I built my first working push-pull, as opposed to single ended, ClassD amp using two 6N137s for level shifting. Those might be fast enough to try and see. It did have the advantage of isolating the outputs from the low-level stage in the reverse direction. I found that a 1.5kohm resistor and a .1uf filter capacitor was all that I needed to provide operating current from the 12v Mosfet drive supply for each opto. There is a good chance that if I build a new amp circuit, I will again use the 6N137, but this time with discrete MOSFET drivers. I find that the opto also provides ease in configuring the signal transfer from the output of the comparator. It is great from the perspective of my p2p addiction as well.
Cheers Sub,
I've pulled out the self oscillating design i had put away and i've been playing with some optos. They work well apart from being a bit slow, but it just so happens i also have some 6N137's laying around. Nice cream cases, according to the spec sheet they can have up to 100n/s delay. Thats on the limit, but i'll try them. Using the slower 4 legged devices i bunged in ( comosK1010 ) i now have two N channel output fets instead of P/N ( Hooray! ) and the capability to use almost any voltage i want in the self osc. I have some magnetic couplers at work somwere, i've been looking for them but i must have filed them in a black hole. They are supposed to be super fast but i'm a bit warey of any coupling that may take place with other things like the output inductor or anything elsr]e with stray flux.. Hmm, If this 6N137 is any good i'll use that, supposed to be designed to run up to 10Mbit sec. We shall see.....
Mad.P
I've pulled out the self oscillating design i had put away and i've been playing with some optos. They work well apart from being a bit slow, but it just so happens i also have some 6N137's laying around. Nice cream cases, according to the spec sheet they can have up to 100n/s delay. Thats on the limit, but i'll try them. Using the slower 4 legged devices i bunged in ( comosK1010 ) i now have two N channel output fets instead of P/N ( Hooray! ) and the capability to use almost any voltage i want in the self osc. I have some magnetic couplers at work somwere, i've been looking for them but i must have filed them in a black hole. They are supposed to be super fast but i'm a bit warey of any coupling that may take place with other things like the output inductor or anything elsr]e with stray flux.. Hmm, If this 6N137 is any good i'll use that, supposed to be designed to run up to 10Mbit sec. We shall see.....
Mad.P
Hi nitrate, I have considered magnetic couplers for use in classD. I have some HCPL-9000 on hand for future experimentation, but have no experience with them at all. They are rated for up to around 100mb/s, 12nS propagation delay--slightly slower than the ones to which Charles linked.
Indeed, self-oscillating is my planned approach. Since I have every intention of using negative feedback, this way tends to use a simpler circuit.
It's great that you have a dual N-ch amp now.
The opto enables the the audio to be referenced to ground while having equal circuit characteristics for producing positive and negative output signals.
One's first impression of the 6N137 could be be that it has internal hysteresis because of its association with digital circuitry. But I examined one in a simple test circuit and concluded that it has linear rather than digital characteristics. Maybe I could put it differently: it displayed limited gain and did not automatically produce an either high or low output. This property should be helpful in preventing dead time distortion without needing high slew signals to overcome hysteresis. Therefore, low signal level circuitry should have fewer design constraints. I think, in fact, that in a suitable circuit, its slowing effect may actually be much less than one would expect so long as it could be held in its linear region.
My feeling is that a good way to get a clean classD amp is to basically use an ultra high speed linear amp. I do see no prohibitive reason why most of the circuitry can't have either a high or low state, but I think it opens the door to trouble with timing errors between the high and low outputs of the amp.
Indeed, self-oscillating is my planned approach. Since I have every intention of using negative feedback, this way tends to use a simpler circuit.
It's great that you have a dual N-ch amp now.
The opto enables the the audio to be referenced to ground while having equal circuit characteristics for producing positive and negative output signals.One's first impression of the 6N137 could be be that it has internal hysteresis because of its association with digital circuitry. But I examined one in a simple test circuit and concluded that it has linear rather than digital characteristics. Maybe I could put it differently: it displayed limited gain and did not automatically produce an either high or low output. This property should be helpful in preventing dead time distortion without needing high slew signals to overcome hysteresis. Therefore, low signal level circuitry should have fewer design constraints. I think, in fact, that in a suitable circuit, its slowing effect may actually be much less than one would expect so long as it could be held in its linear region.
My feeling is that a good way to get a clean classD amp is to basically use an ultra high speed linear amp. I do see no prohibitive reason why most of the circuitry can't have either a high or low state, but I think it opens the door to trouble with timing errors between the high and low outputs of the amp.
Hi.
I agree with Charles that NVE magnetic isolator is superior solution, but it has a minor drawback. Since it uses differentiated pulses to transmit information across the barrier, it needs to be proprerly initalized by input pulse. This is a problem with self oscillating designs, since they need DC feedback to start reliably.
Agilent makes HCPL-7723 which is almost as good as NVE solution, but can also transmit DC information.
Best regards,
Jaka Racman
I agree with Charles that NVE magnetic isolator is superior solution, but it has a minor drawback. Since it uses differentiated pulses to transmit information across the barrier, it needs to be proprerly initalized by input pulse. This is a problem with self oscillating designs, since they need DC feedback to start reliably.
Agilent makes HCPL-7723 which is almost as good as NVE solution, but can also transmit DC information.
Best regards,
Jaka Racman
IIRC there was even a MOSFET driver with built-in magnetic coupler. But I can't remember by which manufacturer.
I also once experimented with 6N137s. But they were only used in experimental circuits so I didn't care much about their signal inversion properties (or where I made the compensating inversion). But if you intend to use them in a circuit for everyday use then use them together with inverting drivers. Or you will encounter unintended turn-on of your output devices in some situations.
Regards
Charles
I also once experimented with 6N137s. But they were only used in experimental circuits so I didn't care much about their signal inversion properties (or where I made the compensating inversion). But if you intend to use them in a circuit for everyday use then use them together with inverting drivers. Or you will encounter unintended turn-on of your output devices in some situations.
Regards
Charles
Hey Chris,
If you're still watching this thread i just thought i'd say i've been studying your schematic that you posted a link to earlier in this thread. I'm very impressed at the way you have sorted out the drive to your output FET's. I am thinking of using somthing similar in my oscillating amp, but coupled to the front end with an opto, as long as you don't mind me using your idea. At the mo i have some small zetex's and an opto driving the fets but zetex's get warm. I'm determined to get this thing going now as i seem to have hit limits with the carrier design ( not that it sounds bad ). I might have a rummage around the pile of old network cards later to see if i can find any fast couplers
Mad.P
If you're still watching this thread i just thought i'd say i've been studying your schematic that you posted a link to earlier in this thread. I'm very impressed at the way you have sorted out the drive to your output FET's. I am thinking of using somthing similar in my oscillating amp, but coupled to the front end with an opto, as long as you don't mind me using your idea. At the mo i have some small zetex's and an opto driving the fets but zetex's get warm. I'm determined to get this thing going now as i seem to have hit limits with the carrier design ( not that it sounds bad ). I might have a rummage around the pile of old network cards later to see if i can find any fast couplers
Mad.P
Hi all,
I've succsefully been using optos for the signal level conversion in my 2nd class D amp. I'm very happy with the results, they work well to remove a lot of problems i was having using capacitive coupling on my old amp. I have ordered some magnetic couplers at work and as soon as they arrive i'll swap them for the opto's. Not that there is anything fundimentally wrong with the opto's, i just find them too slow. Does anyone have any tips about speeding up turn off times of fets? I'm having trouble getting my driver fets to turn off in less than 300n/s. Am i correct in thinking that if a fet turns on with 3 volts on its gate, my putting 12 volts into the gate is just making things worse as i have to discharge all the surplus charge when i come to turn the thing off? Is this what my problem could be and can i reduce turn off times by limiting the gate charge with a zenner? I've been looking at the reference amp drive technique and i can't see what the bat64 shockety is doing... i imagine it has somthing to do with speeding up the trannys turn off time but why? It looks as if it is just going to short out the drive input.
P.S sombody on here once told me that self oscillating designs were better than carrier ones. Well my second design is self oscillating and i have it working properly now. After building both types i have been placed in a position were i can compare both types side by side. I must admit the self oscillating design sounds superior, and that's the one i'm going to stick with from now on. But i am glad i built the carrier one as i set out to find out for myself which type is best and i would not have taken anyones word for it.
Take care all,
Mad.P
I've succsefully been using optos for the signal level conversion in my 2nd class D amp. I'm very happy with the results, they work well to remove a lot of problems i was having using capacitive coupling on my old amp. I have ordered some magnetic couplers at work and as soon as they arrive i'll swap them for the opto's. Not that there is anything fundimentally wrong with the opto's, i just find them too slow. Does anyone have any tips about speeding up turn off times of fets? I'm having trouble getting my driver fets to turn off in less than 300n/s. Am i correct in thinking that if a fet turns on with 3 volts on its gate, my putting 12 volts into the gate is just making things worse as i have to discharge all the surplus charge when i come to turn the thing off? Is this what my problem could be and can i reduce turn off times by limiting the gate charge with a zenner? I've been looking at the reference amp drive technique and i can't see what the bat64 shockety is doing... i imagine it has somthing to do with speeding up the trannys turn off time but why? It looks as if it is just going to short out the drive input.
P.S sombody on here once told me that self oscillating designs were better than carrier ones. Well my second design is self oscillating and i have it working properly now. After building both types i have been placed in a position were i can compare both types side by side. I must admit the self oscillating design sounds superior, and that's the one i'm going to stick with from now on. But i am glad i built the carrier one as i set out to find out for myself which type is best and i would not have taken anyones word for it.
Take care all,
Mad.P
nitrate said:Hey Chris,
If you're still watching this thread i just thought i'd say i've been studying your schematic that you posted a link to earlier in this thread. I'm very impressed at the way you have sorted out the drive to your output FET's. I am thinking of using somthing similar in my oscillating amp, but coupled to the front end with an opto, as long as you don't mind me using your idea. At the mo i have some small zetex's and an opto driving the fets but zetex's get warm. I'm determined to get this thing going now as i seem to have hit limits with the carrier design ( not that it sounds bad ). I might have a rummage around the pile of old network cards later to see if i can find any fast couplers
Mad.P
Hi there,
Sorry I hadn't responsded to your question sooner I think I just missed it amongst the other posts.
Thanks, I too was impressed with that driver circuit when I first saw how simply it can be done.
So what you meant by "if I don't mind you using my idea"..
Is irrelevant to me because I posted it with the intent of people making good use of it to further such non-commercial work as ours.
My permission was given when posted, basically I don't own rights to it, but that all came from Bruno's patent. So if it's the Baker clamp you were referring to (not sure what you meant, it's all a real standard output stage).. The baker clamp is supposed to rob just enough current from the base to keep the BJT out of saturation. Proper device selection you'll find is critical here or it just won't do it's job and might even make it worse.
I found an alternative Baker clamp which uses two small signal 1n914 doides that very effectively cleans up the signals, at least in simulation, and also been used in my homebrew for a long time, actually it is what got the drivers to behave properly, whereas before one mosfet top or bottom would be running hotter than the other.
One Cathode to the collector, the other Cathode to the base. Your bias resistor may connect directly to the base after the diode. Connect both Anodes together, and that is your point to be driven by the comparator output.
In exentisve simulation I found this to be equally good as a properly working schottky based backer clamp, and feel it would be much more robust across all conditions. Doing the clamps this way on my homebrew got my mosfets switching right, and am still using it today.
Yeah I said self oscillating was better, still say so, now you've heard the difference, you can start asking yourself why that is.
If you had a FET that fully enhanced at 3 volts, then you'd hae a working threshold of around 1 volt, pretty well no noise or transient immunity. In other words, you'd have the wrong mosfet.
The mosfet won't fully enhance until at least 7 to 9 volts, afterwhich point overdriving the gate to 12 Volts is simply good engineering practice to ensure that the mosfet is fully enhanced everytime, giving you more/less the same low Ron values everytime.
At three volts Vgs, it's conducting current but still has a very high Ron, working as non linear voltage controlled current source, until it enters the miller region where it at least becomes a linear relation as Rds On becomes to drop with increased gate to source voltage.
These regions are nothing but nuissance to us in class d, and are a part of what using self oscillating designs makes the amp robust against, which is timming errors from everything included within the feedback loop.
You can do that digitally, OK, yeah, but never as good, elegant, or stable.
Regards,
Chris
You mean keep the output stage linear during switch transient? but would this cause shoot through?
About using an ultra high speed linear amp: I have once considered that a classd build around an ultra high speed linear amp may be better. But my transistor based thing still have 1% distortion.
About using an ultra high speed linear amp: I have once considered that a classd build around an ultra high speed linear amp may be better. But my transistor based thing still have 1% distortion.
subwo1 said:
I had the 6N137 in mind as far as the device to be kept in the linear region. The rest of the MOSFET driver would need to have a fast slew rate. I am not sure if I could get a real circuit to work well this way, but as far as I can tell, the simulation works well. I think a good high speed opamp could be gotten to drive both optocoupler diodes.
I think the problem could have something to do with the feedback loop if the amp is able to operate smoothly from the 10s of Hz to several MHz range. There could also be things like circuit layout and noise which begins to detract from clean high frequency operation.
Hello nitrate
Your on the good way I think.
I built class d amplifier for about 5 years and I was on your ways. Visit my web site at www.d-amp.com
I use feedback after the output coil to to maximise sonic performance,,your right, feedback before output coil sound very ''harsh''. Stability is just about zobel circuit at the output and limitation in feedback bandwith...Fews experience will show the ''how too'. The only inconvenient is the phase shift that ''****'' a bit the hight frequency above 10 Khz, on the scope, but I dont know why, distortion analyser show under 1% distortion at that frequency....Some one cane make some try on it?? Anyways, this give a very warm sound...Great? no?
Have a nice day every one! Sleep time for me, too much experiment!
Fred
Your on the good way I think.
I built class d amplifier for about 5 years and I was on your ways. Visit my web site at www.d-amp.com
I use feedback after the output coil to to maximise sonic performance,,your right, feedback before output coil sound very ''harsh''. Stability is just about zobel circuit at the output and limitation in feedback bandwith...Fews experience will show the ''how too'. The only inconvenient is the phase shift that ''****'' a bit the hight frequency above 10 Khz, on the scope, but I dont know why, distortion analyser show under 1% distortion at that frequency....Some one cane make some try on it?? Anyways, this give a very warm sound...Great? no?
Have a nice day every one! Sleep time for me, too much experiment!
Fred
Hi Fred,
We can take all kinds of guesses but you know a few pictures are worth a thousand words. They also do wonders to bridge the language gaps.
Your experience is welcome here. I think so far your stuff is interesting enough to warrant its own thread, you know at least so we can follow it.
Regards,
Chris
We can take all kinds of guesses but you know a few pictures are worth a thousand words. They also do wonders to bridge the language gaps.
Your experience is welcome here. I think so far your stuff is interesting enough to warrant its own thread, you know at least so we can follow it.
Regards,
Chris
IMHO lowering of feedback is not the correct way to achieve stability. In a correctly designed class-d amp the loop gain is still significantly higher than one for frequencies above the output filter's cutoff.
If your output filter has a cutoff frequency of 18 kHz and you are feeding it 10 kHz - any harmonics from 3rd on up will be damped significantly and your measurements will show less THD than the amp actually has.
Regards
Charles
Hi all,
Since more feedback means less gain, reducing the feedbacks bandwidth is going to make the amp high gain towards out of band frequencys and this i gather will de-stableise the amp and possibly contribute to HF oscillations. So providing full bandwith gain should be the way to go to achive not only a linear freq response, but also stability?? Thats what i think anyhow. Maybe i am wrong
Anyway i've lobbed a baker clamp into the driver stage and it has halfed the amount of turn off time i am getting ( now down to 100n/s delay befor turnoff ). I don't fully understand this baker clamp theory however and i hate using stuff i don't know about so i wonder if sombody could describe whats going on here in more detail? Why do you need to use schoketys? is it due to the low volt drop?
Anyway, i'll post the schematics for the new amp soon.
Keep it up folks!
Mad.P
Since more feedback means less gain, reducing the feedbacks bandwidth is going to make the amp high gain towards out of band frequencys and this i gather will de-stableise the amp and possibly contribute to HF oscillations. So providing full bandwith gain should be the way to go to achive not only a linear freq response, but also stability?? Thats what i think anyhow. Maybe i am wrong
Anyway i've lobbed a baker clamp into the driver stage and it has halfed the amount of turn off time i am getting ( now down to 100n/s delay befor turnoff ). I don't fully understand this baker clamp theory however and i hate using stuff i don't know about so i wonder if sombody could describe whats going on here in more detail? Why do you need to use schoketys? is it due to the low volt drop?
Anyway, i'll post the schematics for the new amp soon.
Keep it up folks!
Mad.P
The Baker-clamp simply removes drive current (and voltage) from the transistor's base and therefore prevents saturation of said transistor as soon as it is turned on.
A saturated transistor can't be switched off as fast as an unsaturated one because there is more charge to be removed from the base than necessary.
Shottky diodes have two advantages over ordinary So diodes: lower forward voltage (the most important feature in this case) and better reverse recovery propoerties.
Regards
Charles
A saturated transistor can't be switched off as fast as an unsaturated one because there is more charge to be removed from the base than necessary.
Shottky diodes have two advantages over ordinary So diodes: lower forward voltage (the most important feature in this case) and better reverse recovery propoerties.
Regards
Charles
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.