OT: Class D Amplifier Design

Quite correct, but if you can approximate well enough it works pretty good.

Talking about getting class D amps up to 250 KHz ? Not sure that is a viabl e avenue give CDs are only 44.1 KHz.

Of course there are better formats now. More bits and faster sampling rates , but as you said it is still an approximation.

Audiophiles can rightly ask what could be better than a wire with gain ? Th e audio goes into microphones or whatever, to the mixing board and then to the cutting lathe making the master record and then that makes the stamping plates and then those make the album you listen to eventually. What could possibly be better than that ? Put it all into a computer and hope it comes out alright ? There are alot of people who want to see this country return to paper ballot for elections and I agree. There is good reason.

High fidelity.

Don't get e wrong here, class D and ad all this digital shit is fine for ce rtain applications, but for real high fidelity I would want to avoid it. We ll, maybe to run a subwoofer. But full range, no.

Depends. Charge is quantized, but energy per carrier isn't--in a solid all the conduction and valence band states get smeared out into a continuum.

Cheers

Phil Hobbs

The 250kHz was the rate at which the output switching transistors switched.

The idea of a Class-D amplifier is the output devices spend their time on or off, creating a pulse-width modulated waveform which - after it has gone through a low-pass filter - looks like the music waveform you wan to hear.

If the output devices switch at 250kHz, the minimum on or off period would be 4usec.

This is a bit pathetic, but it has nothing to do with the 44.1kHz sampling rate of CD recordings, which would imply that your pulse-width waveform would be delivered in 22.675 usec wide chunks.

The dumbest way of doing this is to create a series 22.675usec wide chunks, each with just one on-to-off transition. If you wanted 10-bit - 0.1% - amplitude control you'd have to move that transition along in 22nsec increments.

In a 250kHz system, you'd be trying to move them in 4nsec increments (which does happen to be practical, if a bit demanding).

In reality, you futz around with how many transitions you have to move most of the spectrum of the switching noise well above the cut-off of your 20Khz brickwall filter, or what ever real filter you end up going for.

Sigma-delta D/A converters have been doing it for about thirty years now.

Assuming the input to output conversion efficiency of the SMPS plus power amp is around 90% and the SMPS is using some fashion of power factor correction, it looks like it would be drawing around 19 amps from the wall continually looking mostly like a resistive load. 19 amps continually isn't going to blow any 240V 15 amp circuit that I'm aware of, people run their resistive teakettle drawing 3000 watts from 15 amp circuits all the time, it's nice and makes your tea heat up real fast.

It's not gonna melt down the IEC connector or catch the power lead on fire. The wiring might get a little warm.

No, "analog" is just an approximation of the "real" quantum world.

But two channels are contributing so each delivers 55V rms.

The internal DC supply is about +/- 82V.

At 2500 watts (one channel, 4 ohms) I measured the AC draw as 18A rms.

With a simulated music signal ( compressed, band limited pink noise ) the AC draw was around 5A rms. The amp sensibly has limiter that stops peak clipping.

If some user managed to push it too far, a thermal breaker on the back will trip the AC supply long before cables or connectors get hot.

With music programme, there is no issue with AC draw. You also need to remember that loudspeakers are not resistors - the nominal impedance is only seen in the mid band, near 250Hz and rises at lower and higher frequencies.

Plus the copper wire voice coils have a positive tempco and can reach almost double room temp resistance before harm occurs.

.... Phil

Right - which is why it's interesting that speakers are current-mode devices.

Sorry, I got that wrong of course. current produces force, whereas a fixed voltage produces a fixed velocity, due to back-EMF. This is what my giant voice-coil (ex HP disk drive the size of a washing machine): Connect a voltage source, say a fully-charged 1.2V NiCd C cell, and the cone rises at about 2cm/sec constant. It does not accelerate (after the invisibly short transient).

So if the amplifier is a voltage source, it's producing velocity proportionate to the voltage - which doesn't correspond to SPL (it's not current hence not acceleration).

But current is not proportional to voltage (at least not below the primary resonance).

So does the relevant answer lie in the fact that the action is interesting above the resonance?

achieve low distortion and a flat frequency response unaffected by the load . Speaker makers had to produce designs that worked best with such amplifie rs or be left behind in the hi-fi market as it evolved in the 1950s. Luckil y, this was not too difficult and resulted in much better performing speake rs and consistent results with different designs and makers of amplifier.

acceleration. Remember F = ma ?

Applying a voltage to a speaker produces a current in the VC which then acc elerates the cone. Best forget your imaginary analogy with devices that are NOT loudspeakers.

o accelerate in proportion to current flow. Constant current flow produces constant SPL. It also follows that for constant SPL, the excursion of the c one reduces by a factor of four whenever the frequency is doubled.

However, the four times excursion reduction per octave rule applies at any frequency designers try to keep that happening at low frequencies too, at l east in sealed enclosures.

.... Phil

Ok, it's not a loudspeaker. But the simple fact is that a fixed applied voltage (to this voice-coil) produces a fixed velocity. If your argument applied in this case (and I grant it's not a speaker) then the voltage would produce a fixed current, which would create a fixed acceleration... but...

That's not what I observe. Instead, I see "linear motor" behaviour where back-EMF is at play.

Can you explain why this is not what happens in a loudspeaker? The basic linear motor is the same.

Clifford Heath.

Do not come back until you have.

.... Phil

The amplifier is rated for more than 6 kW (2x3.1 kW) output,, so I would expect it to draw 27-28 A from 240 V and up to 30 A from 220 V. A 15 or 16 A, not to mention a 10 A fuse would not last very long, in the best case 10-20 minutes.

Beware, if you blow the fuse with such overload, it is going to be _hot_, so take care of your fingers. When throwing such fuse into water it will hiss :-)

The reason that the 10 A IEC connector will survive is that the amplifier doesn't produce the rated power all the time. Even with highly compressed music, the average level is -3 to -6 dB between maximum output, thus limiting the average heating.