MOSFET Power Loss calculation ?

Thread Starter

AhmedKhouaja

Joined Nov 2, 2019
20
Dear All About Circuits Team,

I hope everyone is doing well,

I have a question about the MOSFET's behaviour, How can I figure out how much heat the Mosfet produces per amp and how much current it can consume when acting as a load?

Below, the Mosfet application shown in the figure, it's an electronic load configuration.

Any suggestions ?
4804-2-FIG05.jpg
 

Janis59

Joined Aug 21, 2017
1,834
You answered by Your own to this question. Take the current through the mosfet, multiply the voltage on it and thats it!. This is the true reason why in high power circuits all are prone to use mosfets in full off/ full on regimes exclusively, aka SMPS.
P.S. If the frequency would be a high (not a DC as here) then add the gate power loss what may be higher (sometimes) than in Drain.
 
Last edited:

crutschow

Joined Mar 14, 2008
34,285
The circuit generates a constant-current through the MOSFET.
Note that the load, whatever it is, may affect the voltage at the MOSFET, so its drain voltage may not be the load supply voltage.
For a series resistive load, the maximum power dissipated in the MOSFET will occur when the voltage at the drain is 1/2 the supply voltage.
 

LowQCab

Joined Nov 6, 2012
4,029
Be careful ........
Most FETs are designed for "Switching-Duty" ( On or Off ),
and are not well suited to being used as a "Variable-Resistor".

Check the Spec-Sheet for the "Safe-Operating-Area", ( SOA ), Graph of the chosen FETs,
look only at the "DC" Graph-Lines, not the "Pulsed" Graph-Lines,
if there are no DC-SOA-Graph-Lines, then choose a different FET.

Use multiple Op-Amp/FET combinations to divide the Heat over a larger area,
You can have as many parallel Op-Amp/FET combinations as You like.
More is better.
.
.
.
 

WBahn

Joined Mar 31, 2012
29,979
For a series resistive load, the maximum power dissipated in the MOSFET will occur when the voltage at the drain is 1/2 the supply voltage.
That's only the case when Rs = 0.

But what holds is that max power dissipation occurs when Vds (not Vd) is half the total supply voltage.
 

Irving

Joined Jan 30, 2016
3,845
For larger electronic loads you should consider using MOSFET devices designed for the purpose such as IXYS' Linear2 series. These have much wider guaranteed safe operating areas (SOA) and a physical construction that eliminates hot spot failure modes more common in devices designed for fast switching when operated in this way.

Heatsinking is critical in this application as the load is continuous.
 

Thread Starter

AhmedKhouaja

Joined Nov 2, 2019
20
First of all, I really appreciate all the answers provided by your side.

To be more specific, I want to put this kit that I bought from AliExpress in an enclosure (that I haven't chosen yet), so my question is how big the heat sink that I need to choose and whether I need to add a fan or not. You can see the kit and the scheme of it, as well as the SOA graph of the Mosfet, in the figures.
Any suggestions on how to choose the right heat sinks for further application of this kind of power electronics?
 

Attachments

LowQCab

Joined Nov 6, 2012
4,029
1)
What, EXACTLY, do You intend to connect to this Circuit ?
2)
How many Seconds, or Minutes, will the Load be applied ?
3)
Why do You want this Circuit ?, What will be its purpose ?
4)
Do You intend to add a Fuse on the Load-Input ?

There is nothing that will limit the amount of Current, or Heat, generated by this Circuit,
and nothing that will monitor the Heat-Sink-Temperature.
.
.
.
 

Irving

Joined Jan 30, 2016
3,845
To be more specific, I want to put this kit that I bought from AliExpress in an enclosure (that I haven't chosen yet), so my question is how big the heat sink that I need to choose and whether I need to add a fan or not. You can see the kit and the scheme of it, as well as the SOA graph of the Mosfet, in the figures.
Any suggestions on how to choose the right heat sinks for further application of this kind of power electronics?
So, breaking down this circuit, the TL431 gives a 2.5v reference, which is fed to the 4 opamps as a 0 - 1.7v value due to R22/R23. The act as comparators with a little bit of hysteresis due to the 220k feedback resistor, comparing the reference with the voltage across the 0.22 source resistors R1 - R4, giving a constant current sink of 1.7/0.22 = 7.7A per MOSFET or 30A total.

Working from the datasheet , the max junction temperature is 175degC but we'll derate that to 150degC for safety and the thermal characteristics for the TO-220/TO-263 packages is 0.63degC/W thermal resistance junction-case. We generally like case temperature not to exceed 75degC so we calculate max wattage as (150-75)/0.62 = 120W. However, the limiting factor is usually the heatsink. Assuming ambient at 30degC the heatsink would need a thermal resistance of (75 - 30)/120 = 0.375degC/W or about 0.3degC/W allowing for the thermal paste. And that is per MOSFET. Now, 0.3degC/W is doable, but its not immediately obvious how. A quick squint at Mouser or Digikey throws up a huge range of heatsinks but choosing 0.3degC/W gives some very expensive options! The trick here is understanding how heatsink performance is documented and every manufacturer does it slightly differently. As a result the tables given by Mouser and Digikey are not very helpful.

So we'll turn the process round and pick a manufacturer, lets say ATS (catalog attached) and pick a heatsink... lets use ATS-EXL116-300-R0, the key details being:

heatsink.JPG

On the face of it this is nowhere near the 0.3degC/W we need. But look closer... those figures are per 1"(2.54cm) of extrusion, so a 12" length, naturally convected, is actually 5.7/12 = 0.46degC/W and even gently fan assisted at 200lfm (linear feet per minute) its 1.9/12 = 0.16degC/W ! And thats at 10W per inch, or 120W.

Now, here's how I would play this, I'd cut 2 off these in half and fit one MOSFET per section; that gives 0.32degC/W though the amount of metal in the heatsink may not support 120W for long-duration runs. I'd then arrange the 4 sections like this:

heatsink2.jpg

Why? Because thats now approx 120mm x 120mm and so we can slap a 120mm fan underneath blowing up. That fan needs to achieve 200lfm per section, or 0.9m3/min or 15litres/sec. A cheap Sunon or RSPro 120mm x 32mm fan will do 2.5m3/min or 600lfm which should get close to 0.2degC/W
 

Attachments

Thread Starter

AhmedKhouaja

Joined Nov 2, 2019
20
So, breaking down this circuit, the TL431 gives a 2.5v reference, which is fed to the 4 opamps as a 0 - 1.7v value due to R22/R23. The act as comparators with a little bit of hysteresis due to the 220k feedback resistor, comparing the reference with the voltage across the 0.22 source resistors R1 - R4, giving a constant current sink of 1.7/0.22 = 7.7A per MOSFET or 30A total.

Working from the datasheet , the max junction temperature is 175degC but we'll derate that to 150degC for safety and the thermal characteristics for the TO-220/TO-263 packages is 0.63degC/W thermal resistance junction-case. We generally like case temperature not to exceed 75degC so we calculate max wattage as (150-75)/0.62 = 120W. However, the limiting factor is usually the heatsink. Assuming ambient at 30degC the heatsink would need a thermal resistance of (75 - 30)/120 = 0.375degC/W or about 0.3degC/W allowing for the thermal paste. And that is per MOSFET. Now, 0.3degC/W is doable, but its not immediately obvious how. A quick squint at Mouser or Digikey throws up a huge range of heatsinks but choosing 0.3degC/W gives some very expensive options! The trick here is understanding how heatsink performance is documented and every manufacturer does it slightly differently. As a result the tables given by Mouser and Digikey are not very helpful.

So we'll turn the process round and pick a manufacturer, lets say ATS (catalog attached) and pick a heatsink... lets use ATS-EXL116-300-R0, the key details being:

View attachment 284018

On the face of it this is nowhere near the 0.3degC/W we need. But look closer... those figures are per 1"(2.54cm) of extrusion, so a 12" length, naturally convected, is actually 5.7/12 = 0.46degC/W and even gently fan assisted at 200lfm (linear feet per minute) its 1.9/12 = 0.16degC/W ! And thats at 10W per inch, or 120W.

Now, here's how I would play this, I'd cut 2 off these in half and fit one MOSFET per section; that gives 0.32degC/W though the amount of metal in the heatsink may not support 120W for long-duration runs. I'd then arrange the 4 sections like this:

View attachment 284020

Why? Because thats now approx 120mm x 120mm and so we can slap a 120mm fan underneath blowing up. That fan needs to achieve 200lfm per section, or 0.9m3/min or 15litres/sec. A cheap Sunon or RSPro 120mm x 32mm fan will do 2.5m3/min or 600lfm which should get close to 0.2degC/W
Good Afternoon Mr. Irving,

This was a really good tutorial to my question, this was very useful, thank you so much for your time, I really appreciate it.


Good Afternoon Mr. LowQCab,

My question was about what Mr. Irving explained, Thank you very much for proceeding with me in my question.

Best Regards,
Ahmed.
 

Thread Starter

AhmedKhouaja

Joined Nov 2, 2019
20
So, breaking down this circuit, the TL431 gives a 2.5v reference, which is fed to the 4 opamps as a 0 - 1.7v value due to R22/R23. The act as comparators with a little bit of hysteresis due to the 220k feedback resistor, comparing the reference with the voltage across the 0.22 source resistors R1 - R4, giving a constant current sink of 1.7/0.22 = 7.7A per MOSFET or 30A total.

Working from the datasheet , the max junction temperature is 175degC but we'll derate that to 150degC for safety and the thermal characteristics for the TO-220/TO-263 packages is 0.63degC/W thermal resistance junction-case. We generally like case temperature not to exceed 75degC so we calculate max wattage as (150-75)/0.62 = 120W. However, the limiting factor is usually the heatsink. Assuming ambient at 30degC the heatsink would need a thermal resistance of (75 - 30)/120 = 0.375degC/W or about 0.3degC/W allowing for the thermal paste. And that is per MOSFET. Now, 0.3degC/W is doable, but its not immediately obvious how. A quick squint at Mouser or Digikey throws up a huge range of heatsinks but choosing 0.3degC/W gives some very expensive options! The trick here is understanding how heatsink performance is documented and every manufacturer does it slightly differently. As a result the tables given by Mouser and Digikey are not very helpful.

So we'll turn the process round and pick a manufacturer, lets say ATS (catalog attached) and pick a heatsink... lets use ATS-EXL116-300-R0, the key details being:

View attachment 284018

On the face of it this is nowhere near the 0.3degC/W we need. But look closer... those figures are per 1"(2.54cm) of extrusion, so a 12" length, naturally convected, is actually 5.7/12 = 0.46degC/W and even gently fan assisted at 200lfm (linear feet per minute) its 1.9/12 = 0.16degC/W ! And thats at 10W per inch, or 120W.

Now, here's how I would play this, I'd cut 2 off these in half and fit one MOSFET per section; that gives 0.32degC/W though the amount of metal in the heatsink may not support 120W for long-duration runs. I'd then arrange the 4 sections like this:

View attachment 284020

Why? Because thats now approx 120mm x 120mm and so we can slap a 120mm fan underneath blowing up. That fan needs to achieve 200lfm per section, or 0.9m3/min or 15litres/sec. A cheap Sunon or RSPro 120mm x 32mm fan will do 2.5m3/min or 600lfm which should get close to 0.2degC/W
Dear Mr @Irving

I was trying to redo all these calculations that you did for another type of Mosfets, this time is TO-247 Package, type IRFP480 from Vishay.

The main objectif is to use 8 IRFP460 Mosfets as Electronic Load to drain 600Wattsin total or lets say 30V, 20A.


After a lot of research, I comes to this conclusion:

Per IRFP460 Mosfet:

  • Total Power Dissipation = 600 watts
  • Power Dissipation per MOSFET = 600 watts / 8 = 75 watts (let’s say 90W for safety reasons)
  • Tjmax = 150°C (we can use 120ºC for safety)
  • Junction-Case resistance = 0.45
  • Estimated temperature rise selected this case is 90ºC so we calculate max wattage as (120-90)/0.45 ≈ 70W
  • Assuming ambient at 30degC ,
    (90 - 30)/70 = 0.85ºC/W

I have lay arround two kind of a heatsink, one with reference and datasheet and other only one information that i will mention after:
• First Heatsink is : V5583K with 3.2ºC/W ( i have 2 of them)
Datasheet: https://www.digikey.pt/en/products/...5583K/3511453?s=N4IgTCBcDaIG4FYEA4DMBrEBdAvkA
• Second Heatsink belong to specific company that they told me is 209W/m-k, means that it's 0.005ºC/W

So the question comes here is, Which of these two heatsinks is the best to mount the 8 Mosfets, or 4 Mosfets per 4 Mosfets (2 Heatsinks) ?
 

Irving

Joined Jan 30, 2016
3,845
So the question comes here is, Which of these two heatsinks is the best to mount the 8 Mosfets, or 4 Mosfets per 4 Mosfets (2 Heatsinks) ?
Taking your calculated required heat-sink performance of 0.85degC/W per MOSFET, your existing heat-sink (V5583K) at 3.2dgC/W is not practical, even force-cooled its unlikely to be good enough even for a single MOSFET.

209W/m-k is the thermal conductivity of the material the heatsink is made from (its an aluminium alloy) but it doesn't equate directly to a degC/W figure for a heatsink. The performance of a heatsink is far more dependent on the physical layout of the heatsink than the material its made from. If you know the physical shape of the heatsink we can estimate its performance but right now you have nothing useful to make a decision with...
 

Thread Starter

AhmedKhouaja

Joined Nov 2, 2019
20
Taking your calculated required heat-sink performance of 0.85degC/W per MOSFET, your existing heat-sink (V5583K) at 3.2dgC/W is not practical, even force-cooled its unlikely to be good enough even for a single MOSFET.

209W/m-k is the thermal conductivity of the material the heatsink is made from (its an aluminium alloy) but it doesn't equate directly to a degC/W figure for a heatsink. The performance of a heatsink is far more dependent on the physical layout of the heatsink than the material its made from. If you know the physical shape of the heatsink we can estimate its performance but right now you have nothing useful to make a decision with...
Dear Mr @Irving

Thanks for replying on my question, it's really usefull informations provided by yourside, but my doubts here are, do we need the thermal resistance of the heatsink less than the calculated thermal Resistance? ( means the Rs-a < 0.85 ?)
And we need to consider the thermal resistance as a normal resistance or not? Means in this case we need a heatsink that provide a number that match Rtotal=8mosfets / Rs-a per mosfet ?
 

Irving

Joined Jan 30, 2016
3,845
Means in this case we need a heatsink that provide a number that match Rtotal=8mosfets / Rs-a per mosfet ?
No, you got it upside down:

Rheatsink(#)=Rs-a per mosfet/# of mosfets
so for required Rs-a per MOSFET = 0.85, Rheatsink(8) = 0.85/8 = 0.1 approx

To achieve the necessary flow of heat flux, required thermal resistances for multiple heat sources on 1 heatsink are effectively in parallel - intuitively this must be so, two MOSFETs have to shift twice the flux for the same temperature rise so the heatsink resistance must be halved...

Also I noticed your calculation is wrong...

600W with 8 mosfet = 75W/ Mosfet
Rj-c for this MOSFET = 0.45, so max P per MOSFET is (120-90)/.45 = 66.6W. This isn't enough for 600W so either need more MOSFETs or could relax Tj-max to 125C giving Pmax = (125-90)/.45 = 77.8W. This is probably ok.

78W @ Tc=90C and Ta = 30C requires Tc-a of (90-30)/78 = 0.77C/W or allowing for the thermal paste a heatsink of 0.68C/W per MOSFET or lower than 0.68/8 =0.085C/W for all 8 MOSFET. Such a heatsink is possible but expensive so here I would probably look to put 2 MOSFETs per heatsink of lower than 0.68/2 = 0.34C/W, and use 4 heatsinks as before.
 
Last edited:

Thread Starter

AhmedKhouaja

Joined Nov 2, 2019
20
No, you got it upside down:

Rheatsink(#)=Rs-a per mosfet/# of mosfets
so for required Rs-a per MOSFET = 0.85, Rheatsink(8) = 0.85/8 = 0.1 approx

To achieve the necessary flow of heat flux, required thermal resistances for multiple heat sources on 1 heatsink are effectively in parallel - intuitively this must be so, two MOSFETs have to shift twice the flux for the same temperature rise so the heatsink resistance must be halved...

Also I noticed your calculation is wrong...

600W with 8 mosfet = 75W/ Mosfet
Rj-c for this MOSFET = 0.45, so max P per MOSFET is (120-90)/.45 = 66.6W. This isn't enough for 600W so either need more MOSFETs or could relax Tj-max to 125C giving Pmax = (125-90)/.45 = 77.8W. This is probably ok.

78W @ Tc=90C and Ta = 30C requires Tc-a of (90-30)/78 = 0.77C/W or allowing for the thermal paste a heatsink of 0.68C/W per MOSFET or lower than 0.68/8 =0.085C/W for all 8 MOSFET. Such a heatsink is possible but expensive so here I would probably look to put 2 MOSFETs per heatsink of lower than 0.68/2 = 0.34C/W, and use 4 heatsinks as before.
Dear Mr. @Irving

Thank you for correcting my calculations, i think your calculation are more accurate than mine after double check my results.

So, can you please give me something affordable for this application from Aliexpress Heatsinks? i know that they're not professional heatsink, but they are more convinient for my pocket in this moment...

Ahmed.
 
Top