phase_accurate said:
OK Charles,
I didn't mean to sound like I was suggesting that these amps sound good because of their damping factor. I was just bringing up the Crown's as an example of an amplifier with VERY high damping factor. There are many other things that go in to these amps that make them sound good. They have many paralleled output devices, overbuilt power supplies, excellent heat dissapation, etc..... I'm sorry for the misunderstanding.
Cheers,
Zach
May I argue adainst voltage drive?
To PRR
> with respect to distortion and compression in the loudspeaker
> Distortion will be unchanged. Voltage, current, same thing just factored by impedance.
Linkwitz wrote:”I have observed that the distortion products in the acoustic output spectrum of a driver are highly correlated with the distortion components of its voice coil current waveform, when driven from a voltage source. Using feedback it should be possible to linearize the voice coil current and thereby reduce distortion to some degree”.
I also find less IMD at least for mid bass and bass drivers at the frequencies higher than 1 kHz. You can see how Z (inductance) varies with cone displacement. With voltage drive we get the correspondent modulation of current, thus higher IMD as comparing with current drive.
To Marcel
>What all of this adds up to, is that with respect to distortion and compression in the loudspeaker, it may not be such a bad idea to equalise the low-frequency part of the loudspeaker's impedance characteristic with one or two parallel LRC tanks and then current drive the whole thing.
Hi, Marcel
The better idea is to use current drive and adjustable DSP filter before the amp to equalize frequency responce, and the best is to use tube amp with output impedance of several Ohm.
Then we get not much boost at lf resonance but the IMD improvement at mids.
To PRR
> with respect to distortion and compression in the loudspeaker
> Distortion will be unchanged. Voltage, current, same thing just factored by impedance.
Linkwitz wrote:”I have observed that the distortion products in the acoustic output spectrum of a driver are highly correlated with the distortion components of its voice coil current waveform, when driven from a voltage source. Using feedback it should be possible to linearize the voice coil current and thereby reduce distortion to some degree”.
I also find less IMD at least for mid bass and bass drivers at the frequencies higher than 1 kHz. You can see how Z (inductance) varies with cone displacement. With voltage drive we get the correspondent modulation of current, thus higher IMD as comparing with current drive.
To Marcel
>What all of this adds up to, is that with respect to distortion and compression in the loudspeaker, it may not be such a bad idea to equalise the low-frequency part of the loudspeaker's impedance characteristic with one or two parallel LRC tanks and then current drive the whole thing.
Hi, Marcel
The better idea is to use current drive and adjustable DSP filter before the amp to equalize frequency responce, and the best is to use tube amp with output impedance of several Ohm.
Then we get not much boost at lf resonance but the IMD improvement at mids.Attachments
From the replies I read, I think I begin to understand what is the point of damping factor. My conclusion is that damping factor is how an amp can control the movement of the speakers. How independent it is. The more damping factor, the power amp is more independent to however the speaker reacts. This amp will only follow the input signal, and the speakers reaction is not important to how the signal will be generated. If this is true, then the simple formula of Zload/Zinternal is not 100% correct. If that formula is correct, you can include cable resistance, connector resistance, etc, that will lower the damping factor.
But if the essence of damping factor is my conclusion above, the damping factor is more independent to any external resistance. It is just how independent the power amplifier to the speakers, no matter how long the cables you use (in rational length). You don't need to put only 20cm cable to reduce DF loss.
Again, if my conclusion is correct, any power amp that have smaller gain (like 10x) will have bigger DF than power amp that have bigger gain (like100x). It is because smaller gain amp will have more control over it's output than the amp with bigger gain, since the sensitivity of a differential transistor is the same value, no matter what closed gain figure we use.
But if the essence of damping factor is my conclusion above, the damping factor is more independent to any external resistance. It is just how independent the power amplifier to the speakers, no matter how long the cables you use (in rational length). You don't need to put only 20cm cable to reduce DF loss.
Again, if my conclusion is correct, any power amp that have smaller gain (like 10x) will have bigger DF than power amp that have bigger gain (like100x). It is because smaller gain amp will have more control over it's output than the amp with bigger gain, since the sensitivity of a differential transistor is the same value, no matter what closed gain figure we use.
Quote from PRR:
"Distortion will be unchanged. Voltage, current, same thing just factored by impedance."
This would only be true if the loudspeaker had a perfectly linear relationship between voltage and current, which it has not, for the reasons explained in one of my previous posts and further clarified by Dimitri.
"Distortion will be unchanged. Voltage, current, same thing just factored by impedance."
This would only be true if the loudspeaker had a perfectly linear relationship between voltage and current, which it has not, for the reasons explained in one of my previous posts and further clarified by Dimitri.
to lumanaum
>If this is true, then the simple formula of Zload/Zinternal is not 100% correct.
The formula is correct. Damping factor is measured at the _amplifier output clamps_ Zload/Zout. Loudspeaker sees Zout+Zcable on his clamps. There is no good to do DF less then 50-100, as it is hard to make Zcable less than 0.1-0.2 Ohm.
DF was invented as the next marketing figure when THD comes to it limit (0.001-0.003%) and manufacturers should continue to compete with each other.
The main problem of DF is that it is measured at only one frequency and only one output current value. If the frequency response of DF, individual harmonic content of DF, how DF varies with output current is taken into account you will got a lot of additional info.
>any power amp that have smaller gain (like 10x) will have bigger DF than power amp that have bigger gain (like100x)
If you consider feedback amp, its output resistance and gain will decrease with increase in loop gain. Your statement is true for given amp with feedback, but _not_ the basis to compare different amps
>If this is true, then the simple formula of Zload/Zinternal is not 100% correct.
The formula is correct. Damping factor is measured at the _amplifier output clamps_ Zload/Zout. Loudspeaker sees Zout+Zcable on his clamps. There is no good to do DF less then 50-100, as it is hard to make Zcable less than 0.1-0.2 Ohm.
DF was invented as the next marketing figure when THD comes to it limit (0.001-0.003%) and manufacturers should continue to compete with each other.
The main problem of DF is that it is measured at only one frequency and only one output current value. If the frequency response of DF, individual harmonic content of DF, how DF varies with output current is taken into account you will got a lot of additional info.
>any power amp that have smaller gain (like 10x) will have bigger DF than power amp that have bigger gain (like100x)
If you consider feedback amp, its output resistance and gain will decrease with increase in loop gain. Your statement is true for given amp with feedback, but _not_ the basis to compare different amps
Measuring DF
Hi PRR,
Although there is no need to argue about which way is best to measure DF, I would like to point out that the current injection method I described is not 'more friendly' to amplifiers than the method of measuring output voltage in loaded and unloaded condition.
It is true that the output voltage of the amplifier under test is fixed at ground level and will not move, but this is not important. What is important is that current is forced to go in and out of the amplifier. In that way it travels along all load conditions, as if it was driven full swing.
A proper class AB push pull amplifier (with emitter follower outputs) that has about 25mV over its emitter resistors will have more or less the same output impedance (and DF) for no signal as for large signals. If the amplifier is over-biased, then it will have a higher DF at zero voltage, just like it will have a lower DF at zero output if under-biased. But the charm of pushing current into and out of the amplifier by a second amplifier (that can be any lousy amplifer; quality is not important here, as long as it is stable) is that the amplifier under test is deliberately forced away of its idle condition. The current ratio of the top and bottom half of the push pull stage is modulated. So, although the output is stable w.r.t. voltage it is modulated in the same way as if it was driving a load resistor itself.
Steven
PRR said:
Hi PRR,
Although there is no need to argue about which way is best to measure DF, I would like to point out that the current injection method I described is not 'more friendly' to amplifiers than the method of measuring output voltage in loaded and unloaded condition.
It is true that the output voltage of the amplifier under test is fixed at ground level and will not move, but this is not important. What is important is that current is forced to go in and out of the amplifier. In that way it travels along all load conditions, as if it was driven full swing.
A proper class AB push pull amplifier (with emitter follower outputs) that has about 25mV over its emitter resistors will have more or less the same output impedance (and DF) for no signal as for large signals. If the amplifier is over-biased, then it will have a higher DF at zero voltage, just like it will have a lower DF at zero output if under-biased. But the charm of pushing current into and out of the amplifier by a second amplifier (that can be any lousy amplifer; quality is not important here, as long as it is stable) is that the amplifier under test is deliberately forced away of its idle condition. The current ratio of the top and bottom half of the push pull stage is modulated. So, although the output is stable w.r.t. voltage it is modulated in the same way as if it was driving a load resistor itself.
Steven
> although the output is stable w.r.t. voltage it is modulated in the same way as if it was driving a load resistor itself.
Yes, I wrote without enough coffee in me.
The current swings. That's usually the biggest problem.
Voltage swing: on modern designs that's not an issue. But I grew up on designs where the stage before the emitter follower was resistor loaded. The gain was very different when the output was +20V or -20V.
I suppose nobody builds them like that any more.
> about 25mV over its emitter resistors
Last night I showed that this is half-right. I am not convinced of my logic and models so I have to ponder some more. But here is what I am thinking:
In a push-pull emitter follower, distortion and Zout flatness (two sides of the same coin) have two minimums. One is "about" 28mV across the emitter resistors. But that does not say what the resistor value or idle current is. And this turns out to be a very narrow minimum: a few mV less and distortion gets gross, a few more mV higher and it gets bad again. With Silicon drifting at 2mV/°C and lots of heat flowing around a power amp, I suspect it is exceptionally difficult to keep an amp on that 28mV minimum. It is only about +/-5mV wide. And any driver impedance shifts the optimum significantly (maybe that is why there is no general agreement, and 15mV is often a good value; it allows for some driver impedance).
The other "minimum" is HIGH current. It actually degrades a hair when you get into Class A: the two sides fight each other. But in the range far above the usual 28mV but below Class A, there is a broad stable minimum.
And the free variable so far is the resistor value. You can get 28mV at any current, by changing the resistor. But lower IS better for distortion in any case, IF bias is either 10-30mV or over 100mV. The low value is very critical and difficult to hold in real life. The high value is nice and stable.
I'm thinking the practical optimum is to set the bias current as high as you can stand, then set the resistors to drop 100-200mV at that current. This is for "hi-fi", not for lowest-dissipation, and will generally run warmer than any mass-market AB design, though cooler than a true Class A design.
BTW: both minimums are obscured by mis-match between NPN and PNP. There are no truly complementary pairs when looking at this scale. To see what was really happening I had to hack-up a SPICE model of a PNP with values from an NPN, to get a "perfect" matched pair.
Yes, I wrote without enough coffee in me.
The current swings. That's usually the biggest problem.
Voltage swing: on modern designs that's not an issue. But I grew up on designs where the stage before the emitter follower was resistor loaded. The gain was very different when the output was +20V or -20V.
I suppose nobody builds them like that any more.
> about 25mV over its emitter resistors
Last night I showed that this is half-right. I am not convinced of my logic and models so I have to ponder some more. But here is what I am thinking:
In a push-pull emitter follower, distortion and Zout flatness (two sides of the same coin) have two minimums. One is "about" 28mV across the emitter resistors. But that does not say what the resistor value or idle current is. And this turns out to be a very narrow minimum: a few mV less and distortion gets gross, a few more mV higher and it gets bad again. With Silicon drifting at 2mV/°C and lots of heat flowing around a power amp, I suspect it is exceptionally difficult to keep an amp on that 28mV minimum. It is only about +/-5mV wide. And any driver impedance shifts the optimum significantly (maybe that is why there is no general agreement, and 15mV is often a good value; it allows for some driver impedance).
The other "minimum" is HIGH current. It actually degrades a hair when you get into Class A: the two sides fight each other. But in the range far above the usual 28mV but below Class A, there is a broad stable minimum.
And the free variable so far is the resistor value. You can get 28mV at any current, by changing the resistor. But lower IS better for distortion in any case, IF bias is either 10-30mV or over 100mV. The low value is very critical and difficult to hold in real life. The high value is nice and stable.
I'm thinking the practical optimum is to set the bias current as high as you can stand, then set the resistors to drop 100-200mV at that current. This is for "hi-fi", not for lowest-dissipation, and will generally run warmer than any mass-market AB design, though cooler than a true Class A design.
BTW: both minimums are obscured by mis-match between NPN and PNP. There are no truly complementary pairs when looking at this scale. To see what was really happening I had to hack-up a SPICE model of a PNP with values from an NPN, to get a "perfect" matched pair.
PRR,PRR said:
I agree that 20-30mV across Re is a narrow optimum that is difficult to keep stable in practice. Turning up the idle current at least moves the point where Gm is halved away. So for small signals Zout is quite stable. But, I'm a little surprised that your simulations shows that it degrades when you get into class A. I would assume that as long as you are in class A without one of the transistors starving for current the output impedance would be mainly determined by Re//Re, which is very stable.
Actually, I don't understand what you mean with 'below Class A'; as long as the amplifier is idling, we are in the class A region.
Steven
PS I'm currently busy with simulations for the 'Spreading the heat in Class A' case. I changed the circuit such that I don't have jumps on the collector of Q1 anymore that invalidates the Class A benefits. I will post the answer this weekend in that thread.
> as long as the amplifier is idling, we are in the class A region.
Point taken about idling (though like the tree in the forest, can we talk of Class when there is no signal?)
I mean: idling at a very high current that allows full-power Class A operation.
Let me check some more things. Unless my pants catch fire, I probably won't post over the weekend (56K modem, new definition of "poverty").
Point taken about idling (though like the tree in the forest, can we talk of Class when there is no signal?)
I mean: idling at a very high current that allows full-power Class A operation.
Let me check some more things. Unless my pants catch fire, I probably won't post over the weekend (56K modem, new definition of "poverty").
I've been away a bit... I don't know if I agree with the following, PRR:
Acoustic output is not a function of cone velocity per se. It is a function of cone displacement which is proportional to net force, which is determined by coil current, which results in a net acceleration a=F/m. At no time does the oscillating diaphragm possess velocity in the constant sense. Unless you mean angular velocity...
And impedance characteristics varying as they do wrt frequency, I repeat my perception that current drive of a speaker appears more natural/logical...
The relationship between applied voltage and flowing current is not linear in a speaker. This being so, it is *only* the current in the coil which generates flux and thence force. The voltage which caused the current is irrelevant.
Acoustic output is not a function of cone velocity per se. It is a function of cone displacement which is proportional to net force, which is determined by coil current, which results in a net acceleration a=F/m. At no time does the oscillating diaphragm possess velocity in the constant sense. Unless you mean angular velocity...
And impedance characteristics varying as they do wrt frequency, I repeat my perception that current drive of a speaker appears more natural/logical...
DrG said:
agreed. But why is that there is few current drive amps out on the market?
I remember that doing similar things (using a current sample resistor in serial to the speaker coil as feedback) in tube days but it seems to have died out (I tried it and other than tighter bass it did nothing audiable).
I guess is that the differernce between a current vs. voltage drive amp is minimal, at best.
I don't know...
Perhaps. But I'm not sure and I'd like to hear from the brains trust out there what the concensus is. Anybody tried it?
I saw a little amp some years back incorporating the speaker into the feedback loop, with a small 0R22 shunt resistor. What are the drawbacks of such an arrangement?
Hi
See some exemples of current amplifiers (transconductance):
http://www.ultrasound-hifi.com/Us_wh_1/main_frame_en.html
Cheers
See some exemples of current amplifiers (transconductance):
http://www.ultrasound-hifi.com/Us_wh_1/main_frame_en.html
Cheers
The answer is "why not!" Seriously though, my personal approach to everything audio is to try to be as unconventional as possible. I figure the chances of creating something really special are greater when choosing the road less travelled by. And admittedly there is also an ego thing... I did it MY way etc. Each to his own.
Thanks for the link Tube_Dude
>I saw an amp incorporating the speaker into the feedback loop.<
The remote feedback system was invented by Shin Nakagawa, and was subsequently used in consumer audio amplifiers from the likes of Kenwood and Toshiba.
>What are the drawbacks of such an arrangement?<
The feedback loop will now include substantially more reactive elements than if it had been kept within the amplifier only, which can lead to a variety of stability and recovery issues. The phase compensation for a remote feedback amplifier will need to be fairly heavy-handed to keep the amplifier under control.
Good approach for a self-powered subwoofer, but if you want a good-sounding full-range power amplifier, I think that there are better ways than the remote feedback approach.
>my personal approach to everything audio is to try to be as unconventional as possible.<
Unless it makes you fee better about yourself, I don't think that it means all that much to be unconventional or conventional. Rather, I suggest that you do whatever is appropriate for the task at hand. If there is a specific engineering or performance goal that you are aiming for, and if you cook up an original and unusual method to achieve that goal, fine.
But if not, I don't see the point in being different just to be different.
regards, jonathan carr
The remote feedback system was invented by Shin Nakagawa, and was subsequently used in consumer audio amplifiers from the likes of Kenwood and Toshiba.
>What are the drawbacks of such an arrangement?<
The feedback loop will now include substantially more reactive elements than if it had been kept within the amplifier only, which can lead to a variety of stability and recovery issues. The phase compensation for a remote feedback amplifier will need to be fairly heavy-handed to keep the amplifier under control.
Good approach for a self-powered subwoofer, but if you want a good-sounding full-range power amplifier, I think that there are better ways than the remote feedback approach.
>my personal approach to everything audio is to try to be as unconventional as possible.<
Unless it makes you fee better about yourself, I don't think that it means all that much to be unconventional or conventional. Rather, I suggest that you do whatever is appropriate for the task at hand. If there is a specific engineering or performance goal that you are aiming for, and if you cook up an original and unusual method to achieve that goal, fine.
But if not, I don't see the point in being different just to be different.
regards, jonathan carr
You needn't. Each to his own, like I said. And you're right, I do feel better about myself when I achieve something via an alternate route. "It works well and I did it my way" does more for my ego that "It works well because I copied someone else's established method." But that's just me.
Jonathan, I admit a problem-oriented approach using the best circuit element for each situation is probably best from an engineering approach. But I'm not a commercial audio engineer.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.