No heatsink needed for secondary diode of 33W offline flyback?

Hello,
We have an idea to do an offline 33W CCM Flyback SMPS, without needing a heatsink to be assembled onto the secondary diode. This is achieved by paralleling three TO220, 600V SiC diodes, CSD01060 (which have positive temperature coefficient).

Is this a good idea? Why does nobody seem to do it?

The saves the cost of somebody having to assemble a heatsink on to the secondary diode.

The Flyback spec:

• VIN = 100-265VAC
• VOUT = 25V
• POUT = 33W
• F(sw) = 100kHz

CSD01060 SiC diode datasheet:

LTspice sim and pdf schem attached.

A diode has a small amount of ohmic resistance. It does not become prominent until you push diode conduction high enough.

So perhaps you're distributing resistance partly via the PN junction, and partly via ohmic? And by paralleling 3 of them, the ohmic resistances turn out to be more or less equal, thus acting as load balancing resistors?

What duty cycle are you running.

- - - Updated - - -

My 2 Cents worth.

33w/25v = 1.32A

1.6vf x 1.32A = 2.112W

Temperature Rise Over ambient for Sic diode.
2.112W x 60 deg C/w junction to ambient = 127 Deg C, with no heat sink.

2.112W x 7 deg C/w + PCB as heat sink = Something less than 127 Deg C.

How much less? usually after 1 SQ inch of pcb heat sink you get diminishing returns. 1 oz copper 1 sq inch about right, and 2 oz copper 1.25 sq inches or so. Vias can also be used to pull heat to the bottom or inner layers.

It looks like duty cycle could add as much as 10 to 15 Deg C also.

From the Cree secret pdf, no longer available.

CPWRAN01A.pdf

"temperature coefficient of forward current allows us to parallel
more than one die in a package, or many in a circuit, without
any unequal current-sharing issues."

"The devices were packaged in plastic TO-220 packages.
These parts are rated for a maximum junction temperature
of 175°C. For a case temperature of up to 150°C, the
junction temperature remains below 175°C at full rated
current."

From some casual searching it looks like 150 Deg C is the maximum junction temperature you should go for in a Sic diode. So you may just need one package of a Sic diode with pcb heat sinking. It will depend on how enclosed your product is.

However a standard Schottky looks like a winner to me. I picked the cheapest one to make a point. This is not necessarily the best choice.

Standard schottky 3A 200V
CDBB3200-HF
15.5 Cents in quantity
No graph for duty
Vf under .8 V


.8V x 1.32A = 1.056 W, half of Sic diode.

Temperature Rise Over ambient.
1.056 W x 55 deg C/w junction to ambient = 58 Deg C, with no heat sink.
Compared to 127 Deg C for Sic diode.

You should probably not exceed 125 Deg C junction for regular Schottky.

- - - Updated - - -

Oops, with the 25 vdc out of the flyback, the 200 vdc Schottky is probably to low a voltage rating. Now i see why you went with the Sic diode.

Perhaps it will be wise to bypass the diodes with a small value capacitor- just to be on the safe side. However, I have not seen this in practice though.

- - - Updated - - -

BradtheRad said:

...the ohmic resistances turn out to be more or less equal, thus acting as load balancing resistors?

Click to expand...

Many people ignore this important aspect: there is no guarantee that the diodes will share the current equally because even a small difference can have a rather large effect. If you add external resistors, you kill the efficiency; if you don't, then you may be taking some expensive chance. And it may become worse after the service when a failed diode is replaced. Load balancing is *very* important!!!

Hi,

Load balancing:
With usual Si diodes the pn voltage drops with rising temperature. This makes it more conductive...and in case of paralleled diodes the hottest diode carries the most current. The current obviously getting out of balance.

I don't know how SiC diodes behave. Check this.

Klaus

Thanks Guys, I know that you all appreciate that the "current hogging" problem doesn't happen for "SiC" diodes in parallel, because they have a "Positive temperature coefficient". -They automatically share the current equally between each other when paralleled.

- - - Updated - - -

What duty cycle are you running.

Click to expand...

For the design in the top post the duty cycle over 85-265VAC is 0.3 to 0.1
-it was deliberately made to be this so as to reduce secondary rms and peak current.

However thankyou for the above 200V single schottky suggestion, if using this then I would change the flyback txfmr turns ratio so as to allow less reflected voltage to the secondary diode, so I will look into that.

- - - Updated - - -

That is a great idea to use just one standard 200V schottky.
Some of the high current, 200V schottkys have superbly low Vf’s, and initially appear great to use, but then you notice they actually have a significant reverse recovery time, or a really high junction capacitance. –So you then try and get round this by just using DCM, but then you notice that the high current 200V schottkys have a really high reverse leakage current, which really increases losses. So we can’t use the really high current , low Vf, 200V schottkys at all without a heatsink…..

Therefore, I then went for a VT5202 200V, 5A Schottky in TO220. This appears to have no reverse recovery time, so I actually opted to use it in CCM..-but then I changed my mind and thought what the heck I’ll use it in DCM anyway because that gets rid of the RHPZ.
So I reckon the VT5202 200V Schottky would dissipate 0.65V*1.3A = 0.845W
To be honest, 0.845W is a bit too much power for a TO220 with no heatsink.
The 52 degC per Watt referred to in the VT5202 datasheet refers to “Free Air”, and of course, the air inside an enclosed case isn’t “Free Air”.

Therefore, I reckon that the multiple paralleled SiC Schottky diodes is the only way to actually avoid the need for a heatsink in this Flyback SMPS.

Did you consider using the PCB as a heat sink?

yes but as you know, that's always limited, and most places don't really allow enough room for much heatsink copper on the PCB.

treez said:

the "current hogging" problem doesn't happen for "SiC" diodes in parallel, because they have a "Positive temperature coefficient"

Click to expand...

Perhaps this feature would be not fully applicable when dealing with higher switching frequencies. If you drive the diode with a continuous current this shouldn't mean any issue, but for pulsed mode, the time spent to heat the substract, although having a small surface on silicon ( meaning a fast temperature rising ), this somehow could retard its PTC effect, particularly because you specified 100kHz.

thanks, but if the high frequency stops the diodes from heating up, then that actually solves the problem.

Well i think he meant heating up evenly because they would not have time to stabilize. But if i wanted to parallel diodes, i would get a package were there were two diodes inside, so they had to share the heat. I am sure there are many of these for the 200V Schottkys but I have not looked for Sic's.

BTW, just using a surface mount part with minimal copper area is still much better than ambient.

- - - Updated - - -

Look at these low Rth in this small package.

Three TO220 packages is enormous overkill for a 33W converter. The output diode should only need to dissipate around 1W.

But thermal matters can't really be discussed without knowing how the supply will be packaged and sited.

There is 1.3A average going through the diode, and that gets you up near the 1 Watt mark for dissipation (with It being a 200v diode).
That's really too much for an SMC diode, so you are talking dpak or to220.....both same price and size, so pick the better thermally capable to220.
The PSU will be of the enclosed adapter type, like a laptop PSU.

treez said:

That's really too much for an SMC diode, so you are talking dpak or to220.....both same price and size, so pick the better thermally capable to220.

Click to expand...

This depends completely on the thermal path to ambient. An SMC on a good ground pour will perform better than a TO220 without a heatsink in still air.

The PSU will be of the enclosed adapter type, like a laptop PSU.

Click to expand...

Are you going to pot everything? Will the PCB be thermally bonded to the enclosure wall?

I appreciate that but there will not be room for a big ground pour.
We won't pot due to cost reasons. Nor will we thermally couple to enclosure, in fact,
enclosure will be plastic.