Using Class D for tracking power supply?

Status
Not open for further replies.
There's a Class A amplifier I have a strong attraction to and I want to use it as a 50W amplifier but I will end up dissipating crazy amounts of heat at the idle current required. I was thinking the only way to keep the dissipation low is to have the power supply track the output signal using a class D amplifier that has a DC offset on it. I have never built a SMPS or class D amplifier, I need a bit of hand holding. For starters is class D properly suited to this task because as far as I know it has rather significant phase shifts at the output. Any thoughts? This is really the only way I am going to be able to use my Class A amplifier so I must solve this design problem.
 
Wait, why would PSRR matter? The class D "power supply" will be following the output signal so it will effectively be acting as a signal cascode to the output devices so they only see DC voltage across their terminals.

Actually what I would do is not use the class D amp directly to power the amp, I would float a linear regulator within the modulated supply between the class D power supply and the amplifier so I'm not at the mercy of the "regulation" of the class D supply. This way I can keep the output devices @ only a few volts DC while allowing for huge voltage swings and high idle current.

then maybe you don't need the Class A at all?
I don't like the sonic performance of class D amps I doubt I can make it sound as good as a class A, especially the one that I have. The output phase shifts of class D make it a real pain to correct the flaws in class D.
 
Last edited:
The Class D amp either correctly follows the signal or it doesn't. If it does then it is good enough to be the main amp. If it doesn't then the Class A amp will see a distorted signal on its supply rail so it will need excellent PSRR. In any case, many Class A amps will have their gain varied by the supply rail voltage so at the very least you risk having significant second-order distortion.

If you don't like Class D then you can't use Class D in this way.
 
You can get most of the goodness with a class H design, and if the output stage is good enough for that, it will be good enough for a class D envelope tracker.

Note that the risetime on the class D output will be important, this needs to ramp up quickly enough to avoid running out of headroom on a transient, possibly easier if you have a little delay in play.

One neat approach is to have a linear regulator supplying the power via a current sense resistor and then rig up a current mode switcher to servo the current in that shunt to zero, all the speed of the linear pass device, with most of the efficiency of a switcher, servo the linear reg to maintain a few volts across the output device.

You might find some of the discussion of "Slow drain modulation" from the RF power world of interest.

Regards, Dan.
 
The Class D amp either correctly follows the signal or it doesn't. If it does then it is good enough to be the main amp. If it doesn't then the Class A amp will see a distorted signal on its supply rail so it will need excellent PSRR. In any case, many Class A amps will have their gain varied by the supply rail voltage so at the very least you risk having significant second-order distortion.

If you don't like Class D then you can't use Class D in this way.

You make a good point. I guess I could wrap the amp in a voltage cascode of its own as a layer of protection against the class D imperfections. Spice is showing that this provides extremely high psrr. It would double the power dissipation in the output stage but it would still be far lower dissipation than it would be otherwise.

So I guess that problem is solved. How extensive are the phase shift issues with class D?

You can get most of the goodness with a class H design, and if the output stage is good enough for that, it will be good enough for a class D envelope tracker.

Note that the risetime on the class D output will be important, this needs to ramp up quickly enough to avoid running out of headroom on a transient, possibly easier if you have a little delay in play.

One neat approach is to have a linear regulator supplying the power via a current sense resistor and then rig up a current mode switcher to servo the current in that shunt to zero, all the speed of the linear pass device, with most of the efficiency of a switcher, servo the linear reg to maintain a few volts across the output device.

You might find some of the discussion of "Slow drain modulation" from the RF power world of interest.

Regards, Dan.
Hmmm. I need a bit to go over what you've said.
 
Last edited:
First time class D design

I want to design a class D amplifier with the main goals of high fidelity and as low phase shift/delay as possible up to 20khz.
I am particularly interested in ensuring that the phase delay is as minimal as possible so I can apply external feedback against the source signal.

As far as I understand the key to both of these goals is using a high switching frequency in the class D amplifier so the filter corner frequency can be high enough as to not pose a shift in frequency response in the audio range.
I am going to assume this also solves the phase delay issue?

I'm also going to assume that a higher switching frequency is better? If that's the case what are the limiting parameters on switching frequency?
How high can GaN fets go for example?
 
Moderator
Joined 2011
I'm also going to assume that a higher switching frequency is better? If that's the case
what are the limiting parameters on switching frequency? How high can GaN fets go for example?

There is a pair of right half plane zeros that limits the useful bandwidth.
The switching frequency is limited primarily by the power dissipation from
switching losses and on resistance.
 
Last edited:
You're completely above my head when you speak of right half plane poles. I tried googling it but it's too arcane for me. The article you linked says that the cutoff frequency should be 20khz but I've read elsewhere if you increase the carrier frequency the cutoff frequency of the filter can be higher which allows for better phase response. What am I not understanding?
 
Moderator
Joined 2011
the cutoff frequency should be 20khz but I've read elsewhere if you increase the carrier frequency the cutoff frequency of the filter can be higher which allows for better phase response.

Both are right, you want at least 20kHz to avoid audible noises. For industrial purposes, often 20kHz is used to minimize losses,
which increase with switching frequency.

I would suggest extensive background reading before any decisions, since this is a very complex subject area.
 
Last edited:
How can they both be right if the cutoff frequency still needs to be 20khz? And how can noise above 20khz be audible in a class D amp?

I think I'm misunderstanding you and you are saying the carrier frequency needs to be at least 20khz? Implying that the cutoff frequency of the filter only needs to be at the carrier frequency?
 
you are saying the carrier frequency needs to be at least 20khz?
Implying that the cutoff frequency of the filter only needs to be at the carrier frequency?

Yes, 20kHz is the minimum switching frequency used in general design with current high speed, low loss devices, mainly due to noise. The output filter is designed to prevent aliasing. The switching frequency must be more than twice the desired bandwidth.

Another good introduction: Class D Audio Amplifiers: What, Why, and How | Analog Devices
 
Last edited:
Status
Not open for further replies.