Supply for pulsed load

[Anno Siegel]

Suppose a supply has to deliver a considerable load, but only for short
pulses with a low overall duty cycle. One could use a standard supply
for the full load and be done with, or try to take advantage of the
low duty cycle.

To be concrete, I'm looking at a supply for around 12 V (voltage isn't
fixed yet) whose load comes in pulses of up to 4 A with 10 - 30 ms duration
and 10% duty cycle, tops. Regulation can be lax and ripple isn't critical,
if it matters.

My immediate idea is to use a first-stage supply with some extra voltage,
say 16 V, that feeds a capacitor large enough not to lose more than 4 V
per pulse. The power of the first stage, at 10% duty cycle, is (ideally)
10% of the full load, so 4.8 W instead of 48 W. A secondary linear regulator
would shave off the excess voltage and deliver 12 V. (Dropout voltage
neglected.)

That would need a capacitor C = 4A*20ms/4V = 20,000 uF, if my arithmetic
is right. That's not out of the world, but it won't come for nothing at
16 V or, likely, more. (Does it have to see 16 V?). The idea is to save
bulk and money, not to spend more. The whole design looks rather clumsy.
Could the functionality be realized in one stage? Are there tried-and-
proven solutions?

Anno

 

[Larry Brasfield]

"Anno Siegel" <anno4000@lublin.zrz.tu-berlin.de> wrote in
message news:d8v0gh$3gl$1@mamenchi.zrz.TU-Berlin.DE...

Suppose a supply has to deliver a considerable load, but only for short
pulses with a low overall duty cycle. One could use a standard supply
for the full load and be done with, or try to take advantage of the
low duty cycle.

To be concrete, I'm looking at a supply for around 12 V (voltage isn't
fixed yet) whose load comes in pulses of up to 4 A with 10 - 30 ms duration
and 10% duty cycle, tops. Regulation can be lax and ripple isn't critical,
if it matters.

My immediate idea is to use a first-stage supply with some extra voltage,
say 16 V, that feeds a capacitor large enough not to lose more than 4 V
per pulse. The power of the first stage (with 10% duty cycle) is (ideally)
10% of the full load, so 4.8 W instead of 48 W. A secondary linear regulator
would shave off the excess voltage and deliver 12 V.

That would need a capacitor C with 4A*20ms/C = 4V, or C = 20,000 uF, if
my arithmetic is right. That's not out of the world, but it won't come
for nothing at 12 V (does it have to see 12 V?).

It would have to operate at 16 V according to your description.

The idea is to save
bulk and money, not to spend more. The whole design looks rather clumsy.
Could the functionality be realized in one stage?

One stage is certainly doable, depending on your actual regulation
and ripple requirements. Does your volume justify a custom design
of the initial converter?

A supply designed to deliver 4 A pulses at 5 W average power
would be considerably smaller than a 50 W supply. Think of a
supply with semiconductors good for the higher current but with
smaller heatsinks than would be required to deliver that current
continuously. I would be tempted to exploit the fact that input
resevoir capacitors in a switching supply can be depleted quite
a bit before regulation suffers. At the timescale you mentioned,
getting a switching regulator to convert (for example) an input
resevoir ranging from 300 V down to 150 V during the pulse
into 12 V with 10+ KHz load response would be doable. By
keeping the efficiency up, supply dissipation could be kept to
less than 1 W, which should get bulk and cost down.

Are there tried-and-
proven solutions?

I've seen supplies custom designed for such requirements.
They are in equipment that is costly to service, and have
not been a problem, so I suppose they are tried and proven.

--
--Larry Brasfield
email: donotspam_larry_brasfield@hotmail.com
Above views may belong only to me.

 

[Anno Siegel]

Tim Wescott <tim@seemywebsite.com> wrote in sci.electronics.design:

Anno Siegel wrote:

Suppose a supply has to deliver a considerable load, but only for short
pulses with a low overall duty cycle. One could use a standard supply
for the full load and be done with, or try to take advantage of the
low duty cycle.

To be concrete, I'm looking at a supply for around 12 V (voltage isn't
fixed yet) whose load comes in pulses of up to 4 A with 10 - 30 ms duration
and 10% duty cycle, tops. Regulation can be lax and ripple isn't critical,
if it matters.

My immediate idea is to use a first-stage supply with some extra voltage,
say 16 V, that feeds a capacitor large enough not to lose more than 4 V
per pulse. The power of the first stage (with 10% duty cycle) is (ideally)
10% of the full load, so 4.8 W instead of 48 W. A secondary linear regulator
would shave off the excess voltage and deliver 12 V.

That would need a capacitor C with 4A*20ms/C = 4V, or C = 20,000 uF, if
my arithmetic is right. That's not out of the world, but it won't come
for nothing at 12 V (does it have to see 12 V?). The idea is to save
bulk and money, not to spend more. The whole design looks rather clumsy.
Could the functionality be realized in one stage? Are there tried-and-
proven solutions?

Anno

If your primary source of power (batteries? power line? Hamster with a
generator?)

Power line.

can take the pulsing you could design a power supply that
would deliver 4A without drooping but size the heatsinks and wire sizes
for the 1.26A RMS current. This is probably the best approach from a
size and economy standpoint.

Thanks. Your keywords about what to dimension accoding to what were what
I needed to solidify the vague idea of "make it for 4.8W *and* for 4A,
it won't overheat".

The only reason I could see to take your two-step approach would be to
improve the power factor on the input side to your supply -- in theory
you could get the input power down to a constant value as you state.

No problem at all (I think). It was the amateurish idea I had to add
something to exploit the duty cycle. The right solution is to leave
some things off, i.e. to exploit it, period. Thanks again.

Anno

 

[Anno Siegel]

Larry Brasfield <donotspam_larry_brasfield@hotmail.com> wrote in sci.electronics.design:

"Anno Siegel" <anno4000@lublin.zrz.tu-berlin.de> wrote in
message news:d8v0gh$3gl$1@mamenchi.zrz.TU-Berlin.DE...

[some snippage]

To be concrete, I'm looking at a supply for around 12 V (voltage isn't
fixed yet) whose load comes in pulses of up to 4 A with 10 - 30 ms duration
and 10% duty cycle, tops. Regulation can be lax and ripple isn't critical,
if it matters.

My immediate idea is to use a first-stage supply with some extra voltage,
say 16 V, that feeds a capacitor large enough not to lose more than 4 V
per pulse. The power of the first stage (with 10% duty cycle) is (ideally)
10% of the full load, so 4.8 W instead of 48 W. A secondary linear regulator
would shave off the excess voltage and deliver 12 V.

That would need a capacitor C with 4A*20ms/C = 4V, or C = 20,000 uF, if
my arithmetic is right. That's not out of the world, but it won't come
for nothing at 12 V (does it have to see 12 V?).

It would have to operate at 16 V according to your description.

Right. I corrected that in a superseding posting, but Usenet (well, my
server) seems to spin faster these days. I used to be able to safely
catch a posting before it left my server. No more, it seems. Sorry
for the confusion, if any.

The idea is to save
bulk and money, not to spend more. The whole design looks rather clumsy.
Could the functionality be realized in one stage?

One stage is certainly doable, depending on your actual regulation
and ripple requirements.

Both are relaxed. The pulses drive solenoids that make a ferromagnetic
peg jump against a metallophone plate to make it sound. Ripple and poor
regulation won't be a big deal.

Does your volume justify a custom design
of the initial converter?

Oh dear. I don't see a huge market for MIDI metallophones
No, if at all, this is going to be a one-off thing, amateur level.
However, that may also justify custom design.

A supply designed to deliver 4 A pulses at 5 W average power
would be considerably smaller than a 50 W supply. Think of a
supply with semiconductors good for the higher current but with
smaller heatsinks than would be required to deliver that current
continuously. I would be tempted to exploit the fact that input
resevoir capacitors in a switching supply can be depleted quite
a bit before regulation suffers. At the timescale you mentioned,
getting a switching regulator to convert (for example) an input
resevoir ranging from 300 V down to 150 V during the pulse
into 12 V with 10+ KHz load response would be doable. By
keeping the efficiency up, supply dissipation could be kept to
less than 1 W, which should get bulk and cost down.

That seems to be essentially the solution Tim Wescott suggested
in the other branch: Get the time constants right and dimension
for full current but reduced power. Very well.

Are there tried-and-
proven solutions?

I've seen supplies custom designed for such requirements.
They are in equipment that is costly to service, and have
not been a problem, so I suppose they are tried and proven.

I should have been clearer about my purpose from the start. No need
for industrial strength, it just has to work. If it can be simple,
I want it simple. I think I over-estimated the power problem. I'm
glad to know it won't be very hard to solve.

Thanks, Anno

 

[Mac]

On Fri, 17 Jun 2005 17:23:02 +0000, Anno Siegel wrote:

Suppose a supply has to deliver a considerable load, but only for short
pulses with a low overall duty cycle. One could use a standard supply
for the full load and be done with, or try to take advantage of the
low duty cycle.

To be concrete, I'm looking at a supply for around 12 V (voltage isn't
fixed yet) whose load comes in pulses of up to 4 A with 10 - 30 ms duration
and 10% duty cycle, tops. Regulation can be lax and ripple isn't critical,
if it matters.

My immediate idea is to use a first-stage supply with some extra voltage,
say 16 V, that feeds a capacitor large enough not to lose more than 4 V
per pulse. The power of the first stage, at 10% duty cycle, is (ideally)
10% of the full load, so 4.8 W instead of 48 W. A secondary linear regulator
would shave off the excess voltage and deliver 12 V. (Dropout voltage
neglected.)

That would need a capacitor C = 4A*20ms/4V = 20,000 uF, if my arithmetic
is right. That's not out of the world, but it won't come for nothing at
16 V or, likely, more. (Does it have to see 16 V?). The idea is to save
bulk and money, not to spend more. The whole design looks rather
clumsy. Could the functionality be realized in one stage? Are there
tried-and- proven solutions?

Anno

I wonder if you could use a smaller capacitor at a much higher voltage,
and use a switch-mode regulator to bring the voltage down. Maybe something
like a pure-hysteresis converter. Then you would need an inductor and a
filter cap, too, obviously.

I think you need to deliver 12V * 4A * 0.03 seconds = 1.44 Joules in your
worst case pulse. (Double-check my notoriously poor arithmetic and even
my basic equation.)

So, if you wanted to go with a Voltage of, say, 36 Volts, you could
drop all the way to, say, 15 Volts, and you wanted that delta V to
correspond to a delta E of 1.44 Joules, you would need a capacitor of
what size?

dE = E2-E1 = (C/2) * (V2*V2-V1*V1)

So:

C = 2 * (E2-E1)/(V2*V2-V1*V1)

C = 2 * 1.44 / (36*36 - 15*15)

C=~ 2700 uF.

This seems potentially workable. I don't know if it would save you money
or not. And you still have to go through the exercise of trying out
different inductor and output filter capacitor values (analytically, I
mean) to see what works, and what frequencies and duty cycles you will get
for the converter.

The advantage of a pure hysteresis converter is that you don't have any
problems with stability or minimum currents or duty cycles. It can go all
the way down to zero duty cycle, and in this case it probably will in
between pulses. But any step-down converter that can handle the range of
input Voltages and output currents should do.

Just my first thought on the subject.

--Mac