11 to 30 volts --> 24 volts

[Jim Weir]

I'd like a device that will let me input anywhere between 11 and 30 volts and
have a nice clean 24 volts (or 20 volts, or 22 volts...) at one amp as the
output.

I asked this question before and was given a pointer to a National Semi device.
What I found with the National devices is that they are as squirrely as Kansas
in springtime. One of the applications engineers actually told me that unless I
followed the layout pattern PRECISELY that the device would be unstable. Hell,
I've been doing groundplane construction up to 2.3 Gigs for thirty years and
haven't had that sort of problem.

Point being...anybody got a boost-buck device that they have had particular good
luck with?

Jim

 

[Richard Henry]

I'd like a device that will let me input anywhere between 11 and 30 volts and
have a nice clean 24 volts (or 20 volts, or 22 volts...) at one amp as the
output.

I asked this question before and was given a pointer to a National Semi device.
What I found with the National devices is that they are as squirrely as Kansas
in springtime. One of the applications engineers actually told me that unless I
followed the layout pattern PRECISELY that the device would be unstable. Hell,
I've been doing groundplane construction up to 2.3 Gigs for thirty years and
haven't had that sort of problem.

Point being...anybody got a boost-buck device that they have had particular good
luck with?

http://para.maxim-ic.com/ss.asp?Fam=DCDC_All&Tree=PowerSupplies&HP=PowerSupplies.cfm&ln=

 

[Jim Weir]

And which one did you have particularly good luck with? I couldn't find one
that would do what I specified, but if you can point me, it would be most
appreciated.


"Richard Henry" <[email protected]>
shared these priceless pearls of wisdom:


->>
->> Point being...anybody got a boost-buck device that they have had
->particular good
->> luck with?
->
->http://para.maxim-ic.com/ss.asp?Fam=DCDC_All&Tree=PowerSupplies&HP=PowerSupplies.cfm&ln=
->

 

[Jim Weir]

As I said, I use my breadboarding technique up beyond 2 Gigs and things work
just fine. At these frequencies I treat fractions of picofarads and nanohenries
as problems.

The basic technique is to use a solid copper substrate and solder all the
grounded parts directly to the substrate with zero length leads. The active and
ungrounded components solder right at the body of the component so that there is
usually less than 3 or 4 mm of stray inductance. What with #20 wire having an
inductance of about a nanohenry per millimeter, I didn't think that was all that
bad.

The apps engineer almost implied that there was some "magic" going on with the
layout as they did it, and that either a better or worse layout might honk up
the works. I am not a happy camper with that sort of baloney.

Jim



"Terry Given" <[email protected]>
shared these priceless pearls of wisdom:


->
->Jim,
->If you treat your smps layout with the same care and attention - explicitly
->think about parasitics (especially L, but C can couple things places you
->dont want too), and keep them as low as practicable - then pretty much any
->smps chip will run fine.

 

[Tony Williams]

I'd like a device that will let me input anywhere between 11 and
30 volts and have a nice clean 24 volts (or 20 volts, or 22
volts...) at one amp as the output.

Jim,

I downloaded SwitcherCAD from Linear Technology
last week, <www.linear-tech.com>. It has a
facility where you plug in your dc-dc specs and
it offers a design suggestion (based of course
on an LT chip, if one is available). I gave it
your numbers above and it offered two circuits,
based on the LT1270 and 1270A. The simulation
allows probing the V and I of every component.
Might be a useful starting point.

 

[Tim Shoppa]

I asked this question before and was given a pointer to a National Semi device.
What I found with the National devices is that they are as squirrely as Kansas
in springtime. One of the applications engineers actually told me that unless I
followed the layout pattern PRECISELY that the device would be unstable. Hell,
I've been doing groundplane construction up to 2.3 Gigs for thirty years and
haven't had that sort of problem.

I don't think that the layout comment is targetted so much at you as it
is towards a neophyte who puts the circuit together on a perfboard with
ground via 30AWG wire-wrap connected willy-nilly.

I've had very good luck with National Simple Switchers, indeed several
of them are put together on perfboard (none of them SEPIC, admittedly).

Tim.

 

[Genome]

|
| I'd like a device that will let me input anywhere between 11 and 30
volts and
| have a nice clean 24 volts (or 20 volts, or 22 volts...) at one amp as
the
| output.
|
| I asked this question before and was given a pointer to a National
Semi device.
| What I found with the National devices is that they are as squirrely
as Kansas
| in springtime. One of the applications engineers actually told me
that unless I
| followed the layout pattern PRECISELY that the device would be
unstable. Hell,
| I've been doing groundplane construction up to 2.3 Gigs for thirty
years and
| haven't had that sort of problem.
|
| Point being...anybody got a boost-buck device that they have had
particular good
| luck with?
|
| Jim
|

I'd be inclined to say 'stuff it' and immediately go for a single switch
'isolated' forward converter. I've just had a play with designing a
sepic converter and I wouldn't put it up my bum.

DNA

 

[Tony Williams]

I'd be inclined to say 'stuff it' and immediately go for a single
switch 'isolated' forward converter. I've just had a play with
designing a sepic converter and I wouldn't put it up my bum.

DNA. I know you have LTSpice/SwitcherCAD and that
does offer a sepic converter (based on the LT1270A)
in response to Jim's spec. I'd be quite interested
in your informed comments/criticisms about LT's
offered circuit.

 

[Joerg]

Hi Jim,

Did you mean the LM3478? That's the one I am currently using in a design. So far
it's ok but I'd appreciate if you could share some experience or feedback from the
app engineers.

Pretty much all boost converters can be configured SEPIC. That architecture only
takes one more inductor and one more cap. Then it lets you go above and below the
input voltage. A flyback is another option but that requires a custom inductor with
two windings. The SEPIC can be done with two regular inductors.

Regards, Joerg

 

[Jan Panteltje]

Hi Jim,

Did you mean the LM3478? That's the one I am currently using in a design. So far
it's ok but I'd appreciate if you could share some experience or feedback from the
app engineers.

Pretty much all boost converters can be configured SEPIC. That architecture only
takes one more inductor and one more cap. Then it lets you go above and below the
input voltage. A flyback is another option but that requires a custom inductor with
two windings. The SEPIC can be done with two regular inductors.

Regards, Joerg

How about his circuit with LTC3440?
www.eetasia.com/ARTICLES/2002APR/ 2002APR08_ICD_PL_PD_MPR_TAC.PDF
Have not tried it, just curious what you think for this sort of application.
JP

 

[Joerg]

Hi Jan,

The LTC3440 is nice and it has the transistors in there. But at the end of the day much
boils down to cost. 100 qty lists above $3 which is kind of high. The LM3478 is about $1
and the extra transistors and inductor are just pennies. I mostly stick to what is low
cost.

Regards, Joerg

 

[Terry Given]

I'd like a device that will let me input anywhere between 11 and 30 volts and
have a nice clean 24 volts (or 20 volts, or 22 volts...) at one amp as the
output.

I asked this question before and was given a pointer to a National Semi device.
What I found with the National devices is that they are as squirrely as Kansas
in springtime. One of the applications engineers actually told me that unless I
followed the layout pattern PRECISELY that the device would be unstable. Hell,
I've been doing groundplane construction up to 2.3 Gigs for thirty years and
haven't had that sort of problem.

Point being...anybody got a boost-buck device that they have had particular good
luck with?

Jim

Jim,
If you treat your smps layout with the same care and attention - explicitly
think about parasitics (especially L, but C can couple things places you
dont want too), and keep them as low as practicable - then pretty much any
smps chip will run fine. Layout is usually the culprit when things dont go,
because most people dont understand it (easy: current flows in loops.
minimise them hence the comment. There are lots of other reasons too,
depending on the chip. spikes on the current sense line can cause short
pulses (the comparator & flip-flop act as a sample-and-hold), the list goes
on (wrong diodes, wrong caps, wrong FETs, etc)

Occams razor suggests the chip does actually work - nat semi makes mistakes
just like everybody else, but not many. Mind you, I once looked at IR 1200V
half-bridge IGBT gate drive ICs. We bought a demo board from IR, and it
promptly blew up. After replacing the drivers 3 times, we gave up and solved
our problem a different way (they were costly anyway). A couple of years
later, IR convinced me to look at them again. Same chip, properly laid out
demo pcb, and they worked fine - we looked at the original PCB (I still had
it) and the layout was appalling, stray inductance was maximised. no wonder
they blew up (we didnt look into the failures at the time, because we
thought IR could design their own pcbs properly, and we were busy). The rep
said that particular applications ****-up cost them quite a few customers.
Alas, we had developed some cute technology last time, and could make a
gatedrive that performed as well, for half the price. so they lost our
custom, permanently. the 200,000+ gatedrives per year we made (still make,
but no me in the we) represented a fair few dollars lost for IR.

Terry

 

[Joerg]

Hi Jim,

FYI: I just tested the converter part of a design, SEPIC with the LM3478. Works like
a champ, no problems. And that was not even on a nice layout but on breadboard. It
goes from zero to wherever I want it to go.

Regards, Joerg

 

[Terry Given]

As I said, I use my breadboarding technique up beyond 2 Gigs and things work
just fine. At these frequencies I treat fractions of picofarads and nanohenries
as problems.

The basic technique is to use a solid copper substrate and solder all the
grounded parts directly to the substrate with zero length leads. The active and
ungrounded components solder right at the body of the component so that there is
usually less than 3 or 4 mm of stray inductance. What with #20 wire having an
inductance of about a nanohenry per millimeter, I didn't think that was all that
bad.

The apps engineer almost implied that there was some "magic" going on with the
layout as they did it, and that either a better or worse layout might honk up
the works. I am not a happy camper with that sort of baloney.

Jim

Right. Clearly you more than know what your doing with layout, so that aint
your problem. And like hell theres magic in their layout - unless they have
a sneaky min. inductance requirement, or something odd like that. Care to
post any more data on the circuit that didnt go? we might learn something.

Cheers
Terry

 

[Genome]

| In article <2pJsc.551$_%[email protected]>,
|
| > I'd be inclined to say 'stuff it' and immediately go for a single
| > switch 'isolated' forward converter. I've just had a play with
| > designing a sepic converter and I wouldn't put it up my bum.
|
| DNA. I know you have LTSpice/SwitcherCAD and that
| does offer a sepic converter (based on the LT1270A)
| in response to Jim's spec. I'd be quite interested
| in your informed comments/criticisms about LT's
| offered circuit.
|
| --
| Tony Williams.

That's a hard question, not sure I can do it justice.

The software, naturally, adjusts the feedback resistor ratio to give the
output voltage required. It also adjusts the inductance so there is more
decicion making going on. I don't now what criterea it may use but a
guess might be something like 20% ripple current in the inductors.

One thing that doesn't seem to change is the compensation components
which appear to be fixed at 1K and 1U which, sort of, implies a slow
loop. I'd have to analyse it further to comment and that sort of thing
bends my head until I come up with a method.

I have downloaded AN-19 for a cursory poke around.

The system is going to be second order up to the zero, 1K/1U, frequency
and there is also a right half plane zero knocking about in there.
Overall it doesn't look like the software optimises the loop. LT's
method is a bit 'suck it and see' so I think you are still left with
that task yourself.

The standard circuit it comes up with is a bit...... 'Fixed' and it's
useful to break it out of the given design mode and copy then paste the
circuit into an ordinary diagram window. Then you get to have a proper
play. I'll mention that the inductor coupling coefficient of 1 is
unrealistic and may be undesireable if you want to achieve ripple
current steering.

The circuit has start up and stability problems and is improved by
adding a 100nF capacitor across the feedback resistor. This renders the
circuit first order overall and the start up and transient seem better.
That's waving a bit of a wet finger in the air. Like I say I haven't
tried to analyse things properly.

Overall I'd say the circuit provided is a starting point but requires
further optimisation to get the final result. I wouldn't expect software
to be able to do that too well anyway. You are still forced to get your
head into what the circuit is doing and make appropriate decisions
yourself.

As I've said LT's application notes aren't to heavy on analysis and
compensation so... they could do with a bit more work in that area. That
sort of stuff is difficult and it's no better and maybe slightly worse
than what I've seen from other sources.

It's not a panacea but 11/10 for having a good bit of software to try
things out. I wouldn't be without it.

DNA

 

[Tony Williams]

| DNA. I know you have LTSpice/SwitcherCAD and that
| does offer a sepic converter (based on the LT1270A)
| in response to Jim's spec. I'd be quite interested
| in your informed comments/criticisms about LT's
| offered circuit.

That's a hard question, not sure I can do it justice.

Thank you for the comments. I'm sure you can even casually
look at such a circuit and more than do it justice.

[snip interesting details. Thank you]

That start-up wobble does hint at an underlying
instability that would need further investigation
before using the circuit. Confirmed by your comments.

Overall I'd say the circuit provided is a starting point but
requires further optimisation to get the final result. I wouldn't
expect software to be able to do that too well anyway. You are
still forced to get your head into what the circuit is doing and
make appropriate decisions yourself.

As I've said LT's application notes aren't to heavy on analysis
and compensation so... they could do with a bit more work in that
area. That sort of stuff is difficult and it's no better and
maybe slightly worse than what I've seen from other sources.

It's not a panacea but 11/10 for having a good bit of software to
try things out. I wouldn't be without it.

LTSpice is my new toy, and I'm still trying things out
to get confidence in the results it produces. All the
circuits tried so far have produced an impressively
accurate analysis (finger trouble excepted).

Thanks,

 

[Aubrey McIntosh]

I'd be inclined to say 'stuff it' and immediately go for a single switch
'isolated' forward converter. I've just had a play with designing a
sepic converter and I wouldn't put it up my bum.

In sepic, are the inductors magnetically coupled?

 

[Genome]

| >
| > I'd be inclined to say 'stuff it' and immediately go for a single
switch
| > 'isolated' forward converter. I've just had a play with designing a
| > sepic converter and I wouldn't put it up my bum.
| >
|
| In sepic, are the inductors magnetically coupled?

They don't have to be but they can be. The advantage is that you reduce
things to one magnetic component. You can also design the thing to place
leakage inductance at the input or output node, generally the input is
chosen and this reduces the ripple current at the chosen node.

When you do couple the inductors you cannot avoid having leakage
inductance and this resonates with the coupling capacitor and can cause
problems if it isn't damped in some way. The usual method is to add a
series RC in parallel with the original to achieve parallel damping.

The resonance itself can be an issue in particular if you try to
regulate input current directly. It is avoided by controlling switch
current which is, supposedly, ultimately equal to input current but the
switch does not lie within the resonant loop so you avoid the associated
phase changes.

One of unitrodes engineers has done a couple of application notes that
deal with the sepic converter. Worth reading but I can't remember the
link at this time, somewhere in the design seminars on TI's website.

Oh.... I got the sepic working having taken my own advice and stopped my
feeble attempts to think in terms of bode plots and just wacked the
linear model into LTspice..... handy bit of soft. I still can't claim to
like it but it does do the job. I ended up with voltage mode control
using a nominal FAN7556. I quite like that IC, pity it doesn't come with
a reference.

A current mode control loop would have broken my head but I suppose I
should give it a bash?


DNA