Re: R:thanks for answering so quickly

On Thu, 16 Mar 2006 05:08:48 -0600, "Tim Williams"
<tmoranwms@xxxxxxxxxxx> wrote:

"Mochuelo" <cucafera@xxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:grdi12dbtkh6gthk6ikosjbjp0scvnbshb@xxxxxxxxxx
I didn't mention any integrator.

It seems to me a LPF integrates.

False. Try dc input.

Whatever you're doing, if you view it in
the time domain, you're integrating PWM'd up and down pulses into a smooth,
slower waveform. In the frequency domain, you're cutting out the carrier
and sidebands, leaving the low frequency band of interest. Same thing.

An integrator has infinite dc gain. An LPF does not.

And you say to LPF it anyway???

Don't know if I understood you. The LPF for the feedback path does not
need inductors, and can be more linear and with tighter tolerances
than the high-power, passive LPF at the output of the switching stage.

Well, the goal is to get excruciatingly linear *output* voltage, isn't it?
So wouldn't the idea be to correct the output voltage itself?

That can yield better performance, but it is more complex to design
the feedback so that the system is stable, and I don't recommend it to
someone that wants to do it himself. Besides, this is analog. To aim
at the highest performance, I would recommend an all-digital solution
(upsampling, uniform-to-natural sampling conversion, noise shaping,
digital pwm generation, MOSFET driver, full-bridge switching stage,
passive low-pass filter and some elaborate feedback).


BTW, if you drop inductors, you can't get as sharp a cutoff response in the
loop, correct? So to get the same amount of filtering of the carrier,
you'll really slow down the loop, possible causing oscillation? Fast op-amp
active filters aside.

You don't need op-amps, but it is not that you cannot use them. You
can, if you want, and know how to.

Best,
.

Quantcast