Why does my power op amp IC (TI OPA549) thermally malfunction? IC works electrically perfectly, no oscillations or anything. By thermal malfunction I mean that Rj-c thermal impedance rises from default value (1.4°C/W) to about 6...7°C/W.

What kind of mechanism can cause such malfunction?

PCB has three identical channels (A, B, C) where two of them work flawlessly (B, C) but one (A) constantly malfunctions after short time of use (I have tried three chips).

Below you can see IC and its three important spots center (red), edge (green) and case (blue). With maximium stress ( ~ 40W), channels heat up as follows (measured with thermal camera):

  • Working ones:
    70°C (red), 55°C (green), 40°C (blue)

  • Malfunctioning:
    +100°C* (red), 59°C (green), 40°C (blue)
    *throttling, junction temperature reaches 150°C limit and stops.

IC thermal spots

Thing I have considered:

  • soldering: I have always screwed the case to the heatsink and after that carefully soldered the pins.

  • mechanical stress: Heatsink is heavy, but firmly attached to PCB.

  • case temperature is always same (40°C), so I can assume that the heatsink is working alright.

  • B and C channels are working flawlessly as I have designed.

Update: 29 June 2017

As Chupacabras brought up, I have noticed that problem has something to do with the Vcc line.

I previously had a common heatsink for all channels, but then I installed individual heatsinks for each chip. This had unexpected results:

  • C channel works flawlessly
  • B channel malfunctions as badly as before
  • A channel is something in between.

I also figured out that when I was using linear lab supply, no malfunctions occurred. By default I have been using switching (65kHz) PSUs connected in series, where middle tap is ground.

However, I'm unable to measure any kind of differences in Vcc pins.

  • 1
    \$\begingroup\$ Perhaps image about PCB would be good, issue might not be with the IC itself, but support circuitry. \$\endgroup\$ – akaltar Aug 4 '16 at 18:29
  • 3
    \$\begingroup\$ Seems like the channels are not really identical. Might be construction rather than design (or something like a hair line short). More info such as schematic and photos - there's really no way to help with info given. \$\endgroup\$ – Spehro Pefhany Aug 4 '16 at 18:38
  • 1
    \$\begingroup\$ Can you add circuit scheme? Do you have those 3 ICs isolated from heatsink, or they are connected? Are you drawing some current via heatsink and via ICs tab? \$\endgroup\$ – Chupacabras Aug 8 '16 at 7:18
  • 1
    \$\begingroup\$ @churpacabras i think we have point here. Thanks to this conversation i started figuring out about tabs being connected to -12V (and heatsink). There is good separate supply lines connected to each channel and maybe enough inductance between -12v pins, so that output current dips. Output currents have 120° phase difference (50Hz). Unfortunately i dont dare to upload full schematic here, but once this is confirmed ill put it to answer. \$\endgroup\$ – jalaffo Aug 9 '16 at 12:30
  • 4
    \$\begingroup\$ I'm voting to close this question as off-topic because the OP hasn't modified the question in months and its a debug question. \$\endgroup\$ – Voltage Spike Jun 12 '17 at 21:20

I'll go out on a limb, and suggest that you look carefully at your heat sink. You label your 3 units A, B and C, and it's A which gives the problem. I'd guess that you have a single heat sink, and the three units line up on a common extrusion. My guess is that, at the A end of the extrusion, the mounting surface is not perfectly flat. So you're getting tight contact at one side of the IC, but not on the other. This gap is large enough that, even with thermal compound or a Silpad (and you are using one or the other, I hope) the thermal resistance across the gap is much too great.

The defect might be a slight protrusion around the mounting screw, or it might be a bend in the surface. Take a good straight-edge to the area in question, and verify that there is no distortion from a perfectly flat surface. Make sure to check in both axes. It might only take a few thousandths of an inch to be a problem, so be careful.

  • \$\begingroup\$ That does not really fit with 40°C at the blue spot, which makes contact with the heatsink \$\endgroup\$ – PlasmaHH Aug 4 '16 at 19:37
  • \$\begingroup\$ @PlasmaHH - It might, if contact is only being made around the mounting screw. The rest of the package has to channel its heat laterally to the hole. \$\endgroup\$ – WhatRoughBeast Aug 4 '16 at 21:11
  • \$\begingroup\$ at least some rise above the 40°C would be expected since the heat will spread from the much hotter part of the tab throughout it a bit \$\endgroup\$ – PlasmaHH Aug 5 '16 at 7:59
  • \$\begingroup\$ Heatsink is common for all three channels. There is graphite sheet between surfaces, and i have checked flatness of the surface several times (this was my initial worriy, as it is yours too) \$\endgroup\$ – jalaffo Aug 6 '16 at 7:41

Not the answer you're looking for? Browse other questions tagged or ask your own question.