I designed for myself a different sort of class D amp. It is two switchers running out of phase with the outputs coupled with inductors. The audio is fed out of phase to the two amps. With no audio input there is no ripple at the output. When there is audio the resultant summed pulse input to the filter appears at twice the switcher frequency making it easier to filter.
Have I invented something or has this been done before?
Have I invented something or has this been done before?
This has been done for power supplies; it's called interweaved. LTC calls it polyphase. I suppose it would work fine for Class D amps as well.
I think you have made yourself a three-level pwm and suggest you read a bit on "filterless class D" and "multilevel pwm". Also, I believe its 'interleaved' rather than 'interweaved'.
The main reason you don't see this in commercial class D amps is mostly cost. Why spend double the amount on output transistors and inductors for less than a 1% improvement in the amount of distortion or efficiency? The only class D configuration that benefits from this is high current PDM. You use lower cost, slower mosfets with low output impedance in a dual or tri phase arrangement and significantly reduce losses. It's common in computer motherboard voltage regulators designs because of this.
My reason for doing it was because I designed my own FET drivers using just simple small signal transistors (2N2222 and 2N2907) My output FETs are taking 250ns to switch making it inefficient to run over 100kHz. I would like to find a simple driver circuit that could speed this up. I am using +/_ 40 volt supply rails. I have used IR2112 drivers before but they only work from a single supply making it necessary to use a H bridge.
I am new to this stuff so any links would be greatly appreciated.
I am new to this stuff so any links would be greatly appreciated.
How is the current sharing between the switching stages of your multiphase (that's the correct name) arrangement?
I am using overall feedback with all DC coupling so there is only a few millivolts offset between amps. The output inductors measure .3 ohms so there should be less than 100 milliamp difference current flowing. I only have this working on a breadboard and will scrap the idea if there better drive circuit do use and just stick with a single output stage.
Since the inductors are effectively in parallel the resistance is .15 ohms. I just wound some air core coils about one inch in diameter that measure 50 uh so in parallel they look like 25.
I plan on getting about 100 watts at 8 ohms out of it.
I am learning this as I go and welcome feedback as to what I am doing wrong. I really need to find a way to drive FETs fast. The rise and fall times on the FET gates is 100ns with a 50ns delay to prevent both being on at the same time. I think there must be a simple way to speed this up. It's being done on a prototyping board so that might be part of the problem.
I plan on getting about 100 watts at 8 ohms out of it.
I am learning this as I go and welcome feedback as to what I am doing wrong. I really need to find a way to drive FETs fast. The rise and fall times on the FET gates is 100ns with a 50ns delay to prevent both being on at the same time. I think there must be a simple way to speed this up. It's being done on a prototyping board so that might be part of the problem.
Hi guzzibreva1,
Be careful if you use low resistance inductors as there can be current imbalance between the phases. (it is equivalent to paralleling of multiple SMPSs)
The multphase buck is a good configuration and can make a really good amplifier if
1) Optimised properly.
The more number of phases, lesser the conduction loss but more switching loss.
But lesser freq maybe used on each phase to reduce switching loss. 🙂
So, deciding on the number of phases itself is something important (or else you maybe spending money and power unnecessarily)
2) Using post-filter feedback
This maybe possibly easier because the lesser ripple in the output allows smaller inductors and capacitors, hence higher bandwidth and lesser delay in output LPF because poles are far away)
A Zobel network (never 10R / 0.1uF but something better 🙂) will make it easier.
Be careful if you use low resistance inductors as there can be current imbalance between the phases. (it is equivalent to paralleling of multiple SMPSs)
The multphase buck is a good configuration and can make a really good amplifier if
1) Optimised properly.
The more number of phases, lesser the conduction loss but more switching loss.
But lesser freq maybe used on each phase to reduce switching loss. 🙂
So, deciding on the number of phases itself is something important (or else you maybe spending money and power unnecessarily)
2) Using post-filter feedback
This maybe possibly easier because the lesser ripple in the output allows smaller inductors and capacitors, hence higher bandwidth and lesser delay in output LPF because poles are far away)
A Zobel network (never 10R / 0.1uF but something better 🙂) will make it easier.
IR2xxx chips are commonly used with success, you just have to level shift the PWM to negative rail and have an auxiliary uply of 12 to 15V related to negative supply.
- Status
- Not open for further replies.