I had an idea for a quick-and-dirty class-H inspired architecture that, rather than using dynamic supply rails, uses separate class-D and class-AB amplifiers in a BTL-type configuration so that the speaker is driven by one amplifier while the other is on the "return path" effectively shifting the ground rather than the supply rails.
The class-D amplifier will have a low pass filter and will provide most of the drive potential while the class-AB amplifier provides the "corrections", adding in the high frequencies and correcting any phase or amplitude distortions from the class-D amplifier. The class-AB amplifier would be driven by the error between the input signal and the class-D output.
The benefits that I hope this design may have are either more accurate sound reconstruction than basic class-D amplifiers or lower power dissipation in the class-AB stage. I am thinking it could be constructed using pre-made class-D amplifiers and hopefully AB amplifiers too with only the error amplifier and biasing circuits being custom-made.
A similar design could be to replace the class-AB amplifier with transistors leading directly to ground, sort of like a push-pull amplifier with both power supply rails connected to ground instead. If the output from the class-D amplifier overshoots then transistors turn off, decreasing the voltage across the speaker, if it undershoots then they turn on, increasing the voltage.
It may even be possible to combine this ground-pull circuit with a class-AB amplifier to make a very odd looking stage where the output can be driven to the positive supply, negative supply, and also ground! With this design the ground pull transistors should spend most of their time in a low-resistance state, improving efficiency. I bashed up a crude diagram of this:
I would be interested to know whether this is possible, can two completely different amplifiers be used in such a configuration?
I am pretty new to amplifier design so don't expect any of this to make too much sense, by the end of my various electronics courses I may realise exactly why this is a silly idea.
The class-D amplifier will have a low pass filter and will provide most of the drive potential while the class-AB amplifier provides the "corrections", adding in the high frequencies and correcting any phase or amplitude distortions from the class-D amplifier. The class-AB amplifier would be driven by the error between the input signal and the class-D output.
The benefits that I hope this design may have are either more accurate sound reconstruction than basic class-D amplifiers or lower power dissipation in the class-AB stage. I am thinking it could be constructed using pre-made class-D amplifiers and hopefully AB amplifiers too with only the error amplifier and biasing circuits being custom-made.
A similar design could be to replace the class-AB amplifier with transistors leading directly to ground, sort of like a push-pull amplifier with both power supply rails connected to ground instead. If the output from the class-D amplifier overshoots then transistors turn off, decreasing the voltage across the speaker, if it undershoots then they turn on, increasing the voltage.
It may even be possible to combine this ground-pull circuit with a class-AB amplifier to make a very odd looking stage where the output can be driven to the positive supply, negative supply, and also ground! With this design the ground pull transistors should spend most of their time in a low-resistance state, improving efficiency. I bashed up a crude diagram of this:
I would be interested to know whether this is possible, can two completely different amplifiers be used in such a configuration?
I am pretty new to amplifier design so don't expect any of this to make too much sense, by the end of my various electronics courses I may realise exactly why this is a silly idea.
It's an interesting idea. Hybrid amplifiers have been built in the past, see e.g. the "current dumping" concept.
Potential pitfalls that I can see:
1. You've got a feedback loop in there. Stability will need to be addressed via control theory.
2. You need to make sure that the AB and ground-pull stage don't decide to work against each other. The latter may also require oversized versions of muting transistors, which are slightly exotic to begin with.
3. D amps are usually operated with a bridge output for good reason - that makes it much easier to get rid of all the high-frequency crap. Offending radiation usually is common mode, and a SE output gives a CM signal of 1/2 (V_out + V_Gnd) = 1/2 V_out.
4. I have my doubts about the "easy" part. 😉
5. And finally, I wouldn't call it "Class H", even if power consumption could be significantly reduced over an AB-only amp.
Potential pitfalls that I can see:
1. You've got a feedback loop in there. Stability will need to be addressed via control theory.
2. You need to make sure that the AB and ground-pull stage don't decide to work against each other. The latter may also require oversized versions of muting transistors, which are slightly exotic to begin with.
3. D amps are usually operated with a bridge output for good reason - that makes it much easier to get rid of all the high-frequency crap. Offending radiation usually is common mode, and a SE output gives a CM signal of 1/2 (V_out + V_Gnd) = 1/2 V_out.
4. I have my doubts about the "easy" part. 😉
5. And finally, I wouldn't call it "Class H", even if power consumption could be significantly reduced over an AB-only amp.
The 'easy' part is speculative, I am hoping that the feedback and ground-pull stage can just be added on to an existing class-AB amplifier. The main thing I am concerned about is whether the entire ground-pull concept will even be possible, and whether there will be zero-crossing noise and such.
The crap from the class-D amp should not be a problem if the design works as I hope. The output from the class-D will be heavily filtered, and the AB-GND stage will correct the poor frequency and phase response introduced by the filtering and the (likely very cheap) class-D amplifier itself.
You are right about it not being class-H, it's more "class-H inspired". A while ago I came up with a variable supply idea, later finding that somebody had already come up with it as the class-H concept. Of course I quickly realised it would be too hard for me to design such a thing myself so I came up with this mess instead 🙂
What exactly do you mean by "muting transistors"? Is it simply a circuit for dragging a signal line to ground or is it some exotic sort of transistor that conducts current symmetrically in either direction? Since one side is grounded I would hope that the biasing on BJTs should be fairly simple.
I think the simplest method to execute this would be having the ground-pull circuit without the class-AB circuit. That way you can just bias the ground-pull transistors in a nearly-fully-on state and modulate them up and down with a op-amp based feedback circuit. Including the class-AB circuit introduces a lot of complexity that I am not up for sorting out.
The crap from the class-D amp should not be a problem if the design works as I hope. The output from the class-D will be heavily filtered, and the AB-GND stage will correct the poor frequency and phase response introduced by the filtering and the (likely very cheap) class-D amplifier itself.
You are right about it not being class-H, it's more "class-H inspired". A while ago I came up with a variable supply idea, later finding that somebody had already come up with it as the class-H concept. Of course I quickly realised it would be too hard for me to design such a thing myself so I came up with this mess instead 🙂
What exactly do you mean by "muting transistors"? Is it simply a circuit for dragging a signal line to ground or is it some exotic sort of transistor that conducts current symmetrically in either direction? Since one side is grounded I would hope that the biasing on BJTs should be fairly simple.
I think the simplest method to execute this would be having the ground-pull circuit without the class-AB circuit. That way you can just bias the ground-pull transistors in a nearly-fully-on state and modulate them up and down with a op-amp based feedback circuit. Including the class-AB circuit introduces a lot of complexity that I am not up for sorting out.
I have revised it to divide up the awkward bias circuits and add blocking diodes to the ground drains so that the ground-pull transistor does not have to turn off when the complimentary pull-up/down transistor turns on, there is an alternate control topology but it would be prone to self-destructing if the bias circuits messed up. The biases will not require any voltages outside of the main power rails so the bias circuits should be relatively simple.
Since this circuit has voltage differential feedback it will inherently be a current amplifier, how do current amplifiers sound? It should cancel any phase shift that the speaker inductance introduces but does that have any noticeable effect?
I have also sketched out the control scheme, the arrows indicate the direction in which the class-D amplifier's output has departed from the input curve. The action is different depending on whether the output is currently negative or positive, I still need to sort out what happens at the crossover but I think it can be done smoothly. The ground-pull transistor is fully on unless the output is too high. For example on this diagram if the output has increased beyond the target then the ground-pull transistor is turned off, reducing the voltage across the output. If the output is too low on a positive cycle then it will turn on the negative pull-down transistor which will increase the voltage difference across the output and the opposite for a negative cycle.
Only one bias circuit and set of transistors will be on at a time, controlled by the bias control circuit which is in turn controlled by the error amplifier, there will be some crossover to prevent zero-crossing jumps but it should be minimised to prevent power dissipation.
Since this circuit has voltage differential feedback it will inherently be a current amplifier, how do current amplifiers sound? It should cancel any phase shift that the speaker inductance introduces but does that have any noticeable effect?
I have also sketched out the control scheme, the arrows indicate the direction in which the class-D amplifier's output has departed from the input curve. The action is different depending on whether the output is currently negative or positive, I still need to sort out what happens at the crossover but I think it can be done smoothly. The ground-pull transistor is fully on unless the output is too high. For example on this diagram if the output has increased beyond the target then the ground-pull transistor is turned off, reducing the voltage across the output. If the output is too low on a positive cycle then it will turn on the negative pull-down transistor which will increase the voltage difference across the output and the opposite for a negative cycle.
Only one bias circuit and set of transistors will be on at a time, controlled by the bias control circuit which is in turn controlled by the error amplifier, there will be some crossover to prevent zero-crossing jumps but it should be minimised to prevent power dissipation.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Ah no sorry, I barely even remember making this post. I just **** ideas out into the internet that I think people might be able to use. People that try my ideas find they usually work well but there are very few that I actually build myself. No guarantees that any of them are good, but I do guarantee that they are at least close to something that is.
This one was made with the idea that analog amplifies can swing much faster and can perform feedback way faster than digital ones and so should be able to correct the errors of one. A very well designed digital amplifier should be close enough though and will not have the low efficiency of a analog circuit. So this idea would only be worth it for making a very accurate very high power amplifier if you did not have the signal processing and control theory knowledge to build a very accurate digital amplifier. Just buy a powersoft amp tbh, I am literally shocked by how good those things are.
I'm more into speaker design, other electronics, and chemistry these days.
This one was made with the idea that analog amplifies can swing much faster and can perform feedback way faster than digital ones and so should be able to correct the errors of one. A very well designed digital amplifier should be close enough though and will not have the low efficiency of a analog circuit. So this idea would only be worth it for making a very accurate very high power amplifier if you did not have the signal processing and control theory knowledge to build a very accurate digital amplifier. Just buy a powersoft amp tbh, I am literally shocked by how good those things are.
I'm more into speaker design, other electronics, and chemistry these days.
Actually looking at this again it is will introduce a lot of zero-cross distortion that might not be able to be compensated for by the feedback circuit and the power of the digital amplifier cannot be leveraged since the entire drive current flows though the analog stage. The analog stage needs to be coupled in by a different method, possibly inductively but that would require major compensation in the feedback circuit, resistively would again loose efficiency although much less and seems the closest to making this practical.
Better idea is get a feedback coil driver, use a digital amp on the main coil, use an analog amp on the feedback coil, and use an inertial feedback and error circuit to drive the analog amp, basically like a high performance galvanometer driver. I'm thinking of subs though now since that is what I help with now. The high end has been perfected by people who know wayyy more about this than me or anybody I know. It's only low end sub systems that are worth DIYing anymore since professional ones are rather expensive for the number of them that you need for an incredible system.
Better idea is get a feedback coil driver, use a digital amp on the main coil, use an analog amp on the feedback coil, and use an inertial feedback and error circuit to drive the analog amp, basically like a high performance galvanometer driver. I'm thinking of subs though now since that is what I help with now. The high end has been perfected by people who know wayyy more about this than me or anybody I know. It's only low end sub systems that are worth DIYing anymore since professional ones are rather expensive for the number of them that you need for an incredible system.
please sam post a schematic diagram with components , will your amplifier be efficient like labgruppens class td?? also can it handle low impendance loads like 1 ohm load??
Wow no sorry, I never made an actual circuit for this to my knowledge, It will not be as efficient as any sort of class D or T due to the analog stage, and the analog correction will have to be too high a voltage for the system to have reasonably efficiency at low impedances. After seeing much more of the technology that is out there I believe this is a pretty pointless idea, it's only purpose would be to add accuracy to a low quality digital amplifier and digital amplifiers are so good these days there really is no need for anything of this type, even if it does work well.
Driving 1 ohm loads is quite hard in terms of making a stable system, it is definitely achievable for hobbyists but I would recommend using an expertly designed circuit for such purposes. I am a dabbler, if I wasn't me I wouldn't build anything I designed at this point. I made that post a long time ago out of curiosity, I mainly wanted feedback on the concept.
The other idea that I just mentioned may work well and be relatively efficient (still not on the level of a fully digital amp though) but would require a custom speaker. I am considering making an inertial feedback speaker system but I will not be using an analog correction stage since that would mean that it would require a custom amplifier and I envisage a system that can be added to any regular amplifier rack. Who knows if I will ever do that though, I have so many projects on the go...
If you really want to make something like this though I may be able to help out but I am not a professional, I just substitute intelligence for skill which only gets me so far.
Driving 1 ohm loads is quite hard in terms of making a stable system, it is definitely achievable for hobbyists but I would recommend using an expertly designed circuit for such purposes. I am a dabbler, if I wasn't me I wouldn't build anything I designed at this point. I made that post a long time ago out of curiosity, I mainly wanted feedback on the concept.
The other idea that I just mentioned may work well and be relatively efficient (still not on the level of a fully digital amp though) but would require a custom speaker. I am considering making an inertial feedback speaker system but I will not be using an analog correction stage since that would mean that it would require a custom amplifier and I envisage a system that can be added to any regular amplifier rack. Who knows if I will ever do that though, I have so many projects on the go...
If you really want to make something like this though I may be able to help out but I am not a professional, I just substitute intelligence for skill which only gets me so far.
- Status
- Not open for further replies.