How Do Class D Amplifiers Work? - Building A Discrete Class-D Amplifier
Вставка
- Опубліковано 14 тра 2024
- Class D amplifiers are perhaps the most efficient type of audio amplifier. But that efficiency comes with a serious cost in complexity.
-- Links --
My Website: sinelab.net
Schematic: sinelab.net/pdf/discrete_clas...
Feedback Paper: sinelab.net/pdf/charles_lehma...
LC Filter Value Calculations: sinelab.net/pdf/lc_filter_des...
Buy me a coffee: www.buymeacoffee.com/thesinelab
Bitcoin Address: bc1qwlfr2ml9km4r5vhpfh7f3m60k80whhcq3ysths
Ethereum Address: 0x7e8B8Ff8D6bD8DfB48b4A139694BB343f040Bb3f
-- - Наука та технологія
Watch part 2 here: ua-cam.com/video/PTng5xN2HkE/v-deo.html
quite a decent effort with discrete circuitry no matter what the result. great mindset!
Exactly. Each sub-blocks were fun.
It was a good challenge
@@SineLab you will never be able to match discrete transistors to the same degree as those on a die but you can try sealing the pair in touch and in an epoxy case (self-heating is optional for more accuracy); also try inserting a current limiting resistor in totem pole configurations and for temperature stability in difference pairs and followers use emitter degeneration - a few 10s Ohms helps a lot; finally you can use a pure class B amp with a high open loop gain op-amp as a driver and negative feedback to linearise the output against VBE with incredible results - the op-amp has to have large slew rate for it to work through out the audio spectrum. take care :-)
@@dimitrioskalfakis That's some good advice. Thank you.
@@dimitrioskalfakis Indeed, it's possible to utilize "NPN/PNP" single package solutions designed specifically for this purpose, as highlighted in the NXP product selection document "75017631.pdf". While the individual began with no prior foundation and faced numerous challenges, resulting in a less than perfect outcome, it's not accurate to categorically state that the use of matched pairs in a single package constitutes a singular flaw or deficiency.
I pretty much took on this exact project with the same approach last year, but with a motor driver. It was so much fun! I prefer to use discrete components as they are significantly cheaper and higher rated than ICs.
The exception is if precision is required (like transistor matching in an op amp), and youre NEVER gonna beat a manufacturer on that.
This is incredible! I really like the fact that you included a discrete design (which, in my opinion, is the most fun thing about electronics)! Looking forward to future videos, and think it would be awesome if you created a discrete radio transmitter and receiver pair (kinda like a remote control)
Thanks for the idea :)
Cool, would totally watch
Mr. Sine Lab I will enjoy!
Maybe @SineLab will make in a future video?? I hope. He makes extremely in depth, well made videos. I would love to see if he could incorporate the amplifier into a radio!!!
Thanx for the explanation.
I do have questions that maybe are explained in your following video's.
most informative video about making diy class D amplifier, it just accidentally showed up in recommended for me.
I love the discrete circuitry! You understand how everything works. Easy to find parts to repair instead of depending on some IC you don't know if you can find. Good work.
Impressive! Every component was introduced for a reason.
Informative!
I'm so glad you provided info about the feedback. When I look at class D amps I always wonder how and where to apply the negative feedback.
I'm glad that it helped you
You're an electronic maniac bro. Impressive !
This is great but I have one thing grinding on nerve lol... surrounding the leading and falling edges...
A leading edge is the departure from zero. The falling (or "trailing") edge is the return to zero. Consider polarity "absolute" here, it doesn't matter if a wave is headed up or down on the scope; the relationship to the reference level is what defines them.
What a timing. Im a TA and some students under me are working on Class D Amp. Was literally discussing stuff with them when this came. This video will be helpful for them! 😊
Btw, I have one request, can you please share some references of all the uncommon topologies you use - Like the TTL type MOSFET Driver.
That is good timing :)
As for the MOSFET driver, it's a modified TTL AND gate.
The output is a push-pull amplifier so that the MOSFET can be driven quickly. The high-side driver also has some bootstrapping circuitry which replaces the traditional power rail.
This webpage may be useful to you: www.allaboutcircuits.com/textbook/digital/chpt-3/ttl-nand-and-gates/
pcb looked impressive, good job explaining the whole thing
I saw some 1N4004 diodes strewn in there ... are you sure their extremely slow turn-off times compared to even an 1N4148 don't create issues?
Thanks!
A lot of work and effort
I can't even imagin how talented you are
Awesome work !......cheers !
Definitely subscribed! Nice one!
Amazing effort ... Subscribed!
Wow, well done for doing this and from basic principles 👍 I studied E&EE and electronics has been a hobby since a kid but I don’t get to practise much other than repairing these days - every day’s a school day as the saying goes 😊
I've found, that a class D amplifier controller is relatively easy to make by use of a few 555 timers and an opamp. The opamp works as an non-inverting amplifier, setting the voltage amplification and generates the idle output voltage at half of the operating voltage of the controller. one of the 555 timers works as an astable multivibrator with a highly asymmetric output signal which generates short low voltage spikes. It's capacitor is charged with a constant current source (I have used a E-101 constant current diode, which gives 0.1 mA). The second 555 timer works as a monostable multivibrator which is triggered by the shprt pulses. It compares the voltage at the capacitor and and the output voltage at the op amp. The output voltage of the op amp is connected to the control voltage pin of the second 555 timer. And that is all.
The input of the power stage can be connected with the output of the second 555 timer. The power stage itself can consist of a half bridge driver (LM5104) the power MOSFETs and the output LC filter.
You can just use one 555. Connect the input to pin 5 through a capacitor and it varies duty cycle. Then just pass the output through a capacitor and the speaker is the inductor.
It seems like 555 timers can be used to make just about anything :)
Thanks for sharing, that does make for an interesting circuit.
Lots of hard work went into this video. I admire your tenacity.
Thank you
Really cool video!
Good information
As someone who loves using discrete components like the venerable 2n3904, this video was a pleasure to watch. Great journey and quite an ambitious project! OpAmps and single purpose ICs are great and make things easy, but there is a certain charm to just sticking to basic jelly bean components and common BJT transistors. It's also a great learning exercise. Looking forward to more of your projects!
Thanks for watching! Using only discrete transistors really challenges your analog design knowledge.
Wow ,this is what i subscribed for ,descret component circuits ,thumbs up
Thanks for watching
LOVE THE VIDEO!!!!
Btw I am your 15,000th subscriber!!
Thanks for watching!
Congrats on being the 15,000th :)
I've seen discrete-MOSFET ALU's with less parts than that! (And none of the black-magic analog voodoo, either.) Great Job!
Nice work
Very cool video, ta.
interesting project
Breadboard circuits behave diffrently because each connection is through a jumper which has a small resistance.
Very nice bro great
Amazing video! But i wonder : how do you calculate the power your amplifier is going too neer ?
I am trying to make a portable speaker but i don't know how to calculate the power my amplifier will absorb
I made one of these a whole lot of years ago but my circuit was somewhat simpler.
The main difference was I used two "long tail pairs" for two comparitors.
Each compare drove one of the two output transistors.
One input of the compares had the signal plus the saw tooth
The other input of the compares had a slight offset voltage.
One had a positive offset and the other had a negative.
The slope of the ramp and amount of offset gave the dead time.
I used feedback to make the circuit more linear.
It sounded good and could put about 3W on a speaker from small transistors.
P=V^2/R
V^2=3W * 8
V = 5V RMS about
5*sqrt(2) = 7V
My supply was 18V split to make 9V each way
I've seen class D amps the size of a postage stamp (smallish ones) that can really deliver quite remarkable levels. My Proton Pack literally thrums through the 8' speaker and everyone hears you coming from a good way off when firing.
I know this paper, charles is a member on diyaudio class-d sub forum, we endlessly spent time looking at the feedback loop and used it in a few noval and high power designs early in its release phase. I can make some recommendations if you want to look at a discrete option rather build the (UM10155 Discrete Class D High Power Audio Amplifier) less parts, better performance and you still get to enjoy a successful build and learn the core stages in a class-d amplifier. which is a low propagation delay comparator with complementary drive, discrete mosfet driver + output stage and finally a self-oscillating UcD feedback loop.
The thought that comes to me in this specific project is the comparator. I'm not exactly sure of the specific mechanism that equates to it in class D, but audibly, that's "crossover distortion"... so I'm assuming the slew rate on the comparator has to be all but ridiculous to keep it out (and I'm not sure how you'd do what happens with AB feedback to reduce it either).
These things, just like all the SMPS topologies, require leaving conventional "DC philosophy" behind somewhat and viewing them from entirely different angles. I'm there in power conversion... class D still confuses, but this is very useful to break down all the blocks to see how it works... and who can't love a sprawled out board of discrete components... beautiful build even if the function is less than ideal.
@@MadScientist267
Crossover distortion, primarily associated with Class-AB amplification, where "class-d distortion" manifests in Class-D amplifiers due to PWM (Pulse Width Modulation) non-linearities. This distortion arises from a loss of control under various conditions due to poor feeback/control theory practice, such as fluctuations in frequency control (we assume a self oscillating modulator) , power-supply regulation, long dead-times or poor PCB layout leading to EMI interference in the modulator/Mosfets. Delving into this topic reveals a complex subject intertwined with adjacent disciplines like Delta Sigma conversion, PCB design, power electronics, and ultimately audio fidelity.
About comparators, this isn't an issue as long as the comparator is able to deal with signals at least at Nyquist bandwidth for the intended audio band or x10, x100 times higher that the sampling frequency example: 550KHz(max), comparator proportion delay adds an inherit delay and used as a block as part of the loop control.
The funny thing is the more you build class-d amplifiers the more you find out none of the challenges are audio related.
Lastly, I design class-d amplifiers in my spare time so the information above is proven knowledge.
I really appreciate the idea of building a class D with discrete components. It's just a beautiful way to help people learn the beauty of electronics. I would really love the idea of building an audio delay effect (echo effect generator) with all the discrete components. That would be real funnn.😁
That is a good idea for a circuit
Nice bro👍👍
Congratulations, an interesting project.
However, you could simplify the project, because the more components you have involved in the operation of a circuit, then the chances of failure increase.
The input preamplifier (with the amplification of 10) can be made with 1 or 2 transistors.
The triangular signal generator can be made with 8 transistors.
The feedback amplifier (Q81-Q91) is not required.
Filtering the supply voltage with only 100uF for audio power amplifiers is small, and the power supply of the preamplifier is normally done by RC filtering.
I hope you make version 2 of the circuit.
Don't give up on this project.
Good luck with all future projects
Thanks for the ideas. Simplification is the best next step.
Hey this is a great video for understanding basics of transistor logic, really appreciate the time and effort you have put it. Its rare to find analog content these days and this video just made my day. Just wanted to ask if the failure of the triangle wave generate was a circuit design issue or a PCB issue ? Thanks
It doesn't seem like a circuit design issue since it worked on the breadboard. It's likely either the PCB or the individual components not working well together.
Great 😊
I concur that i had a similar cable spaghetti phenomenon on my breadboards when working on a buck booster circuit, making all the signals from scratch along with a feedback system(with op amps ofc), anyways great video! Will you introduce PLL(Phase locked loop) ic's as well in the future? They are capable of many things such as a frequency generator, demodulate signals etc.
It seems like the spaghetti is inevitable at times :)
PLLs are a good video topic
Breadboards have extremely large parasitic capacitance and as frequency goes higher things are getting worst.But even in a pcb the problems with parasitic capacitance and noise interfere can be remain.If everything designed perfectly in component level maybe it needs extra copper layers for vcc and ground plane and also the components have to replaced with smd parts.
Great video, I want to build a class D amp also, but I'll stick with reular op amps and mosfet driver IC, to make the building process less difficult.
Op-amp ICs would make it easier :)
You can make this simpler by cutting corners. For example, using resister transister logic and just using a two transister square wave generator with an extra transister turned on by analog input to vary duty cycle. The totem pole outputcan be replaced by a single mosfet since the output capacitor will offset that around 0 creating AC.
There are alot of drawbacks but it would be a really simple class D.
I am with vojtechadame5860.
He thinks the same way as me.
A field transistor is much much more accurate and in my logic it makes can make the circuit more ehm simple.
And mosfet amplifiers sound sooo much better!
Not a bad effort, you’d have a better go with even something as simple as an Intersil HIP4082 driver, use bipolar rails to ditch that coupling cap or do a unipolar full H bridge. Remember your distortion performance is going to be limited by the linearity of your tri-wave so the RC would have never put down good numbers. That said with a decent oscillator an op-amp based integrator 0.1 or 0.01% THD+N should come fairly easily.
That or look at Bruno Putzey’s UcD topology which is self oscillating. It’s free for public use and commercially so long as you use one NXP component in your design iirc.
You could of course cheat and just jump to an IRF2092S as well, that will implement nearly everything short of the feedback loop and a few external parts.
Class B suffers from crossover distortion. You have dead zone on either side of the signal before either transistor turns on (why we use class A still)
Philips ucd has very cool schematic😊
I heard that class D have built-in digital audio over drive processing maybe you could add that to your design, can you sense the position of the cone and tune to enclosure and driver et cetera
I added this to your class A video, and repeat it here since people use your channel for education.
Class A amplifier does not mean one transistor. It normally has the same stages as a class B amplifier: input stage, voltage stage, current stage. The difference is that class A and class B is biased differently.
The misunderstanding is probably due to it being possible to make a class a of one transistor, but that is not usually done in real life. I say "usually" just because there exists some amps called Zen. But most class a you buy have just as many transistors as a class b.
Though you tried making class D block from discrete elements, you cant get this to ever work decent. One of the problems is a pile of transistors of questionable bias, thermal coupling etc... But if you want to learn the other way of making class D amplifiers called 'hysteresis class D' and also make really functional and must say good HiFi audio amp, you should try D class circuit called "UM10155" designed by Philips. It works great, have only 13 bjts and 2 mosfets. Here is no place for me to discuss it any further, but you have bunch of forums where people explained the circuit in detail! :)
Although your goal with all discrete is very cool, I wonder if you designed another spin replacing some of the basic blocks (basic buffer amps, comparator, triangle generator, FET driver) with jellybean ICs if you could get a board that sounds decent at the end. And so prove out your basic design. I guess we can see why all the old boys like designing a powerful AB amp instead of tricky D one :D
It'd definitely be more successful with ICs.
If only class D could be as simple as AB :)
Isn't the whole idea of Class D that some of the transistors are not switched on some of the time? In this example, it seems one or the other transistor is always on? Maybe there could be 2 pwm signals one positive and one negative? Or an 3 level mid low high signal?
If you use RF techniques I think you will get a better and more stable result. I think your troubles are in long track lengths aspdding stray inductances everywhere and there would be stray capacitances too. Your triangle wave in your final build was “fairly wonky”, this is because of insufficient grounding.
It is a bit of a pity you jumpped straight to a PCB from the breadboard. If ou have the time, try to build it “Manhattan Style” on a piece if tinplate. Get a large tin can…prefrably one without any ridgrs in it…here in Australia I’d use a 1.2Kg Milo tin…take the ends off with a can opener and cut down the seam with old scissors ir tinshears. Flatten the tinplate out into a sheet and use that as a groundplane. Just start building directly onto the tinplate by soldering the components directly to it. Keep component leads as short as possible…say you have to ground the emitter of sn NPN transistor, bend up the collector snd base leads and make an “L” shaped “foot” with the emitter lead, keep the upright part of the leg short and the body of the transistor close to the tinplate sheet. Say you then meed a collector load resiator for that transistor, get the said resistor, bend one lead into another “L” close to the body and solder it to the cut down collector lead of the transistor, the long end will be your output. Now, lets assume you need a decoupling capacitor gor that stage, either get a leaded monoblock, or even better a surface-mout 100nF, right st the other end of the collector load resistor solder the SMD 100nFto the dide of it…right close to the body, leace the lomg tial to become the DC rail. Now bend it down close to the tin so the other end of the 100nF touches the tin, “toombstone style” and solder it down. A leaded cap can be used too, but keep the leads real short and the body of the cap right down close to the tin.
Staet wirh, say, the triangle generator…once you have built the entire triangle generator, tested it and it is working in the tinplate sheet…NOW, cut from another Milo tin..of Coffe Tin..a strip of tin about 3/4 of an inch wide, use this to mage a “wall” around the triangle generator, run the output through a hole in the wall ot An arch in the wall. Solder the wall down really solid to the groundplane…steel has a poorer thermal conductivity than copper and your iron should be able to handle it easily.
Now simply keep goung, build the compsrator in abother “doll’s house room” right next to the triangle generator, test it, get it all “Hunky-Dorey” and move on to the next stage.
Take partivular care around the MOSFETS and low pass output filter…anywhere high, pulsy currents flow…treat this part of the circuit like it is wotking st 4GHz, not 400KHz! Keep everything “short and sweet”. If you can get your hands on some old scrapped microwave stuff..like an old GSM Mobile base dtation tranceiver, look at it’s layout and try to emulate it in your curcuit.
Whan finished, it will look like s compact “rat’s nest” but it should work really well.
Cheers, “The Globe Collector”, Tasmania.
If you make a PCB next time give yourself a lot of space and a lot of test points and maybe group things in functional little blocks so if one of the blocks is broken you might be able to patch around it
I mean why wouldn't you just build the PCB in the exact way you designed the breadboard setup, using the same transistors that you literally just used (workingly) in the same breadboard.
maybe u should have ordered PCBA and tried to make it all SMD with a large pack , maybe you shorted something and the triangle voltage generator is overloaded
did you check your solder joints? in the video you hooked up the scope probe to a resistor which was not properly soldered and I noticed that it moved. secondly when your triangle generator broke, you should have checked the schematic and then the board to find out why it broke. you dont just "fix" the same circuit and move on several times. you look for why it broke. You designed the circuit. you know how to find the problem. You dont give up after all that effort. lastly in engineering, PATIENCE IS A VIRTUE. one step at a time.
Might use your design for personal use, but cheat by using opamps as.comparators
❤❤❤🎉
Have you thought about using transistor arrays? It would help a lot for the matched pairs. Technically wouldn't be discrete anymore though...
That's a good idea. It would simplify things
Bro how did you learn this things in electronics ?.
Lots of time spent studying. There's probably quite a few good books out there as well (although not sure how many will be on class D), although you have to read them!
@@ivolol namely
By making all of the videos on this channel. I learn something new every time I make a video :)
@@SineLab can you share material bro. I am interested to learn electronic but where to start is question ❓
@@vigneshkarthi3321 Pick a project and research the parts you don't know how to do. You can search google or read a book. That is how I do things.
You can see the improvement in my skill by comparing my earlier videos to the ones that I make now.
Since you are starting out, pick simpler projects. Look at LED blinking, 7 segment displays, and basic Arduino projects.
I still preffer class AB amplifiers, they have much better audio fidelity
next part ?
Eventually, I will try this again. But first, I am working on other things for a while.
i love to see so many connections on breadboard😅look like grass
It's really satisfying to look at :)
the pcb looked pleasant... it shouldve worked 😢
Hmm, those 'breadboards' have a huge amount of C and L. Can't us them for Class C or any RF work.
That may have contributed to the unexpected results on pcb.
What passes for education in electronics today is trash.
Jump over to Great Scott's channel and see how he prototypes.
👍
It's sad you don't have enough time to track down all problems of this device. I'm sure it can sound way better.
Might want to change the title to "how not to build a class D amplifier"
It's to complicated for me
You'll get there eventually
Breadboard: Build it step by step
PCB prototype: Build it all at once
Discrete Class D or not, for me this does not sound good, and complicated circuitry.