Maker Pro
Maker Pro

switching regulator with mcu

J

Joerg

Jan 1, 1970
0
John said:
[...]
Another option would be to use a CD40106 or something similar as a
Schmitt oscillator, with its VCC capped/zenered around 6-8V. This can
drive a little FET, a simple logic level device like a 2N7002 as long
as doesn't cost much. Pipe Vref out of your MCU (hoping it has that
...) and use a cheap opamp to pull the Schmitt oscillator input "to
the side" when the target voltage is reached. That reduces the duty
cycle as much as needed to maintain regulation, pretty much like the
throttle on a gasoline-powered generator. If the target voltage
doesn't have to be very precise you could also use a NPN plus zener
for that, without a reference source. Probably a TL431-type device
would work as well and those are quite cheap, in the penny range.

Hi Jeorg,

Are you talking about a SEPIC/flyback arrangement here? I.e., 2N7001
drives a transformer? How do you stop the opamp railing during startup
or a load step? The schmitt osc would stop and .. how does it go
... Phut! :)

If you use an opamp a simple trick to avoid this is a diode. That way
the enchilada can pull the Schmitt to one side but cannot push it to the
other. If you are really brazen use a comparator with O/C -> no diode,
saves 1-2 cents. If you use a transistor or TL431 the problem goes away
on its own, sans diode. Now, set the max duty cycle to whatever worst
case demand you calculated, at the minimum expected input voltage, plus
20% margin or so.

Also of course by the time he's actually bought a SEPIC transformer it
could all approach his 1.8 Euros. Or are there some super-cheap ones
now?

Yes, but only when you buy in Asia. For some reason you can't get them
much under a buck inside the US even though I could imagine those are
also made in China. Coilcraft is typically the best deal for coupled
SEPIC coils if it has to be domestic.

Using two individual inductors gets you around the problem.
 
J

Joerg

Jan 1, 1970
0
michael nikolaou wrote:

Best is to post underneath quoted text. Makes it easier to follow for
people new to this thread, and they may also bring in ideas.

I need to make clear so we avoid confusions

The device uses a 2.2" tft lcd . This thing has a backlight and no matter
what
draws 60 ma @ 3.3V .That i can't avoid . So with 60 ma Arm7 consumption
i end up to 120 ma.


Sounds pretty normal. But check whether there are power surges after
start-up or during certain phases in the code execution. I guess you
don't want large electrolytics so the switcher would need proper
reserves to deal with that.

Now as i see whether i have a small switcher with huge inductor and
capacitor
so my board space is too large .
The other solution is super fast switcher that has everything small but
needs 2 euros
at least .
I imagine every engineer designing devices has faced this question .


Yep. And every time we go out and just price stuff out. Very tedious,
but sometimes you end up finding a good deal on a 0.5-1MHz PWM chip.
Check the EU companies as well for that, and definitely Japan, Korea
etc. It's very similar to house hunting ;-)

I'm sure though that a simple cpu controlled switcher could cope with the
large dropout
from 24V to 4-5 volts and then a simple linear regulator could take you to
3.3V so you play
it safe.If things go wrong (cpu malfunction ) thermal shutdown from the
linear regulator would
reset the cpu and that would restart the system
As i see it a simple fet switch with cpu pwm control should satisfy that .
Any opinions or experience on that subject???

Personally I never use LDOs because many of those chips are IMHO not
engineered well enough. Non-LDO regulators would need 6VDC or more,
meaning way too much dissipation for a SC70 package. It will cost a lot
more engineering hours but I'd first scour the market for cheap+fast
switchers and if no luck I'd go straight for the CPU-only method.
Directly from 12-24VDC to 3.3VDC. But the MCU code must be firmly under
your control if you do that. If a rookie came in later and added some
code, blissfully unaware of your hysteretic loop for the 3.3V, ... phut
.... *BANG*

No joke, this happened to me. Long story short: A really brazen designer
had gone a lot farther, taking in rectified mains via a
capacitor/resistor combo, rectified by <gasp> a substrate diode.
"Regulation" was achieved by executing dummy code during idle phases so
the load current was always the same and VCC would not run up. Nobody
had told me about this so when I touched the X1 pin with the scope probe
the oscillator briefly stopped and ... *KABLAM*
 
J

John Devereux

Jan 1, 1970
0
Joerg said:
John said:
[...]
Another option would be to use a CD40106 or something similar as a
Schmitt oscillator, with its VCC capped/zenered around 6-8V. This can
drive a little FET, a simple logic level device like a 2N7002 as long
as doesn't cost much. Pipe Vref out of your MCU (hoping it has that
...) and use a cheap opamp to pull the Schmitt oscillator input "to
the side" when the target voltage is reached. That reduces the duty
cycle as much as needed to maintain regulation, pretty much like the
throttle on a gasoline-powered generator. If the target voltage
doesn't have to be very precise you could also use a NPN plus zener
for that, without a reference source. Probably a TL431-type device
would work as well and those are quite cheap, in the penny range.

Hi Jeorg,

Are you talking about a SEPIC/flyback arrangement here? I.e., 2N7001
drives a transformer? How do you stop the opamp railing during startup
or a load step? The schmitt osc would stop and .. how does it go
... Phut! :)

If you use an opamp a simple trick to avoid this is a diode. That way
the enchilada can pull the Schmitt to one side but cannot push it to
the other. If you are really brazen use a comparator with O/C -> no
diode, saves 1-2 cents. If you use a transistor or TL431 the problem
goes away on its own, sans diode.

Of course, neat. Finally the famed Jeorgian discrete SMPS is starting to
take shape :)
Now, set the max duty cycle to
whatever worst case demand you calculated, at the minimum expected
input voltage, plus 20% margin or so.

Yes, but only when you buy in Asia.

Anyone know how to do this for smaller quantities (<~1k, say?). I've got
plenty of Asian suppliers trying to get our PCB business, but nobody
trying to sell me inductors :(.
 
J

Joerg

Jan 1, 1970
0
John said:
Of course, neat. Finally the famed Jeorgian discrete SMPS is starting to
take shape :)

My first one took shape about 15 years ago, tons of them in use by
consumers right now, still coming off the conveyor belt (in China). I
don't know if it will outlast the VW Beetle in production but it might.
It has already survived the client company's president :-(

Back then I learned a valuable lesson: Inductor prices in the western
world must be taken with a grain of salt. A custom mfg place in Taiwan
outbid a catalog (!) part from a western manufacturer, big time. And we
even got an EMI-savvy toroid for our money versus the cheap open cores
from mainstream distributors. Of course it does help not to have a
middleman in the game.

Anyone know how to do this for smaller quantities (<~1k, say?). I've got
plenty of Asian suppliers trying to get our PCB business, but nobody
trying to sell me inductors :(.

<1k? No, I don't think they'd even talk to you. However, Asia is all
about one thing: Connections. Ask the PCB stuffing places what sort of
deals they can get locally. They might be able to swing it, with a
supplier they already buy lots of other stuff from.

If you have reputable places there that do <1k qty board runs with
decent quality let me (and others here) know. I bet lots of us are
interested.

[...]
 
J

Joerg

Jan 1, 1970
0
John said:
Neither do Digikey, Mouser, Farnell or findchips.com :(

With Coilcraft you have to become a "member". Sign up and then you get
instant pricing access on their site. If you don't want to then request
a quote, comes within minutes via email. Their pricing is actually quite
favorable compared to other domestic suppliers, maybe because they do a
lot of automotive.
 
J

John Devereux

Jan 1, 1970
0
Joerg said:
[...]
Of course, neat. Finally the famed Jeorgian discrete SMPS is starting to
take shape :)

My first one took shape about 15 years ago, tons of them in use by
consumers right now, still coming off the conveyor belt (in China). I
don't know if it will outlast the VW Beetle in production but it
might. It has already survived the client company's president :-(

Back then I learned a valuable lesson: Inductor prices in the western
world must be taken with a grain of salt. A custom mfg place in Taiwan
outbid a catalog (!) part from a western manufacturer, big time. And
we even got an EMI-savvy toroid for our money versus the cheap open
cores from mainstream distributors. Of course it does help not to have
a middleman in the game.


[...]
Anyone know how to do this for smaller quantities (<~1k, say?). I've got
plenty of Asian suppliers trying to get our PCB business, but nobody
trying to sell me inductors :(.

<1k? No, I don't think they'd even talk to you.

I should have said "single reel" since there are typically more than
that on a reel. I guess we could probably make a "strategic" buy of ~5k
if needed. If they're *really* cheap 10k! :)
However, Asia is all about one thing: Connections. Ask the PCB
stuffing places what sort of deals they can get locally. They might be
able to swing it, with a supplier they already buy lots of other stuff
from.

If you have reputable places there that do <1k qty board runs with
decent quality let me (and others here) know. I bet lots of us are
interested.

Sorry, I meant bare PCB not stuffed. They advertise stuffing too - down
to any quantity - but they said once they buy non-free issued parts from
Digikey... :)

I just thought it strange that I can get small qtys of PCBs at 1/2 the
price of local suppliers, but nobody does the same for parts. (Which
ought to be much easier to supply).
 
J

Joerg

Jan 1, 1970
0
John said:
Joerg said:
John said:
John Devereux wrote:
[...]
Of course, neat. Finally the famed Jeorgian discrete SMPS is starting to
take shape :)
My first one took shape about 15 years ago, tons of them in use by
consumers right now, still coming off the conveyor belt (in China). I
don't know if it will outlast the VW Beetle in production but it
might. It has already survived the client company's president :-(

Back then I learned a valuable lesson: Inductor prices in the western
world must be taken with a grain of salt. A custom mfg place in Taiwan
outbid a catalog (!) part from a western manufacturer, big time. And
we even got an EMI-savvy toroid for our money versus the cheap open
cores from mainstream distributors. Of course it does help not to have
a middleman in the game.


[...]
Anyone know how to do this for smaller quantities (<~1k, say?). I've got
plenty of Asian suppliers trying to get our PCB business, but nobody
trying to sell me inductors :(.
<1k? No, I don't think they'd even talk to you.

I should have said "single reel" since there are typically more than
that on a reel. I guess we could probably make a "strategic" buy of ~5k
if needed. If they're *really* cheap 10k! :)

You could talk to companies like this:

http://www.xfmrs.com/

But since 10k is a small qty mostly you'll have to start here:

http://www.globalsources.com/gsol/I...urers/b/2000000003844/3000000184262/22705.htm

Sorry, I meant bare PCB not stuffed. They advertise stuffing too - down
to any quantity - but they said once they buy non-free issued parts from
Digikey... :)

That won't give you much of an upside. I recently had a proto-run fabbed
in the US (Aurora, Colorado), full turn-key, and was pleasantly surprised:

http://www.aapcb.com/

I just thought it strange that I can get small qtys of PCBs at 1/2 the
price of local suppliers, but nobody does the same for parts. (Which
ought to be much easier to supply).

They do, in places like Shenzen, if you let them purchase. But be
careful that they don't substitute, say, an electrolytic for a more
"economical" part.
 
J

John Devereux

Jan 1, 1970
0
Joerg said:
John Devereux wrote:
[...]
I should have said "single reel" since there are typically more than
that on a reel. I guess we could probably make a "strategic" buy of ~5k
if needed. If they're *really* cheap 10k! :)

You could talk to companies like this:

http://www.xfmrs.com/

But since 10k is a small qty mostly you'll have to start here:

http://www.globalsources.com/gsol/I...urers/b/2000000003844/3000000184262/22705.htm

[...]
Sorry, I meant bare PCB not stuffed. They advertise stuffing too - down
to any quantity - but they said once they buy non-free issued parts from
Digikey... :)

That won't give you much of an upside. I recently had a proto-run
fabbed in the US (Aurora, Colorado), full turn-key, and was pleasantly
surprised:

http://www.aapcb.com/

Thanks for the links; bookmarked. Will look into it further.
 
N

Nico Coesel

Jan 1, 1970
0
Joerg said:
But mucho Dolares. I'd try using the MCU if possible if this is a high
volume product. But it'll require lots of nifty engineering. For low
volume, yeah, don't bother and use a chip. Then I'd use the MC34063
which costs under 20 cents. None of those high-faluting newfangled ritzy
ones ;-)

Yes, but the decrease in costs of the chip is compensated by a large
inductor and capacitors.
 
J

Joerg

Jan 1, 1970
0
Nico said:
Yes, but the decrease in costs of the chip is compensated by a large
inductor and capacitors.

If you have the space those are cheap. Unfortunately Michael doesn't
have the space. But he also doesn't have the BOM budget for a modern
switcher chip :-(

IOW he is between a rock and a hard spot so the MCU or homebrew may be
his only options. I don't know if his ARM7 MCU has good timers left in
there. If not he's going to have to roll his own around a Schmitt inverter.
 
L

legg

Jan 1, 1970
0
I need to make clear so we avoid confusions

The device uses a 2.2" tft lcd . This thing has a backlight and no matter
what
draws 60 ma @ 3.3V .That i can't avoid . So with 60 ma Arm7 consumption
i end up to 120 ma.

As described, you intend this 'small' device to dissipate almost 4W
with a linear solution. You should assume a surface temperature rise
of one degree for every milliwatt dissipated by a square centimeter of
surface area for the package. You might want to recalculate
permissible losses on this basis and reappraise some basic design/cost
considerations that include your power source.
Now as i see whether i have a small switcher with huge inductor and
capacitor
so my board space is too large .
The other solution is super fast switcher that has everything small but
needs 2 euros
at least .
I imagine every engineer designing devices has faced this question .
I'm sure though that a simple cpu controlled switcher could cope with the
large dropout
from 24V to 4-5 volts and then a simple linear regulator could take you to
3.3V so you play
it safe.If things go wrong (cpu malfunction ) thermal shutdown from the
linear regulator would
reset the cpu and that would restart the system
As i see it a simple fet switch with cpu pwm control should satisfy that .
Any opinions or experience on that subject???

In any system with a self-regulated supply voltage, the first question
is 'Who's on first?'. Don't turn this into an Abott and Costello
routine.

RL
 
L

legg

Jan 1, 1970
0
1000 kelvin per watt in 1 cm**2? That's pretty pessimistic. Even
SOT-23s do better than that.

Cheers

Phil Hobbs

An SOT23 is not, in itself, the package that claims the Rthja rating;
it counts on conduction through the leadframe to larger external
surface areas (standard coupon patterns) and produces junction/surface
limits greatly exceeding those permissible to commercial external
surface touch or hold limits.

This simple rule of thumb for average surface temperature rise works
at room temperature for packages between matchbook and breadbox sizes,
without forced convection, to an accuracy of 20%.

This is not manufacturer's published bumph; it is not optimism or
pessimism; it is the result of many physical demonstrations over the
years, repeated usually for the benefit of midding-to-highly educated
and highly over-paid doubting Thomases.

It is also demonstrated in the free magnetics and capacitor
application software provided by mfrs such as Siemens, Magnetics Inc
and Cornell Dubilier - with pedigrees that reach back into the ancient
past of (gasp) print literature. Crunch the numbers back yourself on
whatever packages you currently are familiar with. Any software
package (no matter how expensive) has been mis-programmed or
mis-applied, if it predicts otherwise. Ball park compliance is a good
two-second paper and pencil check of the integrity of a convective
simulation.

RL
 
J

John Devereux

Jan 1, 1970
0
legg said:
legg wrote:
[...]
[...]
1000 kelvin per watt in 1 cm**2? That's pretty pessimistic. Even
SOT-23s do better than that.

An SOT23 is not, in itself, the package that claims the Rthja rating;
it counts on conduction through the leadframe to larger external
surface areas (standard coupon patterns) and produces junction/surface
limits greatly exceeding those permissible to commercial external
surface touch or hold limits.

This simple rule of thumb for average surface temperature rise works
at room temperature for packages between matchbook and breadbox sizes,
without forced convection, to an accuracy of 20%.

This is not manufacturer's published bumph; it is not optimism or
pessimism; it is the result of many physical demonstrations over the
years, repeated usually for the benefit of midding-to-highly educated
and highly over-paid doubting Thomases.

It is also demonstrated in the free magnetics and capacitor
application software provided by mfrs such as Siemens, Magnetics Inc
and Cornell Dubilier - with pedigrees that reach back into the ancient
past of (gasp) print literature. Crunch the numbers back yourself on
whatever packages you currently are familiar with. Any software
package (no matter how expensive) has been mis-programmed or
mis-applied, if it predicts otherwise. Ball park compliance is a good
two-second paper and pencil check of the integrity of a convective
simulation.

RL

That's an interesting and useful "rule of thumb", thanks.
 
J

John Devereux

Jan 1, 1970
0
legg said:
On Thu, 22 Oct 2009 10:06:35 -0400, Phil Hobbs
As described, you intend this 'small' device to dissipate almost 4W
with a linear solution. You should assume a surface temperature rise
of one degree for every milliwatt dissipated by a square centimeter of
surface area for the package. You might want to recalculate
permissible losses on this basis and reappraise some basic design/cost
considerations that include your power source.
[...]
1000 kelvin per watt in 1 cm**2? That's pretty pessimistic. Even
SOT-23s do better than that.

Cheers

Phil Hobbs

An SOT23 is not, in itself, the package that claims the Rthja rating;
it counts on conduction through the leadframe to larger external
surface areas (standard coupon patterns) and produces junction/surface
limits greatly exceeding those permissible to commercial external
surface touch or hold limits.

This simple rule of thumb for average surface temperature rise works
at room temperature for packages between matchbook and breadbox sizes,
without forced convection, to an accuracy of 20%.

This is not manufacturer's published bumph; it is not optimism or
pessimism; it is the result of many physical demonstrations over the
years, repeated usually for the benefit of midding-to-highly educated
and highly over-paid doubting Thomases.

It is also demonstrated in the free magnetics and capacitor
application software provided by mfrs such as Siemens, Magnetics Inc
and Cornell Dubilier - with pedigrees that reach back into the ancient
past of (gasp) print literature. Crunch the numbers back yourself on
whatever packages you currently are familiar with. Any software
package (no matter how expensive) has been mis-programmed or
mis-applied, if it predicts otherwise. Ball park compliance is a good
two-second paper and pencil check of the integrity of a convective
simulation.

That's an interesting and useful "rule of thumb", thanks.
 
N

Nico Coesel

Jan 1, 1970
0
michael nikolaou said:
I need to make clear so we avoid confusions

The device uses a 2.2" tft lcd . This thing has a backlight and no matter
what
draws 60 ma @ 3.3V .That i can't avoid . So with 60 ma Arm7 consumption
i end up to 120 ma.
Now as i see whether i have a small switcher with huge inductor and
capacitor
so my board space is too large .
The other solution is super fast switcher that has everything small but
needs 2 euros
at least .
I imagine every engineer designing devices has faced this question .
I'm sure though that a simple cpu controlled switcher could cope with the
large dropout
from 24V to 4-5 volts and then a simple linear regulator could take you to
3.3V so you play
it safe.If things go wrong (cpu malfunction ) thermal shutdown from the
linear regulator would
reset the cpu and that would restart the system
As i see it a simple fet switch with cpu pwm control should satisfy that .
Any opinions or experience on that subject???

You mean some sort of pre-regulator before the linear? That might
work, expecially if you make the the backlight is off until the
pre-regulated voltage is withing range.
 
M

michael nikolaou

Jan 1, 1970
0
Dagmar


that is pretty good IC .Only the 150 uH spoils the picture . That
needs 8x 8 mm minimum board size .
The capacitors are fine. i think i will stick to that
 
J

JosephKK

Jan 1, 1970
0
It sounds familiar. I've looked at these 'low power' ARM chips,
noticed that many seem to average 20-50mA, with the lower figure only
on a good day and many requiring more. It was NOT a shock to me to
see 120mA with ARMs. In fact, I'm just fine with that. But when the
OP writes, "I'm making a mcu based device which i want to be very
small and low cost" and then out the other side of the mouth says
"only 120mA" then I'm truly wondering.

One of the BIG tradeoffs is power __AND__ heat. And by the time you
get anywhere near 120mA, you've often got both problems in spades.
It's a fundamentally different domain.

I guess that's why I just went _white_ when I read those figures and
the OP's language in the same context. The only way 120mA is a
little, these days, is if you are used to x86 processors running at
GHz and requiring multiple power supply rails to help contain heat
problems better.

If you live in that world, I can see it. But that sure isn't MY
embedded world perspective. I consider moving into the 120mA domain
as being akin to a "damn-the-torpedoes, devil-may-care" world. At
that point, you are already spending dollars, not pennies, and have
board room to spare. And if you are chugging 120mA, you NEED space,
anyway.


Hehe. Yeah. If we are talking washing machines, 120mA is no problem.
But room isn't a problem, then, either. There's always a corner, plus
one HUGE heat sink, too. ;)


I'd recommend that Michael rethink 120mA. I mean, jeeez! If you
nearing a watt already, with overhead, you need space and you expect
to spend something on the power supply, too. Or some serious, crafted
time. Or both.

Almost two decades ago, I was worrying over a thermal cooling stack
with two Peltier stages and the bottom stage was consuming half a
watt. I was worried about that much heat. And the tiny micro device
plus analog circuits at the top was burning some 30mW. Now that's the
kind of thing you get with 120mA! 2-stage Peltier coolers AND a
micro. 20 years ago.

This is crazy-making to think about 120mA! Yeah, if you are making an
iPhone or internet interface device with RF and all. I mean, you need
to actually broadcast maybe 1/4 watt or so. So yeah. But "I'm making
a mcu based device which i want to be very small and low cost?"

The OP and I must come from different universes!

Jon

You may be on to something. OP seems to also be considering a linear
regulator from 24 V to 3.3. Talk about power waste.
 
J

Jon Kirwan

Jan 1, 1970
0
You may be on to something. OP seems to also be considering a linear
regulator from 24 V to 3.3. Talk about power waste.

OMG! Are you kidding? (I must have missed reading and/or connecting
the dots on posts you saw.)

Jon
 
J

JosephKK

Jan 1, 1970
0
OMG! Are you kidding? (I must have missed reading and/or connecting
the dots on posts you saw.)

Jon

And it is possible that i misread instead. YMMV.
 
Top