequency band. 20-50Hz. Do you have any measurements before / after?OpenLoop-shaping is all about gaining something at point x, and loosing performance elsewhere. I have had LF problems too, but I decided to go for "safety" over performance by setting openloop-gain to be sub-unity below 25 Hz. I can have 'all' the DC offset from opamps and the accelerometer I want
The reason I did this, was the fact that in the time domain I saw that below say 25 Hz, the measured output of the sensor was not even close to linearity. This therefore leads to the fact that stability by analyzing a linearized system is like fooling yourself. You implement rules that hold for linear systems and implement them on a non-linear system
By doing this, I had to lower my open loop gain, but at a value of over 15db from 60 to 300 Hz, I settled for that.
The enclosure is 135 liters. It does not freak out (much)
Sharing is good, as is also criticism. If we dare to post results we should also
tolrate that
I have decided to create a new thread about my project. You are all invided over to discuss what I have done.
I am suppressing the harmonics, not controlling the LF fundamental. I use a small woofer in a enclosure of around 10 liters. A proper 40 Hz propagation out of such enclosure/woofer-combo in practice would be great. Harmonics of that fundamental, say at 80, 120 160 200 Hz, are the targets to suppress.
I am currently working on before-after measurements from the acoustic measurement.
That sounds about right. With RV3 set at maximum, moving from .15uF to 0.33uF adds some phase margin in the 13Hz oscillation range, but going from 0.33uF to 1.0uF won't add much more. I suggested the 1uF since it was a nice common size
Will add other comments/questions in your newly started thread.
Nothing like a servo controlled 18 inch driver in a sealed enclosure
I am interested in seeing the results from this. Any updates ?One of the issues I see with a big driver such as this is achieving high loop gain over a broad band whilst maintaining a reasonable stability margin. It will be interesting to see if you can pull this off.
regards
Trev
How is your project moving on David?
I did some slow motion video recordings of my sub today. It seems that my mounting of the accelerometer and counter weight leads to rocking of the cone. Proper mounting seems to be crucial!
The video can be seen in post #72
http://www.diyaudio.com/forums/subwoofers/239941-analog-servo-sub-8.html
I did some slow motion video recordings of my sub today. It seems that my mounting of the accelerometer and counter weight leads to rocking of the cone. Proper mounting seems to be crucial!
The video can be seen in post #72
http://www.diyaudio.com/forums/subwoofers/239941-analog-servo-sub-8.html
Last edited:
Hello
I have been busy with other projects so I haven't had much time to devote to this project but have kept my website updated when I have any new information. A while ago I built an 18 inch sub using a Mach-5 IXL-18 driver and did some preliminary measurements which were quite favourable. I posted the results on my website recently.
ServoBass - High definition bass reproduction
cheers
david
Not all drivers are suitable for motional feedback due to modal resonances in the diaphragm. Also the tuning and setup is not a trivial task even with a simplified windows software interface that I have written. This is one of the reasons why I have put it on the back burner as far as a diy project is concerned. At the moment it is more suited to an OEM provider who manufacturers subs using the same driver all of the time. Therefore only a single setup file is needed for that particular box and driver combination and variations in the driver tolerances can be accommodated with some fine tuning.
cheers
david
Marvelous spec. Good presentation of the sub and concept. Can't be long before MFB subs wipe out all other kinds - certainly at this forum.
What roles does the DSP element play? Even just a broad-brush account would be interesting.
Do you delay the signal to the driver so as to account for DSP processing time in the feedback loop (unless I am stupidly asking for some suspension of the laws of physics)?
Ben
What roles does the DSP element play? Even just a broad-brush account would be interesting.
Do you delay the signal to the driver so as to account for DSP processing time in the feedback loop (unless I am stupidly asking for some suspension of the laws of physics)?
Ben
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.