Schottky diodes

[Raveninghorde]

I am working on a 5A buck converter. About 50% of the losses are due
to the schottky diode a 50WQ10FNPbF. So I am lookinig to optimize the
diode.

Looking at the specs I am looking to use a dual 15A part, Sanyo
SBT150-06J. I've chosen this because it has the lowest forward voltage
I can find.

The best efficiency comes from using the diode near it's maximum
junction temperature. Allowing for maximum operating temperature in
the case of 70C I reckon I'll get the forward drop down to around
0.25V. That allows for 80C rise so the TO220 package can be used
without a heaksink.

The datasheet doesn't give a thermal resistance for junction to air
but similar devices have figures of 55C/W.

In normal conditions with an internal case temperature of 40C and a
dissipation of 1.4W that gives a junction temperature of about 117C.

Am I missing anything?

 

[Joerg]

Synchronous rectification!

Just use a sync buck -> no diode. Plenty of driver chips for that
available these days.

 

[Raveninghorde]

Synchronous rectification!

John

OK this is a different project (this one is lower power) but when I
was fault finding a synchronous rectifier in a buck converter:

John Larkin - 18 Feb 2010

/quote

Synchronous rectifiers are often not worth the hassle at output
voltages and currents like this. Try shunting Q3 with a schottky, or
replacing it entirely.

John

/end quote

 

[Hammy]

m:

John Larkin is right - you really do need to think about
synchronous rectification. A decent-sized power MOSFET should
offer a much lower forward voltage drop than 0.25V.

That would be 50 milliohm of channel resistance. You can do an
order of magnitude less than that in a TO-220 packaged MOSFET for
a couple of dollars, if the Farnell catalogue is anything to go
by. They come with a big gate and lots of gate capacitance to
drive, but it would solve the powr dissipation problem.

In order for synchronous rectification to be worth it you need a FET
with about 10 to 15 moHm RDSON @25C preferably 10moHm or less.The fet
conduction losses are the RMS current its subject to not the 5A output
current then their is switching losses and losses associated with the
drive circuit.

Considering his diode is a 100V device FETS with 10-15 moHm RDS are
slim to non-existant with 100Vds. The closest I have is a panasonic
device with Vds of 80V IN lfpak.

http://www.nxp.com/documents/data_sheet/PSMN8R2-80YS.pdf

The only thing about a low Vf schottky is they usually have higher
reverse current; depending on the blocking voltage it can add to the
losses.

 

[Hammy]

@r29g2000yqj.googlegroups.
co m:


In order for synchronous rectification to be worth it you need a
FET with about 10 to 15 moHm RDSON @25C preferably 10moHm or
less.The fet conduction losses are the RMS current its subject to
not the 5A output current then their is switching losses and
losses associated with the drive circuit.

Considering his diode is a 100V device FETS with 10-15 moHm RDS
are slim to non-existant with 100Vds. The closest I have is a
panasonic device with Vds of 80V IN lfpak.

http://www.nxp.com/documents/data_sheet/PSMN8R2-80YS.pdf

The only thing about a low Vf schottky is they usually have higher
reverse current; depending on the blocking voltage it can add to
the losses.

Heres how to estimate diode losses from vishay.

Pd = Forward power loss = IF(AV) x VFM at (IF(AV)/D)
PdREV = Inverse power loss = VR1 x IR (1 - D); IR at VR1 = 80 %
rated VR

So with D=0.5 and a reverse current of 8mA and blocking voltage of
80V the reverse losses are about 0.32W alone.Using typical values
from your Sanyo part.

 

[Raveninghorde]

If you are accusing me of saying that synchronous rectifiers make
sense in some situations, and don't make sense in others, consider me
guilty.

John

It wasn't an accusation. I was just pointing out I had been following
your good advice

Last time it was 12V/8A or 24V/5A switched and it was hassle to get
the sychronous rectifier working properly with a chip designed for the
job.

This time it is 8.4V/5A and I chose a simple switcher, LM22677. All
works well except for the diode dissipation. It is a bit late to
change to synchronous which is why I am looking at optimizing the
diode.

I posted because running a part stinking hot to achieve efficiency
goes against my basic instincts to keep everything as cool as
possible.

 

[Raveninghorde]

What's your input voltage? Assuming enough output inductance, the
diode dissipation is about Io*Vf*duty cycle, which shouldn't amount to
a lot of efficiency loss at 42 watts out. At 50% duty cycle and 0.3
volts drop, that's only 0.75 watts in the diode. Maybe you just need a
little more heat sink. Or let it get hot.

A TO-220 can be heat sunk to a bit of copper pour and dissipate a watt
easily. I posted on a similar situation recently. You can dump heat
into connected pours, or transfer heat through the pcb laminate into a
bigger plane, like the ground plane.

Here's 1 watt, cooled by conduction through 25 mils of FR4 to the
copper on the opposite side, electrically isolated:

ftp://jjlarkin.lmi.net/T750_t1.JPG

ftp://jjlarkin.lmi.net/T750_t3.jpg

ftp://jjlarkin.lmi.net/T750_t5.jpg

On the final layout, we had the big ground plane on layer 2, and added
added a heat spreader on layer 4 to better reduce the hot-spot effect
close in. You shouldn't have to work very hard to dump 3/4 of a watt.

There are also cheap u-shaped heat sinks for use with TO-220s on a pc
board.

John

The max in is 32V, nominal 24V. I'm using more inductance than I need
at 8.4V, 33uH, because the unit can do 8.4V/5A or 16.8V/2.5A.

National web bench suggests a schottky with 0.7V Vf which gives a
calculated Pd of 2.5W. Don't tell me web bench is crap, I know but it
is a quick tool to get to the starting point. It calculates the SMB
package diode Tj as 30C at 2.5W!

As I originally posted a TO220 in air should be fine. I'm allowing for
a small heat sink but don't think I'll need it. I prefer not to dump
the heat into the board as the IC and inductor will be close by and
putting heat in as well.

 

[Raveninghorde]

Just use a sync buck -> no diode. Plenty of driver chips for that
available these days.

I couldn't get the chosen schottky in quantity and all the
alternatives I could find didn't have low enough Vf with low enough
reverse leakage at high temperature.

So I had to redesign for a synchronous rectifier and used the LM3150
which gave me problems on an earlier design. The redesign works
excellently using a pair of Infineon BSC093N04LS power SO8 FETS with a
schottky across the lower FET. The efficiency of the converter exceeds
95%. I'm loosing more power in protection FETS, fuse, and input
reverse protection.

I've always used 2 layer boards but over the last few months moved to
4 layer boards for power conversion. It has dramatically reduced
design time as things just work. Top layer, 2oz, and 1 internal layer
for tracking, an internal layer for power planes and the bottom layer,
2oz, for ground and power dissipation.

The overall efficiency is >90% compared to <85% with the earlier
design.

 

[Raveninghorde]

Be careful if you do that. There's a short non-shoot-through period
when both fets are off, and some body diode is furnishing the load
current. Some of these diode have delusions of being SRDs.

The results can be impressive. Imagine every opamp on the board, even
ones far away, suddenly developing 100 millivolts of input offset.

LM3102:

ftp://jjlarkin.lmi.net/SwitcherRise.JPG


John

I fitted a 1A schottky across the synchronous FET. That handles the
current until the FET switches. No sign of overshoot.

But you just reminded me of something. I allowed for a snubber network
of 33R and 1nF, a hangover form the earlier circuit, and I haven't
checked performance with that disconnected.