SEPIC SMPS design help

[Brooke Clarke]

Hi:

I need a SMPS design and I think a SEPIC topology would be best.
Input Voltage 14.4 to 24 V supplied from 16 each "C" cell Alkaline batteries
Output Voltage 15.0 V @ up to 3.5A (5 A desirable)
Custom coil/transformer is OK, but there is a height and area limits for
the board.
Any thoughts?

Brooke Clarke, N6GCE
[email protected]
http://www.PRC68.com

 

[dan williams]

sounds like something a NS simple switcher would fit. I'm on a "smps
bonanza" latley, and would be verry interested to know what you come up
with and how well it works.

I THINK I have Buck supplies worked out, I can make you up a quick
"design" if you would like, I cant tell you anything about how well it
may or may not work.

ask if you interested, because all I have to do is put the numbers into
my program. ( I would now, but I dont have a copy here)

dan

I think a buck converter would be more efficient (especially is
isolation is not needed), if you would accept about 15.5 volts as a
minimum battery voltage instead of 14.4. I think you will gain more
battery life from the higher efficiency than you get from pulling the
batteries down that last volt( which at 5 amps load will take a few
seconds). It will also probably end up smaller.

--
Dan Williams, Owner
Electronic Device Services
(604) 741 8431
RR8 855 Oshea rd
Gibsons BC Canada
V0N 1V8

 

[Brooke Clarke]

Hi John:

That idea just occured to me also and I'm trying to look at the buck
design, but National has changed their Webench to a new format that I
think has serious bugs.

Brooke Clarke, N6GCE

 

[Harry Dellamano]

Hi:

I need a SMPS design and I think a SEPIC topology would be best.
Input Voltage 14.4 to 24 V supplied from 16 each "C" cell Alkaline batteries
Output Voltage 15.0 V @ up to 3.5A (5 A desirable)
Custom coil/transformer is OK, but there is a height and area limits for
the board.
Any thoughts?

Brooke Clarke, N6GCE
[email protected]
http://www.PRC68.com

I believe that a SEPIC topology would not be best. You will have a MOSFET
operating at >5 watts which is a pain to cool in a low profile environment.
A custom inductor and a coupling capacitor that will be very expensive.
Mosfets are cheap and operated at <1.0 watts easy to cool. Use a full bridge
topology. Disconnect the top drain connection, tie them each to storage
capacitors. One is the input voltage the other is the output. Place a single
inductor winding across the bridge. Pump energy either way across the
bridge. Drive bridge in normal fashion. No diodes, high efficiency, heat
spread for easy cooling, low cost, easy current mode operation, easy short
circuit protection, piece of cake.

spank me with your keyboards!
harry

 

[Fred Bloggs]

spank me with your keyboards!

Well- okay - as long as you asked for it. You have a certain book review
on Amazon, S-MPSSC, that must have been written while you were
inebriated - delete or rewrite it and use the spelling checker.

 

[Brooke Clarke]

Hi Harry:

Can yo point me to a web page with a schematic and some design info for
this topology?

Thanks very much,

Brooke Clarke, N6GCE

 

[Ken Smith]

Hi:

I need a SMPS design and I think a SEPIC topology would be best.
Input Voltage 14.4 to 24 V supplied from 16 each "C" cell Alkaline batteries
Output Voltage 15.0 V @ up to 3.5A (5 A desirable)
Custom coil/transformer is OK, but there is a height and area limits for
the board.

Since you didn't mension a cost limit: I suggest you look at the
inductors Pico sells. They have two sets of windings on them of equal
numbers of turns. They intended the windings to allow the same inductor
to have two inductances by connecting them in series or parallel.

I found that they work very nicely in a SEPIC design. The two windings
have more than 90% coupling so the current in the capacitor isn't as high
as one might expect for independant inductors.

You can look at www.linear.com they have some fairly good app-notes about
the SEPIC converters using their parts. I have found that their swcad-iii
does a good job of modeling the circuits.

If using Linear's models beware: Several of their models have errors with
respect to the power pins. The LT1498 model sources current into your
design and the LT1247 model references things back the the "0" node rather
than the chips ground pin.

At the low voltages you are needing, MOS-FETs with very low on resistances
are available. You should be able to make a switcher run up to about 1MHz
at 5A. The higher frequency allows smaller parts but there is a cost.
You are likely to have more loss in the switching than the steady state ON
condition. Also the core material in the inductors gets more important.
Fortunately for you, some new materials are just now getting commonly
available that should make the inductor easier to do.

 

[Harry Dellamano]

Hi Harry:

Can yo point me to a web page with a schematic and some design info for this topology?

Thanks very much,

Brooke Clarke, N6GCE

Hi Brooke,
No I can't. LT uses this topology in some of their power supply ICs, check it out. Like most newer designs, companies are using it, but it is proprietary. The real good designs are not mainstream. Do I know Vicor's topology in their third generation modules, not really.
I think that I explained enough so that you can see it's simplicity. I am not going to do the design for you but will help.
Hope I didn't upset the spelling wizard too much.

harry

 

[Brooke Clarke]

Hi Dan:

SEPIC = Single Ended Primary Inductance Converter. See the National
LM3478 data sheet. Uses two inductors and a series cap coupling them.
Vout is isolated from Vin by the series cap. Vout can be above or below
Vin. No custom transformer needed. Both inductors can be on one former.

I'll read up on the TL494.

Thanks,

Brooke Clarke, N6GCE

 

[[email protected]]

Since you didn't mension a cost limit: I suggest you look at the
inductors Pico sells. They have two sets of windings on them of equal
numbers of turns. They intended the windings to allow the same inductor
to have two inductances by connecting them in series or parallel.

I found that they work very nicely in a SEPIC design. The two windings
have more than 90% coupling so the current in the capacitor isn't as high
as one might expect for independant inductors.

That brings up a question I've had:
While I see the advantages for the discrete inductor SEPIC, is the
performance of the coupled-inductor SEPIC better than a simple flyback
(transformer isolated buck-boost)?

I've posted a schematic of a coupled-inductor SEPIC in

- In both cases, they use transformers with the same winding phasing
and same connections to the switches (transistor & diode).

- The use of a coupled transformer gives the opportunity to apply a
turns ratio to meet the input/output voltage range requirement.
The issue of the input voltage range crossing the output voltage is no
longer an issue.

- While I do measure some RMS current in the coupling capacitor, it
has DC applied across it for most of the switching cycle. It certainly
doesn't couple energy like the capacitor in the discrete inductor
version.
I think I'd rather let the transformer work a little harder (if at
all), as opposed to adding an aluminum electrolytic capacitor to the
circuit dissipating heat.

What are the advantages of the coupled-inductor SEPIC over a flyback?

You can look at www.linear.com they have some fairly good app-notes about
the SEPIC converters using their parts. I have found that their swcad-iii
does a good job of modeling the circuits.

If using Linear's models beware: Several of their models have errors with
respect to the power pins. The LT1498 model sources current into your
design and the LT1247 model references things back the the "0" node rather
than the chips ground pin.

At the low voltages you are needing, MOS-FETs with very low on resistances
are available. You should be able to make a switcher run up to about 1MHz
at 5A. The higher frequency allows smaller parts but there is a cost.
You are likely to have more loss in the switching than the steady state ON
condition. Also the core material in the inductors gets more important.
Fortunately for you, some new materials are just now getting commonly
available that should make the inductor easier to do.

-----

 

[Ken Smith]

Nice one Pat, you just made a simple proof of why L1 must equal L2 in a
SEPIC.

Only if they are tightly coupled, do L1 and L2 need to be the same. In
the isolated inductor case, the two inductors can be very different.

There is a slight advantage to making L1 greater than L2.


[....]

wider range of output voltage. In a perfect world and a flyback turns ratio
of one they are both the same.

what perfect world?

 

[Harry Dellamano]

Nice one Pat, you just made a simple proof of why L1 must equal L2 in a
SEPIC.

Only if they are tightly coupled, do L1 and L2 need to be the same. In
the isolated inductor case, the two inductors can be very different.

There is a slight advantage to making L1 greater than L2.


[....]

wider range of output voltage. In a perfect world and a flyback turns ratio
of one they are both the same.

what perfect world?

Good Points Ken, but if unequal inductors are coupled how much leakage is
needed?
I would only use equal coupled inductors in a SEPIC and only in supplies
of less than 25 watts. Higher power causes too much power dissipation on the
single FET, similar to a Flyback. The OP has a 45 Watt output so I would
look at other topologies.
A perfect world is my SPICE simulator.

regards
harry

 

[dan]

Ah, well then, my answer was a little off topic, no lolipop for me :-(
If someone would check my design and tell me if its inline, I'd be greatfull.

here's what looks like a good link for designing one...

http://www.maxim-ic.com/appnotes.cfm/appnote_number/1051

(dont reply via this email address, I never check it)
dan

 

[[email protected]]

Nice one Pat, you just made a simple proof of why L1 must equal L2 in a
SEPIC.

And the proof is ... ?

Also, I'm assuming you're still talking about a coupled-inductor
SEPIC.

 

[Harry Dellamano]

And the proof is ... ?

You have two windings on same core with both doted leads tied to a fixed
voltage (which is an AC short) and non-doted leads connected thru a
capacitor (another AC short). So AC wise the windings are shorted together.
Therefore the turns ratio must equal 1.00.

Also, I'm assuming you're still talking about a coupled-inductor
SEPIC.

I was speaking of SEPIC in general but Ken Smith says that there is a slight
advantage in non coupled inductors to make L1 >L2. We need him to explain.

regards
harry

 

[Ken Smith]

Good Points Ken, but if unequal inductors are coupled how much leakage is
needed?

Think of leakage as an inductor L(leak) in series with L1.

Think of L1 and L2 as perfectly coupled with equal turns in them.

In this perfect world, L(leak) has no AC voltage on it and hence there is
no ripple current in it.

Make L(leak) as big as needed for this imperfect world to still get close
to zero ripple current.

Now convert this into a real world set of windings and you will find that
the L1 ends up with more turns than L2 and hence more inductance.

The result is a DC-DC converter with fairly low input ripple current.

I would only use equal coupled inductors in a SEPIC and only in supplies
of less than 25 watts. Higher power causes too much power dissipation on the
single FET, similar to a Flyback. The OP has a 45 Watt output so I would
look at other topologies.

The biggest real problem with the SEPIC converter is the pounds of core
material in the inductors not the pounds of power FETs. The SEPIC and
flyback converters need bigger fet but there is only one so it sort of
"comes out in the wash".

The biggest flybacker I've done was 160W. Less than 10% of the power get
lost in the power devices.

 

[dan williams]

So what did you end up with?

dan

Hi:

I need a SMPS design and I think a SEPIC topology would be best.
Input Voltage 14.4 to 24 V supplied from 16 each "C" cell Alkaline batteries
Output Voltage 15.0 V @ up to 3.5A (5 A desirable)
Custom coil/transformer is OK, but there is a height and area limits for
the board.
Any thoughts?

Brooke Clarke, N6GCE
[email protected]
http://www.PRC68.com

--
Dan Williams, Owner
Electronic Device Services
(604) 741 8431
RR8 855 Oshea rd
Gibsons BC Canada
V0N 1V8