Hi David,
Sorry for the late reply...
> My personal problems with stabilising (I have limited experience in this
> design area) are:
>
> adding regulation means putting an inductor in the output and as a
> consequence getting the phasing/feedback correct for all possible
> conditions: fine for steady no-load up to full load conditions, but for all
> the ssb and cw combinations I would need help. I would be interested to
> know what the extreme combinations were in your design. Did you just make
> the feedback loop longer time constant than the ssb/cw loadings?
My approach to stabilizing power supplies is this:
I first design the low pass filter. This starts by deciding the
acceptable ripple current in the inductor, which depends on cost
considerations for the rectifier and primary element, the cost of the
inductor versus the filter capacitor, and very importantly on the
threshold required to go from discontinuous into continuous mode. Other
people apply rules of thumb and say that the ripple shall be 10% of the
maximum DC current, and that's often quite OK...
Then comes the definition of the filter capacitor, which is done by
calculating the required capacitance and ESR for a given output ripple
voltage, and also the ripple current through the capacitor must be
considered. The latter, and the ESR requirement, often dictate the use
of several capacitors in parallel.
Then, the full frequency/gain/phase response has to be established. MAny
designers calculate or simulate it. I'm more practical, so I let the
power supply run in open loop, say with 50% duty cycle, and superimpose
a test signal on the pulse width modulator input. I use a stereo sound
card to record the PWM input signal along with the power supply output,
over a frequency range from about one Hz to a frequency high enough to
get negligible response, which is usually about half the switching
frequency. With this data I generate the gain/phase shift curve. Then I
choose the necessary control loop response to obtain roughly a 60 degree
phase margin, and calculate the components needed in the error amplifier
to achieve it. I put them in, make the power supply run in closed loop
mode, while I inject disturbance signals into the PWM input and watch
that the output is properly damped at all frequencies.
I also love to do a final test by putting a heavy load in series with a
large MOSFET, and pulsing the MOSFET at frequencies from low to quite
high, measuring the resulting ripple voltage on the supply output. All
power supplies will exhibit some resonances in this test, but one can
keep them small enough to be no problem.
It is crucial to understand what phase margin is, how it affects
stability, and how to implement the desired curve. For me, I remember
that it was rather hard to realize that a combination of capacitors AND
resistors was needed in the feedback network, to achieve the necessary
loop damping! Against intuition, a purely resistive feedback network
does NOT give the required damping, and results in an unstable power supply.
Also a note for beginners: You can add a second stage of low pass
filtering at the supply output, to reduce ripple, and mostly to reduce
RF noise. But DO NOT try to take the feedback signal from this second
stage! The phase rotation over TWO low pass stages is unmanageable with
a simple control loop, and will result in instability at some frequency.
In my 40A supply, I took DC feedback from the output, so I could
compensate for the second choke drop, but the AC component is taken from
the FIRST low pass filter!
Stabilizing power supplies is comparatively easy, because we want a
CONSTANT output voltage! So we can get away controlling only the low
frequency range with the active control loop, and using brute force
filtering for the high range. When making other servo systems instead,
such as switching amplifiers, motor controllers, robots, etc, the issue
is much more difficult, because the output needs to change value
quickly, while remaining stable. Brute force filtering is out there, and
the whole stability has to be provided by the control loop. This
requires well polished PID controllers.
> After all,
> how much ripple is tolerable to a power amplifier?
Commercial amplifiers often work with 10% ripple. Since the ripple is at
100 or 120Hz, it's below the typical SSB passband, and isn't really
audible on the transmitted signal. But if you listen to a CW signal
produced by such an amplifier, with your receiver in AM mode, you will
hear the hum on the signal!
> at higher voltages, there "seems" to be less need for good regulation for
> fixed mains input voltage, more need for stabilising against mains
> variations, eg using a generator or at the end of a long ac distribution
> network at a farm for instance.
I don't see a difference in filtering and regulation requirements, with
varying voltages. It should all be proportional. What is true, is that
filter capacitors with a certain energy storage are smaller at higher
voltages.
> without switching regulation a smaller output capacitance (smaller size and
> weight) is needed as the guard band is very small compared to the triangular
> dip when using an inductor
Sorry, I didn't get this...
> On the subject of insulating the rf and control stages rather than the mains
> voltage stages: I'm very glad to hear someone of your experience suggest
> this; I have been greeted with incredulity and disdain (not on this list).
I have found many people who cannot break free from the idea that tube
cathodes (or grids), and of course MOSFET sources, MUST always be at
earth ground potential. These same people have difficulties
understanding an amplifier that has the anodes tied to ground and the
load connected to the cathodes, with a negative high voltage supply. But
this arrangement is very convenient in tuned cavity amps for microwaves!
Also these people have serious trouble understanding a circuit with
positive ground and PNP transistors! I know that it's often hard to
"think the other way around", but it has to be done in many cases, to
arrive at a cost-effective, good solution! A floating amplifier that has
the absolute potential of the "floating ground" humming up and down
between neutral and phase voltages is a nightmare to these people, but
it works perfectly, and can have a huge cost advantage!
Keep in mind that 50 years ago almost any music radio had a live
chassis, with a transformerless power supply. In those times lives were
cheaper, so there was only one layer of insulation, such as the phenolic
knobs! If the user pulled off a know and touched the shaft, he could get
a nasty shock, depending on which way the (unpolarized) power plug had
been put in! Nowadays we would make such a device use double insulation.
That's safe enough even for modern standards designed for the dumbest
possible users.
> In my work on MW power supplies to the mining and steel industries, it's the
> controllers and final output device (motors, furnaces, etc) which are
> insulated, I don't see why suitable insulation systems cannot be designed
> for rf stages.
They can, and it's quite easy.
> At the other end of the power scale, this was also an
> approach taken in medical monitoring when I was in that field.
Yes. Medical RF generators very often work this way.
Manfred.
_______________________________________________
Amps mailing list
Amps@contesting.com
http://lists.contesting.com/mailman/listinfo/amps
|