Does having high thermal resistance in a diode SMT package reduce efficiency?

Thread Starter

SiCEngineer

Joined May 22, 2019
442
Hi All,

Having an almost philosophical thought regarding poor thermal dissipation of surface mounted Silicon carbide diodes with thermal pads.
I tested my resonant converter and expected >90% efficiency but only get around 70% at full-load. I put a FLIR camera onto the board and noticed that the Silicon carbide diodes heat up to around 58 degrees celcius, with 25 degrees ambient temperature. The only other thing that gets remotely warm is the resonant inductors that are from Coilcraft, so this led me to believe that the majority of the power is being lost in the diodes, due to poor thermal resistance or how I layed out the thermal structure of the board.

My question is, does having high thermal resistance due to poor layout increase power dissipation? I would think it would, but it could be that the diodes are dissipating the exact same amount of power, but because the thermal resistance is higher, the diodes are getting hotter with the same level of actual power dissipation. To illustrate, let's say that a diode dissipates 20W of power.

Thermal resistance = 1.5C/W - would heat the device up 30 degrees C.
Thermal resistance = 3C/W - would heat the device up 60 degrees C

Even though both devices dissipate 20W, one gets much hotter because of the increase thermal impedance from the case to the junction and ambient. In this case, the efficiency of the converter is not affected at all by the increased thermal impedance. But I expect this is not the case and instead the poor connection I have made with the thermal layout, has caused the majority of the dissipation, because I cannot for the life of me see where else the power could be lost within the circuit. The inverter switches are below 30 degrees, and no other component on the board really gets hotter than 35 degrees Celcius. Diodes approach 60 degrees.

TIA for any answers!
 

crutschow

Joined Mar 14, 2008
34,280
No, the thermal resistance doesn't change the efficiency as long as the same power is being dissipated.
Only if he dissipation goes up with temperature, would the efficiency be affected.
Since the diodes have a negative temperature coefficient, their forward drop goes down slightly with temperature, so their efficiency would actually improve.

So your drop in efficiency with load is likely caused by an increase is dissipation from the other components (the switching loses and the inductor resistance are likely suspects), not the diodes.
 
Last edited:

Thread Starter

SiCEngineer

Joined May 22, 2019
442
No, the thermal resistance doesn't change the efficiency as long as the same power is being dissipated.
Only if he dissipation goes up with temperature, would the efficiency be affected.
Since the diodes have a negative temperature coefficient, their forward drop actually goes down slightly with temperature, so their efficiency would actually improve.

So your drop in efficiency with load is likely caused by an increase is dissipation from the other components (the switching loses and the inductor resistance are likely suspects), not the diodes.
I don’t really have a way to see how much power is being dissipated in the diodes, though. I know they are getting hot but because the actual thermal resistance will differ from the datasheet, I cannot quantify it.
I also suspect the inductors. I use Coilcraft inductors but unfortunately there is no data on their core material and losses at my switching frequency of 200-400kHz. But they do not seem to get excessively hot, which is what I would suspect if they were dissipating large majority of the converter losses. I may be better designing a custom inductor at these frequencies and testing the efficiency with that. I know for sure the transformer is not Litz wire wound, but if my concerns were true that the losses in the core/winding are due to high frequency effects, I’d expect to see a large increase in the conducted heat from them. The GaN switches do not rise much above ambient even at 400kHz frequency, but they do have heat sinks attached to them which probably helps a bunch.
 

Papabravo

Joined Feb 24, 2006
21,158
I don’t really have a way to see how much power is being dissipated in the diodes, though. I know they are getting hot but because the actual thermal resistance will differ from the datasheet, I cannot quantify it.
I also suspect the inductors. I use Coilcraft inductors but unfortunately there is no data on their core material and losses at my switching frequency of 200-400kHz. But they do not seem to get excessively hot, which is what I would suspect if they were dissipating large majority of the converter losses. I may be better designing a custom inductor at these frequencies and testing the efficiency with that. I know for sure the transformer is not Litz wire wound, but if my concerns were true that the losses in the core/winding are due to high frequency effects, I’d expect to see a large increase in the conducted heat from them. The GaN switches do not rise much above ambient even at 400kHz frequency, but they do have heat sinks attached to them which probably helps a bunch.
There are several things you can do that may not have occurred to you yet.
  1. You can measure the voltage drop across the diode. You know there is a relationship between the forward voltage drop and the current.
  2. You can measure the temperature to see if the rise over ambient is consistent with the average current in the diode.
  3. Based on your knowledge of the output power and the switching waveforms you should be able to come up with reasonable values for the diode current.
That said, if you find that things are still not to your liking you can search for a diode with a lower voltage drop at the desired current or superior revers recovery characteristics.

I can't tell you where the largest "bang for the buck" will come from, but that is what you should be considering. Maybe you can offer up a schematic and we can help you identify potential problems.
 

michael8

Joined Jan 11, 2015
410
I don’t really have a way to see how much power is being dissipated in the diodes,

Could you run power through the diodes (DC) with all the other power off? Measure the current & voltage
and increase the current until the diodes reach the same temperature...
 

Thread Starter

SiCEngineer

Joined May 22, 2019
442
Do you have access to a differential probe and a current probe?
This last item should be of the Hall effect variety.
I do have a differential probe, that is 200MHz and 750VDC. I study a PhD at university so I assume there will be a Hall Effect current sensor somewhere, I'll have a look around now. I suppose this is for testing the inductors rather than the Diodes? It would be good to know how they are performing, as I believe I will need to design a custom one with better losses at higher frequency that the converter will switch at.
 

Thread Starter

SiCEngineer

Joined May 22, 2019
442
There are several things you can do that may not have occurred to you yet.
  1. You can measure the voltage drop across the diode. You know there is a relationship between the forward voltage drop and the current.
  2. You can measure the temperature to see if the rise over ambient is consistent with the average current in the diode.
  3. Based on your knowledge of the output power and the switching waveforms you should be able to come up with reasonable values for the diode current.
That said, if you find that things are still not to your liking you can search for a diode with a lower voltage drop at the desired current or superior revers recovery characteristics.

I can't tell you where the largest "bang for the buck" will come from, but that is what you should be considering. Maybe you can offer up a schematic and we can help you identify potential problems.
This got me thinking. I am using Silicon carbide diodes that have a larger voltage drop, typically, than Silicon diodes. I am also only testing at a low voltage because my final design should be for 6kV, and my supervisors wanted me to design a low voltage board to test the switching waveforms and layout first. Therefore the efficiency drop due to these diodes at low voltage should be much larger than I'd expect when switching at my full voltage. There may be larger conduction losses too because the parasitic capacitance in parallel with these diodes typically decreases significantly with applied DC bias. Thanks for the heads up on that one!

My issue with measuring the temperature, is that I am concerned that the high temperature is more to do with my poor thermal layout than the diodes dissipating more power than they should. The output current of this test board is about 1.5A, but the diodes are I believe 5A rated, so I would not expect such a high increase in the temperature of the junction/case and to ambient. So what I mean to say is that it is hard to discern whether the diodes are dissipating too much power, causing an increase in temperature, or whether the poor thermal resistance is causing high diode temperature even if power dissipation is low. Kind of a chicken and the egg problem!

Thanks to everyone for their input thus far. Appreciated.
 

Papabravo

Joined Feb 24, 2006
21,158
This got me thinking. I am using Silicon carbide diodes that have a larger voltage drop, typically, than Silicon diodes. I am also only testing at a low voltage because my final design should be for 6kV, and my supervisors wanted me to design a low voltage board to test the switching waveforms and layout first. Therefore the efficiency drop due to these diodes at low voltage should be much larger than I'd expect when switching at my full voltage. There may be larger conduction losses too because the parasitic capacitance in parallel with these diodes typically decreases significantly with applied DC bias. Thanks for the heads up on that one!

My issue with measuring the temperature, is that I am concerned that the high temperature is more to do with my poor thermal layout than the diodes dissipating more power than they should. The output current of this test board is about 1.5A, but the diodes are I believe 5A rated, so I would not expect such a high increase in the temperature of the junction/case and to ambient. So what I mean to say is that it is hard to discern whether the diodes are dissipating too much power, causing an increase in temperature, or whether the poor thermal resistance is causing high diode temperature even if power dissipation is low. Kind of a chicken and the egg problem!

Thanks to everyone for their input thus far. Appreciated.
The diodes may have more than the output current going through them (and the inductor) because the peak current will be higher than the average current and depending on the the duty cycle and the configuration, that peak current may be much higher. If the diode is on the wrong side of the duty cycle (eg conducts for more than 50% of the switching period) then they will tend to heat up fast and cool down slow. Most SMPS designs use Schottky diodes for this reason. They have a lower forward voltage drop for the same current and they have little to no reverse recovery time because the junction is not a p-n junction but a metal to semiconductor junction. May be time to crank up some simulations if you can't get the measurements you need.
 
Top