One Bit Sigma Delta DAC

[Woody]

Having a class D amp without an output filter does look odd doesn't it, but the speaker's impedance at the switching frequency should be so high that the 'filterless' operation can be achieved by some designs (without frying the speaker). There's quite a lot of info on this on the Internet. When I attach an 8Ohm speaker load to my classD output I detect no change in current on the 12V supply (it has 10mA sensitivity). EMI is the real enemy, so the speaker needs to be physically close to the PA to do this but it's certainly possible.

lilltroll is right that the engineer needs to consider how good the supply rail needs to be to achieve the required SNR for an application, and there are certainly system cost trade offs to be made.

[Andy]

The app note giving an overview of how a class D power amp can be implemented on an XCore has been published here:
http://www.xmos.com/published/classdan

Enjoy it :)

I will publish the code too soon.

Looks good. I look forward to trying out the code.

[lilltroll]

Berni: The electrodynamic loudspeaker always becomes inductive at higher freq.*1 + a portion of frequeucy dependent resistance.
The inductance arise in the voice coil - since it is is a coil - the resistance arise due to the fact that the coil is surrounded with ferromagnetic material, generating loss in the magnetic cuircuit higher up in frequency skin-effect*2.
If you remove the DC-resistance of the coil - the electrical phase is typically around 60-80 degrees (for frequencies above the mechanical resonance of the speaker) and its value is approx. frequency independent. :!:
Other types of speaker (piezoelectric for an example- use class G instead) may fry an class D amplifier since it's impedanace decreases with higher frequencies, but the electrodynamic can be connected to the output without output filters.
Adding diods, ferrite beds and/or filters typically increases the output current and distortion, since the impednace seen from the amplifier gets lower at high frequencies, and the look of the transition of the output transistors will become less ideal.

EMI is your enemy - specially over 25 MHz, if you do not use a spread spectrum techniques. Many coutries still use the analouge FM radio - and with long loudspeaker cables you may destroy the reception for the FM-reciever in the same room if a harmonic from the class D corresponds to the FM-carrier frequency.
An (real) example (from a commercial product during final testing before going to the market ) was that your radiostation at 98.0 MHz dissapered, but you can still listen to 95.5 and 104.5 MHz. The class-D amplifier had partials at n*1.00 MHz far beyond 100 MHz. :mrgreen:
(Also you cannot sell your product if you do not pass the FCC and EU tests regarding EMI)

*1 Home made passive crossovers may use a (too large) capacitor+resistance in parrallell with the tweeter. That might not be a good friend with class-D solutions. The main reason for the cap. is that it becomes more easy to calculate and create the crossover that way.
*2It's not about skin effect in the coppar/Al cable in the voicecoil - that happens much higher up in frequency - it's about the skindepth in the magnetic cuircuit.

[lilltroll]

Regarding the stable rail problem- it might get even more tricky if you try to scale up the design for larger power consumtions.

The half-bridge layout can been seen as another common device as well (if you look at it backwards): the LC DC-DC pump.

Connecting a reactive load such a larger loudspeaker may force reactive energy to move from the coil in the speaker to the rail-capacitors if a low-frequency (bass) signal is applied on the audio input.
The charge will pump up the voltage on the rail, which can damage the capacitors due to overvoltage, but also in the open-loop case -> add distortion due to the voltage modulation of the rail.
If the rail is fed from an AC-transformer with rectifier diods - the energy cannot escape back to the "wall". Instead it will be stuck in the "overcharged" rail-capacitors, until the output current has changes it's direction in the voice coil.

A common closed loop design of the rail will not solve that situation.
(A full bridge design will prevent railpumping)

[Berni]

Well yeah but this simplistic design is only practical for up to like 10W so its not that hard to make a PSU as the currents and voltages are relatively low. You can simply use a standard 78xx regulator and some bulk capacitance. It should keep the power rails pretty clean. As for EMI yeah theres lots of that with d-class. Well AM radio is even more sensitive to EMI as the modulation technique is very susceptible to noise.

I am building a D-Class my self but its with a purpose made chip, that dose the feedback and all sorts of protections to keep it from blowing up no mater what you do to it. Luckily the feedback considerably reduces the need for a clean power rail as with the powers that the amp operates it would be very impractical do to linear regulation since the dissipated heat would be huge. And the main point of a D-Class is efficency.

[shawn]

I think my TC2290 samples at 1 million sample per second.
Gives an 18bit db resolution @ 36khz~approx w/>100 db Dynamic 20-20k..

D-Class is on par to A-Class when placed in an alchemist hands.
The best thinks in life are more or less simple.