one phenomenon i see in the pro-audio business is speaker systems that sound dry and dead with high damping factor amps (1000 and above), but sound much better with a medium (100-900)damping factor, and some systems sound boomy and loose with medium damping factor, and sound great with a high damping factor. has anyone come up with a way to create a variable damping factor in an amp? i think i have, but i need to model it first, just to see if it's even do-able.
ok, i modeled this 2 different ways, first with simply an output stage to model the effect of changing the output emitter resistors, the second using a complete amp with 1 ohm emitter resistors to see the effect of feedback.
the first model returned the results i expected, a properly biased output stage with no feedback has an output impedance equal to all of the emitter resistances paralleled, plus all of the on resistances of the output devices paralleled, plus the power supply impedance.
the second model surprised me a bit.
i stepped the Av of the amp in a 1-2-5 series from 1 to 10000, expecting the output impedance to be inversely proportional to the feedback (all i changed was the feedback resistor, from 1k to 10Meg), but it wasn't. it started out at 7 or 8 milliohms, and stayed there until i got Av up to about 100. then the impedance began to change until it was about .5 ohms at Av=10000.
i'll post the actual results tomorrow (sorry, left the data at home.....) along with the schematics of the models.
the first model returned the results i expected, a properly biased output stage with no feedback has an output impedance equal to all of the emitter resistances paralleled, plus all of the on resistances of the output devices paralleled, plus the power supply impedance.
the second model surprised me a bit.
i stepped the Av of the amp in a 1-2-5 series from 1 to 10000, expecting the output impedance to be inversely proportional to the feedback (all i changed was the feedback resistor, from 1k to 10Meg), but it wasn't. it started out at 7 or 8 milliohms, and stayed there until i got Av up to about 100. then the impedance began to change until it was about .5 ohms at Av=10000.
i'll post the actual results tomorrow (sorry, left the data at home.....) along with the schematics of the models.
i was thinking of having the damping factor set by jumper settings on the amp board, to minimize the possibility of problems due to OHS errors (Operator HeadSpace) or at least a variable internal control. the idea being to have the amp feedback variable, and have a preamp opamp with variable gain in an inverse ratio, so the overall gain from input to speaker remains the same. only problem is the amp distortion will also go up, so the open loop distortion has to be super-low to begin with. i can probably do that with a design that has a medium (in the mid thousands) open loop gain to begin with, as long as i can keep distortion to a minimum (less than 1% open loop). the other method would require rather large (physically) resistors and would add a small amount of power loss, but would definitely change the damping factor to one or more preset ratios.
Unclejed613,
Very simple. Take an amplifier with the highest damping factor desired, and simply start adding resistance until you get the lowest. An external series resistance will have exactly the same effect as any amplifier adjustments.
But as Sreten said, there is no way that damping factors above say 50 can have any audible effect. The basic laws of physics preclude that. If such effects were noticed the explanation was elsewhere. And again with all respect, I must ask about the conditions under which such was noticed (heard)/compared.
Just for clarity (although I am sure you know this), the damping factor as currently defined is misleading and nonsense as an amplifier characteristic. It is supposed to be a measure of the "braking power" exerted on a loudspeaker. That occurs because of back emf shorted out by an external short across the loudspeaker terminals. But the current thus generated flows in a circuit (circle), the resistance of which must be all-inclusive. That is, it includes the dc resistance of the loudspeaker, which is unavoidable. For an 8 ohm loudspeaker this is usually >say 5,5 ohm.
Thus, even if the amp output impedance is a dead short, as in a damping factor of a million, the actual damping factor will still only be 8/5,5 or 1,5. One can now see how rediculous the everyday definition of df is as a real factor influencing braking.
To humour the situation: A df of 30 would mean an amplifier internal impedance of 8/30 or 0,266667 ohm. That will give a "braking" df of 8/5.266667 or 1,3873. Take a df of 1000, which indicates amplifier internal impedance of 0,008 ohm. The braking df with that would be 8/5,5008 or 1,4543.
Thus the second figure would indicate an actual physical df inprovement of under 5%. And this is audible?
Apology for the lengthy epistle, but qualitive arguments would not illustrate the effect so vividly. It serves to indicate that such real differences are simply negligible. As said, not calling anybody a liar, but there are just two many other factors to merit a variable df.
Regards.
Very simple. Take an amplifier with the highest damping factor desired, and simply start adding resistance until you get the lowest. An external series resistance will have exactly the same effect as any amplifier adjustments.
But as Sreten said, there is no way that damping factors above say 50 can have any audible effect. The basic laws of physics preclude that. If such effects were noticed the explanation was elsewhere. And again with all respect, I must ask about the conditions under which such was noticed (heard)/compared.
Just for clarity (although I am sure you know this), the damping factor as currently defined is misleading and nonsense as an amplifier characteristic. It is supposed to be a measure of the "braking power" exerted on a loudspeaker. That occurs because of back emf shorted out by an external short across the loudspeaker terminals. But the current thus generated flows in a circuit (circle), the resistance of which must be all-inclusive. That is, it includes the dc resistance of the loudspeaker, which is unavoidable. For an 8 ohm loudspeaker this is usually >say 5,5 ohm.
Thus, even if the amp output impedance is a dead short, as in a damping factor of a million, the actual damping factor will still only be 8/5,5 or 1,5. One can now see how rediculous the everyday definition of df is as a real factor influencing braking.
To humour the situation: A df of 30 would mean an amplifier internal impedance of 8/30 or 0,266667 ohm. That will give a "braking" df of 8/5.266667 or 1,3873. Take a df of 1000, which indicates amplifier internal impedance of 0,008 ohm. The braking df with that would be 8/5,5008 or 1,4543.
Thus the second figure would indicate an actual physical df inprovement of under 5%. And this is audible?
Apology for the lengthy epistle, but qualitive arguments would not illustrate the effect so vividly. It serves to indicate that such real differences are simply negligible. As said, not calling anybody a liar, but there are just two many other factors to merit a variable df.
Regards.
sreten said:
Hi,
To be more correct they had variable gain via varying the amount
of feedback. The lower the gain the lower the output impedance.
/sreten.
Hi Sreten
Not so. They use a mix of voltage feedback and current feedback .
The current feedback come from a small resistor in series with the load.
Using a pot , they change the feedback to the input stage , from the normal low output impedance obtained by voltage feedback, to the high impedance of current feedback .
A classic example.
http://www.audiofanatic.it/Schemi/Tipo/Valvole/finali/pic_finali_PP/6550_HeatkitW6M_PP.jpg
Hi,
In tube amps, there was the old trick of positive current feedback with which you can get more than infinite damping factor, a negative output impedance. This is mentionned in Thiele's JAES paper about bass-reflex loading but there were number of papers about this fascinating technique before, for example by Werner Clements.
You just need three resistors to implement it in a solid state amplifier, by bootstraping the input. It can be useful if you need to obtain a lower Qt for a speaker.
In tube amps, there was the old trick of positive current feedback with which you can get more than infinite damping factor, a negative output impedance. This is mentionned in Thiele's JAES paper about bass-reflex loading but there were number of papers about this fascinating technique before, for example by Werner Clements.
You just need three resistors to implement it in a solid state amplifier, by bootstraping the input. It can be useful if you need to obtain a lower Qt for a speaker.
Ye-e-e-s,
But that method is fraught with traps. The moment you use that sort of positive feedback it becomes dependant on the load current (magnitude and phase), which is different from a simple 'independant" low df topology.
Combined positive and negative feedback in tube amplifiers were used from an early stage (early 50s, e.g. the Connoisseur amplifier), but sometimes caused havoc because of the above fact. The advantage was that it could completely cancel the loudpseaker dc resistance (that was in reality the only way to achieve high "real" df.), but the sensing network ideally had to be the converse of the loudspeaker impedance. In that sense it was best done as a committed circuit for a particular loudspeaker. A very practical application could however be to apply over a limited bandwidth, say from low up to 200 Hz. This did audible improvement for low frequency response. (It was quite exciting to put a finger to the cone of a loudspeaker driven thus - one could feel the "effort" of the cone increasing as one tried to limit its amplitude.)
Regards.
But that method is fraught with traps. The moment you use that sort of positive feedback it becomes dependant on the load current (magnitude and phase), which is different from a simple 'independant" low df topology.
Combined positive and negative feedback in tube amplifiers were used from an early stage (early 50s, e.g. the Connoisseur amplifier), but sometimes caused havoc because of the above fact. The advantage was that it could completely cancel the loudpseaker dc resistance (that was in reality the only way to achieve high "real" df.), but the sensing network ideally had to be the converse of the loudspeaker impedance. In that sense it was best done as a committed circuit for a particular loudspeaker. A very practical application could however be to apply over a limited bandwidth, say from low up to 200 Hz. This did audible improvement for low frequency response. (It was quite exciting to put a finger to the cone of a loudspeaker driven thus - one could feel the "effort" of the cone increasing as one tried to limit its amplitude.)
Regards.
actually, with negative feedback, the voltage across that 5.5r resistance gets "opposed" actively by the amplifier. with my second model, it did not matter what the open loop output impedance of the amp was. once there was enough feedback, the amp maintained a constant .01r or lower output impedance, effectively cancelling any (except for a very small, about 0.008v) applied voltage at the output terminals. even with 1r source resistors...... it's basically applied feedback theory, the amp will do WHATEVER it takes to maintain the inverting and noninverting inputs at the same voltage, even if the error is supplied to the output by source external to the amplifier, it is still treated as a feedback error, just like distortion and other errors, and cancelled out by the amp maintaining the input balance.
Unclejed613,
If I understand you correctly, yes I agree with the operation of nfb, but it cannot cancel what it cannot see. To put it better, let us think in terms of current. Let us accept that the amplifier has zero output impedance by whatever method. The loudspeaker is "braked" (damped) by the current in the circuit as a result of back emf. This current is still limited by the total resistance in the circuit. If the loudspeaker voice coil resistance is 5 ohm, that will be the current limiting element. If the voice coil resistance could be made 2,5 ohms, twice the current would flow and the damping would be better. This has nothing to do with the amplifier.
For the purpose of this argument the output device equivalent circuit is a voltage generator in series with a resistor. The amplifier cannot correct for a voltage across its output terminals that does not exist, because the output impedance is zero. It is equivalent to testing a battery for the short-circuit current, by shorting the terminals with, let as assume, a perfect short. Smaller batteries with higher internal resistance will give a lower current than low resistance batteries. As said, if I understood you correctly.
Regards.
If I understand you correctly, yes I agree with the operation of nfb, but it cannot cancel what it cannot see. To put it better, let us think in terms of current. Let us accept that the amplifier has zero output impedance by whatever method. The loudspeaker is "braked" (damped) by the current in the circuit as a result of back emf. This current is still limited by the total resistance in the circuit. If the loudspeaker voice coil resistance is 5 ohm, that will be the current limiting element. If the voice coil resistance could be made 2,5 ohms, twice the current would flow and the damping would be better. This has nothing to do with the amplifier.
For the purpose of this argument the output device equivalent circuit is a voltage generator in series with a resistor. The amplifier cannot correct for a voltage across its output terminals that does not exist, because the output impedance is zero. It is equivalent to testing a battery for the short-circuit current, by shorting the terminals with, let as assume, a perfect short. Smaller batteries with higher internal resistance will give a lower current than low resistance batteries. As said, if I understood you correctly.
Regards.
Johan Potgieter
---The advantage was that it could completely cancel the loudpseaker dc resistance (that was in reality the only way to achieve high "real" df.), but the sensing network ideally had to be the converse of the loudspeaker impedance.---
Completely removing the DC resistance would make the loudspeaker having a Q of 0, behaving like a differentiator, with an ascending slope of 6 dB/o towards the high frequencies.
A problem with a high value of negative resonance is that the inductance of the voice coil, Le, makes a resonant LC circuit with the Ces of the motionnal impedance. The inductance of the voice-coil having not a stable value, it is quite difficult to find an optimal value of negative impedance which will nullify it.
Short simulations have shown me that for a negative resistance equal to half of the DC resistance of the voice coil, Re, there is nothing needing to be compensated for, and that Le should be compensated for values of negative resistances higher than 2/3 Re.
As these values are easy to implement, I find that negative resistance is a useful tool to divide the Qe of a driver by 2 or 3 to suit particular needs.
---The advantage was that it could completely cancel the loudpseaker dc resistance (that was in reality the only way to achieve high "real" df.), but the sensing network ideally had to be the converse of the loudspeaker impedance.---
Completely removing the DC resistance would make the loudspeaker having a Q of 0, behaving like a differentiator, with an ascending slope of 6 dB/o towards the high frequencies.
A problem with a high value of negative resonance is that the inductance of the voice coil, Le, makes a resonant LC circuit with the Ces of the motionnal impedance. The inductance of the voice-coil having not a stable value, it is quite difficult to find an optimal value of negative impedance which will nullify it.
Short simulations have shown me that for a negative resistance equal to half of the DC resistance of the voice coil, Re, there is nothing needing to be compensated for, and that Le should be compensated for values of negative resistances higher than 2/3 Re.
As these values are easy to implement, I find that negative resistance is a useful tool to divide the Qe of a driver by 2 or 3 to suit particular needs.
Indeed, forr
The "cancel the dc resistance" was over-simplification, more to convey a concept. That is why I included that reality would have to include loudspeaker impedance. Thanks for the other information. My own experience with this was years ago, and I would not have been able to supply the information you just did.
Regards
The "cancel the dc resistance" was over-simplification, more to convey a concept. That is why I included that reality would have to include loudspeaker impedance. Thanks for the other information. My own experience with this was years ago, and I would not have been able to supply the information you just did.
Regards
i am not looking for a NIC. too much opportunity to "let the smoke out". just a way of controlling damping factor without a lot of difficulty. i think the best way to go here will probably be adding resistance between amp and speakers, since controlling it electronically will introduce distortion. i thought of using a MOSFET controlled by DC bias, but that could be a source of distortion as well.
variable output resistance
Such amplifiers already are - http://sakevich.ru/eng_german/eng1200/1200eng.htm
"Reference power amplifier with the variable output resistance SK1200 Studio3"
Such amplifiers already are - http://sakevich.ru/eng_german/eng1200/1200eng.htm
"Reference power amplifier with the variable output resistance SK1200 Studio3"
- Status
- Not open for further replies.