After thinking about the overheating of output transistors in a Class AB scheme, I decided to try instead to build a Class D amplifier, with about 1 W of output (powered by a 12V supply).
Yes, I know I could just buy a TPA3122D2 for about $3, but for some odd reason I'm drawn to trying to build a small Class D system from discrete components (or op-amps).
I am studying these:
formatting link
formatting link
Question:
Is a triangle wave or sawtooth wave preferred? In the first link the "triangle" waves appear to actually be sawtooth waves.
It doesn't matter. The threshold just sets the duty cycle, and a triangle or sawtooth will work fine to do that, as long as you don't have sync problems or anything, but for audio frequencies, I don't think that'd be a concern.
In theory with either a perfect triangle wave or a perfect sawtooth wave it does not matter which one you use.
With non-perfect waves there are some differences. If the slopes of the wave are not straight lines then the output of the amplifier will be distorted.
With a triangle wave, then both the rising and falling slopes need to be as straight as possible.
With a sawtooth wave, the quickness of the transition at the end of the wave reduces the effect of any non-linearity in this slope. Thus more effort can placed on producing a straight ramp.
If you're using stereo, the sawtooth will result in switching transients on both channels at the same time; I'd prefer the triangle.
More important, remember that the input currents are going to be high frequency, and an order of magnitude (or more) higher than the amplifier output currents; the power supply filtering will be a large part of your design problem.
Ok, from what I understand, I will need a sinusoidal audio input fed to the (+) side of a high-speed comparator (no ordinary op-amp), and a triangle generator (frequency at least 20x the max of the highest audio frequency, or 20x 20 kHz) connected to the (-) side of the comparator. The comparator output will be a square wave, varying from
+V to -V, requiring a dual supply.
Tough parts seem to be:
Feeding the resulting square wave into a half-bridge push-pull N- channel and P-channel mosfet pair, without allowing shorting of the power supply due to the mosfet requiring some time to turn off (OFF delay).
Activating the N-channel mosfet's gate, since it wants a higher voltage than +V.
Am I right so far?
Would (1) be partly solved by using XOR gates?
I'm looking at this:
formatting link
and I'm wondering how the XOR gates can help here, since they are before the MOSFET driver itself...
Yes, an H-bridge output stage would be nice... I do not understand though how a series capacitor will help. (Remember I am a beginner at electronics!) The analog input will be positive and negative... should I simply tie the Vs- end of the comparator to ground to get pulses between Vsupply and 0?
N-
Good idea! This H-bridge mosfet driver looks interesting... 250+ kHz switching frequencies...
formatting link
r
Ah ok. I had a feeling a 555 wasn't a good choice anyway... not such a great comparator at high frequencies?
Thing is, they seem to be headphone amps. Not sure how that would help me since I want to take the sound output from the computer and attach that to some speakers via a homemade amplifier.
He seemed surprised that a basic transistor sounded better than his second 555. Second circuit:
turns out the drawing is wrong. In his simulation he has 555 TRIG+THRS connected to R3+C1 instead of to R3+D2 and that makes a lot more sense.
the TC4451 used in this circuit should drive 4 ohms speakers at 20W or so from a 12V supply that's a real 20W continuous sinewave not some marketing hype 20W.
I can't vouch for the theoretical fidelity of this amplifier. The ramp used by comparitor is not linear and the switching frequency is modulated by the audio input, the non-linear ramp is likely to cause distortion the FM could cause problems too.
for the purists there is the 'zen' mosfet amplifier (and space heater*). said to be very good in sound quality. (*gets hot, needs a really big heatsink)
formatting link
formatting link
ther second circuit deals with the PWM frequency problem but not the linearity problem. this pone does that.
formatting link
substitute it for the 555 part of the power amplifier.
Still there's probably going to problems with noise from the powersupply etc... these circuits are good as a learning exercise but in the real world they have problems.
"I remember reading in Harpers magazine (at least I believe it was Harpers) a list of the top problems facing high schools back in the '50s. At the top of the list were stomach-wrenching problems that not even the world's most capable community organizer could hope to solve, such as chewing gum and being in the hallways without a pass. Today=E2=80= =99s list boasts much more tractable problems, as befits the hugely greater investment we make in education today and the '50s lack of a United States Department of Education (created in 1979), problems such as murder, rape, arson, drugs, illiteracy=E2=80=A6 Ah, indeed, those days of o= ld when tubes glowed bold."
Class A as a space-heater... ah, now I get it! Perfect for Russia or Norway I guess, but it gets HOT in my neck of the woods... was in the
90s all this week. Class D it is... should extend the life of 'lectrolytic caps too...
About as normal as using a 555 for an audio amplifier. (in other words, "not really") Whilst TC4451s are not designed to drive loudspeakers they can do it.
2.6A at 12V (the voltage on the schematic) is about 31W, but a sine wave that peaks at 2.6A has a lower RMS current (a factor of 0.71) and there's also losses in the 4451a and the output filters.
my estimate of 20W might actually be a bit on the high side,
For something simple, I found a work-around to the shoot-through problem: You can use 4 N-channel MOSFETs in an H-bridge. The low side MOSFETS are switched at a high frequency. The High Side MOSFETs are switched according to the input signal polarity. This gives you a push-pull amplifier topology.
On positive signal polarity, quadrants 1 and 3 are off, quadrant 2 is full on, and quadrant 4 is switching at a high frequency. The opposite is true on a signal negative polarity. Now the speed of the high-side switching transition can be tuned to the amount of crossover distortion you can tolerate. You may even be able to reduce the shoot-through trouble by placing an inductor in the source of the high-side MOSFETs, thus greatly reducing your crossover margin. That way it will only source average currents through the load, but will not respond to the fast pulse when the signal is near zero, or silence.
The nice thing is this relaxes design requirements for high frequency switching of the high side, and reduces power handling requirements on transistors driving those gates (though for a 1W audio amp, you don't even need gate drivers). Everything I can think about this method works in favor of the designer: You don't have as much power loss due to power dissipated in the gate driver circuits, shoot-through is minimal and can be tuned away, at low signal levels this can be tuned to a slight DC offset to eliminate switching current...
Just an idea. I wish I had a schematic up as that would make it more clear. I haven't been able to find this mode of operation on the web... and probably for good reasons. The concept is so old and outdated that people have forgotten that hobbyists may be interested in trying this. The thing that is popular today are the switching schemes that effectively double the switching frequency (like Crown's "Class I"), or other things you have been looking at.
A really simple ramp generator can be made with a comparator, capacitor, diodes and a few other parts. It's a classic relaxation oscillator. The key to making a linear ramp is to use a transistor as a constant current source charging the capacitor where most schematics only show a resistor. I have made a triangle oscillator by triggering a constant current sink to discharge the capacitor instead of simply dumping it all at once...it seems it was only good at audio frequencies though. Depending on the comparator, this can go to several MHz as a ramp.
--------------------------------------- Posted through
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.