Case Study: Use of Transient Thermal Data to Minimise Size & Weight of a Power Supply

 

The problem

Celab was asked to design and manufacture a power supply for use in a Jet Fighter, which supplies a total of 140W to avionics equipment from a nominal 28VDC supply. However, the input was specified to be in accordance with MIL-STD-704A which specifies operational ranges and transients associated with engine start, bus-bar switching, weapons operation and the like.

The power supply design

Looking at the standard (below) and taking the customer’s specific requirements into account, we see that normal, full-load operation is required from 16 to 30VDC steady-state but also 10 to 80VDC transiently.

Power design challenges

Our first design challenge was to assess the operation at low-line; at 10V input, there will be 17A of input current. Remember that this is an avionics product so the weight of input filter inductors and copper tracks is critical – realistically we expect to lose ~2V in losses through the input filter so the power supply actually has to run on 8V – this now will draw 20A! That is manageable though, so we designed the input filter and tracking to drop 2V at 20A. This dissipates 40W transiently and ~6W under normal conditions.

The knock-on effect of this design choice is that, with a 6:1 input range – which is pretty good for a 1990’s design – we can only really operate up to 50V; what do we do when the input rises above 50V?

A switching pre-regulator would be an ideal choice, but with space and mass at a premium one could simply not be included so a linear pre-regulator was the only practical solution. In principle, this consists of a tiny charge-pump to bias a FET with a 54V zener on its’ gate. However – at 50V, the power supply will be drawing 3.3A which means that during the 80V event, the linear regulator will be dissipating ~100W. Bear in mind that when operating with “nominal” input, the whole Power Supply dissipates less than 25W and we are faced with designing in five times as much heatsink as would be otherwise needed!

Clearly, that is not practical so we looked at details of suitable FETs; The IRF3205 was chosen as it had one of the lowest RDSon figures available at the time – 8mΩ which gave 3.2W maximum dissipation during low line transients. But; with a maximum ambient of 70°C and the Power Supply already heated up to 90°C, we see from the datasheet that, with a thermal impedance RθJC of 0.75°C/W, 100W gives a steady state rise of 75°, i.e a junction temperature of 165°C. This is within the manufacturer’s 175°C limit but does not allow for derating which limits us to TJMAX – 20°C.

 

The solution

There are three factors to consider in designing for thermal management of a silicon part;

• A) the junction-to-case thermal performance, then,
• B) getting the heat out of the device into the local ambient; either junction-to-ambient or case-to-sink figure will apply here and finally,
• C) getting the heat away from the unit altogether which is another area in its’ own right…

Looking at A) and B)

Junction-to-Case: We see from the locus curve in MIL-STD-704A that the 80V transient lasts for something less than 100ms (note that this curve is a locus, not a time-domain waveform). Let’s call it 100ms and consider the FET’s transient thermal impedance chart in the datasheet: That gives the same RθJC of 0.75°C/W at 100ms so that does not help, but shows that the junction will heat up to it’s maximum 165°C by the end of the transient.

What we could really do with is limiting the device’s dissipation to a figure which is manageable without additional heatsinking. If we take the RθJA figure from the datasheet of 62°C/W then we can allow 880mW of steady-state dissipation to keep the junction below 155°C.

Now, we have 660mW maximum of dissipation in the device when under steady-state input conditions, so we can allow a further 220mW. Averaging this out means that the 100W transient can be present for not more than 1/450 of the time i.e one transient every 45s.

MIL-STD-704 does not explicitly give a repetition rate so we took this back to the customer and pointed out that the locii of transient events are drawn up to 50s. We took the view that this implied a maximum repetition rate of 1 event per 50s and that we should design and qualify on this basis. The argument was accepted and a successful product was designed, qualified and manufactured and is still flying today.

Surpassing clients expectations

The drive for “smaller, cheaper and lighter” products is never-ending. This design shows just one example of applying the detail of a requirement so as to allow a product to meet or even exceed a customer’s expectations on size and efficiency. We used a similar approach to create a 2kW boost converter for another avionics power supply which was a little bigger than a matchbox!

Our team of expert power designers here at Celab work tirelessly to create bespoke power supplies and power solutions for customers worldwide. We work across multiple industries such as industrial, military, CATV & telecoms. Contact us today via our online form to discuss your power supply requirements.