Amplifiers go RF

Amplifiers go RFThere are two solid-state options for output transistors in RF power amplifiers, LDMOS and bipolar. EW asked Ericsson which is best for what? Bengt Ahl and Stig Jonsson responded.
There is no easy answer to the LDMOS vs bipolar debate. For some applications, LDMOS is best, for others, bipolar is preferable. Yet for LDMOS, there is still great scope for further development and improvement, so it may well be that in the future LDMOS wins the race.
Stig Jonsson, an application engineer with Ericsson Microelectronics’ RF Power Products, set the questions and Bengt Ahl, an RF power design expert at Ericsson Radio Access provided the answers. Stig Jonsson +468 404 6004
Q Many people are saying that LDMOS technology offers so many performance advantages over bipolar that all new RF power amplifier designs will soon be LDMOS. Is this true?
A Definitely not. In the case of a very broadband RF power amplifier, say from 100MHz to 1GHz, LDMOS would be the choice. Also, LDMOS is excellent for cellular base station power amplifiers that require an extremely linear performance, such as IS-95 CDMA, W-CDMA and Multi Carrier. However, for high power applications at frequencies around 2GHz, bipolars exhibit better efficiency and are easier to match, so these would provide a better solution.
Q It is said that LDMOS is more linear than bipolar, in terms of both frequency range and power output. Is this so?
A At frequencies below 1GHz, yes. LDMOS is a better choice at all frequencies in terms of class AB power dynamic linearity. However, above 1GHz the capacitance in an LDMOS device has to be taken into account, making the design of a high power RF amplifier much harder. Furthermore, if an LDMOS transistor is to be pushed close to saturation levels, the linearity gets worse, and so a bipolar device would be preferable.
Q LDMOS technology makes it easier to build hybrid amplifiers. Will the use of these functional blocks in cellular base stations increase or decrease in the future?
A Flexibility and price are on the negative side for these modules at the moment. Each is dedicated to one application only and the integration level is too low for a real cost reduction. With LDMOS technology, high impedance hybrid transistors (HITs) offer a level of integration that should be easy to standardise for circuit design. In a similar way, LDMOS technology can also be used for multistage monolithic designs.  
Q An interesting point is the design of the bias supply circuit for both types of transistor. Are there any fundamental differences between LDMOS and bipolar in the design of this circuit?
A  Yes, for bipolar transistors the bias supply circuit needs to have a very low impedance, which often creates problems for the designer due to resonance. But in the case of LDMOS, the bias circuit can have a much higher impedance, thereby considerably reducing the risk. The component count for an LDMOS bias circuit is often much less, saving on board space and cost.
Q On the subject of costs, there is a common belief that the packages used for LDMOS should cost less than those for bipolar, because an LDMOS die needs no isolation to the case. Is this true, and what do you think about the cost relation in the future?
A Today the package price is approximately the same – in fact LDMOS packages can be slightly more expensive. However, the fact that no isolation between die and flange is needed opens up the possibility to design new packages for LDMOS which should work out to be less expensive than bipolar junction transistors.
Q RF designers have discussed future cellular systems that will operate at frequencies as high as 3GHz. What technology do you think will be used for these power amplifiers?
A At these frequencies, ‘improved’ bipolar devices such as silicon-germanium types might be the choice. Today, GaAs FETs are used but they are negative on the component cost side. Improved LDMOS is very interesting but there is a long way to go before it will work at 3GHz. Another exciting technology is silicon carbide. It works well in high power applications at 3GHz and above, but manufacturing is difficult.
Q How does the ruggedness of an LDMOS amplifier compare with a bipolar one?
A Because of its temperature characteristics, LDMOS exhibits self-regulation when its output power is increased above a rated level, and so is generally more rugged against load mismatch. Similarly, it also offers greater ruggedness against input overdrive. However, prior to mounting, the device ruggedness of LDMOS is inferior to bipolar due its sensitivity to ESD (electrostatic discharge). As a result, the manufacturing production line has to take account of this.
Q Comparing data sheets, it would appear that LDMOS offers a higher drain efficiency than bipolar. How would the overall efficiency for a 60W GSM base station compare?
A At frequencies below 1GHz, LDMOS offers a higher efficiency because of its higher gain. However, above 1GHz, bipolar is catching up. It is worth noting that transistors are normally compared at the 1dB compression point. But if you compare transistors with the same saturated power you will find that LDMOS compresses earlier.

Leave a Reply

Your email address will not be published. Required fields are marked *