Mosfet, Mesfet vs Bipolar transistors in RF circuits

[artis]

I was wondering in general what types of semiconductors are used in most solid state RF power amplifiers like the ones at cell base stations and elsewhere. Do they use mosfets etc which can only switch a square waveform or are bipolar ones also used that can output sinusoidal outputs that can be directly fed into an antenna?

Do mosfet amplifiers have an output filter much like the lower frequency class D audio amplifiers or does the antenna function as a filter at such high RF frequencies?

I imagine for example an RF amplifier outputting a QAM modulated wave into an antenna so that the signal at the antenna looks like higher and lower amplitude half periods of a sine, but if the amp works on mosfets, does the signal at the transistor output is the same amplitude but square ?One of the reasons I ask this is because I have had limited experience with RF circuits and I am building a est device where I would want to excite an RF cavity that has no capacitance but only inductance associated with it and I would like to drive it with an RF signal that is sinusoidal so I am wondering do bipolar transistors come with high enough power levels at RF frequencies and aren't they too inefficient due to the sinusoidal operation mode vs the on/off of the mosfets etc?PS. at RF frequencies is there any place where a square wave is used or is it all just sinewaves and modulated sinewaves?
I guess one cannot even achieve a square wave radiating from an antenna but how about some special purpose RF circuits?

 

[Baluncore]

Do they use mosfets etc which can only switch a square waveform or are bipolar ones also used that can output sinusoidal outputs that can be directly fed into an antenna?

Can you please provide a reference to that square/sinewave restriction at RF ?

 

[tech99]

I think all devices can operate either in a linear or switched mode. If you have a signal which relies on amplitude variations, such as AM or QAM then the device must be operated in a linear mode. This usually means adjusting the bias so it is slightly "on" with no signal. If the signal is just on or off, or maybe is frequency modulated, then the device can operate more efficiently by using it in a switched mode, or something approaching that (such as Class C). If you use a switched mode, the output filter (or tuned circuit) will remove the harmonics, so we still end up with a sinusoidal carrier. However, we are unable to amplitude this carrier. Your cavity will remove harmonics, so the waveform will be sinusoidal even if the cavity is driven with a square wave carrier.

 

[davenn]

I was wondering in general what types of semiconductors are used in most solid state RF power amplifiers like the ones at cell base stations and elsewhere.

MOSFETS or variations there-of are the common devices used

Do they use mosfets etc which can only switch a square waveform

This is completely incorrect ...
Not sure where you got your idea from ?

Many/most transmitters and receivers these days ( and for quite some years)
use MOSFETS and and perform very well with analog signals
ALL my transceiver gear has MOSFETS for the RF power amplifier stages

cheers
Dave

 

[alan123hk]

I was wondering in general what types of semiconductors are used in most solid state RF power amplifiers like the ones at cell base stations and elsewhere

Looking at the list on the sales page, there are several types of RF transistors, including bipolar, JFET and MOSFET.

https://www.mouser.com/Semiconductors/Wireless-RF-Semiconductors/Transistors-RF/_/N-ax2e7

 

[DaveE]

Looking at the list on the sales page, there are several types of RF transistors, including bipolar, JFET and MOSFET.

https://www.mouser.com/Semiconductors/Wireless-RF-Semiconductors/Transistors-RF/_/N-ax2e7

Yes, but the question was about power amps. This is mostly the domain of MOSFETs now days, unless the frequency is really high (GaAs, GaN, etc.). I've never heard of a power JFET.

 

[davenn]

. I've never heard of a power JFET.

Did you not read any of the data in Alan's link ?
They show many high power, GHz frequencies JFETS
here's just a few of them...

 

[artis]

@Baluncore and @davenn Well I think I misspoke, I did not intend to say that mosfet driven amplifiers can only output square waveforms but I assume that internally the very transistor itself works in a on/off square wave fashion does it not? Because after all that is the very reasoning behind saving energy and heat dissipation.But like @tech99 said the output filters or reactance of rf cavities or antennas will "round off" a signal so that eventually it resembles a sine.

 

[DaveE]

Did you not read any of the data in Alan's link ?
They show many high power, GHz frequencies JFETS
here's just a few of them...

View attachment 271779

Oops, yes you're correct. I tend to put the GaAs, GaN, and SiC devices in a different category, but yes they are JFETs. I was thinking of Si devices.

 

[Baluncore]

Well I think I misspoke, I did not intend to say that mosfet driven amplifiers can only output square waveforms but I assume that internally the very transistor itself works in a on/off square wave fashion does it not?

No it does not. It is more like a voltage controlled variable resistor. A vacuum tube is just a FET with a pilot lamp. They both do the same jobs. Vacuum tubes worked well in linear circuits for 50 years.
It takes only about 0.5 volt to switch a BJT between on and off. A MOSFET requires close to 10 times that gate voltage. I see no reason why you would want to use a MOSFET for square wave switching at RF. MOSFETs have a higher gate or miller capacitance than the base of a BJT or the grid of a VT, so turning MOSFETs all the way on then all the way off at RF, would be quite inefficient.

 

[artis]

@Baluncore hmm I see your making a point here. Well vacuum tubes were used for both non linear and linear operation just that in linear operation like a class A amp the lamp current was never allowed to go below a certain level so that the lamp is always on. In the Mosfet analogy to be honest I've never heard of a mosfet amplifier circuit where the mosfet is always on because the mosfet very function is such that it's either conducting or not as inbetween it has a higher internal resistance which would dissipate a lot of heat if used under high power. I've even learned this from experience when using mosfets in power supplies and the gate voltage goes "bad" and turns the fet only partially on.
Sure a mosfet can be always ON like in a ssd lamp switch but then it's either always off or on not in between.I do see your point about the gate capacitance, because the amount of energy required to charge/discharge that gate capacitance is a function of the capacitance itself and the number of times it is discharged/charged so when the frequency gets really high one would need a substantial power put into "driving" many such parallel fets , is this your point?

PS. I feel like I don't know something here because when speaking about mosfets I'm typically thinking the larger power types used in smps and other switching circuits and all of them work in the fully on/fully off manner, but from your language I suspect that there might be something different in RF circuits with mosfets, can you elaborate?

 

[DaveE]

the mosfet very function is such that it's either conducting or not as inbetween it has a higher internal resistance which would dissipate a lot of heat if used under high power.

This is true for any type of device when operated as a linear amplifier. That is the point, you turn it partly on and it will dissipate heat. This has nothing to do with the structure of the device (BJT, FET, Tube, etc.) it is the design intent in the circuit.

 

[DaveE]

I once designed a very large linear power supply which was made of 48 large MOSFETs (SOT-227) in a series/parallel arrangement. It was water cooled and could (was supposed to) dissipate 5KW of heat. It was very reliable (with N+2 redundancy) and is still in production. Trust me, you can operated MOSFETs as linear devices.

One issue I had was that I designed for a 200KHz bandwidth, but had to deal with some parametric (i.e. sharing) oscillations at frequencies as high as 200MHz with these devices, which were not intended for RF applications.

edit: BTW, there are some large MOSFETs with multiple die that do not work well in high power linear applications. This is about device construction and selection and not intrinsic to the MOSFET dice.

 

[artis]

yes sure @DaveE, I somehow got confused , the mosfet can be operated below saturation just that the current through it must be limited I assume otherwise it could destroy the device , so one probably wouldn't want this mode of operation in a typical power supply or am I wrong ?

 

[Baluncore]

PS. I feel like I don't know something here because when speaking about mosfets I'm typically thinking the larger power types used in smps and other switching circuits and all of them work in the fully on/fully off manner, but from your language I suspect that there might be something different in RF circuits with mosfets, can you elaborate?

When it comes to digital on/off power switching, the power MOSFET is best for switching high currents occasionally, while the BJT wins for high frequency switching.

That is because to saturate a BJT requires excess base current flow continuously, for as long as it is on. A saturated BJT will have a Vce of about 20 mV. PN junctions do not do well in parallel since the negative temperature coefficient causes the hottest patch take the greater current, which, due to positive feedback, results in a hot spot.

To saturate a MOSFET only requires the gate voltage change once, but that change takes time and current to charge the gate capacitance. A saturated power MOSFET will have a Vds of about 2 mV. MOSFETs work well in huge parallel arrays because channel resistance shares the current fairly.

With linear and RF circuits, the BJT and the MOSFET generate similar amounts of heat.
FETs are used in low level RF amplifiers because FET circuits can have lower noise characteristics than BJT circuits.

 

[artis]

I guess the best way to put it would be to say that a mosfet can also work in linear fashion but within just a fraction of it's maximum rated (usable at room temps) current capacity which can only be fully used in the non-linear aka switching operation mode.

@Baluncore so I've heard long ago about the bipolar transistor tendency to form hot spots and thermal runaway etc, but how about mosfets , say if operated within a linear region with rather high currents, can't they also break down from similar problems ?

 

[DaveE]

Every part of the data sheet for devices was written for a reason. You must use devices within their intended limits. A common rookie mistake: not reading and understanding ALL of the data sheet.

 

[Baluncore]

... but how about mosfets , say if operated within a linear region with rather high currents, can't they also break down from similar problems ?

MOSFETs work well in huge parallel arrays because channel resistance shares the current fairly.

The low ON resistance of a power MOSFET array comes from hundreds of MOSFETs, all in parallel. Their individual channel resistance is high, but once in parallel the total is low.

A power MOSFET is designed and rated for switching maximum OFF voltage, and for maximum ON current, but only one maximum at the time. The package decides the power rating.
A BJT is rated for maximum power in a linear application. That may have coloured your misconception that MOSFETs are only good for switching high currents.

 

[DaveE]

MOSFETs work well in huge parallel arrays because channel resistance shares the current fairly.

Yes! As you said, in saturation.

In linear high power operation the gate characteristics also matter (primarily Vgth and it's TC). They work well in linear applications when the gate parameters are matched; of course, you get this for devices paralleled on the same dice (or wafer?). You often don't get good matching for multiple die.

At low frequencies, MOSFETs have an advantage over BJTs for linear, high power, moderate to high voltage applications because they (mostly) lack the second breakdown limits of the SOA. But be careful here, MOSFETs not having second breakdown is a myth, they do, it's just normally much better than BJTs; I think because of the parallel cellular resistive current sharing you described. How this applies in RF designs isn't something I ever looked into, but I think it's the same because it also explains rugged switching performance.

 

[DaveE]

I guess the best way to put it would be to say that a mosfet can also work in linear fashion but within just a fraction of it's maximum rated (usable at room temps) current capacity which can only be fully used in the non-linear aka switching operation mode.

@Baluncore so I've heard long ago about the bipolar transistor tendency to form hot spots and thermal runaway etc, but how about mosfets , say if operated within a linear region with rather high currents, can't they also break down from similar problems ?

Time to learn about the SOA part of the data sheet, I think:
https://www.physicsforums.com/file:///C%3A/Users/4dave/Downloads/application_note_en_20180726.pdf

 

[alan123hk]

Do they use mosfets etc which can only switch a square waveform or are bipolar ones also used that can output sinusoidal outputs that can be directly fed into an antenna?

I think that both bipolar and mosfet can also be used for square wave input or sinusoidal input, just pay attention to the impedance matching at the input.

Do mosfet amplifiers have an output filter much like the lower frequency class D audio amplifiers or does the antenna function as a filter at such high RF frequencies?

Generally speaking, the output of the RF power amplifier should use filters, including the LC filter and the influence of resonant frequency/bandwidth of the antenna itself.

I imagine for example an RF amplifier outputting a QAM modulated wave into an antenna so that the signal at the antenna looks like higher and lower amplitude half periods of a sine, but if the amp works on mosfets, does the signal at the transistor output is the same amplitude but square ?

The carrier of the radio signal must not be a square wave, it is impossible to allow an infinite bandwidth radio signal be transmitted into space, but the modulating signal can be approximated as a square wave, and the mosfet can of cause be used as AM amplifier.

One of the reasons I ask this is because I have had limited experience with RF circuits and I am building a est device where I would want to excite an RF cavity that has no capacitance but only inductance associated with it and I would like to drive it with an RF signal that is sinusoidal so I am wondering do bipolar transistors come with high enough power levels at RF frequencies and aren't they too inefficient due to the sinusoidal operation mode vs the on/off of the mosfets etc?

Bipolar and mosfet can also achieve high power, please check the link https://www.mouser.com/Semiconductors/Wireless-RF-Semiconductors/Transistors-RF/_/N-ax2e7

Regardless of the inductance or capacitive reactance of the RF cavity, as long as this parameter is included in the design of the antenna output matching/filtering circuit.

If you are very concerned about efficiency, you can consider using class E amplifiers, but it seems that class E power amplifiers are still difficult to achieve high operating frequencies, such as higher than 100 MHz.

https://cdn.macom.com/applicationnotes/AN4001.pdf
"A class E power amplifier operating at 81.36 MHz has been designed and built using MACOM MRF151A power MOSFET. Using a 48V power supply, the amplifier yielded 300 watts of output power with better than 82% efficiency .."



In general, I think the design of high-power and high-efficiency RF amplifiers is not easy. In practical applications, you may find that it is more complicated than designing small-signal RF amplifiers.

 

[DaveE]

In general, I think the design of high-power and high-efficiency RF amplifiers is not easy. In practical applications, you may find that it is more complicated than designing small-signal RF amplifiers.

I think this maybe the understatement of the century. These are VERY DIFFICULT designs, IMO.

 

[artis]

Well ok so far so good, but I have another question that might come up as a limit to what I would like to make.
Say I have a bunch of small planar rf coils all stacked in a line some 1 meter long, each coil has it's own small (mosfet or bipolar doesn't matter for the argument) transistor that drives the coil. Then each of the transistors would have to be connected in parallel and driven from a larger transistor.
Assume the transistors are capable of for example 10 Ghz operation, so each coil is small and each transistor is right next to it so far so good, but then the traces connecting all the transistors in parallel are rather long, how would this affect the chance of achieving high frequencies ?I guess I am asking , doesn't the paralleling of devices cause problems in terms of frequency limit if the traces used to parallel the devices get too long? as I would effectively create an antenna which is not part of the plan.

 

[Baluncore]

I guess I am asking , doesn't the paralleling of devices cause problems in terms of frequency limit if the traces used to parallel the devices get too long?

Yes they do. No matter what you do at 10 GHz, (λ = 3 cm), with coils spread over 1 metre, you will have an antenna array. What frequency or waveform do you want to transmit?

You need to explain why you need to arrange a 1 metre long, linear array, of loop antennas. Then we can describe a transmission line length independent solution for your problem.

 

[berkeman]

doesn't the paralleling of devices cause problems in terms of frequency limit if the traces used to parallel the devices get too long?

Are you familiar with how RF splitters/combiners and transmission lines work? The biggest issue is attenuation going through the splitters and loss in the TLs.

 

[davenn]

LDMOS are the standard for high power up into the microwave bands these days
capable of power levels up to 1.8kW and 47GHz ( not necessarily in the same package)
Various manufacturers eg NXP have various ranges available

from wiki

LDMOS
LDMOS (laterally-diffused metal-oxide semiconductor) is a planar double-diffused MOSFET (metal–oxide–semiconductor field-effect transistor) used in amplifiers, including microwave power amplifiers, RF power amplifiers and audio power amplifiers. These transistors are often fabricated on p/p⁺ silicon epitaxial layers. The fabrication of LDMOS devices mostly involves various ion-implantation and subsequent annealing cycles. As an example, The drift region of this power MOSFET is fabricated using up to three ion implantation sequences in order to achieve the appropriate doping profile needed to withstand high electric fields.

 

[artis]

@Baluncore and others, thanks for the input so far, I'm not sure whether it is meaningful for me to try to describe the whole idea here now as that would take up some space and before i get the details I'm not even sure whether it's meaningful as an idea. In a nutshell it's a mechanical RF amplifier in a RF cavity form, where you get the amplifying effect due to Lorentz force within a moving conductor that moves through a magnetic field.
The frequency spectrum would still be limited by the size but I think not as much as in a electron beam/bunch driven klystron cavity because here there is only inductance and no capacitance.

But sure enough I need a magnetic field , it doesn't have to be at all places but in those where my conductor will go. So I am thinking about some pcb or other composite material made where there is both the conducting layer and the small coil arrays close together and the transistors that drive the coils right next.
But for a given coil array there would probably have to be a line that connects each transistors down the line and that line could be from a few meters in length to down to 50 cm or less depending on the frequency range of choice.

@berkeman I'm sort of familiar with how that stuff works.
In PESA radars they use waveguides to deliver the RF power to each individual antenna.
In my case I only would need to deliver transistor drive power to each transistor line.

 

[Baluncore]

I think, if you did described your idea, you would find it physically impossible, or that it was some form of transmission line or distributed amplifier.
https://en.wikipedia.org/wiki/Distributed_amplifier

The frequency spectrum would still be limited by the size but I think not as much as in a electron beam/bunch driven klystron cavity because here there is only inductance and no capacitance.

The idea that there can be inductance without capacitance is unreal. Every inductor is actually an LC loop antenna, or a transmission line with capacitance between sections and the environment.

 

[artis]

@Baluncore I never said there isn't any capacitance , just that it is by orders of magnitude less than compared to the traditional RF cavity. In my case you don't have a central beam hole instead you have a full torus.
It would be comparable to having a single loop/ring of wire. There can't be any capacitance within the loop itself as long as the wavelength is longer than the loop length I think because every place on such a loop is short circuited with every other place on the loop.
But there can be generated current within the loop.

AS far as I'm aware there can't be any charge buildup within an inductor if it is electrically short circuited, only when an inductor is in open circuit can the collapsing B field within it produce charge buildup at the ends of the inductor.
The RF cavity inductor is closed by a capacitor which is the reason for capacitance in the cavity but if one were to short circuit the beam hole walls there would be no capacitance within the loop.