new 30MHz to 300MHz switcher - worlds smallest laptop adapter

[Jamie M]

Hi,

Here's a new SMPS switcher apparently coming out next year for laptops:

http://phys.org/news/2013-12-circuit-laptop-charger-brick.html

It apparently has a "power reclaiming scheme" for higher efficiency,
would this be conventional sync rectification or some other thing?

It looks like power electronics is in for a big (r)evolution once the
300MHz+ designs start to spread!

cheers,
Jamie

 

[Spehro Pefhany]

He got 9 million $ funding, has to sell a lot of cheap converters to break even,
not counting his own pay.
Laptop converters go for 12$ or so on ebay,
And not 2 models are the same.

There are "universal" converters, that apparently more-or-less work
fairly well.

be a while before his chips are on ebay ?

Maybe at 300 MHz just air coupled transformers (just a few turns) would work,
no magnetic losses.

Not sure that follows.

But that could be done with a current sine wave resonant design.
Not sure anybody wants to fork out 9 M$ for me to design one,

You could try crowdfunding if you were interested. Those bricks are a
real PITA. A lot of people would shell out $50 or $100 if it made the
problem disappear. It wouldn't have to be revolutionary technology,
just a different trade-off (and probably some kick-ass styling and
slick videography/photography to sell it).


Best regards,
Spehro Pefhany

 

[Artem]

But that could be done with a current sine wave resonant design.

Ordinary MOSFET at 300 Mhz?

 

[Artem]

On a sunny day (Wed, 25 Dec 2013 04:27:09 -0800 (PST)) it happened Artem

No need for a MOSFEt, transistor would do.

RF transistors usually low voltages.

 

[Artem]

On a sunny day (Wed, 25 Dec 2013 04:46:36 -0800 (PST)) it happened Artem

drivers, although low current, 100 mA at 230 V is already 23 W.

1. It's only for 110v power grid.
2. Bandwidth TV only 6.5 Mhz.
3. Efficiency will be low because transistor in linear mode.

 

[Artem]

Normally everybody screams: 'switcher switcher switcher'.

Switcher transistors have an efficiency problem if they cannot switch fast enough,

power is dissipated during the on and of switching times.



I was proposing (I dunno about that professor), to use a SINE wave converter.

Yes. And main benefit of resonant converter is decreasing switching losses by switching when current is minimum.
For linear amplifier theoretical efficiency only 78.5%
http://en.wikipedia.org/wiki/Amplifier#Class_B

 

[[email protected]]

1. It's only for 110v power grid.

2. Bandwidth TV only 6.5 Mhz.

3. Efficiency will be low because transistor in linear mode.

However annoying patents are, these things are always patented, and
have to be. These guys filed in 2009. Without patents it wouldn't
make sense to do all that work--make all that investment--only
to have it ripped off the nanosecond you ship.

The front end is a capacitive charge pump that charges many capacitors
and switches in series, then flips another set of switches to
discharge the caps in parallel, creating a raw output. A
synchronous buck efficiently regulates the rough voltage
thus created down to the desired output.

It runs the switches and caps at HV in series, so only low
voltage switches are needed. Ditto for the finishing regulator.

http://patft.uspto.gov/netacgi/nph-...50&s1=8212541.PN.&OS=PN/8212541&RS=PN/8212541

The main inventive notion here seems to have been getting the operations
within the range of fast, low-voltage elements.

I haven't looked at it in depth yet..time for Christmas!

Cheers,
James Arthur

 

[Spehro Pefhany]

Yes. And main benefit of resonant converter is decreasing switching losses by switching when current is minimum.
For linear amplifier theoretical efficiency only 78.5%
http://en.wikipedia.org/wiki/Amplifier#Class_B

GaN technology is about ready. Even the drivers are off-the-shelf.


Best regards,
Spehro Pefhany

 

[Spehro Pefhany]

Sounds like yet more MIT vaporware. I doubt that 0.1% of MIT's press-release
inventions ever work.

According to a presentation I saw on licensed inventions from another
university (Kirsten Leute @ Stanford), they've taken in 1.6bn in
licensing fees since 1971 and get disclosures at higher than a daily
basis, and _start_ patent applications on ~50% but they have not had a
new high-income winner in about 18 years. The last big one was
Google-related (1996 $337m as of 2012- they took some equity which
paid off big)... but mostly life sciences/biotech stuff with a couuple
of communications/EE inventions. Before 1996 it was every 2-3 years
between big hits.

I guess one could speculate on whether the pace of home-run inventions
has actually slowed down, or something else is going on.


Best regards,
Spehro Pefhany

 

[[email protected]]

According to a presentation I saw on licensed inventions from another
university (Kirsten Leute @ Stanford), they've taken in 1.6bn in
licensing fees since 1971 and get disclosures at higher than a daily
basis, and _start_ patent applications on ~50% but they have not had a
new high-income winner in about 18 years. The last big one was
Google-related (1996 $337m as of 2012- they took some equity which
paid off big)... but mostly life sciences/biotech stuff with a couuple
of communications/EE inventions. Before 1996 it was every 2-3 years
between big hits.

I guess one could speculate on whether the pace of home-run inventions
has actually slowed down, or something else is going on.

It's my impression that either from conviction or
poor risk/reward conditions, there is less inventing
in the US.
People are spending time mitigating risk and cost,
not seeking it.

OTOH, there's all sorts of opportunity for inventions
and synergy--cheap wireless, FLASH, GHz CPU, etc.

Cheers,
James Arthur

 

[[email protected]]

However annoying patents are, these things are always patented, and
have to be. These guys filed in 2009. Without patents it wouldn't
make sense to do all that work--make all that investment--only
to have it ripped off the nanosecond you ship.

The front end is a capacitive charge pump that charges many capacitors
and switches in series, then flips another set of switches to
discharge the caps in parallel, creating a raw output. A
synchronous buck efficiently regulates the rough voltage
thus created down to the desired output.

It runs the switches and caps at HV in series, so only low
voltage switches are needed. Ditto for the finishing regulator.

http://patft.uspto.gov/netacgi/nph-...50&s1=8212541.PN.&OS=PN/8212541&RS=PN/8212541


The main inventive notion here seems to have been getting the operations
within the range of fast, low-voltage elements.



I haven't looked at it in depth yet..time for Christmas!

Okay, I looked at it a little deeper.

I missed on a couple points.

First, the switched capacitor switches aren't low voltage. There
are only a few of them, so that doesn't fly. Also, when S1 of
Fig. 6 is open, one of the switches S2 has to stand off the entire
input voltage. Let's hope S2 never pops!

Second, I figured out the "energy recirculation" -- it's kind
of elegant. The charge pump efficiency is improved by charging
the series capacitor string *through the synchronous buck*.
That way, the series capacitor strings' peak charge current is
controlled, avoiding inrush charging losses. That's clever.

In effect, the buck runs off the switched-cap converter's ripple
voltage when the cap string is in series, and off the charge
pump caps in parallel during the paralleled time.

It also means you don't need nearly as many switched-cap stages
to get to the roughed output used to feed the sync. buck.

The VHF aspect is confined to the synch. buck finishing regulator.
Its low input voltage allows the use of fast, small geometry
devices.

There, I think that's the gist of it.

No galvanic isolation, which might be changed by substitution of
a suitable isolated "buck."

Cheers,
James Arthur

 

[Maynard A. Philbrook Jr.]

Oh, my! A Marx generator run in reverse. Well, lots of
possible places things could break down, or just lose energy.

Jon

Exactly, nothing new here, just some one trying to bank on
old tech..

Jamie

 

[[email protected]]

So then you need that transformer again, and an other synchr. rectifier,
a simple diode there would ruin efficiency.
May as well go without the cap switches network then...???

You need the cap-switch network. The low input voltage into the
synch. buck is what produces the other opportunities:
1. the cap-switching reverse Marx generator (thanks Jon!)
makes a low, unregulated voltage pretty efficiently.
2. The synchronous buck can use low-voltage, low resistance,
low Qc FETs that scream.

The patent suggests running the cap-switcher at, say, 1Mhz, and
the synch. buck at 5-100x that.

Ramifications: the low voltage differential into the buck
increases the buck's duty cycle, which increases efficiency.
Low voltage differential also reduces the inductance needed
for a given ripple current, which increases efficiency,
reduces volume, and reduces copper losses, which reduces volume
even more.

There's usually a fatal flaw. Haven't seen it yet, but it looks
fragile--lots of switches have to flip reliably, or you smoke it.

Cheers,
James Arthur

 

[[email protected]]

And, he apparently got $9 million for it! Not BAD!
Wish I had great "ideas" like this!

I haven't seen the soft-charging of a cap-switcher trick. Nor
have I seen using a cap-switcher front-end to a buck switcher,
or that people recognized the advantages that flow from that.

I'm not endorsing it mind you, just analyzing.

Cheers,
James Arthur
(P.S. I've not seen soft-charging in a Marx generator either--
that's a good idea

 

[Greegor]

http://free.patentfetcher.com/GetPatentPDF.php?f=Pats/US/82/12/US8212541.pdf

That one retains an archive of it.

 

[[email protected]]

Oh I see lots of problems, this cap switcher, when in series with the buck,
may charge smoothly, but then putting those caps in parallel
on the next step, would require precise equal caps (so they were charged to the same voltage),
else you get big spikes (at that 1MHz) in the switching FETS to equalize the cap voltages,

Agreed--any cap mismatch produces dV(c), making spikey spikes when
switching from series to parallel configuration, same as a
conventional charge-pump.

If on chip caps then they could be equal, in any case what would aging do with external caps?

I haven't bothered with any numbers, but I rather doubt they could
use on-chip caps. The capacitances and voltages needed are too high.
So says my gut, anyhow.

Let's see...if we wanted 60w (input) worth of charge packets at
1MHz at 170VDC input,
c=60W/(.5*v^2*1Mhz)= 4nF for the series string, or 12nF each for
a string of 3, at 57 volts each.

That would be quite a chip.

Ramifications: the low voltage differential into the buck
increases the buck's duty cycle, which increases efficiency.
Low voltage differential also reduces the inductance needed
for a given ripple current, which increases efficiency,
reduces volume, and reduces copper losses, which reduces volume
even more.

Somebody mentioned we need a multi-voltage multi-current design, to grab that market,
that is an other problem here.

If not a multi type output you may as well build it into the laptop,
and provide a power cord connector.

And the fact that any adaptor I have bought lately (last year) says: "110-230V'

may make his chip design a bit more challenging with on chip capacitors and switches....

It should at least stand 240 x sqrt(2) = 340 V DC on the chip input, and as you noticed
at least one of the switches sees that full voltage,

And that is not counting weird high mains, as may still be present in some places in the UK,

UK had 240 V AC.

There's usually a fatal flaw. Haven't seen it yet, but it looks
fragile--lots of switches have to flip reliably, or you smoke it.

Yes, the magic [smoke] may become visible at some point

Cheers,
James Arthur

 

[[email protected]]

Any inductor in series will do that, look at how flash tube circuits work.
the discharge (and what is the difference with charging) is made smooth and lengthened over
time by a small L in series, I have used that.

Of course, we all know that. But I haven't seen the bottom of a
series'd string of switched caps terminated in a synchronous
buck, specifically to reduce losses in the switch-cap charging
and provide a low input voltage to the buck.

You may think it's obvious but the obvious rejoinder is that if
a. it's better, and
b. obvious,
why have billions of laptop adapters been made, without ever
using it?

Nor

That last thing is like saying you can get a patent for adding a 7805 after a switchmode.

If you were the first one to do it, and no one else had thought of
it, that might be patentable too. In 1930.

That is just normal practice, regulators in series (of different type) have been used for years.

If it's normal practice, can you point to anyone anywhere that's done it?
I can't.

????

Controlled output voltage, switching caps regulates NOTHING.

He needed a regulator anyway!

Right, but you're missing the point. He can generate, say, +5v from
+170v with the switched cap thing, so that the switcher can be a
30MHz unit with 10v mosfets. Or +3v and 5v mosfets, etc.

I am against patents for these kind of things,
maybe you could argue using the inductor of the buck as current limit is new,
but using an inductor as current limiter most certainly is not.

Now everybody, if that patent is granted, has to pay that club for something any designer could

come up with if he spend some time, that is what design is!

The existing selection of laptop adapters, representing millions
of man-hours of design, suggest that no one else came up with it.
That's the whole point.

Make new configurations, make new things with the things you have.

Who gave him the 9 M$ was wrong.

That could very well be. If the thing doesn't deliver, they're
screwed. But if they've worked out any defects, it could be a
home-run.

It's reasonable to be skeptical, until then.

Cheers,
James Arthur

 

[[email protected]]

It is the only part of the invention, using the buck inductor to limit
the series capacitor charge current and get some extra power, that is 'new' AFAIK.

The way I read the patent is that it attempts to claim all configurations like that,

The legalese annoys me, but I skimmed the claims. They all 'recite'
(that's the term of art) a switched-cap front end with a dynamically
alterable division ratio, coupled with a conventional switcher that
advantageously uses the lowered input voltage, and produces a regulated
output.

Seems fair to me, and pretty specific.

The 7805? seen it many times...

Not, not that--have you seen anyone using an adaptible-ratio
switched-cap as a front-end to a low-voltage switcher?

Used it myself, may even be on my site.

usually I get a 12 V switching wall wart and make 5 V for PICs from it with yes, an 7805.

He actually does coarse regulation by changing the cap-switching
ratios. I skipped that as not very interesting.

You still need the mains separation, transformer, rectifier in secondary,
Would it really be that more efficient then if you used that directly?

Did not they get better than 87% already?

I did not see any numbers on claimed efficiency by them.

I didn't see any efficiency numbers either, but one infers high efficiency
from the small form factor, otherwise it'd burn up.

Well, they just gobble up the 9 M$ in research.

Same for the research into quantum computers,
the looking for gravity waves, the experiments for break even fusion,
and what not, perpetuum energy sources.


Yep, when in the shops, or a chip I can buy, I will congratulate them.
Especially if it still works after 5 years.

Sounds good.

Cheers,
James Arthur

 

[Peter]

Guys, is it physically possible to transfer energy with just
capacitors for isolation?

I know you can send data that way (because it can be encoded with just
edges, and providing there isn't too much common mode noise, the edges
can be decoded) but I can't see any way to transfer *power* that way.

Inductive components are needed if you want isolation - as in most
consumer appliances.

LED lamps don't need isolation and there is a huge amount of work
being done in that area.

 

[Tim Williams]

Guys, is it physically possible to transfer energy with just
capacitors for isolation?

Signals anyway,
http://www.ti.com/lit/ds/symlink/iso103.pdf

Nothing wrong with it for power, but the capacitances will be small to
ensure galvanic isolation (~nF), suggesting very high frequencies or
inductive reactances to cancel it. The inductive case looks like coupled
resonators, which might be achieved accidentally by proximity of the coils
(in which case you have "wireless energy", which is a slowly rising fad
these days). At that point, isolation distance is only a matter of having
sufficiently high Q factors (and closely matched resonators).

Tim