Maker Pro
Maker Pro

Isolated power supply design

E

Ethan

Jan 1, 1970
0
Hello:

I am just staring on a new design for a medical instrument and need to
figure out the power supplies. I am looking for recommendations on
where to start and what the plusses and minuses are for various
topologies and any general comments on what devices/designs are out
there.

It needs to be a universal input, 85Vac-265Vac, and dc output. The
load will vary between 0 and approximately 30W. For now I am thinking
the output voltage should be a loosely regulated 9V since this is a
convenient starting place for all the other supplies I will need. I am
not really concerned with the low voltage, non-isolated supplies since
there are a lot of straight forward dc-dc ICs and I have a lot of
experience with these.

Most of the reference designs I have seen are for fly back converters.
My concern is that since this is a "patient applied device" (in FDA
terms) the transformer needs to provide reinforced insulation between
primary and secondary. Which means the primary and secondary need to
be physically far apart. A fly back converter requires very low
leakage inductance, so the primary and secondary need to be as close
together as possible. These requirements seem contradictory, so a fly
back might be difficult. Does anybody know what it takes to get a fly
back design to work with reinforced insulation?

I am also looking into various resonant converter designs which should
have no trouble with a large leakage inductance. These topologies
are more complicated, and I haven't seen any ICs that are designed
for this kind of thing. On the one hand: I would have a lot more fun
rolling my own circuit, on the other: this is a real product and low
part count needs to be taken seriously if I can get it to work.

Other big concerns are EMC, radiated and conducted emissions
requirements are pretty stringent, and of course all the other volume
production issues of manufacturability, reliability, part sourcing etc.

What do you guys think?

Ethan Petersen
 
P

Pooh Bear

Jan 1, 1970
0
Ethan said:
Hello:

I am just staring on a new design for a medical instrument and need to
figure out the power supplies. I am looking for recommendations on
where to start and what the plusses and minuses are for various
topologies and any general comments on what devices/designs are out
there.

It needs to be a universal input, 85Vac-265Vac, and dc output. The
load will vary between 0 and approximately 30W. For now I am thinking
the output voltage should be a loosely regulated 9V since this is a
convenient starting place for all the other supplies I will need. I am
not really concerned with the low voltage, non-isolated supplies since
there are a lot of straight forward dc-dc ICs and I have a lot of
experience with these.

Ok.

A 'universal input' supply such as you require means you need to use a
switcher.

30W is definitely in the preferred range for a flyback type.

You can have any voltage you like. Even multiple supplies. If you're
designing it yourself then simply choose the voltages you actually need.

Most of the reference designs I have seen are for fly back converters.
My concern is that since this is a "patient applied device" (in FDA
terms) the transformer needs to provide reinforced insulation between
primary and secondary. Which means the primary and secondary need to
be physically far apart. A fly back converter requires very low
leakage inductance, so the primary and secondary need to be as close
together as possible. These requirements seem contradictory, so a fly
back might be difficult. Does anybody know what it takes to get a fly
back design to work with reinforced insulation?

There is *no* difference in this respect between a flyback transformer and
any other transformer. Reinforced insulation can certainly be used.
Increased leakage inductance isn't especially helpful though but there's a
newish alternative....

If you're concerened about isolation ( sensible in your application ) then
I suggest that you use triple insulated wire in the transformer ( in
addition to any other methods that may be required for medical isolation
requirements ).

Triple insulated wire is great but check the regulatory acceptance for this
in your application.

Here's a link.

http://www.furukawa.co.jp/makisen/eng/product/texe_feature.htm

" Certified as Reinforced Insulation; the combination of any two layers can
resist up to 3,000 ACV for one minute. "

I am also looking into various resonant converter designs which should
have no trouble with a large leakage inductance. These topologies
are more complicated, and I haven't seen any ICs that are designed
for this kind of thing. On the one hand: I would have a lot more fun
rolling my own circuit, on the other: this is a real product and low
part count needs to be taken seriously if I can get it to work.

Other big concerns are EMC, radiated and conducted emissions
requirements are pretty stringent, and of course all the other volume
production issues of manufacturability, reliability, part sourcing etc.

Check out www.powerint.com and read all the app notes. You should be an
expert by then !

You're likely to need to pay special attention to leakage currents I
suspect in your case.

Graham
 
P

PN2222A

Jan 1, 1970
0
Ethan said:
Hello:

I am just staring on a new design for a medical instrument and need to
figure out the power supplies.
It needs to be a universal input, 85Vac-265Vac, and dc output. The
load will vary between 0 and approximately 30W. For now I am thinking
the output voltage should be a loosely regulated 9V since this is a
convenient starting place for all the other supplies I will need.

What do you guys think?

Ethan Petersen
Ethan

Can you clarify something?

Do you need 30W at the patient applied part? Or does the entire instrument
draw appx. 30W
and you need less power at the patient site?

Design of a whole instrument for 10microamps is lots harder than design of a
small sensor or other probe.

Regards
PN2222A

"Been there.
Done that.
Got the 483."
 
A

artie

Jan 1, 1970
0
Ethan said:
Hello:

I am just staring on a new design for a medical instrument and need to
figure out the power supplies. I am looking for recommendations on
where to start and what the plusses and minuses are for various
topologies and any general comments on what devices/designs are out
there.

It needs to be a universal input, 85Vac-265Vac, and dc output. The
load will vary between 0 and approximately 30W. For now I am thinking
the output voltage should be a loosely regulated 9V since this is a
convenient starting place for all the other supplies I will need. I am
not really concerned with the low voltage, non-isolated supplies since
there are a lot of straight forward dc-dc ICs and I have a lot of
experience with these.

Most of the reference designs I have seen are for fly back converters.
My concern is that since this is a "patient applied device" (in FDA
terms) the transformer needs to provide reinforced insulation between
primary and secondary. Which means the primary and secondary need to
be physically far apart. A fly back converter requires very low
leakage inductance, so the primary and secondary need to be as close
together as possible. These requirements seem contradictory, so a fly
back might be difficult. Does anybody know what it takes to get a fly
back design to work with reinforced insulation?

I am also looking into various resonant converter designs which should
have no trouble with a large leakage inductance. These topologies
are more complicated, and I haven't seen any ICs that are designed
for this kind of thing. On the one hand: I would have a lot more fun
rolling my own circuit, on the other: this is a real product and low
part count needs to be taken seriously if I can get it to work.

Other big concerns are EMC, radiated and conducted emissions
requirements are pretty stringent, and of course all the other volume
production issues of manufacturability, reliability, part sourcing etc.

What do you guys think?

Ethan Petersen

I think you should look in the DigiKey catalog where they have units
available off the shelf with the safety/EMC certifications you need.
 
E

Ethan

Jan 1, 1970
0
I expect the whole instrument will draw a maximum of about 30W. Most
of the the circuits are on the same ground reference as the patient
which is where the isolation requirement comes from. The patient site
is only about disipating about 30mW. You can't put too much power
here. It's bad form to burn an unconscious person. The rest of the
power is going to displays and battery chargers and processors and a
bunch of other bells and whistles.

If I don't put all the patient isolation into the mains power supply my
architecture gets really complicated. I have been round and round on
this with my quality assurance and regulatory guys and it really
simplifies things if all the isolation is in one place. Assuming I can
get it to work. If I can't, there are other solutions with multiple
isolation boundries, but these get complicated and I would like to
avoid that.

Regarding Pooh Bear's comments on triple insulated wire, that looks
interesting. My concern is the capacitance between primary and
secondary windings still needs to be low due to leakage currents. If I
remember right, test is less than 10uA-ac-rms at 265V 50Hz, or about
125pF. But still, the wire looks like it is worth looking into for
real.

Thanks,

Ethan
 
P

PN2222A

Jan 1, 1970
0
Ethan said:
The patient site
is only about disipating about 30mW. You can't put too much power
here. It's bad form to burn an unconscious person.

Burning the unconscious is exactly what some of my designs intend!
But it is important to put the burn where the surgeon wants it.
If I don't put all the patient isolation into the mains power supply my
architecture gets really complicated.

Not that complex, really. Standard design for ECG amplifiers includes
a transformer-isolated power supply at a few MW providing the power to the
preamp, then some clever way of sending the signal back across the barrier.
Or at least that was the way we did it a generation ago...

For 30mW, you don't really care about efficiency, just isolation
(4KV standoff MINIMUM)and low,low leakage current. It's a handful of
cheap components costing a few bucks.

Think about how much simpler any supply design is with 30 dB less power to
handle.

I'm also thinking about the capacitance you get for free with a larger
physical
package 'inside the barrier'.

Regards
PN2222A
Tf (fall time) 60 nS
See figure 2
 
R

Roger Lascelles

Jan 1, 1970
0
I think you should look in the DigiKey catalog where they have units
available off the shelf with the safety/EMC certifications you need.

My thoughts too. There are companies who specialise in medical spec
switchmodes - look for US switchmode companies with Google. This removes a
lot of work - do you really want to be a switchmode company ? Unless your
product is very low priced and large volume. You will get to market a lot
faster. Anyway, version 2 can have an in-house switchmode if you feel like
it.

If you don't need medical grade insulation, the Asians can be cheap and
good - I'll bet you could find an Asian 30W switchmode board with UL
certification. Or cased, if you want.

Roger Lascelles
 
Ethan said:
I am just staring on a new design for a medical instrument and need to
figure out the power supplies.

Digi-Key: Elpac "MW" 24 W medical (12, 15, 18, or 24 V) $36/10, $27/100.
If you need more power, the 40 W version is $42/10, $32/100. Mouser has
some similar units at similar prices by Ault, Phihong, and Elpac. These
are all external ones - there are internal ones as well.

Matt Roberds
 
J

Joerg

Jan 1, 1970
0
Hello Ethan,
I expect the whole instrument will draw a maximum of about 30W. Most
of the the circuits are on the same ground reference as the patient
which is where the isolation requirement comes from. The patient site
is only about disipating about 30mW. You can't put too much power
here. It's bad form to burn an unconscious person. The rest of the
power is going to displays and battery chargers and processors and a
bunch of other bells and whistles.

Connecting system ground to the patient is usually not a good thing to
do. Often it is prohibited. Considering that the system can be touched
by personnel, just imagine what could happen if the patient went into a
cardiac situation and would need to be defibrillated.
If I don't put all the patient isolation into the mains power supply my
architecture gets really complicated. I have been round and round on
this with my quality assurance and regulatory guys and it really
simplifies things if all the isolation is in one place. Assuming I can
get it to work. If I can't, there are other solutions with multiple
isolation boundries, but these get complicated and I would like to
avoid that.

Patient isolation at the patient interface connection is usually not
complicated, I do this all the time. What signals need to be
transferred? Frequency range? Noise specs? If you are at liberty to
tell, what's the system supposed to do?

What truly is complicated is procuring the parts to make the isolation
barrier. Agencies usually want all the parts (such as wire insulation)
certified as compliant by the manufacturer. I usually spec in wire where
every single foot produced is HV tested. So, typically the time on the
phone with the various vendors almost matches the design time. This
strategy pays off when doing the agency tests. Labs such as UL often
favor certificates and manufacturer documents over rationale. It's
pretty similar to aeronautics.

If you go that route of isolating at the patient connection you also
need to have a very meticulous assembly procedure and proper training of
the folks doing it (must be documented). Plus Hipot and the usual works.
Regarding Pooh Bear's comments on triple insulated wire, that looks
interesting. My concern is the capacitance between primary and
secondary windings still needs to be low due to leakage currents. If I
remember right, test is less than 10uA-ac-rms at 265V 50Hz, or about
125pF. But still, the wire looks like it is worth looking into for
real.

With a switcher that is going to be hard. As you mentioned you must also
comply with EMC regs. Since you can't have much in terms of Y-caps it
will be tough if using a switcher. With a linear supply you might get
away with it, else it'll likely need a med grade iso transformer (no big
deal). But again, doing a patient isolation solely at the power supply
is something I have never done and would never do.

Regards, Joerg
 
E

Ethan

Jan 1, 1970
0
I would rather not get into the architecture of my system in this
thread. Partly because it would get complicated and hard to explain in
a forum like this in a reasonable amount of time. But mostly because
it is proprietary and my company is pretty sensitive about that kind of
thing these days. Talking about it at all on this list is a little
edgy.

So, for the sake of this discussion I would like to pretend it is a
given that I need to design a universal input power supply with
reinforced insulation. I know it is possible because as a couple
people pointed out I can go to Digikey and buy one. And if they can do
it, I should be able to do it too. Also my manufacturing quantities
are pretty large so developing my own power supply is definitely worth
while from a cost of goods vs NRE trade off point of view, and I am
starting early enough that I have time to do this design without
effecting time to market, etc.

The forward converter designs on the www.powerint.com app notes look
pretty interesting. It is giving me ideas on how to take a fly back
controller chip and put it into a design that is more like a resonant
converter where leakage inductance is not much of an issue.

Does anybody know of app notes on resonant converter designs for
universal input power supplies?

Thanks,

Ethan
 
J

Joerg

Jan 1, 1970
0
Hello Ethan,
Does anybody know of app notes on resonant converter designs for
universal input power supplies?

A resonant converter is the ticket if you absolutely have to have all
the isolation there and must live with a somewhat leaky transformer.
They are also usually lower in EMI. Best source as far as I can tell
would be the old Unitrode app notes, now on the TI web server. They also
have the chips.

Regards, Joerg
 
W

Walter Harley

Jan 1, 1970
0
Ethan said:
[...]
So, for the sake of this discussion I would like to pretend it is a
given that I need to design a universal input power supply with
reinforced insulation. I know it is possible because as a couple
people pointed out I can go to Digikey and buy one.

You might consider buying one and reverse-engineering it to figure out what
kind of topology and techniques they used.
 
J

John Smith

Jan 1, 1970
0
Walter Harley said:
Ethan said:
[...]
So, for the sake of this discussion I would like to pretend it is a
given that I need to design a universal input power supply with
reinforced insulation. I know it is possible because as a couple
people pointed out I can go to Digikey and buy one.

You might consider buying one and reverse-engineering it to figure out what
kind of topology and techniques they used.


Oooooooh, that's cheating...
 
R

Rich Grise

Jan 1, 1970
0
I am just staring on a new design for a medical instrument and need to
figure out the power supplies. I am looking for recommendations on where
to start and what the plusses and minuses are for various topologies and
any general comments on what devices/designs are out there.

It needs to be a universal input, 85Vac-265Vac, and dc output. The load
will vary between 0 and approximately 30W. For now I am thinking the ... ....
Other big concerns are EMC, radiated and conducted emissions
requirements are pretty stringent, and of course all the other volume
production issues of manufacturability, reliability, part sourcing etc.

What do you guys think?

I think you should find somebody who's already making medical
equipment power supplies, and ask them for a quote.

Good Luck!
Rich
 
Top