Just how hot can things get?

[itsmoked]

Just designed a switching charger.

Input: 24VDC
Output: 5-30VDC
Output: 10A

My normal designs nurse nanowatts out of coin-cells and run for years on a battery. This thing funnels 300W thru FETs and blocking diodes etc with 28A thru an inductor.

Anyway, steady state conditions are shaping up to parts on the board under convection cooling hitting 80C. Nothing seems too mad about it but I am not used to feeling heat radiating off my happily running boards.

The hottest part is the blocking diode. It has the worst forward drop and hence the highest dissipation 80C. The part can run the junction @ 150C so I am well below that. I am not worried about torching anything, I am more worried about longevity. On the one hand this is a charger application so I expect maybe 24Hrs charging cycle. Initial time charging some dead battery, running the full 10A, and then tapering later to eventually full cutoff. Then not charging again except to make up again from an engine start every 2 weeks.

My question is where do you guys consider power electronics to be running too hot for reasonable life expectancy? A lot of off-line switchers have to do this endlessly when do you guys start reaching for fans and heat sink forests?

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[VEBill]

If the parts are 80C in a (presumably) 20C environment, then what would they be in a 40C or 50C environment? Ouch.

 

[MikeHalloran]

I have seen diodes run hot enough to desolder themselves.

Not by intent, of course.

Cure was 1cm square brass fins soldered to extra long leads between the diode and the pcb.





Mike Halloran
Pembroke Pines, FL, USA

 

[ScottyUK]

Running a junction at 150C is inviting problems - why push a component to the edge of the envelope when there is almost certainly a similar component which can do the same job cooler, unless this is really cost-critical and longevity doesn't matter? Can you replace the diode with a TO-220 or TO-247 style and bolt it to a heatsink? Better still at the voltage and current you're looking at you might be able to cut your losses in half by substituting a Shottky diode.


----------------------------------

Sometimes I only open my mouth to swap feet...

 

[logbook]

Keith,
if you are saying the case temperature is 80degC but that is ok because the junction temperature is allowed to be 150degC you are missing something. The junction temperature is the case tempertaure PLUS the power dissipation multiplied by the thermal resistance from junction to case (theta j-c).

Suppose 20W dissipation and theta J-C = 5degC/W. That makes the junction 80+ 20*5 = 180degC.

 

[OperaHouse]

I would consider the environment the product is going into. Is the box going to become a shelf for stacking papers or boxes on. Will the environment have a lot of dust that will coat components and thermally insulate them? Will a fan filter be maintained and will the fan just bring more dust into the cabinet. I've seen a lot of products where they stacked the electrolytics right next to hot components, a bad thing in itself, and block airflow. Location can easily lower temperature several degrees at no cost.

 

[MacGyverS2000]

Parallel diodes to cut back on heating each one so much? Just a thought...

Dan - Owner

http://www.Hi-TecDesigns.com

 

[richs]

Hiya-

I too vote for a cooler diode! A blocking diode failure can lead to catistrophic failure down in the design.

As mentioned earlier, the junction temp will be much higher than the case temperature and I'm glad that logbook typed in the equasion, rather than me having to go back and look in my notes.

I can envision the supply "letting go", the batteries getting way over charged and going "pop" at an inconvenient time.

Your call of course. Don't know how closely "tied" to the product you are and what sort of liability you want to assume. Even if the client REALLY wanted to pinch pennies, I'd sure make the "well the cost of repair" arguement to them. For me, I'd be irritated and somewhat nervoue with anything really greater than 40C.

Hope this helps!

Cheers,

Rich S.

 

[BobbyNewmark]

What logbook says is gospel.

However "case temperature" is a slippery concept with surface mount semiconductors, where each terminal has a different thermal resistance and heat sink. Either be very conservative, or use the junction itself as a thermometer (which requires a calibration effort that is likely to outweigh the benefits for a small volume design). Unless the design is very price sensitive, I'd just throw big strong devices at the problem.

Electrical and chemical degradation rates tend to about double for every 10 [°]C increase in temperature, and T[sub]j,max[/sub] is a guarantee of functionality not reliability. Staying toward the cool end of that exponential curve does wonderful things for reliability.

If battery overcharge is expensive and/or hazardous, I'd strongly recommend an overvoltage crowbar on the output. Just give it good RF filters so it won't get set off accidentally.

You'll probably also want a fuse or mechanical circuit breaker too, to protect the charger from the battery and vice versa.

A rectifier on the output might be a good idea too, if anybody might ever plug the battery in backwards.

 

[MikeHalloran]

Yep, logbook is right, you have to _compute_ the junction temperature, and design accordingly.

One other thing about the land of high currents; you can't use the usual vapor- deposited foil on your circuit boards. You must use heavy, heavy foil, and wide traces. At 10A, every wire and every trace is a resistor.





Mike Halloran
Pembroke Pines, FL, USA

 

[itsmoked]

Well thanks for the good input everyone!

This is an SMD design. All the parts are SMT mounted onto relatively large copper planes.

These parts are all D2PAK or TO-236(same thing) essentially a TO-220 with chopped, formed leads for SMD.

A large effort was made to maximize the thermal planes, with dozens of large vias connecting generally the bottom board plane and an inner plane to the top component mounting plane.

VE1BLL: Yes! A serious concern. That 80C is based on a lab temperature of 21C so we are talking a 59C rise. A 50C ambient equates to 50+59 = 109C (Tab temp)

Mike: I haven't seen that but I have often seen brown ditches under resistors that have run hot and long. Occasional charring too.
I wonder if there are heat-sinks you can mount between an SMD and the board to augment the thermal aspects... Never seen one - not sure how the other two pins would like the interjected angle a heat-sink (there) would cause.

Scotty: Already thought of that! It's a dual 10A Shottky to start with. I am seriously considering a TO-220 on a heat-sink.

logbook: (theta j-c) = 2.2C/W worst case as this is for one diode and there are two in parallel in this package.

http://www.st.com/stonline/books/pdf/docs/8001.pdf

P = 0.5V x 10.1A
P = 5.05W
So presumably we have Junc-temp = 5.05W x 2.2C/W + 80C(measured case)
JunctionTemp = 11.11C + 80C = 91.1C

This is 40C below 150C (150C being somewhere we don't want to ever visit..) Wherein lies my question; What's reasonable for not turning boards funny colors and having scary smells occur, and having the system be thermally 'tender'?


OperaH: Classic industrial enclosure. Sadly at the top. Think Hoffman box - no ventilation. CPU controlled fan is being considered. NO GROANING!!! STOP THAT!! This isn't that bad an app for a fan since it would only be used as needed. Remember this is a charger that may not run at all for 95% of the time and when it is running may be dishing out only an amp or two to top up the batteries occasionally, and then a larger load long enough to replace an occasional engine startup debt.

I ran an hour test yesterday. Board horizontal took all temps after S-S reached. Then I stood the board on edge in final mounting profile expecting much better cooling due to both side's convection improvement. I saw about a 2C drop in most locations but surprisingly I saw 2C increases in two places (one being the dang blocking diode!). So I am seeing exactly what you're mentioning. The output cap bank is probably blocking the diode's wind.

Dan: They're already parallel diodes in one case. Which probably isn't what you're thinking.
They do mention that these two diodes can experience thermal runaway. I don't recall, do diodes have a negative tempco on their forward voltage?? That might mean I don't actually have two diodes only one with a fellow passive observer.

Hi Rich: I was having those qualms but you are in the same field as I and most our designs have no detectable heat rises anywhere so having a board covered with parts that are actually 'radiating' makes me queasy - but think about it. Your 40C is only a cold hot tub! I could spend a lot of money on hundreds of boards to battle for that. If it really isn't required what is required? That! Is the question..

Bobby: Welcome to Eng-Tips.
Yep I hear you on the 10C double the reaction rates my Dad impressed that on me long ago. I have fusing and multiple feedback paths including fold-back protection so I'm not concerned with blitzing the batteries - especially since they are large diesel starting batteries with the ability to actually dissipate 150W all day long in some twisted error mode. The blocking diode I'm referring to is the diode to prevent back-feeding the charger or reversing the battery leads.

The job is somewhat price sensitive.

Mike: See above logbook. This is a 2oz layout and most traces are over 1/2" wide usually on both sides.

I'm hoping someone who makes switching supplies or? to say, "we run our 100,000hr MTBF switcher diodes at xxC all the time". That would give us a serious clue as to what the reliability, board neighbor components, and board material can tolerate.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[IRstuff]

Keith,

Often, the copper within the board is used as heat spreader, particularly with lots of thermal vias.

The reliability question is, of course, compounded by the fact that most people don't run their systems at max ambient at all times. Ideally, you need to come up with an operational profile for percentage of time at max ambient and calculate according. However, without something like Relex, accuracy is dubious, although it's dubious even with Relex. MIL-HDBK-217, is extremely conservative for complex circuits, but might be adequate for discrete devices

TTFN

Eng-Tips Policies FAQ731-376

 

[davidbeach]

Keith, you can RF your initial post and ask to have the title edited. I've done it many times, and not always just for my own posts. Site management seems happy to do the editing.

 

[MikeHalloran]

Wow! Surface mount! I wouldn't have imagined it possible.

The last SMD part with which I became familiar was a pass transistor for a servoamp in a printer carriage drive. I became familiar because I was using the printer heavily, and every once in a while, the solder under the transistor would crack, and the transistor would fall off. It was running at 1.25 A.

That was >10 years ago. Maybe SMD is better now.



Mike Halloran
Pembroke Pines, FL, USA

 

[VEBill]

You might consider running one, or a small batch, through some burn-in tests in an oven set to maybe +40C. If it survives, repeat at +50C. To find the weak spot, ramp it up slowly from there until something happens.

It's not an unresonable test since many places really are +40C in the shade.

 

[IRstuff]

+40ºC would be a DREAM environment. Gad, you guys are SO LUCKY!!! We run our reliability predictions for a constant 35ºC ambient. Our latest specs call for +60ºC ambient, which is at the outside of the enclosure, putting the interior ambient at a minimum of +70ºC.


TTFN

Eng-Tips Policies FAQ731-376

 

[electricuwe]

Itsmoked,

since you are using Schottkys you should be aware of thermal runaway due to reverse current:

a Schottky diode operated close to it's maximum rated juction temperature might have a reverse current of several 10 mA, so blocking losses beeing really signicficant

In this case your design will not only have 29 K higher junction temperature when moved from 21°C ambient to 50 °C ambient, it might exceed maximum juction temperature within a few minutues or exibit an increase in junction temperature significantly higher than the increase in ambient temperature.

 

[logbook]

Well you have 10A output from 10A diodes. Not good. I know you have two in parallel but they don't parallel perfectly. You are running them far too hard, regardless of the temperature. It is the forward voltage that is too high. If you use 20A diodes everything should calm down. The power dissipated will be less because the volt drop will be less.

Have you worked out the peak current in the pair due to the duty cycle? I would rate each diode above this peak to make a robust design. ( I have several thousand switchers out in the field to my design.)

 

[itsmoked]

Thanks log and eletric.

I've had problems with Schottkys used for protection on analog INs leaking like a sieve back into the 'protected' inputs. These big diodes do in fact leak horribly at elevated temps, (130mA), I see what you mean about extra heating due to that alone.

I'll check out some other diode options.

<later>
After a couple hours.. I can't even find any non-Schottkys. Sheesh. I'll continue.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[MacGyverS2000]

When forward biased, heating and positive tempco effects can cause current bunching, accelerating failure. High temps/currents over a period of years will affect the impurity diffusion in the device structure and destroy the diode's junction properties.

You may be hard-pressed to find 20A SMD diodes, though without looking myself I probably shouldn't say that.

Dan - Owner

http://www.Hi-TecDesigns.com