Robin Bornoff's blog

Views and insights into the concepts behind electronics cooling with a specific focus on the application of FloTHERM to the thermal simulation of electronic systems. Investigations into the application of FloVENT to HVAC simulation. Plus the odd foray into CFD, non-linear dynamic systems and cider making.

21 October, 2014

Lighting accounts for ~20% of the world’s total energy consumption. This is a staggering statistic. Why so much? Historically a lot of energy had to be consumed to produce the required amount of visible light. Incandescent bulbs, where electric current is used to heat a metal filament, resulted in most of the energy being dissipated as heat. More a hot bulb than a light bulb. Even worse for a candle. LEDs mark a step change in the efficiency by which energy is converted into light, they’re cooler as a consequence and thus more reliable. Win/win/win. Their ubiquity today is due to the breakthrough in creating an LED capable of emitting blue light (red and green having already been achieved). With the ability to create those 3 colours, they could be combined to produce white and all colours in between. The 2014 Nobel prize in Physics was awarded to 3 scientists who ‘invented’ the blue LED in the 1990s, despite the fact that a 1974 US patent indicates otherwise.

A Lumen (lm) is a measure of the energy contained in that part of the EM spectrum that your eye can sense, the visible spectrum. A luminous ‘efficiacy’ of radiation is a measure of how much visible light power there is compared to the power in all frequencies of the light. More the better. LEDs are on their way to becoming x1000 better than historic forms of lighting:


Ironically, LEDs perform better if they are cooler. At higher (junction) temperatures the brightness decreases, the emitted frequency shifts and the reliability is compromised as well. Good thermal management, in terms of the physical characterisation of their optical and thermal properties as well as a simulation based design methodology, is essential. The LED vendor market is coalescing through acquisition, maintaining a competitive product edge is as important as it ever was. Such thermo-optical applications are a real sweet spot for the technology in the Mechanical Analysis Division of Mentor Graphics. The MCAD embedded CFD technology of FloEFD coupled with the T3Ster and TeraLED physical test and measurement equipment is especially well suited to automotive lighting applications where the use of LEDs is now commonplace.

Cluster_LEDsAn interface between T3Ster and FloEFD enables temperature dependent LED optical properties, measured by TeraLED, to be imported into FloEFD whereupon a ‘hot lumen’ prediction can be simulated based on a specified driving current. That, coupled with a choice of 3 optical radiation models in FloEFD and the ability to handle complex MCAD assemblies directly within the MCAD environment, make for a unrivalled industrial strength capability.


Wally Rhines, CEO, Mentor Graphics

One of the authors of the overlooked 1974 US patent “Gallium nitride metal-semiconductor junction light emitting diode was Wally Rhines, CEO of Mentor Graphics. Wally is one of those rare captivating keynote speakers who can take you on a roller-coaster of a ride into the world of EDA and deposit you at the end feeling intrigued yet satisfied. It would have been cool to have a Nobel Laureate as CEO. It’s a bit of a shame it didn’t work out that way this time.

21st October 2014, Ross-on-Wye.

p.s. Wally’s not only a keynote speaker of course, his CEO track record at Mentor speaks for itself. Thought I’d clarify that in case he stumbles on this blog… :)

, , , , , ,

6 October, 2014

“How thick is a leg hair” is not a question I thought I’d be posing when I woke up today. Reading this article about the effect of leg hair on the overall drag of a performance cyclist prompted me to investigate the physics behind this observation. For the sake of expediency I steered clear of a full body MAMIL simulation, focussing instead on just a small patch of a particularly hairy leg subjected to an oncoming air flow.

BoundaryLayerAir will ‘stick’ to any solid surface, even a silky smooth leg. Away from the wall it speeds up until it reaches free stream speeds. That thin sliver of slow moving air, hugging each solid surface, is called a boundary layer, often graphically indicates as a series of velocity vectors, running parallel to a wall, increasing in size away form it. This stickiness offers a resistance, more for treacle, less for air. So, even a shaved leg will offer some resistance to the air flow.

HiarDistributionAdd to the wall (skin) a whole forest of hairs then you’ve added loads of new solid surface area to which the air will stick. I did go looking for a MCAD file of hair but to no avail. Instead I mocked up a random distribution of stubble type hairs of a density pretty high up the Chewbacca scale. Nice and simple, no moving hairs, no curly hairs, no solid fluid interaction. Such stubborn stubble being an extreme case in cycling aerodynamic drag, rather lack of it.

In addition to the increase in surface area and the resulting increased stick of the fluid, the more tortuous a path the fluid has to travel through the hairy obstruction, the more difficult it finds it to do so. This fluid difficulty manifests itself as a pressure drop, effectively the energy required to achieve that flow.

SpeedBetweenLegHairsSo, using FloTHERM (though FloEFD would have been more appropriate, old dog, new tricks etc.), I simulated air approaching the segment of hairy skin to see just how tortuous a path it would be made to follow. The red areas are high speed flow, the blue/green areas low speed, especially obvious in the slow moving recirculating wakes behind each hair. Even more evident if we look at the flow animation:


Each twist and turn of the flow equates to a force required to overcome it. The more hairs you have on your legs, the more force you’ll need to overcome it, the more energy you’d need to expend. For Chewbacca this equates to the difference between changing from a round-tube frame to an aero-style one.

FloEFD has a proven track record for sports applications as this fascinating webinar by Olympian Professor Kristan Bromley shows.

6th October 2014, Ross-on-Wye

p.s. Diameter of a human hair is between 17 and 180 microns (millionths of a meter). For the case above I chose a particularly hirsute 200 microns (0.2 mm).

, , , ,

1 October, 2014

There are a handful of quoted and re-quoted drivers, concepts and technologies in the electronics cooling industry.  Be it a presentation, peer reviewed paper, press release, thermal design guide or webinar the same core material keeps popping up in various guises. This is by no means a bad thing. The challenges faced when cooling electronics are clear, clearly described and need to be constantly reiterated. If, when viewing such presentations, reading the articles or attending a travel budget friendly webinar you might like to partake in this electronics cooling ‘metaphorical’ drinking game to pass the time.

The following items, should they be encountered, will score the following drinking game points penalty… I’ll leave it up to you to decide how those penalty points should be paid.

  • MooresLawMoore’s Law. The number of transistors in an IC doubles approximately every two years. More an observation than a law though it has, until recently, stood the test of time very well. Considering that it is the thermally dissipated power of the transistors that causes the temperature rise and subsequent performance and reliability issues in electronic products, it’s not surprising that this graph is shown again and again. Very common: Penalty Points = 1


  • HeatFlux_vs_YearFollowing on from Moore’s law, the actual heat dissipation divided by the area over which it occurs is the heat flux and has been rising recently, quite dramatically. Analogies are always good, especially when you’re trying to scare the pants off an audience. Although for just a small area (of the order of a few mm^2), the heat flux values in tomorrow’s processors will be ‘as hot at the sun’ (as I’m sure the media will spin it as). Getting to be a more popular graph and it’s fun: Penalty Points = 2


  • ThermalResistancesIn 1822 Joseph Fourier published Théorie analytique de la chaleur in which he postulated that heat flow is proportional to the temperature difference between two points and the thermal resistance between them. Nearly 200 years later and the electronics thermal community thrives on such simple abstractions of what in reality are complex 3D heat flow patterns. Thermal resistor network descriptions of systems are as common as muck and almost twice as useful: Penalty Points = 1


  • Fixing Problems WhenNot unique to the electronics cooling industry, the cost of fixing mistakes at various stages of the product development process is well documented. As any good sales person will be at pains to point out, anything that can front load the design process with risk identification will have a very obvious and evident return on investment. And sure, we’re no different. A graph as valuable as it is overused: Penalty Points = 1


  • Structure_FunctionIn recent years transient thermal test characterisation has become a popular non-destructive approach to investigate the physical construction of a packaged IC or discrete. A thermal equivalent to the use of X-rays, the T3Ster measurement technology is now frequently referenced in the industry. One key aspect of the value of the T3Ster approach is in the determination of a thermal ‘structure function’, a graph that indicates the thermal resistances and capacitances that exist within the package and system in which it is placed. Banded to show what parts of the graph refer to what physical part of the package/system, to understand and appreciate a structure functions is now a prerequisite for all thermal engineers: Penalty Points = 3


  • TSVsTSVs! (through silicon vias) The next big thing. Enabling die to be stacked efficiently, no need for inter-die interposers or peripheral wire bonding. From a thermal perspective you’ve just massively increased the effective heat flux. How to get that heat out is a question that has to be answered before this technology can hit the mainstream. The very high heat transfer coefficients that over-chip liquid microchannel cooling can provide seems the most promising concept. Likely that any material you see or read on this topic will have a picture that looks much like an apartment block with too many elevators.  The future is always fascinating: Penalty Points = 4


  • Equivalent CircuitsThévenin’s equivalent circuit theory is fundamental to many electronics cooling simulation approaches, be they spreadsheet based or modelling methodologies employed for full 3D simulation. Maybe too simple or too well recognised to be described in any detail nowadays, assumed thermal resistances in series or parallel can be used to determine effective resistances and thus, say, effective thermal conductivities.  No different to an electrical circuit and surely everyone knows this stuff by now?: Penalty Points = 0.5

There are many more graphics and phrases you’ll come across when involved in electronics thermal management but the above should provide some amusement next time you’re sitting there, watching presentations late on the 2nd day of a 3 day conference.

1st October 2014, Ross-on-Wye


19 September, 2014

A recently issued patent describes a process by which the critical IC temperature (junction temperature, Tj) can be determined in an end user environment. A critical junction temperature is one that, should the IC temperature go beyond it, will continue to increase in a thermal runaway scenario. In other words the environment in which the IC is operating can’t get the heat out quick enough. The thermal equivalent of an uncontrolled nuclear fission event.

Thermal runaway generally leads to fusing, a melting of the conductor. Temperature increases -> electrical resistivity increases -> Joule heating power dissipation increases -> temperature increases … and so on. As the name indicates, this is precisely how a fuse works. Same effect in digital electrical circuitry can have disastrous consequences.

The power dissipated in a IC comprises two parts, a dynamic part (due to all the turning off and on of the transistors including frequency and capacitive affects) and a static, or leakage, part (that is the power that is drawn when the transistors are in an off state). The static proportion increases with smaller transistor sizes AND is temperature dependent. Temperature first affects functional performance then, as runaway thermal feedback is encountered, it quickly escalates to a full smoking thermal catastrophe. All of which is shown in this (albeit a bit old) video:

P-QThe key to determining the point at which runaway might occur is consideration of both the static power vs. temperature relationship AND the system performance in same terms, determining where they meet, and this is what the patent focusses on. This bears a striking and quite lovely similarity to fan sizing, where the fan P Q curve intersects with the system P Q behaviour is the operating point of the fan, and thus also of the system.

Chip Temperature Distribution Predicted by FloTHERM

Chip Temperature Distribution Predicted by FloTHERM

Be it fan sizing or specification of a temperature dependent power dissipation, FloTHERM can simulate it all. Quickly, early and easily.

This white paper on our chip-package solutions will hopefully provide some additional insights.

19th September 2014, Ross-on-Wye

, ,

8 September, 2014

Develop 3D is a UK based Magazine and Website focussing on the technologies involved in product design. The latest copy had a fascinating article on Dell’s Precision Workstation labs. This blog is being written on a Precision M4500 which I’ve lugged around countless airports and dumped, wearily, on numerous hotel beds. The fact that it works as well now as when I first got it a few years ago is testament to Dell’s design processes and technologies.

The article, , looks into Signal Integrity and Thermal design. Both are key aspects in achieving competitive product performance. From a thermal perspective the overall power dissipation is related to the frequency at which the PC will operate (well, the dynamic portion of the power dissipation is). Increased power leads to increased temperatures. If the temperatures get too high things start to break, the supplier gets a bad reputation, warranty costs spiral, other very bad things ensue. This catastrophe can be diverted by throttling down the power before the temperatures get too high. Downside is that performance is limited. Having a cooling architecture that is sufficiently effective at removing the heat quickly can therefore have a marked and direct effect on the competitive nature of the product (read $$).

As the article quotes, “[We do] simulation way early on, looking at how thermals are going to work, almost in turn figuring out how the whole box is going plug in together and work together before we even build the first one.” An excellent example of doing design at the right time. A mature design process is one where simulation is used at a conceptual, pre ‘concept commit’ stage to determine at a high level what the major design constraints will be. Interesting that at this stage in the design, before the green light has been given and lots and lots of engineers and detailed engineering tools are deployed, there is very little detailed design data. From a simulation perspective this is actually quite liberating. Good engineering judgement can be applied to drive the simulation tool, in the context of little or no IDF or STEP, to answer questions such as “how many vents will we need?”, “one micro blower or two?”, “heatpipes?” etc.

Generic Laptop SectionFloTHERM was adopted early on for telecomms, networking and computing applications and is especially well suited to conceptual design. It’s simple CAD type interface, building block approach to 3D model construction, instant grid calculation and fast+robust CFD solution technology really comes into their own. Here’s some output from an application example we install with FloTHERM. A corner section of a generic laptop with a double inlet/outlet blower, convection cooling a couple of ducted heatsinks that are supplied the heat via a couple of heatpipes connected to a PLCC package.

The animated air flow behaviour enables appreciation of how the design is working, is all the cold air ingested that could be? Is any hot air recirculated back inside?

Generic Laptop Section Top

Generic Laptop Section Below

If you’d like to find out more about FloTHERM this introduction to FloTHERM webinar gives a good overall impression as to what it can do, worth checking out.

One final thought, if you are ‘lucky’ enough to have highly detailed EDA and MCAD design data to hand when using FloTHERM maybe you should ask yourself if it’s a bit late in the day simulating the design for the first time.

8th August 2014, Ross-on-Wye


4 September, 2014

A release so good it has 11 best top 10 features, very Spinal Tap. I thought I’d wrap up this series with a list of more minor, but imo, very useful features and finish off with some words on cultural idioms.

FloTHERM V10.0 and V10.1 together satisfy about 60 software enhancement requests as voted for on the Mentor IDEAS site for Mechanical Analysis Products, a strategy we’ve been committing to for a number of years now. It’s not always the big headline grabbing features that the most interesting though. Here are some of my personal less well known favourites:

  • Indication of whether a loaded model has results or not. Done via a coloured icon in the bottom status bar in the Project Manager. ResultsYesNo No more having to load a model and open the Profiles window to see whether there is any residual history.


  • NomObjectsHow many objects are selected? Now given directly, again in the Project Manager bottom status bar. If no objects are selected this changes to the total number of objects in a model. I benefit from this often in my use of FloTHERM, though to tell you the truth I’m not exactly sure why!


  • AlignI like the simplicity of shape moving and editing in Microsoft products. Needs no training. One thing that is missing from Office (but now not form FloTHERM!) is a double align on both centers, in one operation. Now a one button operation in V10.


  •  Having a RecentProjectslist of recently used files/documents for quick access to recent data is quite standard nowadays. FloTHERM’s had this for a while. However it only worked if the selected project was in the current project solution directory. This has now been rectified. Loading a previous project will automtically force change of the solution directory and load the project.


  • AbortSolver So many times I’ve started a solve only to suddenly realise that I’d forgotten to make some required tweak or modification to the model. The solver stops slowly as it finishes its current iteration then writes out all the solution data to disk. Sure, we could have added an ‘Are you sure?’ dialog each time the GO button was pressed. We’d be lynched though. Instead we’ve added an ‘Abort Solver Action’ feature that will terminate the solver executable immediately, no results written but you can get back and make the change asap. Hands up who remembers ‘kill -9′ on Unix?

Well, that’s it for V10. Development for V11 is well underway. FloTHERM ploughs ahead!

4th September 2014, Ross-on-Wye

p.s. O yes, idioms. Not sure where the phrase ‘odds and sods’ comes from, it’s in common use in the the UK though. ‘Odds and ends’ is the more common variant I suppose. There are a number of phrases that are perfectly acceptable to use in the UK but would raise eyebrows elsewhere, normally in the US. Yep, in the UK it’s perfectly acceptable to go out of the pub and ask someone if you could ‘bum a fag’. Also, a private school in the UK is referred to as a public school (well, it’s open to any member of the public to pay to send their children to!). This and other US/UK travel word tips can be found here.


4 September, 2014

Arguably the most important (and often least well characterised) parameter controlling the temperature of electronic products is their power dissipation. Usually dissipated on the active layer of a die, this heat power seeps through the die, package, PCB, air, chassis all the way to the ambient in which the product sits. The less heat power the better, the lower the temperature rises that occur. The last thing the cooling requirements need is even more power being added to the product but that’s exactly what happens on any sun apparent face. Sun bathing was fashionable in the 70s, it never was for electronics**.

Our own star provides us with a maximum of about 1400 W/m2 of incident solar energy, most of the power coming in the visible spectrum range. This can have a marked effect on the resulting surface temperature of an outdoor pole mounted repeater or base station cabinet. Accurate accommodation of such a power input to a FloTHERM electronics cooling simulation model is often critical.

Solar loading defined and modelled in FloTHERM. The +ve SolarVis field indicates where there are no shadows

Solar loading defined and modelled in FloTHERM. The +ve SolarVis field indicates where there are no shadows. The banded effect of the resulting surface temperature is evident.

SolarSpeedUpFloTHERM’s had the ability to specify the solar effects for a number of releases. The (annual) date and time of the simulation is specified together with the location and compass orientation of the model, FloTHERM does the rest, works out the sun apparent solid faces, the incident angles and imposes the correct resultant power on those faces.

This ‘pre-processing’ calculation used to take some time, days in some cases. It doesn’t any more. Speed improvements comparing V10 to V9 of up to x80 faster are now achieved though a rewrite and added parallel support for multiple processors. Yet another example of our strategy to reinforce the core functions in our software.

4th September 2014, Ross-on-Wye

**Photovoltaics aside of course

, ,

4 September, 2014

FloVENT, FloTHERM’s sister product aimed at 3D CFD simulation of the built environment, has for many years been able to simulate the thermal behaviour of a data center. After all a data center is a built environment, albeit more for servers and the like than humans. We’ve not changed this capability of FloVENT but we have copied the relevant data centric features over to FloTHERM, extending the positioning to cover electronics thermal management from the chip to the room.

FloTHERM 10 splashscreen

The electronics thermal supply chain is linked together with the desire on one side, and the obligation on the other, of assuring thermal compliance in the customer’s application. Thermal performance of the equipment in the data center is a function of density, layout, cooling architecture and operation. Now I don’t know exactly what % of all PCBs designed and manufactured end up in a data center, likely not that many. However I assume that nearly all of the bytes communicated in this IoT age are routed through such rooms.

CRAC operationPushing bytes around the net needs energy. Bytes don’t have much kinetic energy themselves, which is why that energy ends up being dissipated as heat. Leave that heat lying around unattended and things will get too hot and stop working. Ever wondered what a world full of teenagers lamenting the loss of Facebook, Twitter and SnapChat might sound like? Ouch.

FloTHERM’s Rack and Cooler SmartParts allow for the quick and easy definition of these common data center items. That coupled, with FloTHERM’s renowned ease of use and rapid simulation performance, enables the thermal performance of a given data center configuration to be rapidly determined. Check out this video for a look at the typical output from such a simulation.

4th September 2014, Ross-onWye

, , ,

29 August, 2014

Electronics thermal simulation historically focussed on steady state conditions, worse case scenarios where for example a power dissipation is assumed at a maximum and unchanging. Not a direct reflection of reality but the conservative (over)temperature simulation results enabled design margins to be implicitly factored in. Over the years, as margins reduced and the need for ever more accurate simulations increased, the proportion of transient simulations that FloTHERM conducted has increased dramatically. From a FloTHERM development perspective we’ve kept on top of this trend. An overhaul of our transient definition capabilities in V9.1 has now been followed by substantial changes in our transient modelling capabilities in V10.

Beyond standard transient considerations such as the thermal response of a system to fan failure condition, or a cold start power on, thermostatic control is an approach that can be used to both limit operational temperatures and lower power usage. Various user posted feature requests on the Mentor ‘IDEAS for Mechanical‘ site alluded to this, all of which have now been satisfied by a couple of new V10 features.

Transient TerminationThe ability to stop a transient solution, automatically, is now available via a new ‘Monitor point transient termination criteria’ option. One or more monitor points can be nominated, with individual stopping criteria temperatures. During the transient simulation, as soon as one condition is met, the simulation automatically stops. This allows inspection of the model at that time/temperature, results exported or manual modifications made to the model and solved onwards.

The real doozy though is an extension we have made to the transient attribute. The transient attribute is a curve that relates a multiplier value with time. This attribute can then be attached to a power dissipation value, a temperature etc. to vary those boundary conditions (BCs) in time. In V10 we have added the option to also include a multiplier vs. (monitor point) temperature curve as well. In this way a BC can be varied in time AND/OR as a function of temperature.

Themostatic ControlVery powerful, especially when considering the application to die level models where it would allow for a transient variation of the switching power portion and a temperature dependence of the leakage power portion of the power dissipation together.

AndTheresMoreAnd there’s more… FanThermostaticDeratingThermostatic controlled fans are very common, especially in consumer/laptop type products. Ever hear your laptops’ fan come on when the CPU gets too hot running all your FloTHERM simulations? In V10 you can now attach a transient attribute to a fan to affect its RPM derating factor during the transient simulation, either as a function of time or monitor point (hub mounted or remote) temperature.

AndTheresMoreAnd there’s more… In addition to a single curve relating the BC with temperature, a Hysteresis option can be checked to allow two curves to be defined, one controlling the BC when the temperature is increasing, the other to control the BC when the temperature is decreasing.


Accommodation of such hysteresis allows for better tuned thermostatic control methodologies to be investigated and determined in FloTHERM. If you want to leanr more about such control, here’s an interesting piece from Panasonic comparing and contrasting PID and Hysteresis control approaches.

As you can hopefully see, the FloTHERM development team are big fans of transient simulation technologies :)

29th August 2014, Ross-on-Wye

p.s. let’s hear it for Jimmy Cricket!

, , , , ,

18 July, 2014

I’m sure in the (far) future, product design and manufacture will just involve a big box that you can ask to make things for you. “A leg driven transport device with communication, health monitoring and illumination functions that will last me the week. Please Mr. Box” (or ‘Bob the making box’ as I’d call mine). Out it would pop. Think Star Trek Next Generation Replicator, extrapolated. In the mean time product design is far more manual, time consuming and let’s face it, FUN! Time however is money and no matter how much fun it is, it has to be profitable. Waiting around for your FloTHERM thermal simulation to complete is both not fun and time consuming. Maybe not as much now though…

The CFD solver is the core, the nucleus of a CFD software. Everything upstream of it’s deployment is intended to serve it (boundary conditions, geometry, mesh etc.) and nothing downstream can happen (results inspection) until it’s done its business. Just like a watched kettle, the CFD solver can appear to take a long time to finish. More maybe due to the feeling of helplessness, that all you can do is watch those residuals come down.

I’ve been working in the application and product management side of CFD for all my professional career. I know enough about the maths to be dangerous, not enough to be useful. For FloTHERM V10 we have substantially reworked the CFD solver, especially its parallel performance. Using nothing more than words I’ll attempt to g**k porn a description of what we’ve done…

The main aim of this project was to improve the parallel scalability of the CFD solver. Experience is that for a shared memory approach care has to be taken to achieve +ve scaling above 4 cores. Key to this is focussing on load balancing between the cores. To that end we:

  • Modified our data structures to distribute the computational work across the cores more uniformly.
  • Modified our linear solvers to take advantage of the better load-balancing including introduction of a new more scalable preconditioner
  • Memory management is modified to take advantage of the NUMA-architecture (Non-Uniform Memory Access) of modern processors
  • Enhanced memory layout of our data structures to optimize the memory accesses for the modern cache-based processors

There are many ways to present the performance of parallel CFD solvers. I’ve seen enough graphs showing maximum linear scaling (use N cores results in an N times speed up) to start questioning their integrity long ago. We decided to focus on the improvements a FloTHERM user would expect to see when moving from V9 to V10.The following graph shows how many times faster V10 is compared to V9, for a range of different models, on 1, 2, 4 and 8 cores.

v9 vs v10

Being the real world, we found that the relative performance improvement is case specific, the improvement does get better the more cores you use but (surprisingly) the performance is not effected by mesh topology (i.e. the number and layout of localized grid spaces). Some of the changes we made for scalability also had a dramatic effect running even on a single core!

We did tests on up to 8 cores on all models, some we tested up to 32 cores, continued +ve scaling was observed.

Being twice as fast, having to wait half the time, is nothing to be sniffed at. 10 times faster is quite remarkable. Less time to read SF and speculate about the future, more time to get back to the real work of designing good (thermally compliant) product.

18th July 2014, Ross-on-Wye