editor's blog
Subscribe Now

Closing the Thermal Loop

It’s impossible to design an SoC today without good simulation models. Emphasis on “good.” We’ve always simulated, but, back when mask sets cost less than the GDP of a small country, the actual silicon was the true test of whether the design worked.

Of course, using real silicon to iterate the design was never the goal, but the only way to get good models is to test them against actual silicon – it’s that closed loop that gives you confidence in the model. But that means running a lot of silicon to correlate against the model, which takes time, and, if a mask set isn’t too painful to acquire, well, you might take your chances on an early design with an “OK” model rather than waiting for the models to stabilize.

No longer. Process development goes on long ahead of actual first commercial wafers so that the process and the models can be dialed in properly. Which requires a tight loop between silicon characterization and the simulation tools.

As we’ve noted before, the electrical world is impacted by the mechanical – and notably the thermal – world. You do a lot of work to meet thermal requirements. Mostly it’s so that your circuit will perform within a given set of specs. Also it’s so that it won’t wear out early. Or, more dramatically, explode.

With LEDs in particular, there appears to be yet another reason to manage temperatures: as the temp goes up, the LED brightness goes down.

So while you do your part to keep the temperature where it needs to be, your mechanical counterparts have to do the same thing with the package and board and the box it all goes in.

And how do they do that? Just like you do: simulation. We’ve looked Mentor’s FloTHERM before, both in its more limited –IC version and more generally. Tools like this are used to simulate what happens to the heat as it courses through your system. But how do you know it’s accurate? What models do they use?

And this is where the thermal folks have found themselves: in the same place you electrical folks were years back. With the amount of thermal slop we can tolerate going down, it really makes everyone feel a whole lot better if they know that the models being simulated are accurate. And that means closing the loop with actual physical characterization. Which requires a tester. Which Mentor also has.

And here we digress briefly into the world of, “How should we name our product?” (Indulge me; the inner marketer and language aficionado won’t be silenced.) The unit they have for measuring temperature is called T3ster. My first question was, “Is that pronounced ‘tee-three-ster’?” Turns out no, it’s pronounced “tree-ster”. Never mind that pesky “th” thing in there. It interferes with two other clever devices going on here. The formal name is “Thermal Transient Tester” – 3 Ts, or T3. That would be T3 for those of us without access to superscript. Plus… a 3 is E in, what, hackerlandia? Somewhere out there. So “T3ster” is really “TEster”. All of which makes it too tempting to ignore the annoying fact that most English phonetic clues will point you in the wrong direction. It probably makes more sense given that Hungarian (where this originated prior to acquisition by Mentor) has no “th” sound, so that, if German is any clue, “three” probably ends up being pronounced “tree.” OK, maybe not such a brief digression. Their marketing guys are gonna hate me. Anyway…

The thing about T3ster is that it’s very accurate – up to 0.01 °C. It can also register up to 10,000 measurements per minute, allowing it to capture the shapes of temperature changes as they happen. They appear to have a particular focus on LEDs via their TERALED setup.

Mentor has now provided explicit integration between FloTHERM and T3ster for that critical feedback loop that makes it possible to generate abstracted models for various packages and other configurations. According to them, this is an industry first.

Not only does this provide better confidence, but the abstraction also speeds up the simulation time. Essentially, it allows the creation of thermal verification IP, although no one has signed up to get into that business explicitly.

Ultimately, better confidence in the thermal models translates to better confidence that the electrical design will perform as expected.

More info in their press release

Leave a Reply

featured blogs
Jul 18, 2018
I recently talked with Mr Takizawa of TDSC about their use of Cadence's Interconnect Workbench (IWB). You may not recognize those initials. Toshiba split itself into three companies last year and one of them is TDSC, or Toshiba Electronic Devices & Storage Corporatio...
Jul 16, 2018
Each instance of an Achronix Speedcore eFPGA in your ASIC or SoC design must be configured after the system powers up because Speedcore eFPGAs employ nonvolatile SRAM technology to store the eFPGA'€™s configuration bits. Each Speedcore instance contains its own FPGA configu...
Jul 12, 2018
A single failure of a machine due to heat can bring down an entire assembly line to halt. At the printed circuit board level, we designers need to provide the most robust solutions to keep the wheels...