3
\$\begingroup\$

I'm working on a second-generation design for an electronic load. The first generation is described in this prior question if anyone is interested (scroll down to section containing first photo), but this question doesn't bear on it much.

The load has this conceptual schematic. Loop compensation, gate drive, and device choice are neglected here for simplicity.

enter image description here

A comment by Neil_UK on this question got me thinking that the thumbnail-sketch SOA analysis for the MOSFET I did for the first generation might not be quite adequate, so I thought I'd take a more formal approach.

The specs for the load are:

  • 1.5 to 30 V (from connected source)
  • 0 - 5 A (within limits of power dissipation)
  • 30 W maximum power dissipation

One of the devices I've been considering for this second generation is the FQP13N06L (datasheet here). The datasheet contains a nice SOA chart that seems to give me what I need. I've indicated the bounds of my expected operating area with a red line (neglecting effects of \$R_{DS(on)}\$ at low \$V_{DS}\$:

enter image description here

Now, this all seems straightforward to me as I'm reasonably well within the DC boundary. But Neil_UK's comment has been niggling at me and I'm wondering if I'm missing something:

Don't let a 50 watt FET fool you into thinking you can dissipate 50 watts in it like you can a 50 watt BJT. You can if it's saturated, and blasts from on to off in microseconds, you can't if you try to use it in a linear mode, like the pass element for a linear regulator, controlled load, or audio amplifier output device. The Safe Operating Area has timing constraints, the array of cells in the FET unbalances thermally if dissipating for too long. – Neil_UK

The other uncertainty I have is that the datasheets for the IRFZ24N I used in the first generation (and other devices in that family) don't have a DC line. Based on Spehro's answer to this question on SOA I'm figuring I can just draw that in myself (in red below) based on maximum continuous drain current (12 A) and power dissipation limits (45 W):

enter image description here

So my question is: "Am I understanding this correctly? Or is there a long-term heat-dissipation problem with MOSFETs I'm not seeing?"

\$\endgroup\$
  • \$\begingroup\$ Neil has the right comment .For your practical load to be reliable there are things that you can do .Use lateral fets which are really expensive ,Use much higher voltage than needed fets ,Use Fets in series instead of in parallel ,Use old old fets ,Derate everything ! \$\endgroup\$ – Autistic Sep 6 '16 at 3:39
1
\$\begingroup\$

The FQP13N06L's SOA graph is based on a die temperature of 175°c. RTJC is 3.35°c/W so you need to keep the case temperature below 175-(30W*3.35) = 74.5°c. If the ambient temperature is 30°c then you need a heatsink with RTJA of (74.5-30)/30W = 1.5°c/W or lower.

But while the FET is rated to survive at that temperature, reliability will suffer - especially if die temperature cycles up and down a lot. In a constant current circuit you might expect wide variations in power dissipation (and operating temperature) depending on what the load is doing.

I bet your load won't like being connected directly to Ground with no current limiting, so you really don't want your FET to melt down. I would be very conservative and design for a junction temperature below 100°c. That equates to 15W with a 1.5°c/W heatsink at 25°c ambient.

\$\endgroup\$
  • \$\begingroup\$ Ok Bruce, so it sounds like I'm not mistaken about how to place the SOA envelope, just that I need to be mindful that achieving the thermal dissipation that goes with it would need to be attended to closely. Also, importantly, survivability and reliability are two different things, and I'd want to add in margin for that. Have I got that right? In this case, I expect the device to get only occasional use, be operated at maximum dissipation infrequently and for short periods, and would have a thermally triggered turbine fan on a healthy heatsink, so I'm thinking my 30 W max spec is okay. \$\endgroup\$ – scanny Sep 7 '16 at 20:26
  • \$\begingroup\$ Yes. I try to stay well below maximum ratings to give a good safety margin - my rule of thumb for dissipation is halve the power rating and don't let case temperature go above 50ºC. \$\endgroup\$ – Bruce Abbott Sep 7 '16 at 22:56
1
\$\begingroup\$

The long term heat equation uses the Rjc+Rca (jcn to case and case to ambient to achieve the temperature rise you want.

Although it is rated for Tj =175'C abs. max in the SOA curves, you likely want run the case temps around 85'C max or a rise of 60'C and a heat sink of <2 deg C/W.

This heatsink will be large or small with a fan like those used for CPU's.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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