I would like to present the circuit below just as an idea. I have not build it, and it may take a while before I do, if ever. I am actually more interested in discrete circuits.
But, browsing the Chip Amps forum, I wondered if someone had ever put these LM3875's and LM3886's in class A. After a search I did not find such a thing, although some people connected a 1.5k resistor from the output to the negative supply, with the intention to run the amp a little in class A. I think that won't work, because the current sink is far too little, about 20mA. This is close to the actual quiescent current of the chip, leaving the lower output transistor just starving for current. Not good.
I think a better way is to use two amplifiers in parallel connected at the outputs via two 0.22 Ohm resistors. One chip gets a positive offset of around 220mV, the other an offset of -220mV. The DC gain is 1x, so the offset at the non-inverting input is just copied to the output: one output at 220mV, the other at -220mV. In this way the loudspeaker output will be at 0V DC, which is OK, and an idle current of approx 1A will flow from one output into the other. The amplifier with the positive offset will have its upper (positive rail) output transistor running in class A, the other chip will have its lower (negative rail) output transistor in class A. The other output transistors are not conducting at all. Having such an idle current of 1A, a maximum peak current of 2A can be delivered to the load (16W sine in 8Ohm). Using a power supply of +/-20V each chip is dissipating some 20W idle. This is acceptable.
The offset voltages are derived from the power supply by the two transistors. Although the base voltage will change a little with an unregulated supply, the collector voltage will be quite stable, due to the action of the 33 Ohm resistor, which compensates for Vbe modulation. This gives an improvement of about 20 times less ripple on the output (collector). The 22k and 12k resistors and the capacitor bring the collector voltage of 650mV back to 220mV with low noise.
The offset voltage has a negative temperature coefficient, which is good. The amplifier idle current will be slightly lower for increasing temperatures.
The accuracy of the gain determining resistors around the amplifiers should be 0.1% or better, otherwise the the fixed offset of 440mV between the outputs will change with output signal. This 440mV is required to keep the bias current flowing.
To get an accurate 0V at the output it is possible to replace R4 and R15 with a single preset potentiometer with the wiper to ground. Then it is possible to adjust the output exactly to 0V while keeping the bias current current of 1A unaltered (the 440mV will remain stable; only the center moves). An alternative method would be to use only a 220mV bias for the first amplifier and use a servo integrator with its input to the load for the second. This integrator will adjust the offset of the second amplifier (around –220mV) such that the output will be at 0V.
The waveforms shown are the voltage output on both the amplifiers (green and yellow) and at the load (blue) for a low output voltage. Ortherwise the offset is hardly visible.
The second waveform picture shows the output currents of both amplifiers (green and yellow) and the load (blue) at maximum output for class A operation. It is clearly visible that the output current of the amplifiers never cross 0A, that is the output current is unidirectional. In this case this is a condition for class A.
The amplifiers can be any chipamp you like, as long as they act like an opamp with both inverting and non-inverting inputs available. I didn’t draw any other connections, like the ground connection or mute connection for a LM3886. The power supply voltages and bias current has been selected with the LM3886 in mind though.
Steven
But, browsing the Chip Amps forum, I wondered if someone had ever put these LM3875's and LM3886's in class A. After a search I did not find such a thing, although some people connected a 1.5k resistor from the output to the negative supply, with the intention to run the amp a little in class A. I think that won't work, because the current sink is far too little, about 20mA. This is close to the actual quiescent current of the chip, leaving the lower output transistor just starving for current. Not good.
I think a better way is to use two amplifiers in parallel connected at the outputs via two 0.22 Ohm resistors. One chip gets a positive offset of around 220mV, the other an offset of -220mV. The DC gain is 1x, so the offset at the non-inverting input is just copied to the output: one output at 220mV, the other at -220mV. In this way the loudspeaker output will be at 0V DC, which is OK, and an idle current of approx 1A will flow from one output into the other. The amplifier with the positive offset will have its upper (positive rail) output transistor running in class A, the other chip will have its lower (negative rail) output transistor in class A. The other output transistors are not conducting at all. Having such an idle current of 1A, a maximum peak current of 2A can be delivered to the load (16W sine in 8Ohm). Using a power supply of +/-20V each chip is dissipating some 20W idle. This is acceptable.
The offset voltages are derived from the power supply by the two transistors. Although the base voltage will change a little with an unregulated supply, the collector voltage will be quite stable, due to the action of the 33 Ohm resistor, which compensates for Vbe modulation. This gives an improvement of about 20 times less ripple on the output (collector). The 22k and 12k resistors and the capacitor bring the collector voltage of 650mV back to 220mV with low noise.
The offset voltage has a negative temperature coefficient, which is good. The amplifier idle current will be slightly lower for increasing temperatures.
The accuracy of the gain determining resistors around the amplifiers should be 0.1% or better, otherwise the the fixed offset of 440mV between the outputs will change with output signal. This 440mV is required to keep the bias current flowing.
To get an accurate 0V at the output it is possible to replace R4 and R15 with a single preset potentiometer with the wiper to ground. Then it is possible to adjust the output exactly to 0V while keeping the bias current current of 1A unaltered (the 440mV will remain stable; only the center moves). An alternative method would be to use only a 220mV bias for the first amplifier and use a servo integrator with its input to the load for the second. This integrator will adjust the offset of the second amplifier (around –220mV) such that the output will be at 0V.
The waveforms shown are the voltage output on both the amplifiers (green and yellow) and at the load (blue) for a low output voltage. Ortherwise the offset is hardly visible.
The second waveform picture shows the output currents of both amplifiers (green and yellow) and the load (blue) at maximum output for class A operation. It is clearly visible that the output current of the amplifiers never cross 0A, that is the output current is unidirectional. In this case this is a condition for class A.
The amplifiers can be any chipamp you like, as long as they act like an opamp with both inverting and non-inverting inputs available. I didn’t draw any other connections, like the ground connection or mute connection for a LM3886. The power supply voltages and bias current has been selected with the LM3886 in mind though.
Steven
Attachments
Hmm Steven,
Interesting idea, but what about the build-in thermal protection of the LM3886’s (the “SPIKE” protection mechanism)? It will spoil your idea IMHO unless you want not much power from them. Probably you are better off with external power transistors biased in class-A. There was once a driver IC for it, the LM391. It is not in production anymore but the datasheet is still there:
http://www.national.com/ds/LM/LM391.pdf
It is intended for class-AB but with some mods to the bias circuit it can be used for class-A as well.
Cheers
Interesting idea, but what about the build-in thermal protection of the LM3886’s (the “SPIKE” protection mechanism)? It will spoil your idea IMHO unless you want not much power from them. Probably you are better off with external power transistors biased in class-A. There was once a driver IC for it, the LM391. It is not in production anymore but the datasheet is still there:
http://www.national.com/ds/LM/LM391.pdf
It is intended for class-AB but with some mods to the bias circuit it can be used for class-A as well.
Cheers
Hi Pjotr,
The idea is just a general idea, not only for an LM3886. I am normally more interested in discrete amplifiers, but wanted to add something to the Chipamps forum. Otherwise they will never have the benefits of class A.
Nevertheless I think the SPiKe protection leaves room enough for class A opeartion. Notice that the power rails were lowered to a mere+/-20V. At 20V collector-emitter more than 2.5A collector current is allowed for a LM3886, even at Tc=75*C. AT Tc=125*C it is still slightly less than 2A. So a bias current of 1A will not be a problem.
The circuit uses two amplifiers in parallel, remember, for only 16W output.
Steven
The idea is just a general idea, not only for an LM3886. I am normally more interested in discrete amplifiers, but wanted to add something to the Chipamps forum. Otherwise they will never have the benefits of class A.
Nevertheless I think the SPiKe protection leaves room enough for class A opeartion. Notice that the power rails were lowered to a mere+/-20V. At 20V collector-emitter more than 2.5A collector current is allowed for a LM3886, even at Tc=75*C. AT Tc=125*C it is still slightly less than 2A. So a bias current of 1A will not be a problem.
The circuit uses two amplifiers in parallel, remember, for only 16W output.
Steven
Steven said:
I think it will work. I proposed something similar like this on the solid state forum. Essentially, you want to intentially introduce a DC offset to the input to "float" the output and allow it to happen only in one half of the rail to avoid cutting off of output devices -> class a.
Introducing a current source is as good of an idea on paper but in practice it is tougher to get a current source at high amperage.
Another way to configure the same thing is to use two chips, and apply the same DC offset to both chips in exactly the same fashion. Except that in the 2nd chip, you use no signal (or if you want, use the other input for a quasi-balance set-up). This way, the difference between the two chip's output is exactly your desired audio signal -> like one of those X-setup.
Pedja said:
Actually, I don't see that problem. An amplifier like the LM3886 has quite a big common mode input range. For that amplifier it will almost certain make no difference at all whether both inputs are at 220mV (or -220mV) instead of at 0V. If that would be a problem then a difference of 220mV in supply voltage (positive versus negative) also would be a problem, and it does not. The only real difference is that the output is also at 220mV, creating an additional DC output current. That might change the sonic behaviour, but that is by intention, to make it class A.
Steven
Re: Re: Re: Class A gainclone/chipamp idea
yes and no. Creat a big DC offset right on each amp. so each amp will be working on its positive half only (you can do negative half but for discussion purposes just stay with the positive half for now). one amp will have just the DC offset, the other the DC offset + the signal. The load will bridge the two outputs so you get nothing but the signal.
Think of each amp as a transistor in the differential pair. One has the input signal on its base and the other doesn't. And the load is cross the collectors of two such transistors.
Or one of those Zen amps with unbalanced input.
The only purpose to have the 2nd amp is to get rid of the DC offset on the load.
Steven said:
yes and no. Creat a big DC offset right on each amp. so each amp will be working on its positive half only (you can do negative half but for discussion purposes just stay with the positive half for now). one amp will have just the DC offset, the other the DC offset + the signal. The load will bridge the two outputs so you get nothing but the signal.
Think of each amp as a transistor in the differential pair. One has the input signal on its base and the other doesn't. And the load is cross the collectors of two such transistors.
Or one of those Zen amps with unbalanced input.
The only purpose to have the 2nd amp is to get rid of the DC offset on the load.
Steven,Steven said:
It is not a problem with the common mode input range. I said the circuit will work. I have tried to point out the one possible soundwise problem with the configuration you have proposed. Opamps often do sound different with or without or with more or less DC offset. Some decent sounding opamps sound terrible with a few hundreds of mV of DC sent to one of their inputs. At the other hand, certain difference between the positive and negative supply voltages is indeed usually quite benign. I think at this moment I can not give you much good reasons why is this so. Note that I am usually making equipment to please the ear, not the theory. I use the theory to realize what I have heard but I am not always (if ever) quite successful with it. I could be probably more successful with theory, but only deliberately neglecting what I hear.
Pedja
Pedja said:
Pedja,
Thanks for your comment, but I do not understand it. I still think this is just a matter of input common mode range. If you apply let's say 200mV DC to one of the opamp inputs, the other input has to follow because of feedback. Otherwise the output would be sticking to one of the supply rails. Then you have two inputs and the output at 200mV (DC gain is supposed to be unity, as in the circuit shown), which is perfectly OK for an opamp, as long as it fits in its common mode range. If the output would be AC-coupled then the push and pull output currents would be the same as if the ouput would have been at 0V level, or any other DC level. The only difference would be that the maximum voltage swing would be different towards the positive supply rail as to the negative rail. For most opamps the difference in dropout voltage to the supply rails is already more than 200mV between the positive and the negative rail. I think this is not issue here.
As I said in my previous post, the only real difference that a DC offset brings at the output is when the output is not AC-coupled but DC-coupled and a big offset current is flowing. I can imagine that an opamp (chipamp, gainclone) will sound different then, because the push and pull output currents will not be equal anymore on average. But that was exactly the intention of the proposed circuit.
Please note that issue here is current offset, not voltage offset. The 220mV voltage offset has just been applied to get 1A current offset for class A operation, via the 0.22 Ohm resistor. I could have got a 1A current offset by connecting a 1A CCS to the output of the amplifier and kept a DC output voltage of 0V. An offset voltage wil not move an amplifier into class A, only an offset current does.
Steven
Re: Re: Re: Re: Class A gainclone/chipamp idea
Millwood,
I'm sorry but I think you misunderstand how the amplifiers are put into class A. It is not the voltage offset at the output of each amplifier that causes class A operation, it's the current offset that is important. In the case of the proposed circuit the voltage offset is only there to create the current offset in a simple way.
Your proposal to have one ouput at a certain DC voltage level and the other amplifier at the same DC voltage level plus signal applies the signal part to the load if the load is connected between both the amplifiers. That is correct. But none of the amplifiers will be in class A for that matter. The output currents will still be positive and negative, in both amplifiers.
Steven
millwood said:
Millwood,
I'm sorry but I think you misunderstand how the amplifiers are put into class A. It is not the voltage offset at the output of each amplifier that causes class A operation, it's the current offset that is important. In the case of the proposed circuit the voltage offset is only there to create the current offset in a simple way.
Your proposal to have one ouput at a certain DC voltage level and the other amplifier at the same DC voltage level plus signal applies the signal part to the load if the load is connected between both the amplifiers. That is correct. But none of the amplifiers will be in class A for that matter. The output currents will still be positive and negative, in both amplifiers.
Steven
Hi Steven,
Don’t see the offset problem either, the outputs of the two amps are at the same level as their inputs at no signal.. But even then, the output swing is much larger than the input swing. If this would cause problems any op-amp would have a problem.
Think the whole thing works. Only disadvantage is you have a 0.11 ohm resistor in series with the load, reducing damping factor. But will that be a real disadvantage? Curious about the sonic results. Are you intending to build the amp Steven?
Cheers
Don’t see the offset problem either, the outputs of the two amps are at the same level as their inputs at no signal.. But even then, the output swing is much larger than the input swing. If this would cause problems any op-amp would have a problem.
Think the whole thing works. Only disadvantage is you have a 0.11 ohm resistor in series with the load, reducing damping factor. But will that be a real disadvantage? Curious about the sonic results. Are you intending to build the amp Steven?
Cheers
Pjotr,
This is still a damping factor of 80, more than many other good sounding amplifiers. I don't think it is a problem. Anyway, it is possible to lower the voltage offset and use an even smaller resistor, but then the circuit becomes more sensitive to tolerances. Or use even more amplifiers in parallel, the output impedance caused by all these summing resistors becomes smaller too.
At the moment no intention to build, but who knows.
Steven
This is still a damping factor of 80, more than many other good sounding amplifiers. I don't think it is a problem. Anyway, it is possible to lower the voltage offset and use an even smaller resistor, but then the circuit becomes more sensitive to tolerances. Or use even more amplifiers in parallel, the output impedance caused by all these summing resistors becomes smaller too.
At the moment no intention to build, but who knows.
Steven
similar to ?
Steven,
I'm fascinated by your proposal. I'm wondering if this approach has similarities to using tubes in class A applications and in push-pull for greater output.
I've not much to offer as I have little experience. I'm reaching for something I have seen before in order to grasp the concept. Interesting. Please continue.
Sincerely,
Ed
Steven,
I'm fascinated by your proposal. I'm wondering if this approach has similarities to using tubes in class A applications and in push-pull for greater output.
I've not much to offer as I have little experience. I'm reaching for something I have seen before in order to grasp the concept. Interesting. Please continue.
Sincerely,
Ed
Hi Steven,
It's interesting that we should come up with a similar idea at about the same time albeit with different implementations.
I will not try mine for a while. I'm focusing on my class A amps
my idea is here:
http://www.diyaudio.com/forums/showthread.php?postid=327129#post327129
I wonder who's cct will sound best ?
mike
It's interesting that we should come up with a similar idea at about the same time albeit with different implementations.
I will not try mine for a while. I'm focusing on my class A amps
my idea is here:
http://www.diyaudio.com/forums/showthread.php?postid=327129#post327129
I wonder who's cct will sound best ?
mike
Hi Mike,
Yep, both circuits will run in class A, but the circuits are completely different. You use an external CCS, I use the other amp to force a DC output current. Your circuit uses mainly the upper transistor in both bridging amps, mine use the upper in one amp and the lower in the other and both amplifiers are in parallel, not bridged.
Difference in sound? Who knows?
Steven
Yep, both circuits will run in class A, but the circuits are completely different. You use an external CCS, I use the other amp to force a DC output current. Your circuit uses mainly the upper transistor in both bridging amps, mine use the upper in one amp and the lower in the other and both amplifiers are in parallel, not bridged.
Difference in sound? Who knows?
Steven
Class A inverted Gainclone
Hi there,
I implemented a class A gainclone based on the idea by Steven. Each channel comprises of two modules of an inverted gainclone which are paralleled.
One module receives a positive biasing voltage at the non-inverting input, the other module receives a negative biasing voltage. These voltages induce a current across the two LM3875. The two 0.22 Ohm resistors set the idle current to 0.9 A. The arrangement biases the “upper” output transistors in one chip and the “lower” output transistors in the other chip into class A. Since the idle current is 0.9A, class A drive limited to 13W at 8 Ohm and to 6.5W at 4 Ohm.
Instead of a simple feedback resistor I used a T-network as recommended by Joe Rassmussen. The T-network with relatively low value resistors simply sounds better than one resistor of 600k I the feedback loop.
Instead of an input buffer I use an audio transformer. The winding arrangement and the 1k pot at the secondaries result in an input impedance of 4kOhm, which is suitable for most solid state preamplifiers IMHO.
The development of the amp followed three steps:
One module of an inverted gainclone per channel: The simple inverted gainclone delivered a rich and detailed sonic image in combination with my Mission DAD 7 CD player and my SL 600 speakers.
Paralleling the second module without class A biasing: Noticable differences were small. A little more control in the bass and may be a little more depth in the soundstage.
Paralleled modules biased to class A: The chane was clearly audible. Class A biasing added “body and soul” to the sonic image. It sounded less as a transistor amp but without loosing detail.
Listening tests were done blinded by two independent persons. In conclusion it can be resumed that Stevens idea of class A biasing is definitely worth a try.
Cheers
KlausB
Hi there,
I implemented a class A gainclone based on the idea by Steven. Each channel comprises of two modules of an inverted gainclone which are paralleled.
One module receives a positive biasing voltage at the non-inverting input, the other module receives a negative biasing voltage. These voltages induce a current across the two LM3875. The two 0.22 Ohm resistors set the idle current to 0.9 A. The arrangement biases the “upper” output transistors in one chip and the “lower” output transistors in the other chip into class A. Since the idle current is 0.9A, class A drive limited to 13W at 8 Ohm and to 6.5W at 4 Ohm.
Instead of a simple feedback resistor I used a T-network as recommended by Joe Rassmussen. The T-network with relatively low value resistors simply sounds better than one resistor of 600k I the feedback loop.
Instead of an input buffer I use an audio transformer. The winding arrangement and the 1k pot at the secondaries result in an input impedance of 4kOhm, which is suitable for most solid state preamplifiers IMHO.
The development of the amp followed three steps:
One module of an inverted gainclone per channel: The simple inverted gainclone delivered a rich and detailed sonic image in combination with my Mission DAD 7 CD player and my SL 600 speakers.
Paralleling the second module without class A biasing: Noticable differences were small. A little more control in the bass and may be a little more depth in the soundstage.
Paralleled modules biased to class A: The chane was clearly audible. Class A biasing added “body and soul” to the sonic image. It sounded less as a transistor amp but without loosing detail.
Listening tests were done blinded by two independent persons. In conclusion it can be resumed that Stevens idea of class A biasing is definitely worth a try.
Cheers
KlausB
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.