AnsweredAssumed Answered

The thermal resistance of ADI HMCxxx seems unusually high compared to other “similar” parts, why is this?

Question asked by MRichardson Employee on Aug 7, 2017

We get a fair number of questions that revolve around the thermal performance of the ADI HMC frequency generation devices. I'll try to answer this particular question without making it too specific to any particular product.


As far as comparing the Theta j-c published on our datasheets, customers need to be sure that they are comparing similar products. So what are similar products?  Similar products are parts within a product family built on the same process, same package and operating at similar power dissipation levels.  It’s not fair to compare an LNA to a PA or an active multiplier operating to 30GHz to a wideband VCO.  The former is obvious but the latter case may not be so let me explain.


Each product family relies on different technologies (SiGe, GaN, GaAs HBT, PHEMPt,...) and foundries offer various active device geometries for each technology. The maximum junction temperature of the various technologies also varies from about 125°C to 175°C.  Furthermore, what current density did the foundry use as the basis for deriving their MTTF values and how does that compare to the maximum current density for the products we’re comparing? In most, if not all cases, the ADI HMC frequency generation products operate at a significantly lower current density than that used by the foundry to determine the MTTF.  While de-rating guidelines and design rules dictate the worst case limits for current density one designer may opt to use a larger active device (assuming we’re not frequency limited) to handle a given amount of power while another may parallel multiple devices to spread the heat out. Additionally, within a product family (VCO's, or Multipliers for instance) various architectures may be used, each with unique attributes. Amplifiers for instance are biased very differently with varying levels of efficiency depending on their intended use or application.  Next, the thermal conductivity of GaAs itself is non-linear so even within a technology there are factors that need to be considered. Another consideration is that some models are multi-chip modules (MCM’s)  while others are monolithic designs. Finally, the hotspot isn’t always the active device. It might be an emitter resistor. All this to say that there are a lot of variables that can influence the thermal properties of these designs and we haven’t even discussed the package or pcb yet!  


For these reasons, a narrowband VCO operating at 5.0V & 265mA = 1.325W can’t really be compared with a wideband, fundamental VCO design operating at 5.0V, 70mA = 0.35W. However, wideband VCO products operating at similar pdiss levels in the same package using the same technology should have approx. the same Theta jc values.  (e.g. HMC586LC4B  vs HMC732LC4B vs HMC733LC4B).


Finally, for the frequency generation products, the published thermal results in many cases are based on simulation. Typically a few products within the families have measured data to establish some confidence in the models but there can be some variation between simulated and measured data depending on the quality of the model but generally they agree pretty well.

Hopefully this will help you next time you need to compare some of these devices.


Best Regards,