So,this should be easy discussion.I know a lot about D class amplifiers....i built a lot of them.....so i am deeply infected with this 
.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 requires oposite......so whats the middle point solution?Whats minimum switching frequency that will give still perfect sound quality?I want to discuss full bridge and half bridge designs and not multi phase solutions.Any thoughts are welcome.
.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 requires oposite......so whats the middle point solution?Whats minimum switching frequency that will give still perfect sound quality?I want to discuss full bridge and half bridge designs and not multi phase solutions.Any thoughts are welcome.
I just bought my first class D poweramp boards from Parts Express for a battery operated bluetooth boombox I'm building. I actually bought 3 different versions, and 2 of them had bad oscillation problems. The third board (TPA3116D) appears to work pretty well (no osc. and fairly clean transition into clipping).
My contribution here is the suggestion that ferrite beads and maybe even passive low pass filtering be used on the input, if there's any analog circuitry upstream (preamp, etc.). I don't know if there's any significant Rf energy at the input of a class D poweramp, but if there is, it could work it's way into the analog preamp circuitry in a way that could cause slewing related distortion products. If you're after the best possible sound, this could be a small issue.
As far as the switching frequency, it's all about the Nyquist frequency, where anything above half the switching frequency gets turned into a lower frequency. I'm sure you know what I'm talking about. You want to avoid having any aliasing distortion from that. How many dB down do you want the program material to be, when you get to the nyquist frequency? How low do you want the aliasing distortion to be? Personally I like high sample rates (maybe 500kHZ), so the anti-aliasing filter can pass 20kHZ no problem, and have plenty of room to roll off the signal very well before the nyquist frequency (250kHZ in this example). Reconstruction filters that are more than 2 pole can be tricky, and more load Z sensitive.
Another thing I do is to put a zobel across the speaker, so the rising impedance of most speaker drivers only goes so high (a 0.2uF in series with a 15 ohm 5W resistor for example), so the recontruction filter doesn't get mad, and generate high voltage transients.
My contribution here is the suggestion that ferrite beads and maybe even passive low pass filtering be used on the input, if there's any analog circuitry upstream (preamp, etc.). I don't know if there's any significant Rf energy at the input of a class D poweramp, but if there is, it could work it's way into the analog preamp circuitry in a way that could cause slewing related distortion products. If you're after the best possible sound, this could be a small issue.
As far as the switching frequency, it's all about the Nyquist frequency, where anything above half the switching frequency gets turned into a lower frequency. I'm sure you know what I'm talking about. You want to avoid having any aliasing distortion from that. How many dB down do you want the program material to be, when you get to the nyquist frequency? How low do you want the aliasing distortion to be? Personally I like high sample rates (maybe 500kHZ), so the anti-aliasing filter can pass 20kHZ no problem, and have plenty of room to roll off the signal very well before the nyquist frequency (250kHZ in this example). Reconstruction filters that are more than 2 pole can be tricky, and more load Z sensitive.
Another thing I do is to put a zobel across the speaker, so the rising impedance of most speaker drivers only goes so high (a 0.2uF in series with a 15 ohm 5W resistor for example), so the recontruction filter doesn't get mad, and generate high voltage transients.
Gotcha!
You are bored and hope to trigger an amusing battle of believes.
Yup
.... I am "infected" for a very long time.....I tried to achieve.....well its better to count what i didnt try .....And i would like one good discussion very much.I built something that i beleive is good.....but maybe i missed something....So,give it a try....Lets define some tests with scope,signal generator.....different loads etc. .....I would like to test what i have and maybe i missed something.Discussion cant hurt.My sample rate is 400Khz .... with BD modulation its 800Khz on output filter network.....and i have almost completely dead ,zero noise output without input signal...and with it ...Its not bad for start i think
Last edited:
Ferrite beads on input cant do much work becouse they cant react with ultra small curent and here is voltage the issue ....voltage ripple.For me about this ....best working way is to have diferential input and BD modulated output stage...If all is properly designed output noise is absolute zer
f course its not near close to be easy to figure out all instantly and make it to work..... Dont ask me how i know .Zobel across the speaker or any impedance influence on speaker side is good idea....but if we make best possible working amplifier than we should have output impedance independant output stage/amplifier.Any thoughts about how this can be done?
Good point about the ferrite beads needing enough current to work. One of the tradeoffs with a switchmode amplifier is that in order to get the high efficiency, it has to work with ultra fast rise times, with come with Rf freq. spectra, so there's no way to have zero output noise, hence the reconstruction filter. Same story as with a DAC.
If you've got your circuit working as good as you say, then I'd concern myself more with durability with real world conditions. How good are the speakers protected in the event of amplifier failure? How well does it handle a rock guitarist blasting it well into clipping for hours? Does it clip with sharp corners or is there a circuit that keeps it from ever having sharp corners? What's the time constant of that circuit, or does it work in real-time? What about the turn-on (or turn-off) transient; is it safe to directly drive a delicate tweeter in a tri-amp'd system? Does it have a built in level indicator ready to drive external LEDs? How much humidity and ambient temperature range can it handle?
If you've got your circuit working as good as you say, then I'd concern myself more with durability with real world conditions. How good are the speakers protected in the event of amplifier failure? How well does it handle a rock guitarist blasting it well into clipping for hours? Does it clip with sharp corners or is there a circuit that keeps it from ever having sharp corners? What's the time constant of that circuit, or does it work in real-time? What about the turn-on (or turn-off) transient; is it safe to directly drive a delicate tweeter in a tri-amp'd system? Does it have a built in level indicator ready to drive external LEDs? How much humidity and ambient temperature range can it handle?
I have one I built that runs up to 2 Mhz at 42v rails. overload capacity drops off inversely with switching frequency. The sound quality is superb but mounted into a receiver , the FM was noisier that usual.
The other thing is that the overload is hard, if it is exceeded at all the amp will die.
A .47 ohm series output resistor cured the overload shutdown problem.
I do agree that Bruno's UCD design detailed on Phillips mosfet semiconductor sheets is a very good meeting of various comprimises
The other thing is that the overload is hard, if it is exceeded at all the amp will die.
A .47 ohm series output resistor cured the overload shutdown problem.
I do agree that Bruno's UCD design detailed on Phillips mosfet semiconductor sheets is a very good meeting of various comprimises
In my design .... on which i am still working ..... i do have almost dead zero noise output....or with other words i have less hum on output than with any AB class amplifier i had before...i compared Technics and Sharp amplifiers and few i have built which are ultra quiet.....And i figured out how to make absolute zero output noise output which will happen in next month (I am doing designing for living so time is issue a bit).
Speakers are buletproof protected as amplifier it self.Its impossible to couse any damage on amplifier in ANY type of overload.....it will not burn or even get warm if you make dead short on output,on each half bridge to ground,from + power rail to output or in case of overheating.......So speakers are protected very well i think.Hours of clipping are just fine (tested for 3 hours) becouse it will clip on very unique way and is capable to deliver flat top in that case BUT NOT as other amplifiers.I developed NEW PWM algorithm which keeps highest output quality and has higher efficiency than current solutions.Switching frequency and protection responce are NOT in relation to each other and on any switching frequency protection will have same responce.Sound quality i will not describe....i made it and i doubt that anyone will beleive that i will say the truth.....For me personaly this is something i never heard or saw better in my life.
Becouse i did few things VERY differently than anyone till now i achieved to have high definition sound when i have absolute switching phase match on output,which i never found that anyone made before in BD modulation.Every design with BD modulation requires small phase missmatch to work good and have clear output on low power output,but not mine.Also EMI and output noise ripple are very low ..... I have 5 to 10mV ripple and hf noise on output no matter is modulation up to clipping level or its no audio present on output.I cant say anything about humidity and ambient temperature range becouse i didnt test that at all.What i can say is that i dont have any heatsink , that i pulled up to 15A from amplifier without heatsink and its very stable ....There is small heat but nothing that needs heatsink to work.And i am using in prototyping phase IRFB5615 fets.....ringing is not present or its super small becouse of different output stage design.And i will claim one thing for sure.....You dont need eGAN fets to have clear switching.....they work better but they are not a must.Also story about D class is not finished yet nither is all discovered in this area.
It is possible to overcome this and have absolute zero output noise.If i did it someone else will do it for sure too.It took me years to figure out but is possible.
..... i do have almost dead zero noise output....or with other words i have less hum on output than with any AB class amplifier i had before...i compared Technics and Sharp amplifiers and few i have built which are ultra quiet.....And i figured out how to make absolute zero output noise output which will happen in next month (I am doing designing for living so time is issue a bit).Speakers are buletproof protected as amplifier it self.Its impossible to couse any damage on amplifier in ANY type of overload.....it will not burn or even get warm if you make dead short on output,on each half bridge to ground,from + power rail to output or in case of overheating.......So speakers are protected very well i think.Hours of clipping are just fine (tested for 3 hours) becouse it will clip on very unique way and is capable to deliver flat top in that case BUT NOT as other amplifiers.I developed NEW PWM algorithm which keeps highest output quality and has higher efficiency than current solutions.Switching frequency and protection responce are NOT in relation to each other and on any switching frequency protection will have same responce.Sound quality i will not describe....i made it and i doubt that anyone will beleive that i will say the truth.....For me personaly this is something i never heard or saw better in my life.
Becouse i did few things VERY differently than anyone till now i achieved to have high definition sound when i have absolute switching phase match on output,which i never found that anyone made before in BD modulation.Every design with BD modulation requires small phase missmatch to work good and have clear output on low power output,but not mine.Also EMI and output noise ripple are very low ..... I have 5 to 10mV ripple and hf noise on output no matter is modulation up to clipping level or its no audio present on output.I cant say anything about humidity and ambient temperature range becouse i didnt test that at all.What i can say is that i dont have any heatsink , that i pulled up to 15A from amplifier without heatsink and its very stable ....There is small heat but nothing that needs heatsink to work.And i am using in prototyping phase IRFB5615 fets.....ringing is not present or its super small becouse of different output stage design.And i will claim one thing for sure.....You dont need eGAN fets to have clear switching.....they work better but they are not a must.Also story about D class is not finished yet nither is all discovered in this area.
It is possible to overcome this and have absolute zero output noise.If i did it someone else will do it for sure too.It took me years to figure out but is possible.
Last edited:
Patience is a virtue.I will never open mouth and talk if i cant prove something.Its not time yet for show time.It will happen ,you have my word on that.I will not record and show working prototype i have for sure ,once i reach all what i want by my measures and design all on one board.... probably i will make video of it.As i said before i am NOT DIY guy but someone who is making for living by desiging and selling of designs for different industries.So dont expect any info,files,schematics etc.I am interested for now to sell what i made....and in this discussion i expect to see is there something that i missed on my path to perfection.So shoot.....some questions,suggestions,expirience sharings and offers are welcome.
No.You didnt get this right.I discussed things here where i also contributed with results and knowledge from my side.What you said is possible from any discussion on this or any other forum.From any discussion we all can use potential good design suggestions for our benefit.I didnt discuss here about "good design suggestions" but about "Ultimate D class amplifier design in theory" which means a discussion where expirienced people will say what they think about that.I pointed instantly that i am not diy guy and this discussion is about theoretical thing and not about "please tell me how to do what i am doing".Maybe you should read all from start till this post and see what you missed.I dont need any single advice about how to do things....just discussion about things that we should overcome in all whats done till now which should point us to ultimate result.
According to the data sheet these FETs typically have 32mOhm Rdson. At 15Arms output that's 7.2W dissipation in each FET, but it only barely gets warm?
And to add to the "discussion", the ideal class d amp has zero dead time and infinite rise and fall times at the switching node. The closer you can get to that the better. Lower package parasitics get closer to this ideal. Higher switching frequencies are easier to filter, as long as emissions are well controlled through proper PCB design.
Last edited:
Is there any reason to chose the IRFB5615, when there is the IRFB4615 since ten years?
Both are almost identical, but the IRFB4615 has the lower Qrr and a well specified dv/dt rating, while IRFB5615 does not specify max dv/dt.
It looks a little bit like the IRFB4615 is a high end MosFet suitable for hard switching circuits in industrial applications, while the IRFB5615 seems to be the left over from the 4615 and then sold with the label audio/classD...
IMHO there are the following key parameters when watching out for classD MosFets:
Qrr
Max dv/dt when hard switching
Ratio of Crss/Coss (ratio should be low)
Voltage dependency of Coss, the more nonlinear the Coss the more you will struggle to optimize your ZVS and dead time adjustment in order to achieve lowest distortion.
Nonlinearity of Coss has moved to worse in some of the newest MosFet families... ...for the same reason it is not fortunate regarding distortion to use 900V SiC devices in circuits where a 200V MosFet still works. In terms of switching losses and EMI it still can be attractive though. Of course no way in terms of ROI...
Both are almost identical, but the IRFB4615 has the lower Qrr and a well specified dv/dt rating, while IRFB5615 does not specify max dv/dt.
It looks a little bit like the IRFB4615 is a high end MosFet suitable for hard switching circuits in industrial applications, while the IRFB5615 seems to be the left over from the 4615 and then sold with the label audio/classD...
IMHO there are the following key parameters when watching out for classD MosFets:
Qrr
Max dv/dt when hard switching
Ratio of Crss/Coss (ratio should be low)
Voltage dependency of Coss, the more nonlinear the Coss the more you will struggle to optimize your ZVS and dead time adjustment in order to achieve lowest distortion.
Nonlinearity of Coss has moved to worse in some of the newest MosFet families... ...for the same reason it is not fortunate regarding distortion to use 900V SiC devices in circuits where a 200V MosFet still works. In terms of switching losses and EMI it still can be attractive though. Of course no way in terms of ROI...
Well i didnt say 15A of continuous use
.I know the math.Without that what i said you wouldnt be posting here for a long time .And you are right about all you said.Are we limited with speed and quality of components we have to achieve 1ns dead time?Chocoholic you are right in all what you said....But what if amp is made to have almost zero ringing how that should affect all?I will try Fet you mentioned btw.I made optimization to driver circuitry to handle heat on 400KHz-600KHz fine (output sees 800KHz to 1,2MHz).Driver is IR2110.....I have there 10ns dead time that i actualy measured with scope.Shorter we go with dead time closer to disaster we are.....and closer to perfection we are too.It always get exciting beeing at the edge isnt it?
.Is all math and theory same if we have multiphase system....2 phases for one channel and two different phases from first channel on other one (all phases are different which as result has 4 phases actualy)?- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.