[Amps] LDMOS and DMOS RF MOSFETs

[John Lyles]

This was on line, dated 2014. An interesting perspective on LDMOS and 
plain DMOS transistors.
73
John
K5PRO
----------
Reports of the Death of DMOS RF MOSFETs are Greatly Exaggerated
by Mark Vitellaro, Director of Strategic Marketing, Richardson RFPD


The landscape of the RF power transistor market has recently changed due 
primarily to three developments: the emergence of GaN, the maturation of 
the LDMOS process, and the slowdown in the growth of the wireless 
infrastructure market. The performance attributes of GaN are well known. 
The technology offers a highly-valued combination of improved power 
density, high gain, high frequency, wide bandwidth, and high efficiency. 
Another interesting aspect of the emergence of GaN has been the 
appearance of multiple independent GaN fabs, which has opened up 
opportunities for several RF transistor companies that were previously 
unable to develop their own competitive LDMOS strategy. While there is 
certainly time for the GaN market to consolidate, there seems to be 
enough diversification in product development to support several players 
in the medium term. It just so happens that the timing of the GaN 
awakening has occurred as LDMOS is facing challenges from a maturing 
process, as well as the shrinking of the wireless infrastructure market. 
Now dominated by Freescale, NXP and Infineon, LDMOS is on its 9th 
generation, providing diminishing marginal returns from future process 
enhancements. Furthermore, the push for higher efficiency, digital 
baseband with digital predistortion and small cells in the wireless 
market have reduced the LDMOS dollars per amplifier.
The combination of these forces has driven LDMOS suppliers to look 
elsewhere for growth, and the largest area of interest has been the 
industrial, scientific, and medical (ISM) market. This market is 
fragmented and includes many high power HF applications such as RF 
exciters for CO2 laser, RF generators for plasma ignition, and RF 
amplifiers for MRI. Historically these applications have been served by 
RF MOSFETs.

But the current generation of LDMOS was not suitable for ISM, because 
the evolution of LDMOS has been driven primarily by the needs of the 
wireless market—which has a completely different set of technical 
requirements. So Freescale and NXP went back to the drawing board.
Freescale was the first to plant their flag with their “E” series of 
enhanced rugged LDMOS. The new feature set included a very high VSWR 
rating of 65:1 at all phase angles, coupled with a 3 dB overdrive 
feature. Combining those features with their 50V process and wide gate 
withstand voltage introduced a new class of transistor that was suited 
for the high mismatch applications of the ISM market. What’s more, 
Freescale introduced a 1.25kW device that could be operated single-ended 
or push-pull. This new class of transistors set the ISM market on its 
head. Up until this point, most systems were based on combined 300W DMOS 
transistors which required extensive passive matching and combining 
networks. The 1.25kW LDMOS device simplified the architecture and 
eliminated significant supportive circuitry. Furthermore, as the 
efficiency of these circuits were approaching 80%, it was possible to 
utilize surface mount devices. It didn’t take long for NXP to notice, 
and shortly thereafter, it introduced its family of “XR” Extra Rugged, 
RF transistors. The family looked similar on paper, as they included a 
65:1 VSWR rating and 1.2kW and 1.4kW device. But even more telling was 
the 2013 announcement from NXP to close a fab that produced many RF 
transistors, including their DMOS products that were optimized for 
HF/ISM applications. Freescale had already exited the DMOS market more 
than a decade ago when they divested their gold process fab and sold 
that product line to M/A-COM Technology Solutions, which continues to 
produce the line of MRF transistors in their Lowell fab today.

So now you may be asking yourself, “Wait, this blog post began on the 
premise the DMOS was not dead, didn’t it? It sounds like DMOS is 
declining as LDMOS encroaches into their market only to be accelerated 
further now that both Freescale and NXP are focused on transitioning 
customers to LDMOS technology...” Clearly LDMOS has taken share and the 
suppliers are allocating more resources to support the ISM market as 
wireless slows, but the remaining DMOS suppliers are not throwing in the 
towel just yet. As a process technology it would seem that DMOS has 
reached its process limits, but there are incremental improvements that 
can be made specifically to address the ISM market, namely packaging 
innovations, higher voltage, and the pulse characteristics of DMOS.
ST Microelectronics was the first to develop a plastic air cavity 
package as an alternative to costly ceramic packages. Introduced as 
“STAC” packaging, which stands for ST Air Cavity, they successfully 
eliminated the ceramic, which reduced cost and improved thermal 
dissipation and gain. Presently available in bolt-down and direct-solder 
options, several established DMOS transistors are now available in the 
new packaging. Microsemi also developed their own packaging enhancement. 
The flangeless VRF157FL is a near equivalent to the 600W MRF157, but 
Microsemi removed the CuW flange and replaced the ceramic lid with 
plastic. These mechanical changes result in similar benefits as the 
STAC, and hopefully that family grows as much as it has with ST Micro.

Besides packaging, increasing the supply voltage from 50V can simplify 
and shrink the power supply and thus the overall system. In addition to 
a smaller power supply, RF amplifiers for MRI benefit from better SNR, 
which can improve the quality of the imaging. ST and Microsemi both have 
100V conventional DMOS in their portfolio, the SD3933 and VRF3933. ST 
also has a STAC version, the STAC3933. But Microsemi has gone even 
further with their ARF family of DMOS transistors which operate at 150V 
and 250V. These plastic packaged devices operate in highly efficient 
Class D mode and are offered in pairs or with integrated drivers 
(DRF1200). ST is developing their own 150V to 250V “Super Junction” DMOS 
devices, which are in evaluation now.

Finally, DMOS exhibits a much higher peak power rating compared to an 
equivalently sized LDMOS device. Combining that peak power advantage 
with a plastic/flangeless package narrows the cost advantage of LDMOS. 
For instance the STAC3932B is characterized for 3T MRI systems. Under 
1mS, 10% conditions, the part will produce 900W peak power.
In conclusion, the combination of inherent characteristics of DMOS and 
the hard work of innovating suppliers show that the incumbent DMOS 
manufacturers refuse to concede victory to LDMOS. And the ultimate end 
users of ISM transistors are very resistant to change and are not always 
impressed with the next new thing. “Copy Exact” requirements and fear of 
the unknown will ensure a portion of the ISM market continues to use 
DMOS. I, for one, welcome the debate and am curious to see how this 
battle plays out. While everyone else is enthralled with GaN, the LDMOS 
– DMOS battle will be a very interesting one to follow.