power supply heat problem

[Guest]

Hi to all.
Sorry if this is long winded...
I have been tasked with designing a supply 220Vac to 28Vdc , 2.5A (~70W).
The enclosure is plactic 90mm by 90mm with a height of about 4cm at one end
sloping down to
about 1,5cm at the other.The enclosure has NO ventelation (designed to be ant
proof).
The area I have to work with is plenty , but I am having heat issues. Other
constraints are
low cost to make and assemble. Ambient up to 50deg.
The initial power supply I designed was a flyback supply using a uc3842
controller chip. The supply is
running well on the bench , but in the box , under full load is getting to hot.
It has not failed yet , but the
heatsink temp is around 90 deg. I am assuming the junction temp is probably
about 10deg above this.
Add another 30 deg for high ambient temps and I can't see this working for long
:0(
The switching fet has a small heatsink , and I am using a SM rectifieng diode ,
with the board as heatsink.
I'm using quite a large area of the pcb for the diode heatsink. Top and bottom
, with lots of via's to conduct heat
across layers. this seems to be working well.
I'm assuming the supply won't be more than say 80% efficient , so that is about
14W of heat to get rid of.
Quite a lot is a small plastic box with no ventelation...
I have then tried to make quasi resonant convertor using a ncp1207 from ON
semi. The circuit seems to be working well , but the switching fet seems to be
getting hotter than in my other design. Even at low load.I thought a resonant
convertor should be more efficient.I have sent some screenshots of the fet's
drain waveform to ABSE. Maybe someone can see from that what may be going
wrong. The waveforms look like they should as far as I can see.
Maybe I am expecting to much from this type (flyback) of supply. Would I be
better off using a forward converter
type of supply ?

Any tips would be appreciated.
Cheers
Rob

 

[martin griffith]

Hi to all.
Sorry if this is long winded...
I have been tasked with designing a supply 220Vac to 28Vdc , 2.5A (~70W).
The enclosure is plactic 90mm by 90mm with a height of about 4cm at one end
sloping down to
about 1,5cm at the other.The enclosure has NO ventelation (designed to be ant
proof).
The area I have to work with is plenty , but I am having heat issues. Other
constraints are
low cost to make and assemble. Ambient up to 50deg.
The initial power supply I designed was a flyback supply using a uc3842
controller chip. The supply is
running well on the bench , but in the box , under full load is getting to hot.
It has not failed yet , but the
heatsink temp is around 90 deg. I am assuming the junction temp is probably
about 10deg above this.
Add another 30 deg for high ambient temps and I can't see this working for long
:0(
The switching fet has a small heatsink , and I am using a SM rectifieng diode ,
with the board as heatsink.
I'm using quite a large area of the pcb for the diode heatsink. Top and bottom
, with lots of via's to conduct heat
across layers. this seems to be working well.
I'm assuming the supply won't be more than say 80% efficient , so that is about
14W of heat to get rid of.
Quite a lot is a small plastic box with no ventelation...
I have then tried to make quasi resonant convertor using a ncp1207 from ON
semi. The circuit seems to be working well , but the switching fet seems to be
getting hotter than in my other design. Even at low load.I thought a resonant
convertor should be more efficient.I have sent some screenshots of the fet's
drain waveform to ABSE. Maybe someone can see from that what may be going
wrong. The waveforms look like they should as far as I can see.
Maybe I am expecting to much from this type (flyback) of supply. Would I be
better off using a forward converter
type of supply ?

Any tips would be appreciated.
Cheers
Rob

OK, what happens if you get the efficiency up to 90%?
You still got to get rid of the heat


Change the box to metal


martin

 

[Nico Coesel]

Hi to all.
Sorry if this is long winded...
I have been tasked with designing a supply 220Vac to 28Vdc , 2.5A (~70W).
The enclosure is plactic 90mm by 90mm with a height of about 4cm at one end
sloping down to
about 1,5cm at the other.The enclosure has NO ventelation (designed to be ant
proof).
The area I have to work with is plenty , but I am having heat issues. Other
constraints are
low cost to make and assemble. Ambient up to 50deg.
The initial power supply I designed was a flyback supply using a uc3842
controller chip. The supply is
running well on the bench , but in the box , under full load is getting to hot.
It has not failed yet , but the
heatsink temp is around 90 deg. I am assuming the junction temp is probably
about 10deg above this.

Did you look at the Topswitch devices from Power Integrations
(www.powerint.com)? You can choose a oversized device and limit the
current. Also synchronous rectification may be an option.

Add another 30 deg for high ambient temps and I can't see this working for long
:0(
The switching fet has a small heatsink , and I am using a SM rectifieng diode ,
with the board as heatsink.
I'm using quite a large area of the pcb for the diode heatsink. Top and bottom
, with lots of via's to conduct heat
across layers. this seems to be working well.
I'm assuming the supply won't be more than say 80% efficient , so that is about
14W of heat to get rid of.
Quite a lot is a small plastic box with no ventelation...
I have then tried to make quasi resonant convertor using a ncp1207 from ON
semi. The circuit seems to be working well , but the switching fet seems to be
getting hotter than in my other design. Even at low load.I thought a resonant

Maybe the mosfet isn't right. Choosing the right mosfet is a balance
between Rdson and the gate charge (capacitance).

 

[D from BC]

Hi to all.
Sorry if this is long winded...
I have been tasked with designing a supply 220Vac to 28Vdc , 2.5A (~70W).
The enclosure is plactic 90mm by 90mm with a height of about 4cm at one end
sloping down to
about 1,5cm at the other.The enclosure has NO ventelation (designed to be ant
proof).
The area I have to work with is plenty , but I am having heat issues. Other
constraints are
low cost to make and assemble. Ambient up to 50deg.
The initial power supply I designed was a flyback supply using a uc3842
controller chip. The supply is
running well on the bench , but in the box , under full load is getting to hot.
It has not failed yet , but the
heatsink temp is around 90 deg. I am assuming the junction temp is probably
about 10deg above this.
Add another 30 deg for high ambient temps and I can't see this working for long
:0(
The switching fet has a small heatsink , and I am using a SM rectifieng diode ,
with the board as heatsink.
I'm using quite a large area of the pcb for the diode heatsink. Top and bottom
, with lots of via's to conduct heat
across layers. this seems to be working well.
I'm assuming the supply won't be more than say 80% efficient , so that is about
14W of heat to get rid of.
Quite a lot is a small plastic box with no ventelation...
I have then tried to make quasi resonant convertor using a ncp1207 from ON
semi. The circuit seems to be working well , but the switching fet seems to be
getting hotter than in my other design. Even at low load.I thought a resonant
convertor should be more efficient.I have sent some screenshots of the fet's
drain waveform to ABSE. Maybe someone can see from that what may be going
wrong. The waveforms look like they should as far as I can see.
Maybe I am expecting to much from this type (flyback) of supply. Would I be
better off using a forward converter
type of supply ?

Any tips would be appreciated.
Cheers
Rob

I'm no smps expert and without reading into the post too
deeply...here's some bits...
For even cooler operation, use parallel mosfets. But, beware of drive
issues.
Also..
I've read that SiC power diodes can be easily be paralleled for cooler
operation.
Maybe check your magnetics.. I've heard saturation makes mosfets
toasty.

I'm not sure about this but I think if a mosfet does not get proper
heatsinking it will be more lossy.
The switching and conduction heat increase Rdson which causes a
thermal runaway until the mosfet stabilizes at some toasty
temperature.
Cool makes cooler and hot makes hotter..
I'll get around to doing the math on that someday..
D from BC

 

[Tim Williams]

Incidentially, what's up with all that nasty ringing?

As for heat, there's a page somewhere where the guy tested a, I think Dell
laptop power supply brick. There's no way the UL, CSA, etc. markings on the
thing were legitimate in any way. The parts were cooking pretty well at
half load; he was afraid to run it at full ratings!

Tim

 

[Tam/WB2TT]

Hi to all.
Sorry if this is long winded...
I have been tasked with designing a supply 220Vac to 28Vdc , 2.5A (~70W).
The enclosure is plactic 90mm by 90mm with a height of about 4cm at one
end
sloping down to
about 1,5cm at the other.The enclosure has NO ventelation (designed to be
ant
proof).
The area I have to work with is plenty , but I am having heat issues.
Other
constraints are
low cost to make and assemble. Ambient up to 50deg.
The initial power supply I designed was a flyback supply using a uc3842
controller chip. The supply is
running well on the bench , but in the box , under full load is getting to
hot.
It has not failed yet , but the
heatsink temp is around 90 deg. I am assuming the junction temp is
probably
about 10deg above this.
Add another 30 deg for high ambient temps and I can't see this working for
long
:0(
The switching fet has a small heatsink , and I am using a SM rectifieng
diode ,
with the board as heatsink.
I'm using quite a large area of the pcb for the diode heatsink. Top and
bottom
, with lots of via's to conduct heat
across layers. this seems to be working well.
I'm assuming the supply won't be more than say 80% efficient , so that is
about
14W of heat to get rid of.
Quite a lot is a small plastic box with no ventelation...
I have then tried to make quasi resonant convertor using a ncp1207 from ON
semi. The circuit seems to be working well , but the switching fet seems
to be
getting hotter than in my other design. Even at low load.I thought a
resonant
convertor should be more efficient.I have sent some screenshots of the
fet's
drain waveform to ABSE. Maybe someone can see from that what may be going
wrong. The waveforms look like they should as far as I can see.
Maybe I am expecting to much from this type (flyback) of supply. Would I
be
better off using a forward converter
type of supply ?

Any tips would be appreciated.
Cheers
Rob

I gather the heat sink is inside the box. I would put it outside, fastened
to the rear panel if you can't have ventilation holes. You would have to
electrically isolate the heat sink from the components.

Tam

 

[Martin Riddle]

Hi to all.
Sorry if this is long winded...
I have been tasked with designing a supply 220Vac to 28Vdc , 2.5A (~70W).
The enclosure is plactic 90mm by 90mm with a height of about 4cm at one
end
sloping down to
about 1,5cm at the other.The enclosure has NO ventelation (designed to be
ant
proof).
The area I have to work with is plenty , but I am having heat issues.
Other
constraints are
low cost to make and assemble. Ambient up to 50deg.
The initial power supply I designed was a flyback supply using a uc3842
controller chip. The supply is
running well on the bench , but in the box , under full load is getting to
hot.
It has not failed yet , but the
heatsink temp is around 90 deg. I am assuming the junction temp is
probably
about 10deg above this.
Add another 30 deg for high ambient temps and I can't see this working for
long
:0(
The switching fet has a small heatsink , and I am using a SM rectifieng
diode ,
with the board as heatsink.
I'm using quite a large area of the pcb for the diode heatsink. Top and
bottom
, with lots of via's to conduct heat
across layers. this seems to be working well.
I'm assuming the supply won't be more than say 80% efficient , so that is
about
14W of heat to get rid of.
Quite a lot is a small plastic box with no ventelation...
I have then tried to make quasi resonant convertor using a ncp1207 from ON
semi. The circuit seems to be working well , but the switching fet seems
to be
getting hotter than in my other design. Even at low load.I thought a
resonant
convertor should be more efficient.I have sent some screenshots of the
fet's
drain waveform to ABSE. Maybe someone can see from that what may be going
wrong. The waveforms look like they should as far as I can see.
Maybe I am expecting to much from this type (flyback) of supply. Would I
be
better off using a forward converter
type of supply ?

Any tips would be appreciated.
Cheers
Rob

Your efficency should be around 87% with the flyback, that leaves 10W in
losses.
What is the Real efficency that you measured?

Other than that I second the metal case.

Cheers

 

[[email protected]]

Your efficency should be around 87% with the flyback, that leaves 10W in
losses.
What is the Real efficency that you measured?

Other than that I second the metal case.

Cheers- Hide quoted text -

- Show quoted text -

Hi to all and thanks for the replies. I'll try and cover all the
replies.
I can't use a metal box - cost. A big bugbear of mine :0(
I didn't think flyback supplies could achieve efficiencies much over
about 80%. I'll try and do
some measurements and get a real figure.
The mosfet I am using (IRFBE30) is one we use elsewhere , that is why
I have to use that one. I think it is good for the
job though.It still does not answer the question why the QR convertor
is causing the fet to get hotter. The ncp1207 has
a 500mA output stage and the uc3842 is about half that!!

I can't add more diodes - cost. This eliminates parallel diodes and
also synchronous rectification :0(
I've checked the current in the transformer( its a flyback
arrangement , so it's not strictly a transformer) and it is
not going into saturation. Nice linear ramp under all loads.

I can't put the heatsink outside the box. This box fits snugly inside
another enclosure.
The QR convertor should be more efficient, but is not , any ideas.
Waveforms (light , medium and heavy loads) on ABSE

Cheers
Rob

 

[Ross Herbert]

Hi to all.
Sorry if this is long winded...
I have been tasked with designing a supply 220Vac to 28Vdc , 2.5A (~70W).
The enclosure is plactic 90mm by 90mm with a height of about 4cm at one end
sloping down to
about 1,5cm at the other.The enclosure has NO ventelation (designed to be ant
proof).
The area I have to work with is plenty , but I am having heat issues. Other
constraints are
low cost to make and assemble. Ambient up to 50deg.
The initial power supply I designed was a flyback supply using a uc3842
controller chip. The supply is
running well on the bench , but in the box , under full load is getting to hot.
It has not failed yet , but the
heatsink temp is around 90 deg. I am assuming the junction temp is probably
about 10deg above this.
Add another 30 deg for high ambient temps and I can't see this working for long
:0(
The switching fet has a small heatsink , and I am using a SM rectifieng diode ,
with the board as heatsink.
I'm using quite a large area of the pcb for the diode heatsink. Top and bottom
, with lots of via's to conduct heat
across layers. this seems to be working well.
I'm assuming the supply won't be more than say 80% efficient , so that is about
14W of heat to get rid of.
Quite a lot is a small plastic box with no ventelation...
I have then tried to make quasi resonant convertor using a ncp1207 from ON
semi. The circuit seems to be working well , but the switching fet seems to be
getting hotter than in my other design. Even at low load.I thought a resonant
convertor should be more efficient.I have sent some screenshots of the fet's
drain waveform to ABSE. Maybe someone can see from that what may be going
wrong. The waveforms look like they should as far as I can see.
Maybe I am expecting to much from this type (flyback) of supply. Would I be
better off using a forward converter
type of supply ?

Any tips would be appreciated.
Cheers
Rob

My feeling is that you are asking too much of a 70W SMPS inside a
totally enclosed space measuring 90 x 90 x 40 (and tapered).

I can only relate to my experience where I have practical experience
of damaged caused by the heat generated by a typical Astec 40W SMPS
module measuring 5" x 3" x 1-1/4" operating inside a dumb terminal
which has ventilation holes and a heck of a lot more air volume. After
several years of 24/7 service the plastic above the module takes on
quite a distinct shade of burnt brown and a gentle push eith a finger
causes the plastic to disintegrate. Repairing this type of damage is
nigh on impossible when it has not been caught early. A small fan
added to the enclosure certainly helps get rid of the heat and
prevents further damage.

Remember, I am talking only 40W (not 70W) and in a ventilated
enclosure of perhaps 50 times the volume of your enclosure using a
commercial SMPS. Even assuming it was not as efficient as your SMPS it
should still not have been generating heat sufficient to bake plastic.
around it.

 

[Mook Johnson]

I can't put the heatsink outside the box. This box fits snugly inside
another enclosure.
The QR convertor should be more efficient, but is not , any ideas.
Waveforms (light , medium and heavy loads) on ABSE

Cheers
Rob

You problem is getting the heat out of the box. When you say this box fits
inside another enclosure, is this enclosure metal? If not you're screwed.
What is the temperature rating of the plastic enclosure. I do high
temperature electronics design for downhole oil tools that run to 200C
ambient. I have designed power eupplis that will run for 5 years at 150C
ambient and 160C junction. In a lot of cases the datasheet ratings are very
conservative on max temperature.

If the outer enclosure is metal or a theramlly conductive material, you may
be in business. Glue an aluminum or copper plate on the bottom of your box
to act as a heat spreader, and thermally attach your hot components to it.
The heat will spread through the copper and better be able to conduct
through the plastic into the outer box.

Remember, in a sealed box you have no convection so all of your heat must be
removed by conduction.


Do you have heavy wires or connectors going in and out of the box. Use
those as heatsinks too.

 

[Paul Mathews]

Hi to all.
Sorry if this is long winded...
I have been tasked with designing a supply 220Vac to 28Vdc , 2.5A (~70W).
The enclosure is plactic 90mm by 90mm with a height of about 4cm at one end
sloping down to
about 1,5cm at the other.The enclosure has NO ventelation (designed to be ant
proof).
The area I have to work with is plenty , but I am having heat issues. Other
constraints are
low cost to make and assemble. Ambient up to 50deg.
The initial power supply I designed was a flyback supply using a uc3842
controller chip. The supply is
running well on the bench , but in the box , under full load is getting to hot.
It has not failed yet , but the
heatsink temp is around 90 deg. I am assuming the junction temp is probably
about 10deg above this.
Add another 30 deg for high ambient temps and I can't see this working for long
:0(
The switching fet has a small heatsink , and I am using a SM rectifieng diode ,
with the board as heatsink.
I'm using quite a large area of the pcb for the diode heatsink. Top and bottom
, with lots of via's to conduct heat
across layers. this seems to be working well.
I'm assuming the supply won't be more than say 80% efficient , so that is about
14W of heat to get rid of.
Quite a lot is a small plastic box with no ventelation...
I have then tried to make quasi resonant convertor using a ncp1207 from ON
semi. The circuit seems to be working well , but the switching fet seems to be
getting hotter than in my other design. Even at low load.I thought a resonant
convertor should be more efficient.I have sent some screenshots of the fet's
drain waveform to ABSE. Maybe someone can see from that what may be going
wrong. The waveforms look like they should as far as I can see.
Maybe I am expecting to much from this type (flyback) of supply. Would I be
better off using a forward converter
type of supply ?

Any tips would be appreciated.
CheersIf
Rob

If you want to understand the sources of dissipation, you need a good
current probe such as TCP202 for Tek scopes and a high voltage
differential probe. The diff probe goes across power dissipating
components, with the current probe simulataneously monitoring current.
If your scope allows you to do so, observe the product of current and
voltage. If not, use your imagination. Anyhow, you'll quickly find out
where the efficiency losses are occurring and whether anything can be
done to improve efficiency.
Often, the biggest efficiency losses in ACDC switchers with medium
output voltages are associated with the primary switch turn-off, but
you could be having problems with turn-on, as well. Often, changing to
a low gate-charge low Rds on MOSFET will work wonders. Slowly
recovering rectifiers can aggravate such problems. If you have low
voltage high current outputs, the output rectifier(s) can be the
source of the problem. Schottky rectifiers and/or synchronous
rectification can help you there. Proximity and skin effect losses in
the magnetics can also be a problem....high temperature of these
components will be your tell-tale there. If you have the interior
enclosure volume, you can consider operating at a lower frequency to
improve efficiency for all of the above. Sometimes, lowering the
switching frequency allows you to get rid of filter components to save
cost and create more room in the enclosure. The filter magnetics
dissipate heat, too.
Paul Mathews

 

[joseph2k]

I gather the heat sink is inside the box. I would put it outside, fastened
to the rear panel if you can't have ventilation holes. You would have to
electrically isolate the heat sink from the components.

Tam

And ground the heat sink if possible. Advantages in getting safety certs
that way.

 

[Ross Herbert]

Hi to all and thanks for the replies. I'll try and cover all the
replies.
I can't use a metal box - cost. A big bugbear of mine :0(
I didn't think flyback supplies could achieve efficiencies much over
about 80%. I'll try and do
some measurements and get a real figure.
The mosfet I am using (IRFBE30) is one we use elsewhere , that is why
I have to use that one. I think it is good for the
job though.It still does not answer the question why the QR convertor
is causing the fet to get hotter. The ncp1207 has
a 500mA output stage and the uc3842 is about half that!!

I can't add more diodes - cost. This eliminates parallel diodes and
also synchronous rectification :0(
I've checked the current in the transformer( its a flyback
arrangement , so it's not strictly a transformer) and it is
not going into saturation. Nice linear ramp under all loads.

I can't put the heatsink outside the box. This box fits snugly inside
another enclosure.
The QR convertor should be more efficient, but is not , any ideas.
Waveforms (light , medium and heavy loads) on ABSE

Cheers
Rob

It seems to me that you are being asked to come up with a design based
on unreasonable and impossible to satisfy demands based solely on
cost. SMPS design using readily available components (to keep cost
down) will have finite limitations on what is achievable insofar as
efficiency and reliability for a given volume. Unless you resort to
specialised techniques such as thick film modular construction, planar
transformers etc you just can't fit the required power into such a
small volume as you are required to. Naturally, you won't get these
items for low cost, so that is out of the question. You can't use a
metal enclosure because of cost so you are stuck. Tell the client that
you can do miracles but the impossible is a little more difficult.