Actions
Forum Thread Details
  • State Verified Answer
  • Replies 12 replies
  • Subscribers 301 subscribers
  • Views 4348 views
  • Users 0 members are here
Related
See a helpful answer?

Be sure to click 'more' and select 'suggest as answer'!

If you're the thread creator, be sure to click 'more' then 'Verify as Answer'!

Can someone verify my thermal calculations?

Andrew J

Maths is not my strong point but I think I've got basic calculation correct.  So: 4 MOSFETs connected in parallel, on a heatsink with a Sil-Pad isolator.  Assume that the MOSFETS are connected correctly and are sharing the load equally.

Thermal Calculation
Max Power 150 W From design limit
Package Dimension 1.648 cm^2 From datasheet for MOSFET, case and tab
Package Dimension 0.255 in^2 Converted
Parallel MOSFETS 4 From design
Power/MOSFET 37.5 W Power shared equally
Max J Temp 175 C From datasheet for MOSFET
Ambient 35 C Inside Case, assumed when running hot
Rjmb 0.5 C/W From datasheet for MOSFET1
Thermal Paste 0 C-in^2/W Sil-Pad doesn't require paste
Isolating Pad 0.3 C-in^2/W Sil-Pad K-6 datasheet
Heatsink 1.4 C/W Natural convection, from datasheet
Junction to case 18.75 C Power per mosfet * Rjmb
Case to Heatsink 2.86875 C Power per mosfet *((paste + pad) * package dimension)
Heatsink 52.5 C Power per mosfet * heatsink
Ambient 35 C
Total 109.11875 C
%Max J Temp 62.35%

MOSFET datasheet: https://4donline.ihs.com/images/VipMasterIC/IC/NEXP/NEXP-S-A0002881467/NEXP-S-A0002882763-1.pdf?hkey=6D3A4C79FDBF58556ACFDE234799DDF0

Sil-Pad K-6 datasheet: https://www.farnell.com/datasheets/90833.pdf 

The "tricky" bit to watch out for is the area calculation for the Sil-Pad to reach the case to heatsink impact, particularly as I'm converting square cm to square inches.  K-6 seems to be pretty good, but I have seen some quite high C/W figures for silicon pads and this calculation is trying to convert from an area-based thermal resistance to a C/W thermal resistance to keep the units the same through the calculation. EDIT: I'm aware the 0.3 figure will be based on pressure applied and the overall area used during manufacturer's testing, unlikely to be TO-220 sized!  However, I do need to include something in the calculation with the intention of seeing how close to maximum allowed I get.  I'll confirm with real life testing of course and I'm also trading that 0.3 value against the heatsink's 1.4 value as I will in fact be using a fan to bring that figure down.

Alse:

1 The MOSFET datasheet table 5/Fig.4 on page 4, covers the Junction to case temperature but it isn't clear.  It states a MAX of 0.5K/W (0.5 C/W) but is this a max allowed or a max it can be - on the basis that it can be lower if the power through the MOSFET is pulsed more quickly?  According to the graph, 0.5 is based on a pulse duration that never exceeds half the cycle (single shot through to theta = 0.5) once that cycle reaches approx. 0.06seconds but these MOSFETs will be used continuously as part of a DC load, in other words theta would be 1 but this is not shown on the graph.  Therefore, 0.5 doesn't seem appropriate and also seems inordinately low for a TO-220 package.

Parents
  • +1

    I think you have an error in the case to heatsink line, you have 0.255 sq in of sil pad, at 0.3 C/W per square inch so it's 1.176 C/W for the pad. Case to heatsink is 37.5 * 1.176 = 44.1C.

    If there is just one heatsink there is a second error - the heatsink temp rise = 150 * 1.4 = 210 so I'm assuming you have a heatsink per device.

    SO on that basis the temperature rise junction to ambient = 18.75 + 44.1 + 52.5 = 115 so you have zero margin at 35C ambient.

    You might get a small improvement by losing the pad and doing thermal grease directly to the heatsink (feasible with a heatsink per device)  but you won't get it much better - your best bet is to blow the heatsink.

    That graph is showing you that the max thermal impedance tends towards 0.5 K/W as the pulse length increases. So 0.5 K/W is the rating you should use. If you were using very short pulses with low mark/space ratio you could assume a lower thermal impedance.

    MK

  • 0 in reply to michaelkellett
    michaelkellett said:
    f there is just one heatsink there is a second error - the heatsink temp rise = 150 * 1.4 = 210 so I'm assuming you have a heatsink per device.

    Ah, of course!  There is just one heatsink actually and I'm also using it for a Schottky Diode, a Sense Resistor and a Temperature Monitor (all TO-220 packages) so it's likely to be significantly worse.  i do have a fan though.

    Thanks for the feedback on this and clarity on the MOSFET specification.  That figure of 1.176 C/W is much closer to the figure I was expecting for silicon pad.  I do have some thermal grease and mica pads so I will use those instead - TBH, I was considering the Sil-Pads as they were cheap and it avoided messing around with grease. 

    Rather than try and build 'up' I'm going to revise down to a more manageable Power level.  I'm glad I posted this and appreciate the feedback.

  • 0 in reply to Andrew J

    Typically active loads in the 150W - 300W class use tube style heatsinks with blowers - but these are a bit pricey as standard parts.

    You can expect the heatsink performance to improve by a factor of between 2 and 4 with a decent airflow, so you need to go for about 0.6 K/W convection rated heatsink with a fan. This will result in a heatsink temperature rise of < 45K and a junction max temperatuer of140C.

    Farnell have 3813322 at £28 which is rated at 0.113 K/W with fan and better than you need. They have the slightly cheaper 3813292 for about £20 which is rated at 0.4 K/W convection. It should give you no more than 30K above ambient for 150W.

    If it's for your own use you can protect yourself by adding a thermal cut out.

    If you are going to put 4 devices on one heatsink I suggest you stay with the SIL pads.

    Using a MOSFET in a TO247 case will improve the case to heatsink coupling. I'm pretty sure that the 3813322 heatsink has a better surface finish than the 3813292 and that may be worth the extra cost !

    MK

  • 0 in reply to michaelkellett
    michaelkellett said:
    Typically active loads in the 150W - 300W class use tube style heatsinks with blowers - but these are a bit pricey as standard parts.

    I was originally looking at the Fischer Electronic ones which were funky and expensive!  I saw a recent teardown of a DC Load that had these, or similar, inside.

    Any particular reason to stay with Sil Pads, I've not actually bought them yet?  Mica seems better - I'm struggling to find data that isn't represented as W/mk which I can't work out how to convert - but they definitely seem easier to use than Mica and I think if the difference between pass and fail is the thermal resistance of the insulating pad then it's a definite fail in any case!

Reply
  • 0 in reply to michaelkellett
    michaelkellett said:
    Typically active loads in the 150W - 300W class use tube style heatsinks with blowers - but these are a bit pricey as standard parts.

    I was originally looking at the Fischer Electronic ones which were funky and expensive!  I saw a recent teardown of a DC Load that had these, or similar, inside.

    Any particular reason to stay with Sil Pads, I've not actually bought them yet?  Mica seems better - I'm struggling to find data that isn't represented as W/mk which I can't work out how to convert - but they definitely seem easier to use than Mica and I think if the difference between pass and fail is the thermal resistance of the insulating pad then it's a definite fail in any case!

Children