Hello Zei,
Thank you very much for saving the front of my head and face
You were right!!! I measured the correct pins and got 30.1 volts on each op amp! Now I can figure the calculations to know which value resistor!
Any suggestions on which type of resistor or will a Radio Shack cheap one do?
I'll probably be back asking how to check to see if it is operating at what Carlosfm suggested, 5 to 10 ma.
Arjen, I am interested also to the answer to your question.
Regards//Keith
Thank you very much for saving the front of my head and face
You were right!!! I measured the correct pins and got 30.1 volts on each op amp! Now I can figure the calculations to know which value resistor!
Any suggestions on which type of resistor or will a Radio Shack cheap one do?
I'll probably be back asking how to check to see if it is operating at what Carlosfm suggested, 5 to 10 ma.
Arjen, I am interested also to the answer to your question.
Regards//Keith
OpAmp Class-A Biasing -- a Different Approach
Has anyone seen this before? I'm sure someone already invented this before I did...
A pretty simple concept, basically. Run two identical OpAmp in parallel, symmetrically offset them a little -- by injecting offset currents into the -INs --, and combine the outputs with current sharing resistors. As long as less current than +-Iq is drawn from the output, OP1 will always source current and OP2 will always sink current. When more current is drawn, only one amplifier's output current changes polarity, running through its crossover section. Choosing Ibias and Riso requires some thought in an actual implementation.
Although only a simple closed-loop inverting example is shown, this can be easily extended to high AC gains as a quasi-open-loop building block, e.g. for composite amplifiers.
I've tried this arrangement on both small-signal duals/quads and power OpAmps with great success. It can be further extended to more than two amplifiers and also to different class A to B crossover breakpoints / different quiescent currents for each one, resulting in a gradual class A to B transition for the whole circuit. Thus, output characteristics are greatly smoothed when only one amplifier at a time changes its output current direction and when every amplifier runs at a different internal operation point, current-wise.
Among other options, one can make the biasing somewhat signal dependent (e.g. sensing the output current and/or input voltage), reducing idle state power consumption in power amplifiers.
pros:
- high output current
- wide class-A range
- smoother, piecewise class A to B transition
- lower noise
cons:
- decreased output voltage headroom
- increased Zout (though this is seldom a real problem)
- requires multiple and matched networks/amplifiers
regards,
Klaus
Has anyone seen this before? I'm sure someone already invented this before I did...
An externally hosted image should be here but it was not working when we last tested it.
A pretty simple concept, basically. Run two identical OpAmp in parallel, symmetrically offset them a little -- by injecting offset currents into the -INs --, and combine the outputs with current sharing resistors. As long as less current than +-Iq is drawn from the output, OP1 will always source current and OP2 will always sink current. When more current is drawn, only one amplifier's output current changes polarity, running through its crossover section. Choosing Ibias and Riso requires some thought in an actual implementation.
Although only a simple closed-loop inverting example is shown, this can be easily extended to high AC gains as a quasi-open-loop building block, e.g. for composite amplifiers.
I've tried this arrangement on both small-signal duals/quads and power OpAmps with great success. It can be further extended to more than two amplifiers and also to different class A to B crossover breakpoints / different quiescent currents for each one, resulting in a gradual class A to B transition for the whole circuit. Thus, output characteristics are greatly smoothed when only one amplifier at a time changes its output current direction and when every amplifier runs at a different internal operation point, current-wise.
Among other options, one can make the biasing somewhat signal dependent (e.g. sensing the output current and/or input voltage), reducing idle state power consumption in power amplifiers.
pros:
- high output current
- wide class-A range
- smoother, piecewise class A to B transition
- lower noise
cons:
- decreased output voltage headroom
- increased Zout (though this is seldom a real problem)
- requires multiple and matched networks/amplifiers
regards,
Klaus
By lucky accident, I found someone who did this stuff before, and guess who? Nelson Pass! Sort of a best-case reference, I think...
And yes, I *did* search the forums before I wrote the previous post.
http://www.diyaudio.com/forums/showthread.php?postid=423526#post423526
(see circuit #4)
Regards,
Klaus
And yes, I *did* search the forums before I wrote the previous post.
http://www.diyaudio.com/forums/showthread.php?postid=423526#post423526
(see circuit #4)
Regards,
Klaus
Hello,
My question about post 56 did not yet yield a response, so another try: Joe Rasmussen seems to have included a picture in that post. Somehow I cannot see that picture. Attachments of other posts are visible. Am I the only one that cannot detect it in post 56? If so, can someone mail it PM?
Best, Arjen.
My question about post 56 did not yet yield a response, so another try: Joe Rasmussen seems to have included a picture in that post. Somehow I cannot see that picture. Attachments of other posts are visible. Am I the only one that cannot detect it in post 56? If so, can someone mail it PM?
Best, Arjen.
Hi Terry & Andrew,
Actually I don't know if these audio legends (Mr.Self and Mr.Neve) did this but I strongly speculate they did...
I was thinking about the offset problems with paralleled chip amps (and thus the necessity of current sharing resistors in the outputs) and then I went hey, why not use offset deliberately to get the outputs into class A? With chip amps one needs to parallel quite a few anyway, to drive low impedance speakers during transients (like Andrew has pointed out elsewhere).
Further thinking lead me to use different points where the individual amps change their output current polarity which will result in a rather smooth class A to B transition when output current demand increases above pure class A levels.
Finally I realized that class A, from a PSRR standpoint (often critical with chip amps / op amps), is only beneficial in a fully balanced mode (single ended only the sum of supply currents is constant in class A but not the individual currents). Again the transition points give a smooth characteristic when going into class B in small increments, not all at once. As GND becomes a pure reference potential a single supply is sufficient (which has some other nice effects).
One might call this "Balanced Piecewise Class A" or "Balanced Segmented Class A".
Regards, Klaus
Actually I don't know if these audio legends (Mr.Self and Mr.Neve) did this but I strongly speculate they did...
I was thinking about the offset problems with paralleled chip amps (and thus the necessity of current sharing resistors in the outputs) and then I went hey, why not use offset deliberately to get the outputs into class A? With chip amps one needs to parallel quite a few anyway, to drive low impedance speakers during transients (like Andrew has pointed out elsewhere).
Further thinking lead me to use different points where the individual amps change their output current polarity which will result in a rather smooth class A to B transition when output current demand increases above pure class A levels.
Finally I realized that class A, from a PSRR standpoint (often critical with chip amps / op amps), is only beneficial in a fully balanced mode (single ended only the sum of supply currents is constant in class A but not the individual currents). Again the transition points give a smooth characteristic when going into class B in small increments, not all at once. As GND becomes a pure reference potential a single supply is sufficient (which has some other nice effects).
One might call this "Balanced Piecewise Class A" or "Balanced Segmented Class A".
Regards, Klaus
Hi Kstr,
forcing an opamp into ClassA is only used due to the inside of the ClassAB output stage being inaccessible to tweaking.
The output stage can be converted to single ended by shutting down either the upper half or the lower half of the ClassAB push pull pair. Often the output stage is a quasi stage and designed for optimum IC characteristics.
When one converts the stage to single ended we end up with a very inefficient ClassA stage that has heat directly proportional to ClassA bias current.
Assume the opamp normally sends a maximum of 2Vac (2.8Vpk) into a 5k load. This equates to 0.56mApk output current. A single ended stage must have bias set to exceed the output current for all normal signals. If you want some headroom then bias must be set to some factor over the normal maximum.
Lets adopt Ib=2mA giving 11db of headroom.
The dissipation inside the IC is now +-Vsupply * 2mA ~=30*0.002=60mW on top of what is dissipated in the earlier stages. We have probably got close to doubling the dissipation in the IC for a Class A output of 7Vac into 5k and no capacitance on the load.
Some of the advantages that make opamp ClassA output also apply to a chipamp so your question is not unreasonable.
Now do the same calculations for a chipamp.
Wait a very short while for it to blow up.
forcing an opamp into ClassA is only used due to the inside of the ClassAB output stage being inaccessible to tweaking.
The output stage can be converted to single ended by shutting down either the upper half or the lower half of the ClassAB push pull pair. Often the output stage is a quasi stage and designed for optimum IC characteristics.
When one converts the stage to single ended we end up with a very inefficient ClassA stage that has heat directly proportional to ClassA bias current.
Assume the opamp normally sends a maximum of 2Vac (2.8Vpk) into a 5k load. This equates to 0.56mApk output current. A single ended stage must have bias set to exceed the output current for all normal signals. If you want some headroom then bias must be set to some factor over the normal maximum.
Lets adopt Ib=2mA giving 11db of headroom.
The dissipation inside the IC is now +-Vsupply * 2mA ~=30*0.002=60mW on top of what is dissipated in the earlier stages. We have probably got close to doubling the dissipation in the IC for a Class A output of 7Vac into 5k and no capacitance on the load.
Some of the advantages that make opamp ClassA output also apply to a chipamp so your question is not unreasonable.
Now do the same calculations for a chipamp.
Wait a very short while for it to blow up.
Thanks for commenting on this, Andrew. Because I designed a lot of electronics with power op amps at work (ATE stuff) I'm quite aware of the dissipation/SOA related problems, especially with reactive loads. I'm not about to run 1A bias with LM3886's on +-35V supplies...
The actual amp concept will use about a dozen of them per channel, running on a low supply and with (partly) dynamic biasing. It's not class A over the full output range, as I feel the benefits of class A is mainly in the low level range. And it's a composite amp (the chip amps only act as precision output transistors of sorts for an grounded input 2 OpAmp diff with current mode xformer signal coupling), not the typical GainClone/PA/BPA design.
Regards, Klaus
The actual amp concept will use about a dozen of them per channel, running on a low supply and with (partly) dynamic biasing. It's not class A over the full output range, as I feel the benefits of class A is mainly in the low level range. And it's a composite amp (the chip amps only act as precision output transistors of sorts for an grounded input 2 OpAmp diff with current mode xformer signal coupling), not the typical GainClone/PA/BPA design.
Regards, Klaus
Class A chip
I have run an LM3875 class A with a power JFET on a low volt supply +/- 16VDC and had good outcomes, surprisingly 1st harmonic distortion increased, 3rd decreased and 2nd harmonics increased, but the overall sound quality was somewhat more laid back with good control, I dare say more "tube like"taking the edge off. The LM3875 liked 1.0 to 1.2A bias, I slide from as low as the chips internal to as high as 1.5A output is low around 10W.
My overall thought on biasing a chip into A was not worth it, measurements are worse, dissipation is "HOT" and power output low. The LM overture series all use a well designed internal Quasi output, and pushing them into one side operation is just not worth the drawbacks, IMHO.
I have run an LM3875 class A with a power JFET on a low volt supply +/- 16VDC and had good outcomes, surprisingly 1st harmonic distortion increased, 3rd decreased and 2nd harmonics increased, but the overall sound quality was somewhat more laid back with good control, I dare say more "tube like"taking the edge off. The LM3875 liked 1.0 to 1.2A bias, I slide from as low as the chips internal to as high as 1.5A output is low around 10W.
My overall thought on biasing a chip into A was not worth it, measurements are worse, dissipation is "HOT" and power output low. The LM overture series all use a well designed internal Quasi output, and pushing them into one side operation is just not worth the drawbacks, IMHO.
@Tolu:
I think those very good LME497x0's won't profit much from output bias in such a low gain setup. One surely will need a high noise gain circuit (like x1000) to see a change with normal measurement equipment. I would prefer a setup with two OpAmps as I described to passive biasing with a resistor/CCS.
The benefits of class A will become greater the higher a gain we use -- making composite amp front-ends running (almost) open loop the most interesting candidates (AFAIR people like Walt Jung frequently use(d) this).
@tiltedhalo:
Yes, the only way to find out if it works is to try it out. Thanks for your valuable comments on the LM3875.
Regards, Klaus
I think those very good LME497x0's won't profit much from output bias in such a low gain setup. One surely will need a high noise gain circuit (like x1000) to see a change with normal measurement equipment. I would prefer a setup with two OpAmps as I described to passive biasing with a resistor/CCS.
The benefits of class A will become greater the higher a gain we use -- making composite amp front-ends running (almost) open loop the most interesting candidates (AFAIR people like Walt Jung frequently use(d) this).
@tiltedhalo:
Yes, the only way to find out if it works is to try it out. Thanks for your valuable comments on the LM3875.
Regards, Klaus
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.