All is now clear to everybody.
I'd just comment on Art's very helpful and clearly presented post that at hellokitty's higher treble the degrees start to add up and you'd need a massive steel structure to secure the mic .75-inches from the dust cap.
Being too poor to own my own laser tape-measure, I wonder if they do put out a continuous signal with a 1/200,000 second latency that can be fed continuously and without any interruption what so ever into the amp? And we are talking about wonderfully fast analog gear, not gear with bits slowly rummaging about?*
B.
*My best idea ever was to use a DSP in the motional feedback loop (since there is some signal shaping that is really, really worth having in that feedback signal). Too bad such gizmos are way too slow to provide feedback even at puny audio periods, even for a sub.
I'd just comment on Art's very helpful and clearly presented post that at hellokitty's higher treble the degrees start to add up and you'd need a massive steel structure to secure the mic .75-inches from the dust cap.
Being too poor to own my own laser tape-measure, I wonder if they do put out a continuous signal with a 1/200,000 second latency that can be fed continuously and without any interruption what so ever into the amp? And we are talking about wonderfully fast analog gear, not gear with bits slowly rummaging about?*
B.
*My best idea ever was to use a DSP in the motional feedback loop (since there is some signal shaping that is really, really worth having in that feedback signal). Too bad such gizmos are way too slow to provide feedback even at puny audio periods, even for a sub.
Last edited:
They're called PSDs. Here is the datasheet for one I was considering
https://www.mouser.com/ds/2/313/DL100-7-PCBA3-500744-586528.pdf
Last edited:
Great little device. But would I have to play music in the dark while wearing laser protection goggles? If I turn on my reading lamp, would the bass drop?
Obstacles to home use aside, these laser things are sure more fun in research settings than lycopodium powder and a StroboTac*, as in days of my youth.
B.
*free hint: StroboTac is great for setting ignition timing on motorcycles
Last edited:
Good questions. I have no idea. I have a couple of projects on my to-do list before I drop the $350 on the laser to find out.
I'm thinking you can probably just have the laser behind the speaker and then invert the phase so you don't have to worry about eye damaging radiation and light interference.
Back in the late 1990s, I was trying to make some amplitude and phase measurments on a loudspeaker, using a microphone positioned some distance away.
The distance was as small as was compatible with accurate frequency response measurements, but that was already enough to produce vast amounts of total phase due to the time delay.
Someone on the team had the bright idea to use an Alesis digital delay (I think it was a Midiverb) in the reference channel, and set it to provide an equal amount of time delay. In effect, that would let us subtract off the large amount of phase caused by travel through the air.
When I tried to make the measurements, I found that the Midiverb was in fact introducing several milliseconds of delay - even when the knobs were set to zero delay!
That's when I discovered that digital data buffers are a standard part of A/D conversion...and introduce appreciable time delays, whether you like it or not!
-Gnobuddy
This is not feedback, but pre-distorting the signal in the hope that the pre-distortion will cancel out with the distortion in your transmission channel (speakers, in this case.)
There are many historical precedents for optimistic pre-distortion (but it rarely works out as well as hoped-for.)
For example, simple geometry dictates that a record player stylus - which has radiused edges to avoid destroying the vinyl record - cannot possibly follow the same path as the knife-edged cutter that originally cut the groove into the record.
This results in vast amounts of harmonic distortion, inherent to the record cutting / playback mechanism.
What to do? In the spring of the vinyl record era, some record manufacturing plants took to pre-distorting the grooves in the record, in such a way as to (maybe) partially cancel the distortion subsequently caused by the playback stylus. They came up with catchy advertising terms to describe this (I've forgotten them, but use your imagination.)
Unfortunately, the world is complex, and pre-distorting the record groove has a number of catches:
1) You don't know what tip radius the eventual playback stylus will have, and your pre-distortion will make things worse for some people.
2) Record wear re-shapes the groove after one single playback, so your pre-recorded correction evaporates into thin air.
3) The amount of distortion varies with the radius of the track on the vinyl, so there is more distortion near the centre, where the groove velocity is low and recorded wavelengths are shortest. Which radius shall we pre-distort for?
For example if you set up your room and sit in a designated spot, then set up the mic in front of you, then play a song and digitally record the audio on the mic.
Then you save that "setting" for your next playback and align the recorded feedback to match the original recording so they are in phase. Tadah localized acoustic feedback
1) What happens if you move your head an inch? (1.7 centimeters of head movement equals a full 360 degrees of phase shift at 20 kHz. One inch is 548 degrees of phase shift at 20 kHz! )
2) What happens when the room temperature or humidity changes, along with the speed of sound in air, and therefore, the phase shift?
3) What happens when the room temperature changes, stiffening (or softening) the loudspeaker cone, surround, spider, dustcap, and the glues and damping compounds holding it all together?
4) What happens when you bring in that nice new piece of furniture, altering the acoustics of the room?
5) What happens as the speakers age, and their mechanical and acoustical properties change?
Actual feedback corrects, to a certain extent, for all of these things, in real time. Pre-recorded "best guess" correction does not.
For many years now, there have been home theatre setups that use a microphone at the listening position to attempt to equalize away big peaks and dips caused by listening room acoustics. These systems try to correct gross amplitude errors, occasionally with some success, but they never try to correct phase errors - why do you suppose that is?
-Gnobuddy
At one time, I worked in a lab with a powerful argon-ion (green) laser. If you didn't take appropriate precautions and looked into the beam, the laser was more than powerful enough to blind you in milliseconds, long before you had time to realize you were staring at a very bright light.At least I still have my lasers to play with
To help keep everyone on their toes, there was a sign on the outside of the lab door that read "Do not stare into laser with remaining eye."
-Gnobuddy
RCA "Dynagroove".* Some of the Fritz Reiner Chicago records were pretty good (Lt. Kije) but maybe my chi-chi Decca stylus was closer to RCA's template.
If only somebody as profoundly lucid as Gnobuddy was running the.....
B.
*had to google it
Last edited:
Ben, have you seen the Grimm LS1s-dmf subwoofer? https://www.grimmaudio.com/pro-products/loudspeaker/ls1s-dmf/
It uses MFB and DSP.
It's designer (RMS) is also on mfblabs.nl.
I remember you once asked if it was possible to improve already good woofers by a margin:
Lots of interesting reads over at Grimm. Take a look at RMS as well: Intro RMS-Acoustics & Mechatronics
Grimm, RMS, Hypex all have something in common, and it's not hard to spot.
It uses MFB and DSP.
It's designer (RMS) is also on mfblabs.nl.
I remember you once asked if it was possible to improve already good woofers by a margin:
Put on your marketing-talk removal glasses and look for the technical stuff.
Lots of interesting reads over at Grimm. Take a look at RMS as well: Intro RMS-Acoustics & Mechatronics
Grimm, RMS, Hypex all have something in common, and it's not hard to spot.
What do you spot?
Yes, Professor Schmidt's write-ups are enlightening (bolserst gave me the link not long ago). But have they actually produced any speakers?
But 30dB improvement is unlikely: you'd need 30dB of feedback.
Schmidt's advocacy of current-device amps seems misplaced since it gets odd when talking about negative output impedance amps when motional feedback is present. Testing a motional feedback amp is like playing with Dr. Frankenstein's monster; you want to scream, "It's ALIVE". (Ordinary audio power amps with high damping factors are very, very dead, by contrast.)
B.
The Grimm sub runs with 30db feedback, this is only possible with a low latency dsp in the feedback loop (adau1772). This dsp has a new brother which runs 768K samplerate. But it is a 0,5 pitch BGA and I designed a small 4 layer board for it, will be in tomorrow:
It is 2x2cm.....
It is 2x2cm.....
Attachments
Ah, "the plot thickens".... Thanks for update.
My earlier disparaging remarks about DSP speed should have been limited to general purpose DSP boxes. Fast, purpose-built DSP is exactly what is needed, but not usually in the DIY realm of design or cost. Having said that, I recently bought a Raspberry Pi and started learning about system and if it could work for MF. Anybody have thoughts?
But what I'd really like to know is what tricks does the ds23man DSP perform? Esp, how does it tailor the feedback loop (such as bandwidth limitation and phase finagle within the passband) so that you can achieve 30 dB feedback without howl back?
B.
My earlier disparaging remarks about DSP speed should have been limited to general purpose DSP boxes. Fast, purpose-built DSP is exactly what is needed, but not usually in the DIY realm of design or cost. Having said that, I recently bought a Raspberry Pi and started learning about system and if it could work for MF. Anybody have thoughts?
But what I'd really like to know is what tricks does the ds23man DSP perform? Esp, how does it tailor the feedback loop (such as bandwidth limitation and phase finagle within the passband) so that you can achieve 30 dB feedback without howl back?
B.
Last edited:
It would be really nice if you would describe the manner in which the DSP is employed to modify the feedback loop as to passband, phase adjustment, timing, or whatever other functions (AKA "tricks") it performs.
With speakers, very challenging to achieve high levels of feedback correction due to the wonky characteristics of subs. (Which is why MF is so applicable for subs.) And which is why DSP could be very valuable in wrangling the feedback loop.
Or do you consider that proprietary information?
B.
Yes, the LS1s-dmf has been on the market for a while…since mid-2016 I believe.
Hopefully you took the initiative to find that out on your own.
One thing that probably helps achieve the 30dB of feedback is the Hypex power amp used is DC coupled, improving the phase margin for stability. Measurements are shown in Fig 17 of white paper #6 of the open-loop response peaking out just above 30dB. Distortion comparison with/without feedback is included with my “calibrated eye” showing ~27dB reduction in 3rd harmonic of 20Hz where the open-loop gain is ~28dB.
He is advocating for current source amplifier rather than voltage source. This is not a negative output impedance but rather a very large positive output impedance. Current feedback is used but it is negative feedback, not positive feedback as is needed to generate negative output impedance. His argument is that the force generated by the voice coil is proportional to current not voltage so why not use the appropriate tool for the job? Measurements in Fig 4 of white paper #2 show that using a current source amplifier reduces inductance related distortion above resonance by 6dB - 10dB. Notice that this is exactly the area where the loop gain is falling so feedback’s ability to correct distortion is falling as well. Other than that, I don’t see a particular benefit to using a current source rather than more traditional voltage source for the MFB amplifier.
A plot showing the feedback loop tailoring is provided in Fig 16 of white paper #6 along with description in the text as to why each adjustment was made. It looks like a localized boost in the loop gain just above resonance in combination with the traditional integrator circuit is what was used to achieve the 30dB loop gain figure. Phase margins of 45deg at LF and HF indicates very good stability. Note that there is no tailoring of phase with FIR filters as that would involve way too much latency. Since all of the tailoring is minimum-phase the corrections shown could be performed with analog filters. Although looking at the complexity, DSP is much more convenient.
Unfortunately, nonlinearities at the edges of the magnetic field and voice coil, as well as nonlinearities in the spider and surround, ensure that the force actually applied to the speaker cone is not proportional to the current when voice coil excursions are large.
In other words, the claim that force is proportional to current fails to hold true under exactly the conditions where MFB is most useful!
If in fact force at the speaker cone was perfectly proportional to current, we wouldn't need MFB at all. Just go back to no-feedback pentode amplifiers (kidding!), or wire your solid-state amp to work as a current-source, and hey presto, perfect sound!
If only it was that easy!
-Gnobuddy