Hi David, and all,
> If Manfred reads this, I hope he jumps in and gives you further
> advice.
Oh well, I was trying to stay out! So much for that...
Making a properly regulated switchmode power supply isn't really hard,
but it requires a good understanding of feedback loops. If buzzwords
like PID control and phase margin don't ring very clear and loud bells,
then it will be necessary to learn about such matters, before having a
chance at success designing a stable, well-behaved switchmode supply.
The basic procedure is measuring (or calculating) the response of the
entire power part of the power supply, from the PWM voltage to the
output voltage, and then design the feedback loop with the proper phase
and gain characteristics to make the whole system respond as fast as
possible to disturbances, with minimal overshoot, but still maintaining
unconditional stability. It's not hard, once you have learned how to do
it! Until then, it looks like black magic.
Shortly after I published my 13.8V 40A switchmode supply in QST, a ham
from the US contacted me, wanting me to develop a high voltage high
power switcher for his project. It would be a lightweight tube type amp.
He was a nice guy, offered very good conditions, but I couldn't accept
for three reasons: I was a bit scared of unexpected problems handling
the very high dV/dT occuring at the diodes; I had a full-time job that
required exclusivity (so I could not do paid work outside this main
job); and I didn't have much spare time. But the project certainly
looked interesting, and under different circumstances I probably would
have done it.
I believe this is the project that ended up as the WattsUnlimited power
supply, but I'm not sure of this connection. If I had done it, it would
have been regulated.
Larry:
> What is the actual maximum 120 ripple permissible for a supply like
> this?
For a typical switchmode power supply with proper feedback, the
acceptable ripple can actually be quite high. It's a design tradeoff.
The switcher part is designed to be able to work from a DC input from a
certain minimum to a certain maximum. The minimum is given by the peak
of the lowest expected line voltage, minus the drop due to resistance,
minus the ripple. And the maximum is given by the peak of the highest
expected line voltage, plus some safety overhead. So, by accomodating
for a wider DC voltage range at the switching converter, a smaller
filter cap becomes acceptable.
But accomodating a wider DC voltage range increases the cost of the
switching converter. The lower the voltage, the higher current it has to
handle. To operate at lower voltages, the transformer needs to have more
turns on the secondary, and that in turn causes higher voltage (at much
shorter duty cycle) when the input voltage is nearits maximum. So, all
the parts, specially the transistors and diodes, need to widthstand both
higher voltage and higher current, if you want more DC input voltage
range. In extreme cases there can also be stability issues, because the
loop gain of the switcher changes when the input voltage voltage
changes, making it increasingly hard to optimize the control loop.
I would say that typically it would be reasonable to work with an input
ripple of up to 30% of the nominal input voltage.
> I would like to be able to start/stop the supply with the PTT switch
For that you don't need to charge and discharge the input filter cap!
You leave the input section (rectifier and filter) running all the time,
and just start and stop the switcher proper. The output cap of a
switcher can be very small, and it's not difficult at all to start it up
and stop it in one millisecond or so.
> The "H-bridge" seems to be the choice these days. Is that for
> efficiency?
Not mainly. It's mostly because it gives the best performance/price
ratio and robustness, by far. At least in the power class from 1kW up.
> My "classic" supplies just used a push-pull arrangement with a center
> tapped primary. Seems simpler, but surely there must be a reason why
> the larger supplies don't seem to use push-pull
There are several reasons! I will list them:
1) In a push-pull circuit you need two transistors that each can work at
the full input current, twice the input voltage, and in addition there
can be spikes far exceeding that voltage. In an H-bridge you need 4
transistors, but while these need the same current rating, they need
only to be rated for one time the input voltage, and there can be no
spikes, so their rating can be quite tight! And four transistors for,
say, 20A at 400V are a lot cheaper than 2 transistors for 20A at 1200V,
with comparable loss, which would be about the minimum acceptable rating
for a push-pull 230V offline switcher!
2) Push pull circuits are prone to flux decentering and consequent core
saturation. They really need active flux centering to be safe. Bridge
circuits instead can use a simple DC blocking cap in series with the
primary to make them totally immune to this problem.
3) The transformer is simpler for a bridge, more efficient because the
full winding is always used, and works at lower end-to-end voltage.
4) The switching losses are much lower, because switching takes place
over half the voltage swing (at the same current!), and because snubbers
can easily feed back energy into the primary circuit. Push pull snubbers
can only feed back a part of the energy.
5) Bridge circuits generate less QRM, due to the lower switched voltages.
> although all the smaller ones sure do.
Push-pull circuits are mostly only used when the input voltage is so
small that the conduction losses of a H bridge would be excessive, and
anyway the voltage rating for the devices is no issue. Typical case is
switchers with 12V input. But among off-line switchers, I have never in
my life seen a push-pull one! It would make no sense.
Off-line switchers too small to justify an H-bridge usually are of the
half-bridge type (PC power supplies), even smaller ones are single-ended
forward switchers (common in the 100W class), and the smallest are
fly-back types (cellphone chargers).
> Seems like it would be a good idea to make smaller "stackable"
> supplies. Say something like 700 VDC at 2.0 amps max, then stack them
> in series to get what you need, 1400V, 2100V, 2800V, 3500V. Like
> that.
Yes, indeed that's a good way to do a high voltage supply. Either
stacking the entire supplies, or for economy stacking just the
secondary/rectifier/filter groups, using a common feedback and drive
circuit, with the primaries in parallel.
> I see no reason for feedback for voltage regulation. My old
> T-supplies were very stiff and the supply in the article claims 50
> volts drop from ZSAC to full load. Arguments for or against?
I'm in favor of regulation. While the feedback control circuit adds some
complexity, it adds, extremely little additional cost. For that tiny
additional cost, it's great to be able to operate from unstable gasoline
generators, backcountry AC lines, etc. And then, the savings start: As
commented at the start of this ragchew, the feedback can save a lot of
filter capacitance! And large filter caps ARE expensive! So, usually a
well designed supply with feedback is BOTH better AND less expensive
than an unregulated one!
> Simpler is always better as far as I'm concerned.
Normally yes, but not when "simpler" leads to "much more expensive". I
try to simplify my designs down to the point where any additional
simplifications would cost too much in performance, cost, or reliability.
> Is there a good reference for design of this type of switcher
> someplace.
The best I have seen is contained in data sheets and application notes
for switchmode control ICs, MOSFETs, IGBTs, and fast diodes. But many
topics are not properly covered there. Unless you have the luck of
finding a really good book (I haven't), you have to piece together your
knowledge from many sources, plus your own deductions and
experimentation. It helps to get your hands on surplus (but please,
MODERN!) power equipment, and look how the pro's are doing it!
Manfred.
========================
Visit my hobby homepage!
http://ludens.cl
========================
_______________________________________________
Amps mailing list
Amps@contesting.com
http://lists.contesting.com/mailman/listinfo/amps
|