Ultra high spec. opamp MC/MM phono, warp "elliptic" filter, line, headphone amps.
- Thread starter wyn palmer
- Start date
wyn palmer
Addicted Member
Seems good to me, and should be an interesting project. I'm happy to do the layout etc.
wyn palmer
Addicted Member
So, it appears that the composite design is, essentially, modular, with the ability to add push-pull pairs of DC offset corrected LM3886s to the basic unit in order to increase the power, or at least the current, output. This provides a way forward to eventually enable the unit to supply 7, 14, 21, or 28 amps per side in a dual mono configuration. This should facilitate the evaluation for a DIY build.
The design, at least in simulation, is stable into any load that speakers present, but is limited by the LM3886 power supply maximum of roughly +/-28v.
The datasheet maximum is 68W/unit- single ended into 4ohms., which represents about 4.5A RMS or about 5.8A peak. This would be equivalent to 136W into 8 ohms in the push pull case, as the current limiting would be the same, and would fall into 4 ohms.
More parallel stages would improve the current drive and the thermally limited output from the amp but would not necessarily increase the maximum available swing.
For example, two output modules would allow for approximately 136W per side, for a total of 272 W, but this would be into a 4-ohm load, with the 8 ohms case being essentially unchanged.
This offers the opportunity for a prototype to be constructed to test out the concept without having to go for a full design.
The prototype need not even have the final supply, as long as a 10A +/- 30 v DC bench supply is available.
It would include a +/-15v regulator module, a DC coupled, offset corrected balanced /SE input stage and a push-pull input amp/phase compensation section for a composite amplifier plus a single push-pull LM3886 output driver module with connections to support other push-pull output driver modules wired in parallel.
It would also include the on timer and a mute function.
The gain would be x40- or 32dBs, with a full power sensitivity of about 500mv rms, which is compatible with many Audio preamps etc, if not with professional audio gear.
If this is of interest and @hirscwi is willing to lay it out/build it as a first stage to a full amplifier, then I will post the schematics etc.
The design, at least in simulation, is stable into any load that speakers present, but is limited by the LM3886 power supply maximum of roughly +/-28v.
The datasheet maximum is 68W/unit- single ended into 4ohms., which represents about 4.5A RMS or about 5.8A peak. This would be equivalent to 136W into 8 ohms in the push pull case, as the current limiting would be the same, and would fall into 4 ohms.
More parallel stages would improve the current drive and the thermally limited output from the amp but would not necessarily increase the maximum available swing.
For example, two output modules would allow for approximately 136W per side, for a total of 272 W, but this would be into a 4-ohm load, with the 8 ohms case being essentially unchanged.
This offers the opportunity for a prototype to be constructed to test out the concept without having to go for a full design.
The prototype need not even have the final supply, as long as a 10A +/- 30 v DC bench supply is available.
It would include a +/-15v regulator module, a DC coupled, offset corrected balanced /SE input stage and a push-pull input amp/phase compensation section for a composite amplifier plus a single push-pull LM3886 output driver module with connections to support other push-pull output driver modules wired in parallel.
It would also include the on timer and a mute function.
The gain would be x40- or 32dBs, with a full power sensitivity of about 500mv rms, which is compatible with many Audio preamps etc, if not with professional audio gear.
If this is of interest and @hirscwi is willing to lay it out/build it as a first stage to a full amplifier, then I will post the schematics etc.
Yes, I'll be glad to lay out the PCB. I can also do the prototype build but am happy to let phonomac do that first build.
wyn palmer
Addicted Member
I'm going to need input/help in this process. DIY power amps have not been my thing for a long time, so any suggestions etc. would be appreciated.
Despite this being a development that was suggested by an AK member or three, it probably is more of a DiyAudio thing, but I would prefer to keep it here even though I'm certain that more "helpful" input might be forthcoming on that forum, especially since LM3886 based amps are common there.
If staying on Ak is OK, then perhaps we should start a new thread on a different = sub forum?
Incidentally, I've got a "current dumping"/feedforward design working.
However, I don't see any improvement in the simulated distortion relative to the composite design when using the LM3886/OPA1656 and the composite is well, more straighforward.
I'm in the process of doing something that I was reluctant to do- including the THX patent improvements to see if they really are better.
As of this time the composite amp seems like the way to go.
Despite this being a development that was suggested by an AK member or three, it probably is more of a DiyAudio thing, but I would prefer to keep it here even though I'm certain that more "helpful" input might be forthcoming on that forum, especially since LM3886 based amps are common there.
If staying on Ak is OK, then perhaps we should start a new thread on a different = sub forum?
Incidentally, I've got a "current dumping"/feedforward design working.
However, I don't see any improvement in the simulated distortion relative to the composite design when using the LM3886/OPA1656 and the composite is well, more straighforward.
I'm in the process of doing something that I was reluctant to do- including the THX patent improvements to see if they really are better.
As of this time the composite amp seems like the way to go.
@wyn palmer - I agree that you will gett more helpful input on that forum for the reasons you've stated and since board layout and countless other details are important to optimize. I don't think that precludes input from folks here. If you decide to keep it here only, a different thread would be good in my view.
Last edited:
j!m
Super Member
And I think there’s value in testing the “THX circuit” to see how it measures up under the same test conditions. Then you’ll have apples to apples in simulation.
Obviously it can’t be sold, and I’m not suggesting that either. But relying on the data of others often leads to disappointment…
Obviously it can’t be sold, and I’m not suggesting that either. But relying on the data of others often leads to disappointment…
Having spent a lot of time restoring, upgrading and modifying Sansui amps of the late 70s /early 80s era which included feedforward and BTL designs, the feedforward versions didn't achieve the sort of improvements that were touted. So for me the composite OPA1656/LM3886 approach should be more fruitful.
I think keeping things on AK, with a separate thread/sub-forum as appropriate, sounds like a good idea. On DiyAudio it is all too easy to drown in thousands of different views about how to do things - all well-meaning but can take a long time for things to crystallize.
I think keeping things on AK, with a separate thread/sub-forum as appropriate, sounds like a good idea. On DiyAudio it is all too easy to drown in thousands of different views about how to do things - all well-meaning but can take a long time for things to crystallize.
wyn palmer
Addicted Member
A "working" THX design has been simulated. The real novelty was in the intention to replace the inductor in the current dumping design with a resistor/capacitor combo.
This breaks the DC feedback from the output to the input so a secondary feedback loop has to be added, but this can be designed to also increase the distortion error feedback and increase the overall distortion reduction.
Yes, it does work, although getting a flat frequency response takes some finagling. Assuming an ideal low-level amplifier it gives more than 20dB additional distortion cancellation relative to the composite amp.
However, the low-level amp would have to potentially support the full output swing at HF as it's effectively capacitively coupled into the output, and it also becomes the dominant source of distortion. The design also works on the assumption that the imperfect output "amp" is essentially unity gain
The LM3886 is not a good choice for this as it really needs to operate at a gain of 10 or higher to be acceptably stable. This can be fudged, but it hardly seems worth the effort as the other problem is that an opamp like the OPA1656 cannot be used, unenhanced, for the output driver.
In the proposed composite design, the LM3886 will operate at a gain of 26dB and will have about 45dB of additional feedback for correction at 20kHz.
Also, the OPA1656 can be used in essentially a unity gain mode, with a need to only support 1/20th of the output swing voltage, so the distortion, even though it is amplified by 26dB is still extremely low- by my estimation in the noise floor.
So, it looks like the composite amp, rather than the error feedback or the THX style will be the design of choice.
It does appear that my approach to the composite is, well, unusual, based on the schematics I can find on DiyAudio etc. so perhaps I'm missing something.
I'm continuing to search the "literature" and simulate. I've run square wave and dc control loop simulations into tough loads, as well as the AC simulations, and all seems well.
I've also modified the LM3889 Spice model to better reflect the worst-case situations, at least those that I'm aware of, but probably building a prototype is the best way to go.
This breaks the DC feedback from the output to the input so a secondary feedback loop has to be added, but this can be designed to also increase the distortion error feedback and increase the overall distortion reduction.
Yes, it does work, although getting a flat frequency response takes some finagling. Assuming an ideal low-level amplifier it gives more than 20dB additional distortion cancellation relative to the composite amp.
However, the low-level amp would have to potentially support the full output swing at HF as it's effectively capacitively coupled into the output, and it also becomes the dominant source of distortion. The design also works on the assumption that the imperfect output "amp" is essentially unity gain
The LM3886 is not a good choice for this as it really needs to operate at a gain of 10 or higher to be acceptably stable. This can be fudged, but it hardly seems worth the effort as the other problem is that an opamp like the OPA1656 cannot be used, unenhanced, for the output driver.
In the proposed composite design, the LM3886 will operate at a gain of 26dB and will have about 45dB of additional feedback for correction at 20kHz.
Also, the OPA1656 can be used in essentially a unity gain mode, with a need to only support 1/20th of the output swing voltage, so the distortion, even though it is amplified by 26dB is still extremely low- by my estimation in the noise floor.
So, it looks like the composite amp, rather than the error feedback or the THX style will be the design of choice.
It does appear that my approach to the composite is, well, unusual, based on the schematics I can find on DiyAudio etc. so perhaps I'm missing something.
I'm continuing to search the "literature" and simulate. I've run square wave and dc control loop simulations into tough loads, as well as the AC simulations, and all seems well.
I've also modified the LM3889 Spice model to better reflect the worst-case situations, at least those that I'm aware of, but probably building a prototype is the best way to go.
wyn palmer
Addicted Member
That sounds fine to me. Is there a volunteer to set up the thread? (@Nashou66?)
j!m
Super Member
Honestly, with how the phono pre was received over there, I personally wouldn't want to.
But it is Wyn's decision of course.
That said, I agree we should start a dedicated thread on the amplifier (which I am happy to do here) to keep things better organized than my brain is... Do we have a name for the project?
But it is Wyn's decision of course.
That said, I agree we should start a dedicated thread on the amplifier (which I am happy to do here) to keep things better organized than my brain is... Do we have a name for the project?