SMPS topology selection

[Michael]

Here is what I have: high power pulses (I~25A=constant, V~50-80V, t
=300us, rep. rate=100Hz; the pulses come in bursts, long term average
power is low). Input voltage is 10V.
Flyback converter is working fine recharging energy storing capacitor
at the moment. It's hand held instrument, I doubt I can fit much more
than 100uF capacitor, converter output is 200V now.
I was asked to increase power by factor of 4-6 (2x current, 2x
frequency...). ~250W power bursts, that is.
I am looking for the solution right now. Any suggestions how to
recharge the high voltage capacitor from 10V power rail quickly
(professional photoflashes/strobes may be doing something like it)
will be appreciated.
Thank you!
Michael

 

[MooseFET]

Here is what I have: high power pulses (I~25A=constant, V~50-80V, t
=300us, rep. rate=100Hz; the pulses come in bursts, long term average
power is low). Input voltage is 10V.
Flyback converter is working fine recharging energy storing capacitor
at the moment. It's hand held instrument, I doubt I can fit much more
than 100uF capacitor, converter output is 200V now.
I was asked to increase power by factor of 4-6 (2x current, 2x
frequency...). ~250W power bursts, that is.
I am looking for the solution right now. Any suggestions  how to
recharge the high voltage capacitor from 10V power rail quickly
(professional photoflashes/strobes may be doing something like it)
will be appreciated.
Thank you!
Michael

Does the load fully discharge the capacitor? If not what voltage are
you starting up from?

The flyback is likely to be the best if you are working over a large
range. Changing the frequency as the voltage rises helps.

 

[legg]

Here is what I have: high power pulses (I~25A=constant, V~50-80V, t
=300us, rep. rate=100Hz; the pulses come in bursts, long term average
power is low). Input voltage is 10V.
Flyback converter is working fine recharging energy storing capacitor
at the moment. It's hand held instrument, I doubt I can fit much more
than 100uF capacitor, converter output is 200V now.
I was asked to increase power by factor of 4-6 (2x current, 2x
frequency...). ~250W power bursts, that is.
I am looking for the solution right now. Any suggestions how to
recharge the high voltage capacitor from 10V power rail quickly
(professional photoflashes/strobes may be doing something like it)
will be appreciated.

Can the source deliver the 250W power level comfortably? Is the source
included in the hand-held device.

Is the load voltage self-regulated, providing that the current is
constant?

What actual physical volume is permitted?

RL

 

[Michael]

Does the load fully discharge the capacitor? If not what voltage are
you starting up from?

The flyback is likely to be the best if you are working over a large
range. Changing the frequency as the voltage rises helps.

The pulse energy is in the order of 1J now, it needs to go up to 2J+.
I have 100uF cap that discharges from ~210V to ~160V with every pulse.
I do not think I can increase capacitance (size) as I have to use
photoflash/strobe capacitors (high current).

... Changing the frequency as the voltage rises helps.

Can you think of a controller that does it?

 

[Joerg]

The pulse energy is in the order of 1J now, it needs to go up to 2J+.
I have 100uF cap that discharges from ~210V to ~160V with every pulse.
I do not think I can increase capacitance (size) as I have to use
photoflash/strobe capacitors (high current).

Then your only options are to replenish from the supply at much higher
power or draw it down some more and boost via a MHz-switcher. Either
would most likely consume more space and BOM budget than a larger cap.
IOW you are between a rock and a hard spot.

Can you think of a controller that does it?

You'd have to roll your own.

 

[Michael]

Can the source deliver the 250W power level comfortably? Is the source
included in the hand-held device.

Is the load voltage self-regulated, providing that the current is
constant?

What actual physical volume is permitted?

RL

The load is ugly (from EE point of view), it's spark discharge. It is
somewhere between 50 and 100V.I think you know how it works: the marketing wants the product to be
as small/light as possible with infinite battery life and <$1 BOM. You
start negotiating and get something reasonable. Seriously: we need to
make it as small/light as REALISTICALLY possible.

 

[Joerg]

The load is ugly (from EE point of view), it's spark discharge. It is
somewhere between 50 and 100V.
I think you know how it works: the marketing wants the product to be
as small/light as possible with infinite battery life and <$1 BOM. You
start negotiating and get something reasonable. Seriously: we need to
make it as small/light as REALISTICALLY possible.

So, how small does the cap have to be?

http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=P10124-ND

 

[legg]

Can you think of a controller that does it?

This is a misconception drawn from the characteristics of simple
saturation-limited blocking oscillators (self-oscillating flyback with
complete energy transfer).

The audible frequency varies as reset intervals decrease due to the
increasing storage voltage. It is a byproduct, not a design 'feature'.

RL

 

[legg]

Yes, it can. It's a123systems.com battery pack.

At your pulsed power level, their design life is 1000 discharges, or
100 hours, with 6 minutes of of actual operation per full charge.

The 'flash' capacitor you are using, at 100Hz discharge frequency,
will reach it's 5000 x discharge design life in one hour, or 10
battery recharge cycles.

The load is ugly (from EE point of view), it's spark discharge. It is
somewhere between 50 and 100V.

Who or what has determined that the load current must be constant?
Surely this is a simpler energy requirement that could be passively
characterized, within acceptible performance-related limits?

I think you know how it works: the marketing wants the product to be
as small/light as possible with infinite battery life and <$1 BOM. You
start negotiating and get something reasonable. Seriously: we need to
make it as small/light as REALISTICALLY possible.

You might consider separating partsthat must be actually in-hand, from
parts that provide, process or store the energy. Product endurance and
operating life may be a more important issue, as I can envisage few
consumer products that would require or could safely employ the enegy
characteristics that you describe.

Nail down some real requirements.

RL

 

[Michael]

So, how small does the cap have to be?

http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=P101...

I use two United Chemi-Con photoflash (PH series) caps. 10mm diameter,
27mm long. I am not sure I can use anything bigger. The existing board
might take couple more of these, but I haven't taken into account
other parts size increase...
(

 

[Joerg]

I use two United Chemi-Con photoflash (PH series) caps. 10mm diameter,
27mm long. I am not sure I can use anything bigger. The existing board
might take couple more of these, but I haven't taken into account
other parts size increase...
(

Ok, that looks pretty tapped out. You might have to defer other part
size increases, migrate the circuitry to 0201 parts and so on.

Might want to look at chips like the LT3485 series for charging but
check the availability situation first since this stuff is prone to go
unobtanium at times:
http://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1003,C1042,C1098,P13555,D9557

 

[Michael]

At your pulsed power level, their design life is 1000 discharges, or
100 hours, with 6 minutes of of actual operation per full charge.

The 'flash' capacitor you are using, at 100Hz discharge frequency,
will reach it's 5000 x discharge design life in one hour, or 10
battery recharge cycles.


Who or what has determined that the load current must be constant?
Surely this is a simpler energy requirement that could be passively
characterized, within acceptible performance-related limits?


You might consider separating partsthat must be actually in-hand, from
parts that provide, process or store the energy. Product endurance and
operating life may be a more important issue, as I can envisage few
consumer products that would require or could safely employ the enegy
characteristics that you describe.

Nail down some real requirements.

RL

The caps are rated 5000 pulses into Xenon lamp (<1Ohm). I hope that
life time goes up with current going down faster than linearlyI need to control something. Voltage is out of question (it's plasma).
What is the other option?

 

[legg]

Ok, that looks pretty tapped out. You might have to defer other part
size increases, migrate the circuitry to 0201 parts and so on.

Might want to look at chips like the LT3485 series for charging but
check the availability situation first since this stuff is prone to go
unobtanium at times:
http://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1003,C1042,C1098,P13555,D9557

Charge time is measured in seconds for this part.

RL

 

[legg]

The caps are rated 5000 pulses into Xenon lamp (<1Ohm). I hope that
life time goes up with current going down faster than linearly

Hope is not a recommended design tool.

I admit that I have no detailed application info on these parts,
beside the general application guidelines published by UCC/NCC.

The 'special design' of photoflash capacitors is not one specifically
addressing discharge current levels - it only addresses the
electrolytic capacitor's tendency to internally reform (producing
reduced capacitance) under the influence of DC current, in the absence
of extreme pulses.

Any reliability evaluation performed assuming one pulse per second is
probably useless under pulse repetition rates that are two orders of
magnitude higher. A thermometer might produce more meaningful data.

With access to the actual circuitry, you are more able to determine
the nature of part reliability issues than anyone else, once you are
aware of them. You must not assume, however. If there is some urgency
at producing a completed product, you will find things go much more
smoothly if you use data for parts that have already been assessed for
an intended application.

I need to control something. Voltage is out of question (it's plasma).
What is the other option?

The voltage control being free, pulse forming, non-dissipative or
recoverable energy transfer is popular, but usually at higher voltages
where suitable switches are hard to come by.

The PWM output control method, which you suggest is intended or is
already being used, could deliver the intended increased energy
without increasing the current, simply by running longer (>300uSec)
during the intended discharge interval, if the energy is there to
deliver. I assume this has been determined to be ineffective.

Direct energy transfer by a boost circuit is not out of the question,
although the source voltage is low. An inductor could store .6J (80v x
25A x 300uSec)and deliver it to the load at 100Hz rep rates, without
intermediate storage. A 480uH inductor requires 3mSec to develop 50A
peak from an 8V source, and will fully discharge into an 80V clamp in
300uSec. This uses a single switch and rectifier, with peak current
and rep rate being the controllable quantities, once hardware is set
in stone.

While consolidating the parts into one humungous switch and choke may
not seem like a forward step, the thermal and current issues with the
components involved are well known and less subject to crippling life
factors.

Reducung the magnetic storage component size might be possible if the
source and load would tolerate a 'scrubbing' HF current without
malfunction vs the present pulsating DC being employed.

RL

 

[Joerg]

Hope is not a recommended design tool.

Hey, we built our whole church on that ;-)

I admit that I have no detailed application info on these parts,
beside the general application guidelines published by UCC/NCC.

The 'special design' of photoflash capacitors is not one specifically
addressing discharge current levels - it only addresses the
electrolytic capacitor's tendency to internally reform (producing
reduced capacitance) under the influence of DC current, in the absence
of extreme pulses.

Any reliability evaluation performed assuming one pulse per second is
probably useless under pulse repetition rates that are two orders of
magnitude higher. A thermometer might produce more meaningful data.

With access to the actual circuitry, you are more able to determine
the nature of part reliability issues than anyone else, once you are
aware of them. You must not assume, however. If there is some urgency
at producing a completed product, you will find things go much more
smoothly if you use data for parts that have already been assessed for
an intended application.

While I also would not use flash caps here, we did as kids (in Germany).
Being on a shoestring budget but wanting to build those honking kilowatt
amps we took the 230V mains in, sans transformer. Rectifier cascade to
900VDC, bingo (don't tell TUEV I did this...). I secured a bargain of
nice professional and large Siemens caps. The others found a sale of
flash caps, much smaller and prettier. "Zweite Wahl", loosely translated
"Fell off the dump truck". We beat the dickens out of those amps.

One fine day ... KABLAM! One of my Siemens caps decided to mutate into a
rocket. None of the flash cap dudes had one blow, ever. That really
miffed me. But still I would not use flash caps for this kind of
continuous duty application.

The voltage control being free, pulse forming, non-dissipative or
recoverable energy transfer is popular, but usually at higher voltages
where suitable switches are hard to come by.

The PWM output control method, which you suggest is intended or is
already being used, could deliver the intended increased energy
without increasing the current, simply by running longer (>300uSec)
during the intended discharge interval, if the energy is there to
deliver. I assume this has been determined to be ineffective.

Direct energy transfer by a boost circuit is not out of the question,
although the source voltage is low. An inductor could store .6J (80v x
25A x 300uSec)and deliver it to the load at 100Hz rep rates, without
intermediate storage. A 480uH inductor requires 3mSec to develop 50A
peak from an 8V source, and will fully discharge into an 80V clamp in
300uSec. This uses a single switch and rectifier, with peak current
and rep rate being the controllable quantities, once hardware is set
in stone.

While consolidating the parts into one humungous switch and choke may
not seem like a forward step, the thermal and current issues with the
components involved are well known and less subject to crippling life
factors.

Reducung the magnetic storage component size might be possible if the
source and load would tolerate a 'scrubbing' HF current without
malfunction vs the present pulsating DC being employed.

However, if this is supplied from a regular battery I guess it would
require a similar size cap at the battery voltage level.

 

[legg]

legg wrote:

While I also would not use flash caps here, we did as kids (in Germany).
Being on a shoestring budget but wanting to build those honking kilowatt
amps we took the 230V mains in, sans transformer. Rectifier cascade to
900VDC, bingo (don't tell TUEV I did this...). I secured a bargain of
nice professional and large Siemens caps. The others found a sale of
flash caps, much smaller and prettier. "Zweite Wahl", loosely translated
"Fell off the dump truck". We beat the dickens out of those amps.

One fine day ... KABLAM! One of my Siemens caps decided to mutate into a
rocket. None of the flash cap dudes had one blow, ever. That really
miffed me. But still I would not use flash caps for this kind of
continuous duty application.

However, if this is supplied from a regular battery I guess it would
require a similar size cap at the battery voltage level.

It's a pity that 'flash' characteristics aren't offered in different
voltages and temperature grades, or that they aren't issued basic cap
ratings, for comparison. But, hey, the manufacturers are the ones on
the line when it comes to dead inventory and some do offer 'custom'
formulations.

Local ripple decoupling need not be as hairy an issue. Those ARE some
really nice batteries - with the discharge rates permitted. If they
work out, they're obvious replacements for nicad-only apps. I 'hope'
it turns out that lithium has no serious side effects, besides the
current known pharmacological ones. (Those might even help some folks


RL

RL

 

[MooseFET]

The pulse energy is in the order of 1J now, it needs to go up to 2J+.
I have 100uF cap that discharges from ~210V to ~160V with every pulse.
I do not think I can increase capacitance (size) as I have to use
photoflash/strobe capacitors (high current).>... Changing the frequency as the voltage rises helps.

Can you think of a controller that does it?

Nearly any of the ones that use an external RC for the frequency can
be made to do it by adding one extra resistor over to the output.

 

[Joerg]

Nearly any of the ones that use an external RC for the frequency can
be made to do it by adding one extra resistor over to the output.

In very small ranges, usually. And beware the limits. Running into the
Isense dead time can lead to molten solder splattering about.

 

[przemek klosowski]

The pulse energy is in the order of 1J now, it needs to go up to 2J+. I
have 100uF cap that discharges from ~210V to ~160V with every pulse. I
do not think I can increase capacitance (size) as I have to use
photoflash/strobe capacitors (high current).

Energy is 1/2 C V^2, so you are delivering 0.925 J from .5 * 100 micro F
* ((210V)^2+-(160V)^2); if you discharged to zero volts you'd get full
2.2 J . You probably have to up the voltage, which is easier than upping
the capacitance because voltage comes in squared (you just need 40% extra
to double the energy, etc).

 

[MooseFET]

In very small ranges, usually. And beware the limits. Running into the
Isense dead time can lead to molten solder splattering about.

Take a look at the LT1246 for example. You can easily vary its
frequency over a 3:1 range with no risk of splattered solder.

http://www.linear.com/pc/productDetail.jsp?navId=H0,C1,C1003,C1142,C1137,P1158

You just run a resistor from pin-4 to the output of the switcher. The
resistor looks, near enough, like a constant current source to the
oscillator circuit so other than making the ramp more linear it
doesn't bother the operation of the part.