After following some discussions in http://www.diyaudio.com/forums/showthread.php?s=&threadid=25005 and http://www.diyaudio.com/forums/showthread.php?s=&threadid=36528, I think there is a collective need to have something to point to for people getting interested in class D amplifiers. Just as there seems to be rough consensus about pointing to the gainclone for a classic linear design...
This should be something that would allow someone with some knowledge and tools to get a "good" sounding amplifier working quite easily. (Again, comparable to a simple gainclone setup.)
So, a "reference starting point" rather than "reference high-end sound" in the first spin.
As for requirements, I guess moderate power abilities would be sufficient, somewhere in the range of 50-100W? Single-voltage power supply is of course nice, if it is possible to design the amplifier to cope with that. But I would say that low amplifier part count and simplicity in theory and operation should be equally important.
I have seen some references to do-it-all-chips that might satisfy the requirements above. But only chips that are readily available for DIY purchases should qualify.
Alternatively, a UCD design with as little as one op-amp, a comparator, a mosfet driver circuit plus two mosfets could do the trick. (Add some passives too
)
Personally, I think the latter path would be more instructive, as you expose the interesting operation better that way.
(In any case, we don't have to worry about patents for a DIY design!)
With this little introduction, I'd like to open the floor for suggestions. My intention with starting this thread is to let it flow, and see if we can develop rough consensus, first on requirements, and then on a concrete design...
Regards / Johan Sörensen
This should be something that would allow someone with some knowledge and tools to get a "good" sounding amplifier working quite easily. (Again, comparable to a simple gainclone setup.)
So, a "reference starting point" rather than "reference high-end sound" in the first spin.
As for requirements, I guess moderate power abilities would be sufficient, somewhere in the range of 50-100W? Single-voltage power supply is of course nice, if it is possible to design the amplifier to cope with that. But I would say that low amplifier part count and simplicity in theory and operation should be equally important.
I have seen some references to do-it-all-chips that might satisfy the requirements above. But only chips that are readily available for DIY purchases should qualify.
Alternatively, a UCD design with as little as one op-amp, a comparator, a mosfet driver circuit plus two mosfets could do the trick. (Add some passives too
)Personally, I think the latter path would be more instructive, as you expose the interesting operation better that way.
(In any case, we don't have to worry about patents for a DIY design!)
With this little introduction, I'd like to open the floor for suggestions. My intention with starting this thread is to let it flow, and see if we can develop rough consensus, first on requirements, and then on a concrete design...
Regards / Johan Sörensen
Hi,
Again, Good idea!
I think the power might be a bit much for a "newbie" design? Some who try will be experienced.....others might be high school kids..do they know of deadtime and the like? My thoughts are, keep the explosions small.....
Let's keep it ultra simple from the very start and just set +-12V rails as a requirement (op amps...etc)
A 1 comparator, with buffered output could possibly drive a simple AC coupled N+P channel half bridge.
While the patented UCD type drive circuit features some excellent characteristics with what the end goal was for it, I think in this situation a more robust "break before make" buffered, AC coupled N+P channel implementation would be most suitable, if the delay of it weren't too bad. This would help eliminate alot of possible problems with it working properly.
I agree you can learn more from the UCD, a nice self oscillating example for everyone not involving the complications of hysteresis, and simple as it gets.
I'd like to keep this whole thing simulationable as well....and protoboardable. It should sound good, and be simple enough to make them fall in love with class D, in about ten minutes, and go from there.
Hopefully if/when we get the class D forum we'll move this and several other good relevant threads to it, this should turn into a nice sticky for it.
Then we'll work on another for the ultimate quality, no compromise, colaborated patentedless diy class d design
Regards,
Chris
Again, Good idea!
I think the power might be a bit much for a "newbie" design? Some who try will be experienced.....others might be high school kids..do they know of deadtime and the like? My thoughts are, keep the explosions small.....
Let's keep it ultra simple from the very start and just set +-12V rails as a requirement (op amps...etc)
A 1 comparator, with buffered output could possibly drive a simple AC coupled N+P channel half bridge.
While the patented UCD type drive circuit features some excellent characteristics with what the end goal was for it, I think in this situation a more robust "break before make" buffered, AC coupled N+P channel implementation would be most suitable, if the delay of it weren't too bad. This would help eliminate alot of possible problems with it working properly.
I agree you can learn more from the UCD, a nice self oscillating example for everyone not involving the complications of hysteresis, and simple as it gets.
I'd like to keep this whole thing simulationable as well....and protoboardable. It should sound good, and be simple enough to make them fall in love with class D, in about ten minutes, and go from there.
Hopefully if/when we get the class D forum we'll move this and several other good relevant threads to it, this should turn into a nice sticky for it.
Then we'll work on another for the ultimate quality, no compromise, colaborated patentedless diy class d design
Regards,
Chris
Let's give my newbie's opinion
If you want something really easy to build, not too powerful (but still more than 20W)l, and gainclone like, that's what you need:
http://www.national.com/pf/LM/LM4651.html
But I don't see the advantage in using a class-D amp for something like 50-100W. Yes, maybe for 75-100 it becomes interesting, but for 50-75 the efficiency (heat) of a class AB amp isn't relly a problem.
I, personally, am looking for 200W in 8 Ohms. But I won't ask this to be the output power of the reference design. If most of us want this, that's ok, but if the majority is satisfied with less, I'll agree.
An easy project was my first goal, but if we design all together an UcD amp (with a PCB), that's nice! Like someone else said, soldering components on a PCB isn't really a difficult task.
Now, something more aimed to the design itself: what about a full bridge? In some tutorials, I've read about the advantages of a full mosfet bridge. Is it incompatible with the UcD concept?
If you want something really easy to build, not too powerful (but still more than 20W)l, and gainclone like, that's what you need:
http://www.national.com/pf/LM/LM4651.html
But I don't see the advantage in using a class-D amp for something like 50-100W. Yes, maybe for 75-100 it becomes interesting, but for 50-75 the efficiency (heat) of a class AB amp isn't relly a problem.
I, personally, am looking for 200W in 8 Ohms. But I won't ask this to be the output power of the reference design. If most of us want this, that's ok, but if the majority is satisfied with less, I'll agree.
An easy project was my first goal, but if we design all together an UcD amp (with a PCB), that's nice! Like someone else said, soldering components on a PCB isn't really a difficult task.
Now, something more aimed to the design itself: what about a full bridge? In some tutorials, I've read about the advantages of a full mosfet bridge. Is it incompatible with the UcD concept?
I would build it
I've been interested in class d for a while, but they are unfamiliar territory, and look very complicated. I thought that about smps though, and now I'm familiar with them.
A major stumbling block for a beginner (such as myself) is the inductor, so for this design an easily available core that is simple to wind and not-so-critical would be another requirement.
I've been interested in class d for a while, but they are unfamiliar territory, and look very complicated. I thought that about smps though, and now I'm familiar with them.
A major stumbling block for a beginner (such as myself) is the inductor, so for this design an easily available core that is simple to wind and not-so-critical would be another requirement.
Is a core required for a class d output filter coil? We can throw in the info required to wind an air core one easily enough, no one will be faced with a problem of attempting it with different/unknown cores this way. Cheap quick and easy!
Johan I should tell, you a few weeks ago I hit your site and discovered your newest project, I wired up one of the circuits from it...version A probably. Took me 10 minutes to get it simulating.....it did simulate.....not well but for 10 minutes worth of work on it..
It then took me 5 minutes to wire it on a cheap plastic protoboard with parts I made myself (air core inductor) a few caps I cut off from an old motherboard and whatever else I had around.
It didn't work great, I blame my cheap implementation of it of course, it wouldn't play the full audio band, actually the mid range just died....highs and lows were ok, and it had a bit of a crackle at times. I also couldn't find a cap value low enough for the integrator, or had any testing equipment to try and tune it with.
It still had a nice class D type sound which I like, I hooked it up to one of my cerwin vegas and it was the first time in about a year I got to hear it!
Anyway, I think I might have misunderstood the intentions of this thread. Is it to produce a good reference design to make it easy to have a good sounding class d for anybody, or, so people can learn from it easiest. If it's the prior, standard UCD circuit as known on this forum, with an integrated comparator+input buffer.
Still has the potential for very high power.
Either choice you make, unless maybe the mueta comes out tomorrow, I think the "all in" IC's should be out, as for the controller aspect, you learn nothing by it, as per the output stage aspect, seriously limitting the circuit by it, and again, learn nothing. Let's stay out of lego land on this one.
Johan I'll try to help you with any simulation work you require, I get better with it every day.
Regards,
Chris
Johan I should tell, you a few weeks ago I hit your site and discovered your newest project, I wired up one of the circuits from it...version A probably. Took me 10 minutes to get it simulating.....it did simulate.....not well but for 10 minutes worth of work on it..
It then took me 5 minutes to wire it on a cheap plastic protoboard with parts I made myself (air core inductor) a few caps I cut off from an old motherboard and whatever else I had around.
It didn't work great, I blame my cheap implementation of it of course, it wouldn't play the full audio band, actually the mid range just died....highs and lows were ok, and it had a bit of a crackle at times. I also couldn't find a cap value low enough for the integrator, or had any testing equipment to try and tune it with.
It still had a nice class D type sound which I like, I hooked it up to one of my cerwin vegas and it was the first time in about a year I got to hear it!
Anyway, I think I might have misunderstood the intentions of this thread. Is it to produce a good reference design to make it easy to have a good sounding class d for anybody, or, so people can learn from it easiest. If it's the prior, standard UCD circuit as known on this forum, with an integrated comparator+input buffer.
Still has the potential for very high power.
Either choice you make, unless maybe the mueta comes out tomorrow, I think the "all in" IC's should be out, as for the controller aspect, you learn nothing by it, as per the output stage aspect, seriously limitting the circuit by it, and again, learn nothing. Let's stay out of lego land on this one.
Johan I'll try to help you with any simulation work you require, I get better with it every day.
Regards,
Chris
coils / ref design
Air coils are dangerous stuff in class D. They were used successfully in the TacT Millennium but that thing didn't have feedback, so the magnetic field emanating from the coil didn't have a control loop to upset.
I used a "quasi air coil" in one design. It's an air coil of 20uH but I enclose it in P-core halves with the inside ground out completely (so I get a real hollow pot). The pot acts as shielding and increases the inductance by only 50%.
Needless to say that this construction is more exotic than a normal ferrite cored coil.
An interesting take could be writing a piece on how to make best out of any ferrite core salvaged from smps or bought on a dump. Once you know how to estimate current capability and optimum air gap for a toroid using little more than a caliper and some basic maths (and have figured how to grind in the air gap!), it starts becoming fun.
Alternatively someone could lay hands on a minimum ordering quantity of core halves (with pre-ground air gap) suitable for the 50-200W power range and sell unit quantities e.g. through ebay.
A practical "design cell" could indeed be a full-bridge power stage with a comparator input. All sorts of control loops can be fitted to that. I can provide some nice ones, except unfortunately the latest versions of UcD because people are dishing out neat sums for licenses. Even though it's covered by the patent, it's a healthy idea not to give out everything lest the patent become hard to enforce.
Air coils are dangerous stuff in class D. They were used successfully in the TacT Millennium but that thing didn't have feedback, so the magnetic field emanating from the coil didn't have a control loop to upset.
I used a "quasi air coil" in one design. It's an air coil of 20uH but I enclose it in P-core halves with the inside ground out completely (so I get a real hollow pot). The pot acts as shielding and increases the inductance by only 50%.
Needless to say that this construction is more exotic than a normal ferrite cored coil.
An interesting take could be writing a piece on how to make best out of any ferrite core salvaged from smps or bought on a dump. Once you know how to estimate current capability and optimum air gap for a toroid using little more than a caliper and some basic maths (and have figured how to grind in the air gap!), it starts becoming fun.
Alternatively someone could lay hands on a minimum ordering quantity of core halves (with pre-ground air gap) suitable for the 50-200W power range and sell unit quantities e.g. through ebay.
A practical "design cell" could indeed be a full-bridge power stage with a comparator input. All sorts of control loops can be fitted to that. I can provide some nice ones, except unfortunately the latest versions of UcD because people are dishing out neat sums for licenses. Even though it's covered by the patent, it's a healthy idea not to give out everything lest the patent become hard to enforce.
Re: coils / ref design
Hi Bruno,
I presume simple shielding of the air coir coil (tinfoil....gum) wouldn't be sufficient to keep the high frequency field from leaking out? That ruins my design..well..I can still use the foil and gum anyway. That brings up the question of core saturation...I'll research that.
Core properties can be determined experimentally as well but from what I've seen it's alot of work /time. I do like the salvaging idea, but your bulk buy idea is the better one for this topic, might be best to buy them all made up if possible, saw some just yesterday but for SMD.
I think we all have no difficulty understanding and respecting the position you're in regarding the patents. Your help is not at all expected for that very reason, yet highly regarded, and most welcome when given.
We just need to nail down the design goals for this thing.
Regards
Chris
Bruno Putzeys said:
Hi Bruno,
I presume simple shielding of the air coir coil (tinfoil....gum) wouldn't be sufficient to keep the high frequency field from leaking out? That ruins my design..well..I can still use the foil and gum anyway. That brings up the question of core saturation...I'll research that.
Core properties can be determined experimentally as well but from what I've seen it's alot of work /time. I do like the salvaging idea, but your bulk buy idea is the better one for this topic, might be best to buy them all made up if possible, saw some just yesterday but for SMD.
I think we all have no difficulty understanding and respecting the position you're in regarding the patents. Your help is not at all expected for that very reason, yet highly regarded, and most welcome when given.
We just need to nail down the design goals for this thing.
Regards
Chris
I've been wanting to have a play with class-D (and SMPS) for a while, so I'd be interested. Also, I may be able to help with the core situation as my dad used to be a partner in a small transformer manufacturer, he's now retired but still in contact with the old partner.
I'd be looking for >100W to make it more attractive than class-B, single-sided would be nice so that I can make my own boards easily.
I'd be looking for >100W to make it more attractive than class-B, single-sided would be nice so that I can make my own boards easily.
Trying to shield an air coil with tin foil is not a good option... It constitutes a shorted winding! You'd need to put the foil at several cm distance from the coil to make sure most of the magnetic field is enclosed. That would be some construction.
Interestingly, the exact properties of the ferrite material are not very important in gapped cores. Nearly all saturate at similar levels (300mT) and the inductance is determined by the air gap (effective mu << ferrite mu). Core losses are a different thing, but unless you're extremely unlucky it won't be a big issue.
So it really comes down to geometry.
Making class d amps on single-sided PCBs is not something you should do if you're new to class D. It's also not something you would do if you're experienced. It's possible up to 10 or 20 watts using an IC solution, but that's as far as I've gone.
Have a look at the prices of www.thepcbshop.com . Making your own pcbs has become a waste of time since they arrived.
Interestingly, the exact properties of the ferrite material are not very important in gapped cores. Nearly all saturate at similar levels (300mT) and the inductance is determined by the air gap (effective mu << ferrite mu). Core losses are a different thing, but unless you're extremely unlucky it won't be a big issue.
So it really comes down to geometry.
Making class d amps on single-sided PCBs is not something you should do if you're new to class D. It's also not something you would do if you're experienced. It's possible up to 10 or 20 watts using an IC solution, but that's as far as I've gone.
Have a look at the prices of www.thepcbshop.com . Making your own pcbs has become a waste of time since they arrived.
Hi,
Sorry, can't help it, I have to ask!
How about counter wound air core coils in series, I've seen it used in a few amps. Worth doing?
Hi richie00boy, I think we're looking at a near manditory dual layer board anyway, helps ensure it works well, especially when crawling up to the power levels people seem to want out of this.
Sorry, can't help it, I have to ask!
How about counter wound air core coils in series, I've seen it used in a few amps. Worth doing?
Hi richie00boy, I think we're looking at a near manditory dual layer board anyway, helps ensure it works well, especially when crawling up to the power levels people seem to want out of this.
classd4sure said:
That would provide some relief in the far field but the kind of dimensions we're talking about for a 100W amp (6x6cm?) is not "far field". The Q of the inductors would leave something to be desired because of their partial coupling, mutually and towards the ground plane.
I wouldn't settle for anything less than a closed magnetic circuit.
Hi,
Yeah don't we all...I just checked out the two I have again...I don't think you'll find a matched set of them on it should you want 2 channels later. Same with the motherboards they powered, all different sizes.
Going from a large manufacturer of them that's been around a long time is a great idea, bulk buy or otherwise....I'll just cut one off the psu for now
Yeah don't we all...I just checked out the two I have again...I don't think you'll find a matched set of them on it should you want 2 channels later. Same with the motherboards they powered, all different sizes.
Going from a large manufacturer of them that's been around a long time is a great idea, bulk buy or otherwise....I'll just cut one off the psu for now
These are either iron powder or ferrite. In the former case, they are invariably a "soft saturation" kind which produces tons of distortion even at low levels. In the latter case, an air gap needs to be ground in.Bricolo said:
The cores Farnell has have no air gaps so some work is needed there too.
While trying to find info on how to machine ferrite I even heard large companies telling me it's impossible unless you have very specialised machines and tooling.
In the end I managed to order diamond coated burrs in a local shop (20 euros apiece or so. Manufacturer is called Pferd). They work fine for making air gaps in core halves. My drill stand, at 3000rpm is a bit slow though, which results in chipping.
For toroids I use grinding discs from Dremel (1mm thick or so) in a drill stand at slow speeds. It takes about 5 minutes to get through a toroid but it makes a clean cut.
I once thought about using an "air-cored" toroid. The disadvantage would be the high amount of turns needed (=> skin-effect losses).
I once posted the formulae I used for determining the air gaps within my thesis. Maybe it is quicker to search them on my PC than to find the right thread on the forum. The accuracy of the derived inductance relied on the core type. It was better with RM cores than with EFD cores. It should work quite nicely with toroids. I guess this was due to stray fields etc. You will have to measure inductance anyway (not that dificult at all) and add or remove some turns. Simple as that.
Edit: forgot to mention that I introduced the air gaps by the insertion of insulating material between the two core-halves.
In the meantime I found the post:
http://www.diyaudio.com/forums/showthread.php?postid=196536#post196536
The schematic diagram of my thesis' amp is found here. Under today's circumstances it is neither sexy nor refined but should serve well as an easy example for the understanding of PWM amps. But one has to add that it was developed when there was not much literature on class-d around and driver ICs weren't as good as today (neither were affordable OP-AMPs). The IR 2110 was already available but it had worse figures than today's versions, otherwise I'd have gone with that (I actually did experiment with it).
http://www.diyaudio.com/forums/showthread.php?postid=194684#post194684
Regards
Charles
I once posted the formulae I used for determining the air gaps within my thesis. Maybe it is quicker to search them on my PC than to find the right thread on the forum. The accuracy of the derived inductance relied on the core type. It was better with RM cores than with EFD cores. It should work quite nicely with toroids. I guess this was due to stray fields etc. You will have to measure inductance anyway (not that dificult at all) and add or remove some turns. Simple as that.
Edit: forgot to mention that I introduced the air gaps by the insertion of insulating material between the two core-halves.
In the meantime I found the post:
http://www.diyaudio.com/forums/showthread.php?postid=196536#post196536
The schematic diagram of my thesis' amp is found here. Under today's circumstances it is neither sexy nor refined but should serve well as an easy example for the understanding of PWM amps. But one has to add that it was developed when there was not much literature on class-d around and driver ICs weren't as good as today (neither were affordable OP-AMPs). The IR 2110 was already available but it had worse figures than today's versions, otherwise I'd have gone with that (I actually did experiment with it).
http://www.diyaudio.com/forums/showthread.php?postid=194684#post194684
Regards
Charles
I fully agree about double-sided PCB - it's a neccessary evil 
.
Regarding filter coil(s), that's certainly an interesting area, where suitable recommendations will have to be developed. Possibly both pointing to ready-made parts, and instructions on how to DIY.
In my experience, air-wound cores can work with good results, at least up to the power levels I've been tinkering with (my current amp. uses +/- 55V for the power stage - unregulated of course).
That said, I am not advocating them, and I'm hoping to achieve equal or better results, with a much smaller size, using toroid cores instead.
Doesn't the UCD design counteract the non-linearity in the filter core to a large degree? Or is the core material still critical?
.Regarding filter coil(s), that's certainly an interesting area, where suitable recommendations will have to be developed. Possibly both pointing to ready-made parts, and instructions on how to DIY.
In my experience, air-wound cores can work with good results, at least up to the power levels I've been tinkering with (my current amp. uses +/- 55V for the power stage - unregulated of course).
That said, I am not advocating them, and I'm hoping to achieve equal or better results, with a much smaller size, using toroid cores instead.
Doesn't the UCD design counteract the non-linearity in the filter core to a large degree? Or is the core material still critical?
johanps said:I fully agree about double-sided PCB - it's a neccessary evil
Doesn't the UCD design counteract the non-linearity in the filter core to a large degree? Or is the core material still critical?
We could use the components side as a pure ground plane, so it wouldn't cause much trouble for diying double sided boards (having to align the 2 sides...)
As for UcD and non-linearity. See it as feedback on a linear amp.
Try to correct a heavilly non linear amp (or stage) with some feedback: it will work, but won't be as good as a much more linear stage corrected with the same amount of feedback
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.