By feeding back a voltage (positive) derived from a resistance inserted into the ground leg of the speaker outputs.
Like removing back hair with a bull whip... cure worse than the disease... you can see that I gave up! This is trading non-existant errors from one resistance, the cable, for the genuine errors in any feedback loop.
Like removing back hair with a bull whip... cure worse than the disease... you can see that I gave up! This is trading non-existant errors from one resistance, the cable, for the genuine errors in any feedback loop.
Hi !
My guess is, lightspeed is way too low to get this working good...
Simple calculations show that a 3meter cable (6 meter because of forth and back) create a delay of 20ns. Using a resonancefrequency of 7mhz (high BW amplifier) you introduce a phasehift of 50.4°. Assuming a lightspeed of 0.5c in copper you get 100.8° phasehift... BANG !
But, you might use a small cap to short out HF out of the external feedbackloop ?
Mike
My guess is, lightspeed is way too low to get this working good...
Simple calculations show that a 3meter cable (6 meter because of forth and back) create a delay of 20ns. Using a resonancefrequency of 7mhz (high BW amplifier) you introduce a phasehift of 50.4°. Assuming a lightspeed of 0.5c in copper you get 100.8° phasehift... BANG !
But, you might use a small cap to short out HF out of the external feedbackloop ?
Mike
Hello,
This is my first post here, so bear with me in case I press any of the wrong buttons!
I was drawn to the title of this thread as it reminded me of the Kenwood/Trio 'Sigma Drive' system of the early 80's, as already menioned by a couple of other folk here. Being sufficiently intrigued I'd thought I'd have a go and mod my power amp (a Mike Creek Mosfet jobby, aka Hitachi app note, aka Maplin PA).
Given the limited state of my electronics know-how (both then and now) it can't have been too difficult to do, though I can now only remember a few things about its implementation: the existing (primary) feedback loop stayed more or less intact within the amp itself so nothing blew-up if you forgot to attach the extra sense wires. I remember using bog-standard screened cable for the two sense wires which attached to the (externally mounted) crossover brick via a phono connection.
How did it sound? Er, this is where my memory completely fails me. This could indicate that it just sounded different rather than better, or more likely that everything upstream and downstream of the PA wasn't particularly up-to-scratch in those days - I vagely recall using a Connoiseur BD1 at the time (the PA remains in use to this day, however!).
Of possibly much more use is that I've just managed to dig out the August 81 HFP which has an article all about this system and which I would have undoubtedly used for my efforts. If folks are interested I can scan it in and post up here somewhere.
cheers,
Tony
This is my first post here, so bear with me in case I press any of the wrong buttons!
I was drawn to the title of this thread as it reminded me of the Kenwood/Trio 'Sigma Drive' system of the early 80's, as already menioned by a couple of other folk here. Being sufficiently intrigued I'd thought I'd have a go and mod my power amp (a Mike Creek Mosfet jobby, aka Hitachi app note, aka Maplin PA).
Given the limited state of my electronics know-how (both then and now) it can't have been too difficult to do, though I can now only remember a few things about its implementation: the existing (primary) feedback loop stayed more or less intact within the amp itself so nothing blew-up if you forgot to attach the extra sense wires. I remember using bog-standard screened cable for the two sense wires which attached to the (externally mounted) crossover brick via a phono connection.
How did it sound? Er, this is where my memory completely fails me. This could indicate that it just sounded different rather than better, or more likely that everything upstream and downstream of the PA wasn't particularly up-to-scratch in those days - I vagely recall using a Connoiseur BD1 at the time (the PA remains in use to this day, however!).
Of possibly much more use is that I've just managed to dig out the August 81 HFP which has an article all about this system and which I would have undoubtedly used for my efforts. If folks are interested I can scan it in and post up here somewhere.
cheers,
Tony
jneutron said:
I see loads of SMPS with sense wires to give superb regulation over long cable lengths, but many times I've seen 100ohm (approx) safety resistors totally burnt out because the main cable got disconnected but the sense wires didn't. All the O/P volts go through the sense resistors which can make for an enormous dissipation in a puny 1/4w resistor. Not so critical for an audio amp as it spends much of it's time at low O/P power, but still something to bear in mind at the design stage.
bi-wiring
How about adapting the idea to a bi-wired system and including only one pair (LF or HF) of cable in the feedback ? ) If I understand correctly, higher frequencies can cause trouble, or am I wrong ?
I'm not sure about the benefits of the approach, what is actually supposed to be improved, bass or treble ?
Cheers
How about adapting the idea to a bi-wired system and including only one pair (LF or HF) of cable in the feedback ? ) If I understand correctly, higher frequencies can cause trouble, or am I wrong ?
I'm not sure about the benefits of the approach, what is actually supposed to be improved, bass or treble ?
Cheers
I would also like to address the question from the point of whether there is a problem here at all. Measuring the resistance of all but bell-wire (and unless you are using tens of meters), the "loss" is of the order of 0,1 dB or less. I also hope nobody is suggesting that inductance and capacitance play a role. Those parameters only start to count upwards of several 100 KHz to 1 Mhz. Neither does damping factor come into it.
I miss one point, and that is that sensing in this way introduces loudspeaker current feedback; the same action that is used to achieve motional feedback. Uncontrolled, i.e. by just sensing at the loudspeaker side without more ado, can be detrimental instead of advantageous, unless the cable resistance is very low - in which case it does not matter in the first place.
For the possibility of added problems and cost I really see no use for this kind of thing.
Regards.
I miss one point, and that is that sensing in this way introduces loudspeaker current feedback; the same action that is used to achieve motional feedback. Uncontrolled, i.e. by just sensing at the loudspeaker side without more ado, can be detrimental instead of advantageous, unless the cable resistance is very low - in which case it does not matter in the first place.
For the possibility of added problems and cost I really see no use for this kind of thing.
Regards.
You don't need to be an historian to know that a main driver of hi-fi progress has been the extension of the feedback loop. The loudspeaker is the last great frontier of feedback. Yet for all the struggle for better loudspeakers, it would seem the obvious place for R&D and a major gap.
Motional feedback, mentioned a few times previously, compares voice coil back-EMF and makes it true. This is meaningful in sealed boxes but not in resonant boxes.
The Kenwood Sigma Drive is kind of a light-weight version. Like some precision bench power supplies, it senses the voltage at the proper point of reference, the speaker box terminals. Ordinarily, even the small resistance of speaker wires shreads the damping factor. That is why those old Kenwood amps have DF like 300!
Yes, the concept of motional feedback goes back a long ways (JASA 1957, Werner) and negative output impedance back to early 1940's. While a few manufacturers have tried it, it tends to be too unstable for broad commercial use and always requires close integration of amp and speaker. But making it yourself leads to the most amazing pulse behaviour from your loudspeaker that you can barely imagine and fabulous tight bass. The Kenwood may do just a bit of this.
I hope this wraps together a number of concepts floating about in this thread.
Footnote: feedback using an accelerometer or mic? If you think getting the loudspeaker voice coil back-EMF into the loop is challenging, totally nutsy to think of using a 50-cent accelerometer (like Philips) to compare to your DC-to-light electrical amp signal.
Motional feedback, mentioned a few times previously, compares voice coil back-EMF and makes it true. This is meaningful in sealed boxes but not in resonant boxes.
The Kenwood Sigma Drive is kind of a light-weight version. Like some precision bench power supplies, it senses the voltage at the proper point of reference, the speaker box terminals. Ordinarily, even the small resistance of speaker wires shreads the damping factor. That is why those old Kenwood amps have DF like 300!
Yes, the concept of motional feedback goes back a long ways (JASA 1957, Werner) and negative output impedance back to early 1940's. While a few manufacturers have tried it, it tends to be too unstable for broad commercial use and always requires close integration of amp and speaker. But making it yourself leads to the most amazing pulse behaviour from your loudspeaker that you can barely imagine and fabulous tight bass. The Kenwood may do just a bit of this.
I hope this wraps together a number of concepts floating about in this thread.
Footnote: feedback using an accelerometer or mic? If you think getting the loudspeaker voice coil back-EMF into the loop is challenging, totally nutsy to think of using a 50-cent accelerometer (like Philips) to compare to your DC-to-light electrical amp signal.
Well almost, Ben ... But it is perhaps good to have further opinions about this topic.
Yes; well done there. But there is another basic concept that needs inclusion. This is about ...
Not the right 'season' to rain on the parade, but (as has been done in the past), damping factor needs to be re-defined as a meaningful parameter. I am afraid 'loudspeaker impedance/amp + cable impedance' is meaningless, especially at higher so-called DFs.
The meaningful 'braking impedance' is the total resistance (keeping it simple) in the circuit. As we know, this is the resistance that limits the 'braking current' and thus the braking force on the loudspeaker cone proper. The basic equivalent circuit includes amp. output resistance, cable resistance and loudspeaker voice coil resistance - the latter does not sudddenly become zero! For an 8 ohm driver that is usually about 5,6 ohm. Thus the real-life damping factor seen by the loudspeaker can never be more than 1,6. Numerous tests have confirmed the negligible effect of so-called DFs above say 15 - 20.
This is the basis of the argument that loudspeaker cable resistance under normal domestic conditions makes no difference (one is not talking about bell-wire), thus the earlier opinions that sensing at the loudspeaker terminals instead of the amplifier output is 'much ado about nothing'. (Only 'active' positive feedback can negate the driver wire resistance - that has been mentioned, with its accompanying pitfalls.)
Unless one has dealt with this basic fact, the rest of the thread bears little resemblance with reality.
Thank you, Johan; all right and nicely expressed too, except as noted below.
Actually, that old Sigma Drive amp claims a DF of 1000. I think that might be possible given the standard of measurement (resistor not loudspeaker, etc.).
And you are certainly right, silly to be that high in light of voice coil resistance. I hope in my earlier post I didn't seem to advocate high DFs.
As far as the Basic M1 amp, the question is still whether it puts the feedback into a bridge (which gets unbalanced by speaker "errors") or not. I have the schematic but not sure I have the mental where-with-all to judge. My intuitive impression, however, is that sensing at the end of the speaker leads will result in certain corrective feedback relative to speaker motion.
Meaning in plainer language, a bit of motional feedback, as earlier suggested. What do you think?
Few people alive today care much about that old amp. But the issue of extending the feedback loop around the speaker couldn't be more relevant today as every other place for errors in reprodution (except recording methods) moves beyond human scale detection.
Actually, that old Sigma Drive amp claims a DF of 1000. I think that might be possible given the standard of measurement (resistor not loudspeaker, etc.).
And you are certainly right, silly to be that high in light of voice coil resistance. I hope in my earlier post I didn't seem to advocate high DFs.
As far as the Basic M1 amp, the question is still whether it puts the feedback into a bridge (which gets unbalanced by speaker "errors") or not. I have the schematic but not sure I have the mental where-with-all to judge. My intuitive impression, however, is that sensing at the end of the speaker leads will result in certain corrective feedback relative to speaker motion.
Meaning in plainer language, a bit of motional feedback, as earlier suggested. What do you think?
Few people alive today care much about that old amp. But the issue of extending the feedback loop around the speaker couldn't be more relevant today as every other place for errors in reprodution (except recording methods) moves beyond human scale detection.
negative impedance at the output is possible with only local feedback - you could tune the negative impedance to cancel expected wiring impedance over some frequency range - or even cancel some voice coil Rser to change Theil-Small parameters
I have closed negative feedback loops over 50' of cable with ~200 KHz gain intercept
but there doesn't seem any real reason to do either for domestic loudspeakers
I have closed negative feedback loops over 50' of cable with ~200 KHz gain intercept
but there doesn't seem any real reason to do either for domestic loudspeakers
Last edited:
Hi Bentoronto!
I am at a disadvantage here; my Googling skills do not seem to go beyond having to jump through a number of hoops in order to get to a schematic of the KA 900, whichever source I call up. Not to belabour this too much, but my interest is aroused by the (incomplete) descriptions I see of what they might have done: Sigma? bridge? You say you have a schematic; will it be possible, copyright laws allowing, to put it up here or PM it to me?
Otherwise my reply can only be general, and much as before. Yes, an amount of corrective feedback would exist depending on what circuit exactly is involved, but again, with normal cable resistance the 'improvement' would be merely academic. If this is not so, then my previous argument would fall through because cable (or otherwise) resistance would not be negligible - in contrast to the original assumption of negligible cable plus whatever other series resistance (compared to loudspeaker impedance/resistance).
[This is the problem with qualitive statements made in claims/promotion (not referring to anything you posted). As soon as one puts real practical figures in, some claims turn out to be of no consequence whatsoever.]
I am at a disadvantage here; my Googling skills do not seem to go beyond having to jump through a number of hoops in order to get to a schematic of the KA 900, whichever source I call up. Not to belabour this too much, but my interest is aroused by the (incomplete) descriptions I see of what they might have done: Sigma? bridge? You say you have a schematic; will it be possible, copyright laws allowing, to put it up here or PM it to me?
Otherwise my reply can only be general, and much as before. Yes, an amount of corrective feedback would exist depending on what circuit exactly is involved, but again, with normal cable resistance the 'improvement' would be merely academic. If this is not so, then my previous argument would fall through because cable (or otherwise) resistance would not be negligible - in contrast to the original assumption of negligible cable plus whatever other series resistance (compared to loudspeaker impedance/resistance).
[This is the problem with qualitive statements made in claims/promotion (not referring to anything you posted). As soon as one puts real practical figures in, some claims turn out to be of no consequence whatsoever.]
Follow -
I see Jcx has posted simultaneously. His post is of course true, but I suspect that that is not what Kenwood did. As Jcx said, one has to match that sort of thing - especially when there is enough to change T-S parameters - to a specific condition (loudspeaker). What he did not say (and I am sure will confirm!), is that it is not easy and if done incorrectly can have disasterous effects. I don't think we are talking of that with 'feedback-from-the- loudspeaker-terminals' in general.
I see Jcx has posted simultaneously. His post is of course true, but I suspect that that is not what Kenwood did. As Jcx said, one has to match that sort of thing - especially when there is enough to change T-S parameters - to a specific condition (loudspeaker). What he did not say (and I am sure will confirm!), is that it is not easy and if done incorrectly can have disasterous effects. I don't think we are talking of that with 'feedback-from-the- loudspeaker-terminals' in general.
Johan -
Here's the Kenwood Basic M1 schematic and service manual... from a post at DIYAudio
www.audio-circuit.dk Free Downloads
Sorry that I am not up to the engineering concepts, but I can say that with two wires coming out of the amp, you can put the speaker into the leg of a bridge and derive a feedback signal according to the error of the voice coil motion. Two other legs are adjustable resistances (for initial static balance of the bridge) and the last leg is the "generator" (the amp output signal). Hence, it is possible the Kenwood Sigma Drive is doing a bit of correcting, beyond "remote sensing" of wire losses like a precision bench power supply can do. And it is possible that that wire is functioning as one leg of the bridge if you can figure it out.
About jcx who says, "but there doesn't seem any real reason to do either for domestic loudspeakers" - ho, ho, ho. Today's speakers have specs which would make a 1957 single-sided triode amp blush. Other than vaguely linear mechanical restoring forces and the power of prayer, there ain't no feedback. Big waste of effort (and money) trying to make mechanically perfect drivers when a little feedback is what you need.
As someone earlier said, if you push on a motional feedback cone with your hand, it pushes back. You have to see it to believe it.
Footnote: a second voice coil winding can be used for feedback instead of a single voice coil in a bridge. Not sure the relative merits of each approach.
Here's the Kenwood Basic M1 schematic and service manual... from a post at DIYAudio
www.audio-circuit.dk Free Downloads
Sorry that I am not up to the engineering concepts, but I can say that with two wires coming out of the amp, you can put the speaker into the leg of a bridge and derive a feedback signal according to the error of the voice coil motion. Two other legs are adjustable resistances (for initial static balance of the bridge) and the last leg is the "generator" (the amp output signal). Hence, it is possible the Kenwood Sigma Drive is doing a bit of correcting, beyond "remote sensing" of wire losses like a precision bench power supply can do. And it is possible that that wire is functioning as one leg of the bridge if you can figure it out.
About jcx who says, "but there doesn't seem any real reason to do either for domestic loudspeakers" - ho, ho, ho. Today's speakers have specs which would make a 1957 single-sided triode amp blush. Other than vaguely linear mechanical restoring forces and the power of prayer, there ain't no feedback. Big waste of effort (and money) trying to make mechanically perfect drivers when a little feedback is what you need.
As someone earlier said, if you push on a motional feedback cone with your hand, it pushes back. You have to see it to believe it.
Footnote: a second voice coil winding can be used for feedback instead of a single voice coil in a bridge. Not sure the relative merits of each approach.
Last edited:
Indeed, Eva!
Seems like one might want to use acoustic feedback, rather than something inside the speaker, and try to correct for everything at once. But by the time the sensor can hear the output, it's probably too late to affect it (or that part of it, anyway), although a "learning" or adaptive type of feedback control system could still be extremely valuable, with that type of sensor.
Sensing speaker cone or actuator motion should be do-able, for nearly-instantaneous feedback that could be used to shape the output before it's output, using classical feedback control system theory. If an existing electrical signal couldn't be used directly, then (just off the top of my head) a close-range sonar, optical, or radar speaker displacement position-sensing system should be able to be made fast-enough to give feedback at many times the highest audio frequency, and could probably even be made to automatically correct for any relative motion/vibration induced between its mounting or sensing point and the speaker mounting plane, if things couldn't be made rigid-enough.
I also wouldn't discount the possibility that a MEMS-type accelerometer chip could be used. Analog Devices (Analog Devices, Inc. |*Converters Amplifiers Processors MEMS A/D Converters Analog to Digital Video Converter Temperature Sensors Analog Device RF Amplifiers Differential Amplifiers Digital Signal Processing Thermal Management D to A Converters Micro) already has at least one part that goes from DC to 22 kHz, although it goes to 70 g's and might not have enough sensitivity in low-g situations. But more and better ones will be developed soon.
Then again, although I love feedback control systems, maybe one could just use an "active" speaker in which the active amplifier part was nominally just a buffer/follower (possibly with gain), except it could have a digital control/correction system that could be taught the characteristics of the driver/speaker (and the enclosure, and its own amplifier, and the room), possibly using something like the displacement and acoustic feedback systems alluded-to above, once, and automatically correct for all of them forever after. (Of course, without feedback, it would probably have to be re-trained, periodically, as the system aged, or when other things were changed, and would lack other real-time benefits of feedback.)
Cheers,
Tom
Seems like one might want to use acoustic feedback, rather than something inside the speaker, and try to correct for everything at once. But by the time the sensor can hear the output, it's probably too late to affect it (or that part of it, anyway), although a "learning" or adaptive type of feedback control system could still be extremely valuable, with that type of sensor.
Sensing speaker cone or actuator motion should be do-able, for nearly-instantaneous feedback that could be used to shape the output before it's output, using classical feedback control system theory. If an existing electrical signal couldn't be used directly, then (just off the top of my head) a close-range sonar, optical, or radar speaker displacement position-sensing system should be able to be made fast-enough to give feedback at many times the highest audio frequency, and could probably even be made to automatically correct for any relative motion/vibration induced between its mounting or sensing point and the speaker mounting plane, if things couldn't be made rigid-enough.
I also wouldn't discount the possibility that a MEMS-type accelerometer chip could be used. Analog Devices (Analog Devices, Inc. |*Converters Amplifiers Processors MEMS A/D Converters Analog to Digital Video Converter Temperature Sensors Analog Device RF Amplifiers Differential Amplifiers Digital Signal Processing Thermal Management D to A Converters Micro) already has at least one part that goes from DC to 22 kHz, although it goes to 70 g's and might not have enough sensitivity in low-g situations. But more and better ones will be developed soon.
Then again, although I love feedback control systems, maybe one could just use an "active" speaker in which the active amplifier part was nominally just a buffer/follower (possibly with gain), except it could have a digital control/correction system that could be taught the characteristics of the driver/speaker (and the enclosure, and its own amplifier, and the room), possibly using something like the displacement and acoustic feedback systems alluded-to above, once, and automatically correct for all of them forever after. (Of course, without feedback, it would probably have to be re-trained, periodically, as the system aged, or when other things were changed, and would lack other real-time benefits of feedback.)
Cheers,
Tom
Last edited:
I'm with gootee, who seems to have his feedback theory down pat. Various ways to "track" the cone even if there are monstrous issues of phase and bandwidth to address.
On the other hand, barely possible to make much useful gross frequency room acoustics compensation (well, some small improvement is possible with a lot of fooling around with a multi-band parametric equalizer separately for each channel). So dealing with the acoustic output linearity by mounting microphones into your armchair (or some other dream-world idea) is, today, wholly impossible.
Which leads to dhaen's comment. True, motion of say, the dust cap or voice coil may not be exactly what propagates sound to my armchair. But it sure beats any other conceivable means of reference, today.
Thought Experiment: does using a secondary voice coil provide a good feedback reference source or does it simply double the errors of the first voice coil?
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.