I am posting this in an attempt to generate a discussion on various class d ideas we may want to share. it doesn't need to proven or working. just need to be unique and interesting (and weird too).
with the discussion we had in the tas5261 thread, I thought I would open up a new thread and talk about various non-traditional options in going class d.
in its simplest form, class D is no more than a voltage-controlled pwm signal generator followed by an output stage, with or without feedback. let's talk about each of them:
a) signal generator: the simplest is a comparator. UcD used a differential pair, IRAudamp used a comparator IC. Other choices include small-power class D chips (tpa2000 in the tas5261 thread), or in my case pwm controllers. and I am sure other ways exist as well.
they need to run fast: 500Khz for hi-fi applications and probably lower for subwoofer amps. that is not a problem with most of them, as some of the pwm controllers go as high as 1.5mhz.
they need to do 100% duty cycle: not a problem with most of them. the tpa2000 solution has the added advantage of having differential output. and most of the pwm controllers don't do that.
b) output stage: many choices: discrete by ucd; driver IC + discrete output devices by IRAudamp; or fully integrated by tas5261.
the discrete solution offers the best flexibility and independent timing control for upper and lower devices. they may be cheaper to build but they are harder to build due to high part count. However, they can be made to very high power.
driver IC+discrete output devices: lower part count, easier to build (unless you are me), and has good flexibility on output power and timing control, depending on the chips you use.
fully integrated output: lowest parts count but also offers the least flexibility.
In terms of topology: two key points: half or full bridge output, feedback or openloop.
feedback helps maintain operating point stability, critical for DC coupled output, and lowers the requirement on power supply. Personally, I think this is the way to go. However, it can be taunting to add feedback due to stability / topology issues.
openloop is the easiest to build. however it is tough to hold its operating point stable. One way to do it is to use full differential output (tas5261 for example) / full bridge to null the common DC content. it requires one more output filter but is more supply friendly (bus thumping or SE power supply).
in my test set-up, I am using a ucc2803 configured as a pwm generator to dirve a lm5104 driver chip + mosfet output devices. the use of the power controller is purely incidental as I have a tube of them. they are good for low-fi applications as they all tend to have large dead time (75ns or larger). But for a proof of concept, that's sufficient.
with the discussion we had in the tas5261 thread, I thought I would open up a new thread and talk about various non-traditional options in going class d.
in its simplest form, class D is no more than a voltage-controlled pwm signal generator followed by an output stage, with or without feedback. let's talk about each of them:
a) signal generator: the simplest is a comparator. UcD used a differential pair, IRAudamp used a comparator IC. Other choices include small-power class D chips (tpa2000 in the tas5261 thread), or in my case pwm controllers. and I am sure other ways exist as well.
they need to run fast: 500Khz for hi-fi applications and probably lower for subwoofer amps. that is not a problem with most of them, as some of the pwm controllers go as high as 1.5mhz.
they need to do 100% duty cycle: not a problem with most of them. the tpa2000 solution has the added advantage of having differential output. and most of the pwm controllers don't do that.
b) output stage: many choices: discrete by ucd; driver IC + discrete output devices by IRAudamp; or fully integrated by tas5261.
the discrete solution offers the best flexibility and independent timing control for upper and lower devices. they may be cheaper to build but they are harder to build due to high part count. However, they can be made to very high power.
driver IC+discrete output devices: lower part count, easier to build (unless you are me), and has good flexibility on output power and timing control, depending on the chips you use.
fully integrated output: lowest parts count but also offers the least flexibility.
In terms of topology: two key points: half or full bridge output, feedback or openloop.
feedback helps maintain operating point stability, critical for DC coupled output, and lowers the requirement on power supply. Personally, I think this is the way to go. However, it can be taunting to add feedback due to stability / topology issues.
openloop is the easiest to build. however it is tough to hold its operating point stable. One way to do it is to use full differential output (tas5261 for example) / full bridge to null the common DC content. it requires one more output filter but is more supply friendly (bus thumping or SE power supply).
in my test set-up, I am using a ucc2803 configured as a pwm generator to dirve a lm5104 driver chip + mosfet output devices. the use of the power controller is purely incidental as I have a tube of them. they are good for low-fi applications as they all tend to have large dead time (75ns or larger). But for a proof of concept, that's sufficient.