20-60A adjustable ~200vdc current mode buck design

[John Barrett]

I'm looking to build a power supply for a krypton arc lamp, and need a
design that can deliver an adjustable 20-60 amps into a LOW resistance load
(around 5 ohms). I'm a programmer for a living and this is my first power
electronics project (previous electronics projects have mostly been
microcontroller based stuff)

I've done quite a bit of reading, but am running into a brick wall scaling
up some of the various designs that I've found online. I'm working with
LTC's SWCad3 to test concepts, but I'm having problems with input currents
spiking up way too high since my target output is so close to my current
source (AC 220v 60A single phase rectified to 320vdc), and the 120hz line is
just too slow to keep the input capacitor bank charged under heavy load
without some serious high current spikes to recharge.

Any suggestions to smooth out the current draw from the line would be MUCH
appreciated !!

 

[Rene Tschaggelar]

I'm looking to build a power supply for a krypton arc lamp, and need a
design that can deliver an adjustable 20-60 amps into a LOW resistance load
(around 5 ohms). I'm a programmer for a living and this is my first power
electronics project (previous electronics projects have mostly been
microcontroller based stuff)

I've done quite a bit of reading, but am running into a brick wall scaling
up some of the various designs that I've found online. I'm working with
LTC's SWCad3 to test concepts, but I'm having problems with input currents
spiking up way too high since my target output is so close to my current
source (AC 220v 60A single phase rectified to 320vdc), and the 120hz line is
just too slow to keep the input capacitor bank charged under heavy load
without some serious high current spikes to recharge.

Any suggestions to smooth out the current draw from the line would be MUCH
appreciated !!

John,
such a power supply is rather tricky. The power
level is rather high, and the nonlinear negative
dynamic impedance load doen't make it simpler.
You'll have to use a switching approach.
Not recommended.

Rene

 

[[email protected]]

John,
such a power supply is rather tricky. The power
level is rather high, and the nonlinear negative
dynamic impedance load doen't make it simpler.
You'll have to use a switching approach.
Not recommended.

Rene

Must agree here.
At this power level , I'd recommend you get a professional power
supply designer
to tackle this.
DNA , you around :0)
If you still decide to tackle this yourself , I recommend some eye
protetion of some sort
and prepair yourself for lots of bangs and smoke :0(
Rob

 

[[email protected]]

and the 120hz line is
just too slow to keep the input capacitor bank charged under heavy load
without some serious high current spikes to recharge.

The old-fashioned
way:

----------- ~ + ------ L ------------ +
|
bridge ===
---
|
----------- ~ - --------------------- -

'L' smoothes the charge peaks, needs an air gap, and will
basically be a very heavy big piece of iron with copper.

The output of this circuit is somewhat lower then 220V DC,
so more current.
See:
http://tpub.com/neets/book7/27f.htm

 

[John Barrett]

Ohh I'm quite prepared for a bit of mess Even got a few spare arc tubes
just in case And a welding helmet for eye protection

As stated... I'm planning a current mode buck regulator, though it has been
suggested that I'll need a boost circuit in front of it to smooth out the
line input current surges (to push up the voltage so that charge is
transfered into the caps more effciently ??) . I've already got a spice
model created that is close (need to work on the control loops a little
more, and a better soft start circuit., and beat these 120+ amp surges that
I'm getting)

"nonlinear negative dynamic impedance load" -- I'm not an EE so I could use
a bit of explanation I'll probably need to understand that as I'd like to
create a spice model of the arc tube for the later stages of testing.
(resistance decreases as current increases ?? am I even close ??)

(cummon guys !!! so far everyone who has responded, here and in alt.lasers
[because this is for a KW output ND:YAG laser] has had nothing to hand out
but warnings !! I'm committed, it will get built !! )

 

[bernado]

IPM - International Perforation Management
high-tech engineering China-Germany-Thailand
fax : 0049 - (0) 1212-5-375-17-531
http://www.microperforation.com
http://www.microperforation.com.cn
http://www.deguodaguan.com/ipm/
http://www.dk3qv.de
Email : [email protected] - [email protected]

We are using AC/DC converters and special DC constanter with 300 V/DC
30 up to 60 Amps for nano or micro perforation machines since more
then 20 years.

There are several companies very well equiped - have a look to
http://www.kfactor.it or http://www.irem.it

Such AC/DC STABLIZED power supplies are to purchase by around 4000 -
6000 bucks by one year guarantee so that it makes not sense to invest
several 10,000 bucks in developments for NEW primary AC switching and
secondary DC regulated units.

Let me know if you need more or specific details - I'm glad to help
and support you.

 

[John Barrett]

IPM - International Perforation Management
high-tech engineering China-Germany-Thailand
fax : 0049 - (0) 1212-5-375-17-531
http://www.microperforation.com
http://www.microperforation.com.cn
http://www.deguodaguan.com/ipm/
http://www.dk3qv.de
Email : [email protected] - [email protected]

We are using AC/DC converters and special DC constanter with 300 V/DC
30 up to 60 Amps for nano or micro perforation machines since more
then 20 years.

There are several companies very well equiped - have a look to
http://www.kfactor.it or http://www.irem.it

Such AC/DC STABLIZED power supplies are to purchase by around 4000 -
6000 bucks by one year guarantee so that it makes not sense to invest
several 10,000 bucks in developments for NEW primary AC switching and
secondary DC regulated units.

Let me know if you need more or specific details - I'm glad to help
and support you.

The only detail that could possibly help me at those kinds of prices is "do
you give free samples ??"

I makes very little sense to invest that kind of money when I've got less
than $500 into this laser so far, and I dont think its going to cost me more
than about $1000 more to bring it online. I'm not being funded by a
corporation with deep pockets In any case, for that kind of money, I
could afford to pump this rod with 808nm laser diode bars and skip the
massive current requirements that the krypton arcs are forcing me to deal
with. ($6,000 times the 3 supplies I need pays for a LOT of laser diode bars
!!)

And I seriously doubt its going to cost me 10K or better to to design my
own -- even if I count my time invested -- I'm pretty close to something
workable now with less than a week invested, and between the folks here and
what I'm finding online, working out the rest of the issues shouldnt take
all that long. And besides -- I get to learn something new in the process !!
<BigGrin>

[Flame On]
So save the commercial hype for someone that doesnt want to be bothered with
the nitty gritty details !!
[Flame Off]

 

[[email protected]]

dude, your insane trying to build a 6Kw. lamp supply from scratch,
those lamps blow if ya sneeze.
I know I just blew one trying to build a 2 kW version. I have one
left.

contact bruce @ contact @ second source lasers do com, he has two
eski lamp supplies in stock. BTW the series 6 pins and two connector
bodies you asked for shipped yesterday.

Steve Roberts

 

[John Larkin]

I'm looking to build a power supply for a krypton arc lamp, and need a
design that can deliver an adjustable 20-60 amps into a LOW resistance load
(around 5 ohms). I'm a programmer for a living and this is my first power
electronics project (previous electronics projects have mostly been
microcontroller based stuff)

I've done quite a bit of reading, but am running into a brick wall scaling
up some of the various designs that I've found online. I'm working with
LTC's SWCad3 to test concepts, but I'm having problems with input currents
spiking up way too high since my target output is so close to my current
source (AC 220v 60A single phase rectified to 320vdc), and the 120hz line is
just too slow to keep the input capacitor bank charged under heavy load
without some serious high current spikes to recharge.

Any suggestions to smooth out the current draw from the line would be MUCH
appreciated !!

Three phase AC!

But what you're doing is like deciding it would be cool to climb a
mountain for the first time, and buying a ticket to Nepal.

John

 

[D from BC]

Three phase AC!

But what you're doing is like deciding it would be cool to climb a
mountain for the first time, and buying a ticket to Nepal.

John

It's a mountain all right..
I'm very new in smps..
From near scratch, my little 130Watt (90% eff) converter project has
gone on for > 2 months.
I had to read on:
mosfet switching losses
trace inductances
fast power diodes
feedback control theory
control topologies
smps simulation
paralleling power devices
reducing EMI
creepage
ground planes
capacitor losses
magnetics design
line filtering and ov protection
snubbers

But, I got it working.
Just have to fix some ringing....
Score:
smps: 1 quality of life: 0

D from BC

 

[Rene Tschaggelar]

Ohh I'm quite prepared for a bit of mess Even got a few spare arc tubes
just in case And a welding helmet for eye protection

As stated... I'm planning a current mode buck regulator, though it has been
suggested that I'll need a boost circuit in front of it to smooth out the
line input current surges (to push up the voltage so that charge is
transfered into the caps more effciently ??) . I've already got a spice
model created that is close (need to work on the control loops a little
more, and a better soft start circuit., and beat these 120+ amp surges that
I'm getting)

"nonlinear negative dynamic impedance load" -- I'm not an EE so I could use
a bit of explanation I'll probably need to understand that as I'd like to
create a spice model of the arc tube for the later stages of testing.
(resistance decreases as current increases ?? am I even close ??)

Exactly.


Rene

 

[Winfield Hill]

John Larkin
Yep.

It's a mountain all right..
I'm very new in smps..
From near scratch, my little 130Watt (90% eff) converter project
has gone on for > 2 months.
I had to read on:
mosfet switching losses
trace inductances
fast power diodes
feedback control theory
control topologies
smps simulation
paralleling power devices
reducing EMI
creepage
ground planes
capacitor losses
magnetics design
line filtering and ov protection
snubbers

But, I got it working.
Just have to fix some ringing....

That's a mere foothill, and a small one at that.
But certainly keep on climbing.

 

[John Barrett]

dude, your insane trying to build a 6Kw. lamp supply from scratch,
those lamps blow if ya sneeze.
I know I just blew one trying to build a 2 kW version. I have one
left.

contact bruce @ contact @ second source lasers do com, he has two
eski lamp supplies in stock. BTW the series 6 pins and two connector
bodies you asked for shipped yesterday.

Steve Roberts

Thanks for getting those out, Steve !!

and OF COURSE I'm insane.. even more so than you think because I need 8KW !!
(7mm ID bore needing 195-205 volts !!)

However -- the design is progressing -- I've got a model up and running that
(without current regulation) stabilizes at a very smooth 48 amps into the
tube (presuming 5 ohms tube resistance) after startup... its got a bit more
ripple voltage than I would like, and its spiking the AC input hard, but at
least I got the soft start and main current path worked out

having a real problem with the current sense -- trying to use the switch
model to trigger when the ISense voltage passes a preset threshold, but I
must have it modeled wrong some how because it is not switching when the
tube current spikes. (I know -- its a kludge, but I'm too lazy to add in the
full pwm driver until I got everything else close to working as intended)
I'll prolly see what I can do about putting at least a comparator in and see
if that helps

 

[John Popelish]

Thanks for getting those out, Steve !!

and OF COURSE I'm insane.. even more so than you think because I need 8KW !!
(7mm ID bore needing 195-205 volts !!)

However -- the design is progressing -- I've got a model up and running that
(without current regulation) stabilizes at a very smooth 48 amps into the
tube (presuming 5 ohms tube resistance) after startup... its got a bit more
ripple voltage than I would like, and its spiking the AC input hard, but at
least I got the soft start and main current path worked out

At this high output current, I would probably thinking about
a 2 phase parallel current mode output stage (or pair of
stages. This is a fairly simple way to divide the current
load equally between two parallel paths, while having
something like 80% or 90% of the current ripple cancel. Toy
can work with one of the halves, till you pretty much get it
working at half current and then build a second one. The
trick is to come up with the dual complementary triangle
wave to drive the error amplifiers, so that the output
pulses are always almost precisely 180 degrees out of phase.
I have made fairly high current, low ripple current mode
laser supplies this way, though the voltage was much lower,
and the load better behaved.

having a real problem with the current sense -- trying to use the switch
model to trigger when the ISense voltage passes a preset threshold, but I
must have it modeled wrong some how because it is not switching when the
tube current spikes.

I think you may need a more continuous current sense, rather
than a peak trip.

 

[John Barrett]

At this high output current, I would probably thinking about a 2 phase
parallel current mode output stage (or pair of stages. This is a fairly
simple way to divide the current load equally between two parallel paths,
while having something like 80% or 90% of the current ripple cancel. Toy
can work with one of the halves, till you pretty much get it working at
half current and then build a second one. The trick is to come up with
the dual complementary triangle wave to drive the error amplifiers, so
that the output pulses are always almost precisely 180 degrees out of
phase. I have made fairly high current, low ripple current mode laser
supplies this way, though the voltage was much lower, and the load better
behaved.


I think you may need a more continuous current sense, rather than a peak
trip.

I know I need continuous feedback -- but I'm still trying to get a handle on
the main current path before I add complexity -- just wanted something
simple in there to provide SOME regulation so I'd have an idea if I was
getting close.

I have made a deciscion over the last 24 hours -- I'm going with a 2 phase
boost front end for active PFC and an internal 400v DC bus. One phase will
be a non-inverting boost for the positive halves of the 2 line inputs, the
other will be inverting for the negative halves, eliminating all the
problems that I'm having because my negative input to the buck is a half
wave pulse instead of a solid ground.

I'm gonna spend today modeling the non-inverting boost supply.. see if I can
get that working -- nice thing is -- it doesnt have to be strongly
regulated... -- the not-so-nice thing is I'm going to have to do the full
PWM setup because crude regulation methods wont work to drive the boost
switch.

hehehe at this point I would kill for a PWM controller spice model generic
enough to be used for this Gonna have to search around and see whats out
there !!

 

[Genome]

hehehe at this point I would kill for a PWM controller spice model generic
enough to be used for this Gonna have to search around and see whats
out there !!

John was talking about interleaving the buck stage but interleaving the
boost stage wouldn't hurt either.

Here you go....

http://www.genomerics.org/stuff/pfcfullb.asc

It's supposed to model a UCC3817....

http://focus.ti.com/docs/prod/folders/print/ucc3817.html

Actually it's not really that super brilliant but you'll get the idea.

DNA

 

[John Barrett]

John was talking about interleaving the buck stage but interleaving the
boost stage wouldn't hurt either.

Here you go....

http://www.genomerics.org/stuff/pfcfullb.asc

It's supposed to model a UCC3817....

http://focus.ti.com/docs/prod/folders/print/ucc3817.html

Actually it's not really that super brilliant but you'll get the idea.

DNA

Thank YOU !! Very much appreciated !!

I was getting a little put out with half a dozen people telling me
CANT/DONT/BUY !!!

Thanks again for what I came here for -- a little education on SMPS design

 

[Genome]

Thank YOU !! Very much appreciated !!

I was getting a little put out with half a dozen people telling me
CANT/DONT/BUY !!!

Thanks again for what I came here for -- a little education on SMPS design

Oh..... OK, fiddle fiddle fiddle. Try one of these,

http://www.genomerics.org/laser/pfcpar.asc

That's two stages operated in parallel with the clock waveforms shifted. The
idea is sort of that the current error amplifiers receive the same demand
signal from a single or master voltage error amplifier and program the same
current in their respective power stages. You have to figure out how to tie
the available parts in the ICs together to do the job.

One thing you are forced to do is use current transformers to recreate the
inductor current. At high powers this would be done anyway but at low powers
you would use a sense resistor in the return lead to the input rectifier.
Unfotunately when you try to operate things in parallel you can't do that
any more.... they just won't connect together proper like.

I've added some source sense resistors to the 'mosfets' but not implemented
a current limit with them. Problem is that the drain current transformer is
operated at high duty cycles and might saturate. You lose the signal and
things go bang. Resistors don't, lose signals, so it's a kind of 'oh shit!'
protection.

The peak limit is simplified a bit. It really needs slope compensation but
the IC does not make that easy to add, that's why those transistors were in
the previous one. Here I've just thrown in some appropriate current sources.

I've diddled it to do 1KW per section from a 220V line..... I suppose you
will need 8 of them because you are some sort of cross dressing freak. There
are ways of synchronising these ICs by shoving a pulse up their oscillator.
I'm sure if you read some data sheets you'll get some ideas.

In the old days a 4017 Johnson Counter clocked at N (stages) times the
desired base frequency with the reset pin on the appropriate tap did the
job. These day I suppose you would use a PIC.

Really all you are getting is something to practice your sums on. You will
probably find some of the values I've left in there are sub-optimal. You can
try putting in 'real' components to get some guesses about power losses. I'd
recommend you grab hold of some SiC diodes for the main boost diodes.

Have Fun and try not to go too fruit pie.

DNA

 

[John Barrett]

Oh..... OK, fiddle fiddle fiddle. Try one of these,

http://www.genomerics.org/laser/pfcpar.asc

That's two stages operated in parallel with the clock waveforms shifted.
The idea is sort of that the current error amplifiers receive the same
demand signal from a single or master voltage error amplifier and program
the same current in their respective power stages. You have to figure out
how to tie the available parts in the ICs together to do the job.

One thing you are forced to do is use current transformers to recreate the
inductor current. At high powers this would be done anyway but at low
powers you would use a sense resistor in the return lead to the input
rectifier. Unfotunately when you try to operate things in parallel you
can't do that any more.... they just won't connect together proper like.

I've added some source sense resistors to the 'mosfets' but not
implemented a current limit with them. Problem is that the drain current
transformer is operated at high duty cycles and might saturate. You lose
the signal and things go bang. Resistors don't, lose signals, so it's a
kind of 'oh shit!' protection.

The peak limit is simplified a bit. It really needs slope compensation but
the IC does not make that easy to add, that's why those transistors were
in the previous one. Here I've just thrown in some appropriate current
sources.

I've diddled it to do 1KW per section from a 220V line..... I suppose you
will need 8 of them because you are some sort of cross dressing freak.
There are ways of synchronising these ICs by shoving a pulse up their
oscillator. I'm sure if you read some data sheets you'll get some ideas.

In the old days a 4017 Johnson Counter clocked at N (stages) times the
desired base frequency with the reset pin on the appropriate tap did the
job. These day I suppose you would use a PIC.

Really all you are getting is something to practice your sums on. You will
probably find some of the values I've left in there are sub-optimal. You
can try putting in 'real' components to get some guesses about power
losses. I'd recommend you grab hold of some SiC diodes for the main boost
diodes.

Have Fun and try not to go too fruit pie.

DNA

I've already located inexpensive isolated current sensors with models for 50
or 100 amps and 3000v isolation, so I can easily do high side current
sensing -- just gotta do a spice model for it, which looking at your models
has helped me figure out (a voltage source defined via a formula referencing
the load current)

I'm seriously looking at the PFC chip you mentioned, and some others (I've
already tried to model out one using the LT1248 but its not working right
yet)

You want fruit pie ?? here is fruit pie -- what would it take if I wanted a
true earth ground on the output of this mess -- I've considered running
seperate boosts on the poitive and negative outputs from the bridge, using
the isolated current sensor to get EVERYTHING out of the ground path.. one
of them inverting to get a positive voltage (add more in pairs if 2 wont get
me enough current at reasonable ripple)

is that fruit pie enough for you ?? or is there a better way ??

 

[Genome]

I've already located inexpensive isolated current sensors with models for
50 or 100 amps and 3000v isolation, so I can easily do high side current
sensing -- just gotta do a spice model for it, which looking at your
models has helped me figure out (a voltage source defined via a formula
referencing the load current)

I'm seriously looking at the PFC chip you mentioned, and some others (I've
already tried to model out one using the LT1248 but its not working right
yet)

You want fruit pie ?? here is fruit pie -- what would it take if I wanted
a true earth ground on the output of this mess -- I've considered running
seperate boosts on the poitive and negative outputs from the bridge, using
the isolated current sensor to get EVERYTHING out of the ground path.. one
of them inverting to get a positive voltage (add more in pairs if 2 wont
get me enough current at reasonable ripple)

is that fruit pie enough for you ?? or is there a better way ??

Ermmmmm...... Current sense transformers, something like

http://www.cd4power.com/data/magnetics/kmp_5600.pdf

Each PFC section needs to control its own inductor current otherwise they
won't share the total load. It's tied up with the volt-second balance,
inductors integrate the voltage placed accross them as the current through
them. If there is a mismatch in the drive to the inductors then one of them
will hog the current.

The local current error amplifiers force them to share it but you need
separate sense signals for each stage so one honking big current sensing
thing on the input won't do. You also need a reasonable frequency response
from the devices if you want to compensate the current loops correctly.


Ahaaaa, LT1248, shows you how crap I am for not checking if they did one or
if LTSpice had the model. I knew they did but there is something wrong with
my head. I suppose I'm fixated on TI because I grew up with Unitrode. It's
all much of a muchness and once you get happy with one you can see that a
lot of the internals are the same in all the others.


If you really really need to ground the output of your beast....... Uhm,
right, your side of the pond has strange electricity with something like
110V balanced either side of ground. My head hurts but I think you might
only get away with it if you use half wave rectification which is going to
be dirty and waste half of the capacity of your supply.

Since you seem to enjoy pain you might as well go the whole banana and
dangle a transformer isolated coverter of the end of your PFC stages.
Something like a half or full bridge. That will make any (well most)
concerns about earth disappear.

DNA