Ultimate D class amplifier design in theory.... :)

[ChocoHolic]

But what if amp is made to have almost zero ringing how that should affect all?

It's all about the load situation which leads to hard switching.
In this situation the MosFet which is being turned on has to take over the full load current + remove the Qrr before the voltage of the halfbridge starts to slope. During removing Qrr the current in the MosFet can reach much higher values than the load current and at the same time the poor MosFet still is stressed with high Vds. This leads to a short but respectable loss pulse.
Furtheron the di/dt at the end of Qrr removal is extremely high.
Also dv/dt is high in this moment.
The larger the Qrr and the less soft reverse recovery characteristic the body diode shows, the more tough it will be keep resonances small in this load situation. There is no way to have zero resonance in this operating point, it can just be small enough that we do not see it anymore.

I will try Fet you mentioned

Most likely it will be very similar. In case the difference is really just a selection and dv/dt guarantee in the spec, then depending on volume share between both types there might be even no difference in many pieces.

 

[BWRX]

Well i didnt say 15A of continuous use

A 1206 size resistor will also dissipate 15W... but I'll only tell you under what conditions after you ask

If you've come up with a magical amp design then manufacture and sell it. It should sell very well if it's as good as you claim.

 

[29285]

A 1206 size resistor will also dissipate 15W... but I'll only tell you under what conditions after you ask

If you've come up with a magical amp design then manufacture and sell it. It should sell very well if it's as good as you claim.

Just another thread boasting of "my ultimate amplifier design".
No working prototype, no measurements, no numbers, no design detail, no contribution at all.
Nothing to see here.

 

[BWRX]

For good reverse recovery characteristics the GaN FETs are currently the best choice. I posted this over two years ago but this module from TI would still be interesting to build a class d amp around: TI LMG5200 half bridge power stage

 

[Mikelandjelo]

There will be difference in driver heating.I dont need any single difference in output quality.I am just fine with what i have.What about multiphase system ?Any thoughts?This IC LMG5200 is great thing.....What i am missing with it is fun .Its all done there...you need pwm.....feedback or no feedback and you have amplifier....Same cookie that you get after a fight has always better taste.Am i right? .....

@voltwide ..... Bud i never said MY ultimate .....nither i am selling something here.And if you have some expirience in D class designing and not only waiting for someone to do your homework you are welcome to share some ideas.So far discussion is about what all we know.My design is not ultimate and yes i will make sure that it will be.....in next decade of years .....I dont hurry at all.It cant hurt to discuss once all what we CANT achieve and WHY.

 

[sq225917]

Nice thread, great story bro.

 

[ChocoHolic]

This IC LMG5200 is great thing.....What i am missing with it is fun .Its all done there..

For low power LMG5200 opens nice options.
Do you feel design fun is located in the core of the switching stage only?

 

[Mikelandjelo]

Fun is in everything that will make me sweat .....once that was output stage,other time that was pwm,then protections.....now i have about 5 more things that i have fun with .... Best fun is in what i cant achieve .... so lets talk about whats not achieved in D class design if thats ok.....Probably that way we will define one ultimate design way better.

ChocoHolic my question for you is : If i have absolutely zero THD sinewave on amplifier output (20Hz-20KHz) and flat line response will i have on commercial three way speaker best output quality through whole 20Hz-20KHz range......or only one,any single speaker.....will that quality of wave guarantee best possible output quality? Why do you think that?

 

[RestAssured]

In my mind ultimate D class amplifier is something which has absolutely zero output noise without any signal on input,lowest possible distorsion,to be impossible to burn out,zero heat and it should have lowest possible quiescent curent....this last requires low switching frequency and lowest distorsion

Well, you have just described the imaginary perfect amplifier of any type. None of these requirements have anything specific to do with class D.

Since everyone is speculating, and no specific circuit has been discussed, let me say that like everything else in the known world, a class D amp has its strong and weak points. With class D, you get more output power with less heating than with a comparable class AB amp. The trade-off is that a class D amp is more complex to design and build, has more distortion and greater artifacts, and has imperfect translation from the switching stage back into audio. Which one of these drawbacks would you like to focus your design skills on today?

The class D output stage more closely resembles an RF transmitter followed by a pi-network impedance matching circuit. In order for the network to match impedances or in this case act like a low-pass filter, you must know a lot about the characteristics of the load. Therefore in my opinion the first and easiest step in class D output improvement is to design the circuit for a specific known load. Even then, you are asking a "tuned circuit" to work over a span of 3 decades or more, something that no RF engineer would consider possible.

As far as residual switching noise in the output, this is usually at frequencies of 100KHz and above, but it has the habit of becoming audible due to mixing and intermodulation, etc., not to mention that it does consume power to deliver an HF signal into the load. One method of reducing residual switching noise is "ripple steering" which makes sense, but increases the complexity of the circuitry by at least a factor of 2.

In short, you start with a basically good idea for reducing weight and improving efficiency, with a tradeoff for higher distortion levels. Then start adding costly and highly technical tweaks to the basic design. This gives you an expensive and complex device that can't easily be built at home, and usually cannot be repaired if it fails. Class D is a different solution, not a better solution.

A class D amp is really just a switching power supply that is expected to work at any audio frequency, not just one most efficient frequency. It is difficult enough to design a reliable switching power supply. Add the requirements of a class D amp and you have your hands full. That should be a sobering thought for amateur designers. Everything depends upon which benefits and which drawbacks you are willing to live with.

RA

 

[ICG]

The ultimate amplifier is also linear. And load independ. At the same time. These two things are still the biggest problems with class d amplifiers and yet nobody seemed to care much about that so far. Loudspeakers fluctuate in the impedance in a very wide impedance range (yes, even the same speaker!) and also shift the phase. And class d amps react with very poor frequency response at the upper end. Impedance compensation of the speakers are out of the question because you waste there more money on parts and a lot of time, not possible in every case either.

If you think about the ultimate amplifier, that's not a thing you can ignore. If that can't be improved vastly, the amp is not ultimate, not even close.

 

[Mikelandjelo]

ICG you are right about what you are saying partialy ..... NCore Hypex model is claimed as load independant .... flat line linear responce amplifier.About money / result ratio you are right too.What i decided is NOT to care about money ( i know it sounds stupid ) that i will spend .... becouse i want something and i will get it .....without enough money probably i will get it latter but i will....with enough all will be just fine.Good response at upper end of line is possible.....I achieved that.Older products were with poor response......Now its easy to have good reponse up to 50KHz.

What i see as drawback in D class is feedback loop ....Phase shift should be zero and its not.Gain and frequency response are in relation which should not be the case.All d class models are more or less load impedance dependant which i dont like too.We have state of the art components....yet we dont have ultimate solution so thats saying a lot about us engineers.

If i start from feedback loop my first idea was to work without it which i did (tube amps are working just fine without it).And i hit the wall .... o boy that was very funny .....and it took me a lot of time to figure out many things with a lot of spent time....which i could save if i just wanted to read something about D class before i started to design one of them .....Anyway , feedback loop is important and it looks that all whats done is not enough.Issue here is not that we dont have enough good components but ideas.So different feedback solution is needed.

If all i said about feedback is done then load dependance will be solved with it too.What i can say is that even best ever made amp will not be enough.....All about amplification is not related to what we have on output but is related to ORIGINAL sound source.....So only without feedback to ORIGINAL UNTOUCHED signal source all we have is just bad replica.In my point of view amp can be ideal only if we have man thats speaking in to microphone in one room and we can hear it on amp in other room exactly as it is sounding in his room...i mean real time transfer....any store-replay means failure for now...this concept is closest to true audio testing.Binary time domain will kill instantly any source quality no matter which digital solution is used....vinyl is great thing but we have noise.....all whats left is pure analog path....from start till end,and i mean here on input signal transfer from input to output stage.My voice never sounds from any record as me.....never...ever....I tested this on super expencive studio equipment too.So we are almost there,but not there yet.

Why we still cant record emotion in sound but we can record something thats 90% of it?We can record it but cant reproduce it maybe?Its about a time for different thinking based on all we know but with result that will give us best in relation to original source and not in relation to something that some company made and claimed as best.Dont you think same?

From here i can say something interesting.....I have will to start from everything related to signal source till its reproduction...i have all resources i need for this..to change all what i need untill i get to result i want.....I will make new kind of speakers if needed,new concept for preamplifier circuitry and same for amplifiers......This means enough fun for one life i think .When i hear my voice from something that will reproduce it and i hear my self i will have my development done.....how hard this can be ?

With other words....ALL what we have is close to original signal source but far away from where it should be.We are banging heads for this existing solutions without single thinking that we are so far away from truth which original signal source is.I decided to stop here and bang head with whats real and instantly i am at the begining and its great.

From here,money is not included in result ..... its impossible to buy non existing solution......and all the fun is there.We all have same chances so dont waist your time if you are interested in this development but lets start with work.

 

[Mikelandjelo]

"So only without feedback to ORIGINAL UNTOUCHED signal source" .... Not without but WITH ...

 

[ICG]

If all i said about feedback is done then load dependance will be solved with it too.

No, it will not. If you take the feedback from after the filter, you are feeding back the phase shift. And that's not the total phase, it's the partial phase caused by the filter. Since the amp cannot know which impedance is present at which frequency (and that can change anyway, which means a measurement beforehand doesn't solve anything), you cannot know how much the phase is shifted. That is on a purely resistive speaker, which is so far an illusion. The speaker itself will change the phase even more, you don't know how much and you don't even know in which direction. And you're feeding back the harmonics which can really quickly turn the music listener to an active broadcasting radio station. So you'd have to implement an impedance measurement, phase measurement and EMI filter which does not interfere with all that, which then all have to run in realtime and without influencing the speakers, the sound, the phase, groupdelay and can work with all existing passive speakers.

I assume it will be quicker and cheaper to research a different amplifier principle/class which got the same efficiency but works without de-facto mandatory output filters than to do the R&D till you get it working to close-to-perfection.

 

[Mikelandjelo]

"So you'd have to implement an impedance measurement, phase measurement and EMI filter which does not interfere with all that, which then all have to run in realtime and without influencing the speakers, the sound, the phase, groupdelay and can work with all existing passive speakers."

I can measure phase shift coused with output filter and i can measure phase shift coused with speaker which is sum of whole phase shifts.One cure for that shift is phase shift in feedback signal which is same as its on output of the amplifier....Impedance measurement is also not hard to be done as you described it.I designed that and have it working.You have very valid points and idea about other topology is great one.....Research and developing of new one will be the choice if i decide to quit on D class topology.

 

[ICG]

I can measure phase shift coused with output filter and i can measure phase shift coused with speaker which is sum of whole phase shifts.

No, you can't. With the different impedance the filter changes the frequency and that changes the phase angle. Yes, you can calculate the one and maybe even the second but you can't calculate which phase shift is normal for the speaker, you can only correct for a certain phase you think might be right. Speakers don't have a 'right' phase or a fixed phase. You will deviate from the behavour of i.e. a class AB amp, that is unlinearity and since the speaker will change the impedance while running (heat in vc, crossover coils etc., ferrofluid, ptc, tweeter atennuator (pot), whatever else an if it's the sun starts to shine on it), it will even change the sound within minutes. Heck, you even can create interferences between the drivers because of the wrong phase correction, which wasn't there before. Yes, that can happen, especally with supertweeters in place, you can hear the phase shift and the interferences change the frequency response very much, dips and peaks arise much quicker than you think, just play a bit with a supertweeter, a dsp and the phase. You will be quite surprised.

There is simply no way to calculate that on the fly in such a setup. At least not yet and it doesn't look like that will change anytime soon.

Impedance measurement is also not hard to be done as you described it.I designed that and have it working.

While the speaker plays music and without interfering at all? Now I'm really curious, please show me how you do that.

 

[Mikelandjelo]

"With the different impedance the filter changes the frequency and that changes the phase angle."

Did you mean on switching frequency here?If you did .....that can be solved with clocked design.If you mean on audio frequency thats passing through the filter that also can be measured if you have proper reference for comparation.As far as i can see we are talking about phase difference between input signal and output signal in real time.....If you have circuit that can compare in real time that two signals you will have instantly phase difference converted in to something that you may need to feed back in to main amp circuitry.Art is in "how to" knowledge from here.....This is very complex to understand properly and use it to work for quality.....but not impossible.

Impedance measurement is not hard thing at all....Ohms law works here just right if you have proper design which will work for you in real time.This leads to something which turns amp in to completely controled system.Once this is achieved (impedance measurement) amp will "know" exactly at which operating point is and what to do with gain to perform different things like flat frequency response,output power limit and zero phase shift.....and few other things.Once i decide to show what i am working on i will demonstrate this .... it will be fun

Mostly we are not even trying becouse we "know" whats impossible ..... take more time to think deeply about all and after good brainstorming things from impossible will become "oh my God" simple.So far all as you defined as impossible i can say is not,which is good so far.

What i see as kick in the *** with D class is switching frequency paradox.....higher we go with it THD will go higher becouse of dead time constant....lower we go efficiency goes up but filtering will not work good if we want ultra low EMI on output.Dead time of 1ns can do the job....and IR2110 has 10ns.I mentioned that i am using 400KHz in my design and NO ONE noticed that that number is very wrong becouse of automaticaly injected THD.So one thing as i see as solution to this is novel filter design or novel driver design....

Amp without any filter on output will give absolutely same quality sound on output from 30Khz to any higher number in BTL mode with BD modulation.....I tried and i am sure .....Once we include filter game will change completely.I dont like this and this is something to think about.Without filter all whats included in output sound as result is duty cycle change and nothing else....all is independant from switching frequency.

 

[ICG]

"With the different impedance the filter changes the frequency and that changes the phase angle."

Did you mean on switching frequency here?

No. I was talking about changes in the impedance because of power->heat. That's much more relevant than you think, not because power compression (okay, that's relevant too but not what I mean), I speak about ferrofluid in the tweeters, I don't want to repeat myself over and over so I cut it there.

If you mean on audio frequency thats passing through the filter that also can be measured if you have proper reference for comparation.

You can measure it before. You can measure it after. But not while playing music. Using a profile for speaker X can work but the electronics for that become extremely complex, expensive and takes huge efford (=time). If you go that far, it's much less work and much more gain in sound quality to build an active membrane positioning correction.

As far as i can see we are talking about phase difference between input signal and output signal in real time.....If you have circuit that can compare in real time that two signals you will have instantly phase difference converted in to something that you may need to feed back in to main amp circuitry.Art is in "how to" knowledge from here.....This is very complex to understand properly and use it to work for quality.....but not impossible.

No, that doesn't work. That's how you adjust the phase to a certain level/degree. Don't you understand the speaker got a phase shift itself? No, it seems you don't. YES, the drivers, the crossover, it also influences the output, the phase, the impedance! You don't know which phase for the speaker is normal, the correct one and you don't want to force the speaker to another phase, it will sound different. That's not the job of an amplifier, you only want to correct the phase error of the output filter. Do you understand that?

 

[Mikelandjelo]

No. I was talking about changes in the impedance because of power->heat. That's much more relevant than you think, not because power compression (okay, that's relevant too but not what I mean), I speak about ferrofluid in the tweeters, I don't want to repeat myself over and over so I cut it there.

Ok.Clear.We have to change tweeter design then.Right?

You can measure it before. You can measure it after. But not while playing music. Using a profile for speaker X can work but the electronics for that become extremely complex, expensive and takes huge efford (=time). If you go that far, it's much less work and much more gain in sound quality to build an active membrane positioning correction.

I agree here too.What i would say is that we have to make it simpler and less money demanding then.Impossible?

No, that doesn't work. That's how you adjust the phase to a certain level/degree. Don't you understand the speaker got a phase shift itself? No, it seems you don't. YES, the drivers, the crossover, it also influences the output, the phase, the impedance! You don't know which phase for the speaker is normal, the correct one and you don't want to force the speaker to another phase, it will sound different. That's not the job of an amplifier, you only want to correct the phase error of the output filter. Do you understand that?

Yes ser i do understand that.What i know how to eliminate is phase shift of the filter it self.Do you understand that?Any phase shift on speaker it self can be tuned with geometry of speakerbox and all whats made in front or behind or around the speaker.Amplifier has to provide on output all whats on input and that should be done without any phase shift for best performance.So far i can solve this.

 

[ICG]

Ok.Clear.We have to change tweeter design then.Right?

You have to modify the speakers for the amp?! No. An 'ultimate' amplifier has to be able to deal with any existing and working speakers, at least in a reasonable range (impedance, power requirements i.e.). Plus, you can already do the same to any existing class-d amp. If you do that, there's no point anymore to develop yet another not-any-better-amp.

I agree here too.What i would say is that we have to make it simpler and less money demanding then.Impossible?

Yes.

Yes ser i do understand that.What i know how to eliminate is phase shift of the filter it self.Do you understand that?

You can do that on a fixed impedance. If you got fluctuations on the load side, you only can improve it but you can't make it exact. It's influencing each other. See, the filter isn't a wall which shields off any load deviation and influence from the amplifier, it's more like a courtain, goes back and forth, letting some through, other things not. That's the reason you can't just correct one with a bit feedback.

Any phase shift on speaker it self can be tuned with geometry of speakerbox and all whats made in front or behind or around the speaker.Amplifier has to provide on output all whats on input and that should be done without any phase shift for best performance.So far i can solve this.

If you do that, you are building a speaker precisely fitting for the amp. That's not an ultimate amp, that's not even a good amp which needs that.

 

[Mikelandjelo]

You have to modify the speakers for the amp?! No. An 'ultimate' amplifier has to be able to deal with any existing and working speakers, at least in a reasonable range (impedance, power requirements i.e.). Plus, you can already do the same to any existing class-d amp. If you do that, there's no point anymore to develop yet another not-any-better-amp.

Here i was talking about tweeter it self in no relation to amp discussion.If there is something that amp cant solve related to it then i will change tweeters design.

Yes.

So its impossible.Yes means that.And here i will not agree at all.

You can do that on a fixed impedance. If you got fluctuations on the load side, you only can improve it but you can't make it exact. It's influencing each other. See, the filter isn't a wall which shields off any load deviation and influence from the amplifier, it's more like a courtain, goes back and forth, letting some through, other things not. That's the reason you can't just correct one with a bit feedback.

At first place i didnt say that i will correct this with a "bit of feedback" but with complex science based design.Simplifying things is good way to take fast and global look on what someone said....but i didnt say that anything is simple.I said that this is possible to be done (in real time with any impedance change) and now i will say that i will have that done after i finish what i am working on right now.

If you do that, you are building a speaker precisely fitting for the amp. That's not an ultimate amp, that's not even a good amp which needs that.

[/QUOTE]

This is wrong interpretation of what i said.I said THE OPOSITE.Its possible to design AMP to work in autotune mode on any load change with zero phase shift on output....So amplifier will folow all whats on output and not the oposite.