Sign In
Forgot Password?
Sign In | | Create Account

Mind Your Thermal Management To Improve Reliability

John Parry

John Parry

Posted Jul 16, 2009
6 Comments

It’s been a while since I posted on the BBQ. It’s not laziness per se, rather a combination of laziness and working on simulating flow around the BBQ.

This is just a quick post to point out a recent article in Electronic Design Magazine on the link between thermal management and product reliability which you can find here.

Traditionally companies have sought to improve product reliability by reducing steady-state temperature. Within the range of interest most component failure mechanisms are not driven by steady-state temperature. Hopefully the article will help explain why good thermal design is important and how it can improve overall product reliability.

As always, I’d be interested in your comments on the article, so feel free to post.

Reliability, Thermal Management, Electronics Cooling

More Blog Posts

About John Parry

John ParryI started my career in the consultancy group at CHAM Ltd., using PHOENICS for a variety of CFD applications. From the consultancy group I moved into support, helping customers debug models, and figuring out how to model new applications. That broadened into delivering training courses and creating training material. I was invited to join Flomerics when it started in 1989 to head up Customer Services, and I jumped at the chance to work for a startup. After a few years supporting customers using FloTHERM I moved across into research, developing thermofluid models of common electronic parts, like fans and IC packages, later managing the DELPHI and SEED projects. More recently I worked with Flomerics’ Finance Director on the acquisition of MicReD, helping to integrate MicReD’s business into Flomerics Group which was great fun. Since Flomerics acquired Nika, I’ve been responsible for promoting the FloEFD suite in education, and moved into marketing. I now work as part of the Mechanical Analysis Division’s Corporate Marketing group, responsible for ElectronicsCooling Magazine and the division’s Higher Education Program. Expertise: I’m a chemical engineer by training and did a PhD in reactor design before getting involved with CFD more than 25 years ago. Visit John Parry’s Blog

More Posts by John Parry

Comments 6

Post a Comment
On the subject of heatsinks... I think there's a lot of misunderstanding (or over-optimism) about what a heatsink can or do. In my experience, attaching a heatsink to the top of a device will do one of three things; 1) Make the device run cooler 2) Make no difference 3) Make the device run hotter Exactly what happens in a given case depends very much on the nature of the heatsink, the device to which it is attached and the PCB to which the device is soldered. 2) and 3) above are counter-intuitive observations, but I can assure you that I have seen real-life situations where both have happened (and subsequent simulations have demonstrated the same behaviour). Simply adding a few thermal resistances together in series and parallel DOES NOT generally yield the correct answer. To judge the effectiveness, or otherwise, of a heatsinking arrangement you really need to build a prototype or carry out a competent thermal simulation. As St. Kordyban once remarked; "Heat can't flow forever into a chunk of aluminium and just disappear, or migrate into a parallel universe..."

Chris Hill
9:47 AM Jul 20, 2009

Hi Chris, thanks for commenting. I agree with you. Thermal design tends to appear simple, particularly to those that don't understand it ;). Adding a heat sink as an afterthought to fix a poorly designed product is a risky business. One reason that components run too hot is poor air flow. Under such circumstances adding a heat sink just makes the air flow worse, giving your 3) above. People also underestimate the resistance of the thermal interface material. If the heat sink only reduces the air-side resistance by this amount the result is your 2). We advocate pushing thermal design as high up the design flow as possible.

John Parry
10:57 AM Jul 20, 2009

I think the reasons why a heat sink may not work are quite subtle and require some careful thought... Imagine we have a surface-mount component which is dissipating heat and is soldered to a PCB (no heatsink at this point) under natural convection conditions. The heat energy from the device will primarily be conducted into the PCB and then pass into the "local ambient" by convection (I'm ignoring the lesser contribution of radiation here). As a result, the air immediately surrounding the PCB+device will experience an increase in temperature i.e. the air immediately adjacent to the PCB is not at "ambient" temperature. Now suppose we place a heatsink on the top of the device. We know that temperature difference is the "motive force" which drives the flow of heat energy, BUT the heatsink is already being heated by the air rising by convection from the PCB surface i.e. the heatsink is not at ambient temperature. Hence the thermal gradient between device surface and heatsink is much reduced and so the flow of heat energy into the heatsink is not as great as might be expected. Even worse, as you indicate above, the heatsink might interfere with the airflow from the PCB to such an extent that overall the situation is made worse (I have seen this happen). I think the effectiveness or otherwise of a heatsink depends very much on the PCB underneath, and you certainly cannot rely on the manufacturer's K/W figure to predict what will happen. Another slightly different way to look at this... The assembly is losing its heat energy primarily by convection to the surrounding air. The air has a limited capacity to absorb heat energy per unit time as air has mass, inertia and specific heat. There is therefore an upper limit on how much heat energy/unit time the assembly can lose to the surrounding air, irrespective of whether we fit heatsinks or not. You can't have your cake and eat it twice.

Chris Hill
6:04 AM Jul 21, 2009

Essentially this is why we advise simulation from an early stage. As you say, the effects can be subtle. One of my first ventures into this area was modeling a heat sink on the first Intel Pentium package, A CPGA dissipating 16W - approx 4 times its 486 predecessor. To cut a long story short, what I found was that heat was leaving the package under the die, passing through the ceramic and TIM, and entering the heat sink. Obvious enough. However, as heat could spread more effectively in the heat sink than the package, which was rather large, something like 40mm I think, heat passed back out of the heat sink into the package! The performance of the heat sink was improved if the contact area with the package was reduced. When you see the heat flux lines it all becomes clear but it's not something you'd expect to happen. One of the real benefits of CFD is the insight it brings.

John Parry
8:36 AM Jul 21, 2009

Hah! I like that one! Even sneakier than the stuff I deal with, and a great example of another counter-intuitive result. On the subject of flux, I often find it useful to include one or more collapsed regions at critical points, so that I can look at the predicted heat flux through those planes. Of course, I have no way of verifying that Flotherm's flux predictions are correct, but I think it's a reasonable assumption that if the monitor point temperatures are coming out right then the underlying fluxes are also right (in the case of a "calibration simulation").

Chris Hill
9:26 AM Jul 21, 2009

The fluxes are summations of the cell face fluxes, being the conserved quantities that the CFD solves for, so yes they will be right!

John Parry
10:02 AM Jul 21, 2009

Add Your Comment

Please complete the following information to comment or sign in.

(Your email will not be published)

Archives

Tags

 
Online Chat