MRF300: How Do I Estimate Junction Temperature?

[Dennis4RF]

I'm building a 600 W, prototype HF amplifier using a pair of MRF300s. For reliability and thermal testing, I am using a type of modulation that transmits a carrier for 15 seconds, then the amplifier goes into standby (no bias current) for 15 seconds, then repeats the same on/off cycle over and over again.

I used a reduced power output of 400 W for my thermal testing. After 15 minutes or so of this on/off cycle, the heatsink became slightly warm to the touch, and I then went probing with an infrared thermometer to see how hot things were near the MRF300s.

I found that at the very tail end of the 15 second transmit cycle, the top side of the MRF300s (the black plastic resin case) would reach 60 - 65 degrees C. The heatsink area immediately adjacent to the devices was much cooler at only at 35 degrees C. After the amplifier stopped transmitting, the resin case temperature dropped within 5 seconds to the same temperature as the heatsink = 35 degrees C. 

The NXP specifications give a 0.55 degree C / Watt thermal resistance for junction-to-case. Does this spec refer to the resistance between the MOS junction and the metallic underside of the case, which is touching the heatsink? If that is true, then I would estimate the max junction temperature in my testing to be 35 C (heatsink temp) + (200 W x 0.55)  = 145 C?  I use 200 W for my calculations, as there is a pair of devices delivering a total power output of 400W. Are my calculations correct? If so, I believe the devices are operating well within the max junction temperature of +175 C. 

There is also a thermal IMPEDANCE specification given, but I believe this only comes into play for pulse-type operation, and not the type of CW service that I am using...correct?

If I am not estimating junction temperature properly, will an NXP engineer please correct me so that I can redesign the cooling system to prevent device overheating and possible thermal failure? 

Thank you for your assistance.

[LPP]

Please, refer to the application note "Thermal Measurement Methodology of RF Power Amplifiers":
https://www.nxp.com/docs/en/application-note/AN1955.pdf

The case temperature is the temperature of the of the thermal pad surface. The datasheet values was measured for a device attached to effective water cooled hit sink that provides uniform (cold plate) temperature profile at the contact.

Your calculation is not quite correct. The max junction temperature calculation must use thermal power (not RF power) that is dissipated at the device. Heat power depends on output power and the power efficiency of the amplifier. Power efficiency depends on the design.

For example, efficiency of NXP MRF300AN 13.56 MHz reference circuit is E=79.7% @ Pout=320W.
So, the heat power dissipation is Pdis=Pout*(1/E-1)=81.5Watts.

NXP provides Ztheta-jc thermal impedance spec for pulse operation and Rtheta-jc thermal resistance for CW mode.

15sec stands for CW operation. So, use Rtheta-jc=0.55'C/W.

PS. Use reliable value for heat sink temperature. 35'C is quite the usual ambient temperature. Most likely, heat sink temperature is noticeably higher. Your calculations should include thermal resistance of the heat sink. For fan cooled heat sink, use heat sink to ambient thermal resistance from heat sink datasheet. For natural convection mode, evaluate heat sink performance from the size of the heat sink.

https://info.boydcorp.com/hubfs/Thermal/Air-Cooling/Boyd-How-to-Select-a-Heat-Sink.pdf

[Dennis4RF]

Dear Pavel:

Thanks very much for the quick and detailed response. I made a mistake in assuming that NXP's thermal  specifications had already factored in the average efficiency of the devices. My HF amplifier efficiency is very close to the figure you stated, but I'll use a lower 75% just for the sake of discussion. So, if two devices are outputting 400 W total, input power is 533 W with 133 W of waste heat, then divided by two devices = 66.5 W per device. Let's hold onto that figure.

I can use the thermal resistance of the heatsink as a starting point to estimate junction temperature, but there is a slight complication here. The sink is fan cooled, but not by a constant velocity fan. The fan is temperature-controlled, and increases in speed as the heatsink warms up. This is an attempt to keep fan noise to a minimum, while at the same time provide adequate cooling for the devices.

So, I seek to estimate if my junction temperature is within acceptable limits by a combination of empirical testing and simple thermal calculations. I can test the amplifier under normal conditions, measure heatsink temperature very close to the devices, and then use thermal calculations to estimate junction temperature. I can manually adjust the fan speed to see what effect this will have on heatsink temperature, and in turn calculate the junction temperature by using Rtheta-jc=0.55'C/W. 

You mentioned using a "reliable" temperature value for my heat sink. The 35' C temperature I measured IS the actual temperature, measured on the heatsink immediately adjacent to the devices!  My heatsink is massive, 30 cm long x 18 cm wide x 4 cm deep...PLUS has additional fan cooling! Placing my finger adjacent to the devices - I can barely tell if there is any warmth! I would expect that the actual temperature right underneath the device might be a degree or two warmer, but not much more than that. Probing the heatsink 2 - 5 cm further away from the device showed only a slight variation in heatsink temperature. 

So again, going back to the calculations using a "empirical method" with temperature measurement, 66.5 W of waste heat x 0.55' C/W = 36.57 degree rise at the junction. My 35' C heatsink + 36.57 = a junction temp of 71.57' C. If I'm correct, this is very far below - just 50% - of the maximum allowable junction temperature, and provides a lot of tolerance for estimation errors. 

One final question: I mentioned to you that at the very end of my 15 second transmit cycle, I measured between 60' - 65' C on the TOP of the case, on the plastic resin. Is this considered normal, or typical when the junction temp is approximately 70' C? I assume that the resin case is a part of the 150'C rating for the case per the NXP specifications, correct?

Thank you very much again for all of your advice and the links you sent me.

Regards,

[LPP · Accepted solution]

Your consideration is correct.

I recommended using "reliable" temperature value for heat sink since it must account for worst case condition. In practice, ambient temperature could be as high as 35-40'C plus the amplifier is enclosed into a case that could add ten degrees above. I would start from 60'C.

The maximum temperature of the plastic case is 150'C. Temperature at the top of the package must fit this requirement. The package poorly cooled from the top and it is normal that the top temperature is noticeably higher than the heat sink temperature.

Pavel
NXP CAS

[Dennis4RF]

Thank you again Pavel, all your help is greatly appreciated. I am now satisfied that my junction temperature is well below allowable maximum. 

Regards,