Maker Pro
Maker Pro

Switching power supply behavior

S

Scott Miller

Jan 1, 1970
0
Didn't we tell you how to dump that asinine MOV crap in a previous thread
you started about this p.o.s.? -PLONK

Yes, I asked for suggestions on replacing the shunt regulator. I got some
great input on using op amps to buffer the feedback signal. Turns out
that's not really the tough part. I've had decent results feeding the MCU's
ADC directly from a voltage divider. The more difficult part is making
sense of those readings and determining how to change the PWM signal.

I agree that the MOVs as a shunt regulator is a POS. Unfortunately just
about every hobbyist design I've ever seen for a Geiger counter uses the
same sort of thing, sometimes using neon bulbs or lots of zeners. I've been
experimenting with using the MOVs to trigger a PWM change because I know
they work, which eliminates one more possible thing to screw up as I start
making changes to the design.

I'm not asking anyone to design this thing for me. I'm just asking for help
understanding what's going on. I've gotten some great feedback on things
like the diode switching speed. I think I'm well on the way to a reasonably
efficient, inexpensive, non-crappy design, and I appreciate all the help
I've gotten here.

Scott
 
J

John Popelish

Jan 1, 1970
0
Scott said:
Yes, I asked for suggestions on replacing the shunt regulator. I got some
great input on using op amps to buffer the feedback signal. Turns out
that's not really the tough part. I've had decent results feeding the MCU's
ADC directly from a voltage divider. The more difficult part is making
sense of those readings and determining how to change the PWM signal.

Once you get the rectification process better under control, this will
get easier. You may want to parallel the resistive divider with a
similar capacitive divider (put a large, low voltage cap in series
with the ground end of the final supply storage capacitor). Your ADC
sample and hold will be better behaved.

A simple proportional control may be all you need. This would involve
having an output pulse duty cycle varied inversely with the ADC
measurement, with some max. pulse width to handle start up or overload
situations.
You turn the proportionality constant up till the output starts to
wobble, and then cut it is half. You can temporarily use a second ADC
input with a pot across the supply as the gain adjustment. After you
decide what gain is appropriate, store it in the program, permanently.
I agree that the MOVs as a shunt regulator is a POS.

It ain't so bad. At such low current, it is quite functional if
wasteful of battery power. It also takes less understanding to get it
working.
Unfortunately just
about every hobbyist design I've ever seen for a Geiger counter uses the
same sort of thing, sometimes using neon bulbs or lots of zeners. I've been
experimenting with using the MOVs to trigger a PWM change because I know
they work, which eliminates one more possible thing to screw up as I start
making changes to the design.

You can also make the MOV part of the voltage divider to the ADC to
increase the effective signal in the range of interest, though the
temperature coefficient will be higher.
 
S

Scott Miller

Jan 1, 1970
0
Once you get the rectification process better under control, this will
get easier. You may want to parallel the resistive divider with a
similar capacitive divider (put a large, low voltage cap in series
with the ground end of the final supply storage capacitor). Your ADC
sample and hold will be better behaved.

Ok, I'll have to give that a try. I should have the high-speed diodes in a
few days - it'll be interesting to see how they affect the output.
A simple proportional control may be all you need. This would involve
having an output pulse duty cycle varied inversely with the ADC
measurement, with some max. pulse width to handle start up or overload
situations.

Pretty much what I was trying before...
You turn the proportionality constant up till the output starts to
wobble, and then cut it is half. You can temporarily use a second ADC
input with a pot across the supply as the gain adjustment. After you
decide what gain is appropriate, store it in the program, permanently.

That's a good idea, might save me a lot of reprogramming. Assuming I can
free up another ADC. I only used this particular part because I buy them by
the hundreds.
It ain't so bad. At such low current, it is quite functional if
wasteful of battery power. It also takes less understanding to get it
working.

Yeah, battery power is the main issue. This thing sucks a 9v alkaline dry
in no time. I know there are some designs that'll run for a month on one.
The MCU and LCD will add a few ma, but I want to be able to get at least 12
hours of runtime.
You can also make the MOV part of the voltage divider to the ADC to
increase the effective signal in the range of interest, though the
temperature coefficient will be higher.

I'd rather eliminate the MOVs completely if possible, but I'll keep
experimenting.

Thanks,

Scott
 
K

Ken Smith

Jan 1, 1970
0
So would a higher ESR help reduce the effect?

No, higher ESR increases the effect.
It's been a couple of years since I did the original version, but I think I
just assumed that the peak voltage would indicate the most efficient use of
the transformer.

This is somethimes true and sometimes not.
 
J

John Popelish

Jan 1, 1970
0
Scott Miller wrote:
(snip)
Yeah, battery power is the main issue. This thing sucks a 9v alkaline dry
in no time. I know there are some designs that'll run for a month on one.
The MCU and LCD will add a few ma, but I want to be able to get at least 12
hours of runtime.

The resistor you have in series with the transformer primary is
soaking up a lot of energy, also. Once you have the feedback control
working (and the maximum on time limit that safely limits current in
the event of output overload) you might replace it with a ferrite bead
on a wire or other small inductor. C3 should be big enough to supply
an entire power pulse with only a little sag. Say, a few microfarads
of electrolytic in parallel with .1 to 1 uf of film or ceramic for the
high frequencies. This will give you more voltage across the primary
during the on time and more energy in the core to be released during
the off time (narrower pulse needed).
 
S

Scott Miller

Jan 1, 1970
0
The resistor you have in series with the transformer primary is
soaking up a lot of energy, also. Once you have the feedback control
working (and the maximum on time limit that safely limits current in
the event of output overload) you might replace it with a ferrite bead
on a wire or other small inductor. C3 should be big enough to supply

What exactly does the ferrite bead accomplish? I've got plenty on hand to
try it with...
an entire power pulse with only a little sag. Say, a few microfarads
of electrolytic in parallel with .1 to 1 uf of film or ceramic for the

Looks like it's 220 uF right now.
high frequencies. This will give you more voltage across the primary
during the on time and more energy in the core to be released during
the off time (narrower pulse needed).

Ok. I'll play with that some more when I've got the rectification taken
care of.

Thanks,

Scott
 
R

Rich Grise

Jan 1, 1970
0
Yes, I asked for suggestions on replacing the shunt regulator. I got some
great input on using op amps to buffer the feedback signal. Turns out
that's not really the tough part. I've had decent results feeding the MCU's
ADC directly from a voltage divider. The more difficult part is making
sense of those readings and determining how to change the PWM signal.

I agree that the MOVs as a shunt regulator is a POS. Unfortunately just
about every hobbyist design I've ever seen for a Geiger counter uses the
same sort of thing, sometimes using neon bulbs or lots of zeners. I've been
experimenting with using the MOVs to trigger a PWM change because I know
they work, which eliminates one more possible thing to screw up as I start
making changes to the design.

Here's a change you can make with some confidence - use a couple of these
in series:
http://www.vishay.com/docs/88301/15ke.pdf pick the correct voltage of
units, of course. You can probably also get them from Digi-key. And, of
course, that's assuming you're going to insist on using transient
suppressors. ;-)

Then again, they do have a much sharper knee, and since you're going to be
regulating the supply anyway, you could select for just overvoltage, so
they never conduct. :)
I'm not asking anyone to design this thing for me. I'm just asking for
help understanding what's going on. I've gotten some great feedback on
things like the diode switching speed. I think I'm well on the way to a
reasonably efficient, inexpensive, non-crappy design, and I appreciate
all the help I've gotten here.

Another thing you could do is take the flashes out of a couple of
disposable cameras and put them in series. ;-)

Good Luck!
Rich
 
J

John Popelish

Jan 1, 1970
0
Scott said:
What exactly does the ferrite bead accomplish? I've got plenty on hand to
try it with...

Deep breath:

The current consumption of the transformer occurs in pulses (there is
current when the fet is on and none when it is off). This pulsing
load causes the 9 volt battery voltage to bounce because it does not
have a zero internal resistance, and there is inductance in the wiring
between the fet/transformer and the battery. Any circuits at the
fet/transformer end of that wiring will see both the battery bounce
and the additional inductive bounce caused by the pulses. The
capacitor connected directly between fet source and transformer (you
do have it very closely connected to those two points, I assume) acts
as a small local battery that soaks up a lot of that pulsation, so
that it does not all have to come from the battery through the wiring
inductance.

But a capacitor can provide and absorb current only if it changes
voltage. The formula that relates capacitive voltage to current is
I=C*(dv/dt), with I in amperes, C in farads, dv/dt in volts per
second. So if you want the battery to supply the average current
while holding steady voltage and also let the cap bounce up and down
with each pulse a bit, so it can supply the pulses and recharge
between them, you need some elasticity between the battery and the
cap.

That is what the resistor is doing. But it is also dropping voltage
at least equal to the average current from the battery to this
circuit.

An inductor instead, will provide the elasticity that allows the cap
to soak up the pulses while the battery does not, but can waste very
little voltage if its resistance is low. A ferrite bead (or a few in
series) will only separate the harmonics of the pulse (the edges) from
the battery, but that may be enough to make the wiring inductance
fairly insignificant, so at least only the battery internal resistance
is involved in any bounce that other circuits see. A small wound
inductor will give better low pass filtering that separates the cap
voltage changes from the battery. A bigger cap with lower equivalent
series resistance and inductance will have to bounce less to soak up
the transformer pulses, regardless of what size inductor you use.

Looks like it's 220 uF right now.

That is enough to supply essentially an entire pulse with very little
sag, but it will wiggle a bit because of its internal inductance. The
small high frequency capable cap in parallel damps more of that.
 
S

Scott Miller

Jan 1, 1970
0
and the additional inductive bounce caused by the pulses. The
capacitor connected directly between fet source and transformer (you
do have it very closely connected to those two points, I assume) acts

Yep, it's as close as possible...
as a small local battery that soaks up a lot of that pulsation, so
that it does not all have to come from the battery through the wiring
inductance.

Understood that part, that's why the cap is there.
That is what the resistor is doing. But it is also dropping voltage
at least equal to the average current from the battery to this
circuit.

An inductor instead, will provide the elasticity that allows the cap
to soak up the pulses while the battery does not, but can waste very
little voltage if its resistance is low. A ferrite bead (or a few in
series) will only separate the harmonics of the pulse (the edges) from

Ok, I get it. Didn't the the inductance in a ferrite bead would be enough
to make a difference. I use them for filtering out RFI (100 MHz and up) and
never looked much at their characteristics beyond that. Some of this is
starting to come back to me from my ham radio licensing long ago...
That is enough to supply essentially an entire pulse with very little
sag, but it will wiggle a bit because of its internal inductance. The
small high frequency capable cap in parallel damps more of that.

This sounds like a good setup to learn LTSpice with. I found something
similar in the example files. I've also got some dual ferrite cores I can
go throw in the circuit and see what that does to my duty cycle
requirements.

Thanks again,

Scott
 
J

John Popelish

Jan 1, 1970
0
Scott said:
This sounds like a good setup to learn LTSpice with.

Yes. LTSpice has provision to add specified resistance and inductance
ot each capacitor so you can test the effect of various amounts of
these parasitics.
 
K

Ken Smith

Jan 1, 1970
0
[... me ...]
As I said, it can be regulated as long as the N*Vin is considerable
less than the desired voltage, so that adjusting the flyback voltage
component is enough.

The OP needs a good rule for "considerable". I think he is safe so long
as N*Vin is less than about 50% of the desired. He should be sure to
limit the duty cycle of the PWM to not exceed this same percentage.
 
Top