Thanks for the illustration and explanation. For some reason since it was a direct response to Mr. Eraths feedback unit(when they started voiding warranties on their amps if Mr. Earths was added) I figured it would be more sophisticated than that.
This is a very interesting project and I will be following silently..
This is a very interesting project and I will be following silently..
Armand's work is too good to drop from view.
I notice in post #1, you have a miniDSP 2x8 in your toolbox. Seems to me (speaking only speculatively, not from experience), the place for DSP is inside the feedback loop. It can be used to band-limit (10 to 300 Hz), tweak naughty phases, and hand-tune any troublesome anomalies. Indeed, motional feedback in the past has been hobbled by having these tricks done feasibly (sharply, at several frequencies, within the loop, and at low frequencies) prior to DSP gizmos becoming available.
For using bridge-derived feedback, DSP also can convert velocity control into acceleration control, or do I have that backwards?
What would be a grand next step to this thread would be seeing a final sub using a quality sealed box (except also for IB, no other type of enclosure is possible), high quality driver, and quality amp....and mic traces with and without feedback.
Ben
I notice in post #1, you have a miniDSP 2x8 in your toolbox. Seems to me (speaking only speculatively, not from experience), the place for DSP is inside the feedback loop. It can be used to band-limit (10 to 300 Hz), tweak naughty phases, and hand-tune any troublesome anomalies. Indeed, motional feedback in the past has been hobbled by having these tricks done feasibly (sharply, at several frequencies, within the loop, and at low frequencies) prior to DSP gizmos becoming available.
For using bridge-derived feedback, DSP also can convert velocity control into acceleration control, or do I have that backwards?
What would be a grand next step to this thread would be seeing a final sub using a quality sealed box (except also for IB, no other type of enclosure is possible), high quality driver, and quality amp....and mic traces with and without feedback.
Ben
Last edited:
Bad me. You're right. But, I speculate, a DSP in the loop, capable of sharp non-ringing cut-offs, can correct for some of the Nyquist criteria (gain and phase) as expeditiously, I imagine, as the typical analog op-amp design with a very large bandwidth.
We are talking only of quite low frequencies, say below 100 Hz. But it is beyond my math ability to judge what speed is too slow. For sure, 1 m seems too slow. But what's your ballpark estimate?
Thanks for your correction.
Ben
Last edited:
At 1 kHz, that is an open loop delay/phase of 360 degrees. 500 Hz, 180 degrees. 250 Hz, 90 degrees.
If I want to design a controller, in the continous time domain, my unity gain crossover should have (way) more phase margin than the delay added by latency.
Try and design such controller and see how large your THD suppression becomes.
If I want to design a controller, in the continous time domain, my unity gain crossover should have (way) more phase margin than the delay added by latency.
Try and design such controller and see how large your THD suppression becomes.
I am baffled in trying to parse those sentences. Can you please say again in other words.
So, if we are talking about subs playing south of 125 Hz, our 1 ms latency inside the motional feedback loop due to the miniDSP becomes just 45-degrees, at the worst freq, Moreover, the heart of the motional feedback action (and the region where the speaker behaviour starts to go wild) takes place two octaves below that with a phase angle delay of just 11-degrees.
My guess is that we wouldn't sweat that delay.
Another question: if there is no off-the-shelf program for a miniDSP suitable for providing the tricks needed within the motional feedback loop, is it feasible to write one (or is all the software written inside the miniDSP factory)?
Ben
OK, down to the short strokes. Here are my vague impressions but I hope more knowledgeable others will speak up.
If you wanted to put a DSP inside the feedback loop to replace other component bits and pieces, you'd have it do the following tricks.
1. using very sharp slopes, ensure that nothing materially outside the speaker passband appears in the feedback loop
2. depending on various factors of driver size and whether you are using acceleration or velocity feedback, equalize and/or slope the signal to the driver
Ben
negative feedback has problems with "sharp slopes" - they are always accompanied by large phase shifts
in well behaved systems it is possible to close digital loops with latency ~ 1/5 of the loop gain intercept frequency
but there's no excuse for only 1 kHz digital loops today - even running a "real operating system"; linux on GHz processors ($45 BeagleBone Black) can manage >10 kHz sampling/feedback loops
the problem may be I/O - USB uses 1 ms frame rate
figure 11 specifically shows loop gain shaping to deal with a flexible structure resonant mode just beyond loop gain intercept
in well behaved systems it is possible to close digital loops with latency ~ 1/5 of the loop gain intercept frequency
but there's no excuse for only 1 kHz digital loops today - even running a "real operating system"; linux on GHz processors ($45 BeagleBone Black) can manage >10 kHz sampling/feedback loops
the problem may be I/O - USB uses 1 ms frame rate
link does behave oddly - try https://www.google.com/#q=System+Trades+using+the+Asymptotic+Bode-Step+Method
New woofer. New results
Since last time I have mounted an accelerometer on a dirt-cheap Chinese-made 15" PA-woofer. I know, I know. You are asking why?! Well, the reason is that I have a long term goal of making DBA system Double Bass Array ? Wikipedia For that to work I need a lot of woofers and since a lot of woofers is expensive I was tempted to see what I could achieve with cheap drivers. So I bought 16 of these on sale a while ago for $100 each. I was hoping that with the small x-max of 3.25mm these drivers would not have too much distortion in that range.
I was completely wrong!
It turns out that in a 65 liter box with only 3mm excursion at 25Hz (about 35W power) the harmonic distortion is 33%!!
However, I still wanted to give them a chance with accelerometer feedback to see what I could achieve.
The cones are fairly light with only 65 grams so I had to try to keep the added mass as low as possible. The mounting platform is made of carbon fibre and mounted with glue inside the voice coil. It weighs in at 10 grams. The accelerometer is 4 grams, and with about 3 grams of glue the total is about 17 grams. The removed original dust cap was 6 grams, so the effective added mass is then about 11 grams. Carbon fiber is an incredible material and the platform is very stiff. There is no sign of breakup or resonance.
After some tuning a was able to get the THD down from 33% to 4.6% at 25Hz. In other words about 18 dB down.
At 40Hz the distortion went down from 13,4% to 1,8% with 35W power. Also with about 3mm excursion.
I have to say that I am very pleased with the sound. A double bass really sounds like a double bass
.
All measurements are done with microphone so this is real THD reduction. The SPL in the charts is also calibrated. The mic was 35cm from the cone.
I have saved the results as two links. I recommend that you open them side by side so you can compare the results before and after. Note how awful the microphone waves look on the scope pictures at 25Hz !!!
The two links are at the bottom of this page together with som pictures from the project: PA1550_ServoSub
I realize that 65 liters is to small for these cheap drivers, so the next thing I am going to do is to make two boxes of about 100 liters and try to compensate the other woofer from one feedback system.
Armand
Since last time I have mounted an accelerometer on a dirt-cheap Chinese-made 15" PA-woofer. I know, I know. You are asking why?! Well, the reason is that I have a long term goal of making DBA system Double Bass Array ? Wikipedia For that to work I need a lot of woofers and since a lot of woofers is expensive I was tempted to see what I could achieve with cheap drivers. So I bought 16 of these on sale a while ago for $100 each. I was hoping that with the small x-max of 3.25mm these drivers would not have too much distortion in that range.
I was completely wrong!
It turns out that in a 65 liter box with only 3mm excursion at 25Hz (about 35W power) the harmonic distortion is 33%!!
However, I still wanted to give them a chance with accelerometer feedback to see what I could achieve.
The cones are fairly light with only 65 grams so I had to try to keep the added mass as low as possible. The mounting platform is made of carbon fibre and mounted with glue inside the voice coil. It weighs in at 10 grams. The accelerometer is 4 grams, and with about 3 grams of glue the total is about 17 grams. The removed original dust cap was 6 grams, so the effective added mass is then about 11 grams. Carbon fiber is an incredible material and the platform is very stiff. There is no sign of breakup or resonance.
After some tuning a was able to get the THD down from 33% to 4.6% at 25Hz. In other words about 18 dB down.
At 40Hz the distortion went down from 13,4% to 1,8% with 35W power. Also with about 3mm excursion.
I have to say that I am very pleased with the sound. A double bass really sounds like a double bass
. All measurements are done with microphone so this is real THD reduction. The SPL in the charts is also calibrated. The mic was 35cm from the cone.
I have saved the results as two links. I recommend that you open them side by side so you can compare the results before and after. Note how awful the microphone waves look on the scope pictures at 25Hz !!!
The two links are at the bottom of this page together with som pictures from the project: PA1550_ServoSub
I realize that 65 liters is to small for these cheap drivers, so the next thing I am going to do is to make two boxes of about 100 liters and try to compensate the other woofer from one feedback system.
Armand
Great to have Armand posting experiments again. Thanks.
What Armand has done is in the great historical tradition of motional feedback publications: you take a poor woofer and you make impressive improvements. In this case, big improvements are below the resonance in a sealed box (or even the published free-air spec).
If the forum of adherents of the Philips system don't mind me saying so, I can acknowledge their joy in taking a poor driver and getting good bass from it. Nice accomplishment. But my personal interests are related to seeing what happens when you take a pretty good driver and make it better than anything you could buy anywhere.
That's my personal preference.
Ben
What Armand has done is in the great historical tradition of motional feedback publications: you take a poor woofer and you make impressive improvements. In this case, big improvements are below the resonance in a sealed box (or even the published free-air spec).
If the forum of adherents of the Philips system don't mind me saying so, I can acknowledge their joy in taking a poor driver and getting good bass from it. Nice accomplishment. But my personal interests are related to seeing what happens when you take a pretty good driver and make it better than anything you could buy anywhere.
That's my personal preference.
Ben
Last edited:
15" and 3mm max excursion should indicate that this woofer is not designed for anything under 50 Hz
Why did you pick this driver? A 10 inch with 10 mm Xmax should be capable of the same acoustic power.
Yes, I would get the same SPL from a 10 (or 9mm) x-max 10" woofer. But I thought that with only 3.3mm x-max the woofer would have less distortion than a 10" with 9mm. With 8 drivers in front I would have enough SPL anyway. But I was wrong. Distortion is all about magnet motor quality
It is a little wast of time doing all this work on cheap drivers. On the other hand I would not feel very comfortable cutting the dust cap off a $800 driver
Anyway, if my DBA project is succesful I can always upgrade later.
Nicely done!
No 13hz (rocking/cantilever/????) mode in sight Of course that means you were limited to loop gains of "only" 15dB-20dB, more typical of analog MFB systems.
Your distortion reduction factor(~7) is essentially spot on what theory would predict for 15dB - 16dB loop gain.
Definitely.
Like we have discussed before, unless your motor system is extremely linear within the desired excursion limits, an accelerometer is the better sensor choice.
With the same sensor and electronics/loop gain, the same distortion reduction factor(~7) would apply to a more linear woofer driver, limited only by the distortion products of the accelerometer itself which is typically about (0.1%)
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.