Choosing transistors for radio frequency applications

[apis]

I've been trying to figure out how to select transistors for radio frequency circuits but not being able to find much information on the web. There are many radio circuits and plenty of guides of how to bias transistors and different amplifier configurations but not so much on selecting the right components for high frequency applications. A lot of the information that is out there is contradictory and some people just recommend to experiment and see what works (but that is a little bit cost prohibitive and you don't learn that much unless you have an idea about what is going on).

I finally found this PDF guide that seems to cover the basics of how to choose a transistor for HF and VHF/UHF at least. 

I thought id post it here if someone else is interested: http://www.aoc.nrao.edu/~pharden/hobby/HG_DS1.pdf

[German_EE]

Look in the data sheet for a property called ft which is the theoretical maximum frequency of operation. I normally pick a device with an ft that's twice the frequency I'm working at.

[vk3yedotcom]

A few common ones to get you started (some are quite old but you'll come across them in old circuits).

2N3904, BC548, 2N2222A - small signal transistors OK for RF  up to maybe 50 MHz (though an FM bug at 100 MHz will also be OK).  2N2222 a bit more powerful than the others.

MPF102, 2N3819, 2N5459 etc - small signal FETS. OK for RF amps up to maybe 150 MHz.

BF981, MPF121, MFE131 - good for VHF/UHF preamps

2N2369 - good for up to maybe 200 MHz - puts out a few tens of milliwatts - good for low level tx.  2N5179 also common in some countries.

2N3053, BFY50, 2N3019 - good for about a watt output in the lower HF range (up to about 10 MHz)

BD139 as above but nearer to 2w output

2N3866, 2N4427 - good for about a watt output up to maybe 200 MHz

2N3553 - as above but more like 1.5 to 2w out

IRF510 - highly capacitive input but can be made to operate on 7 or 14 MHz - or even higher.  Can give up to 3 - 5w at 12v and much more with 25v. Easy to blow up if bias is excessive.

A  basic low power transmitter for 3.5 or 7 MHz will have a few 2N3904 for the low level stages, then a 2N2222 then one or two 2N3053, BD139 or IRF510 for the final amp.  VHF might be 2N2222 (for say a crystal oscillator at a lower frequency), 2N2369 for freq multipliers or low level RF in the tens of milliwatts region, then 2N4427 in the hundreds of milliwatts, then 2N3553 for a couple of watts.

An HF or VHF receiver might have an MPF102 in the front end as an RF preamp, or even a dual gate MOSFET.  A receiver for the lower HF bands may have a 2N3904 instead, if an RF preamp is needed at all. 

These walkthroughs of simple transceiver circuits show where some of the above devices are used. 

[T3sl4co1l]

Note that fT is peculiar to the method in which it is measured: output [AC] current (for AC grounded collector) divided by input current (don't care about base voltage).

You can still obtain useful voltage or power gain at frequencies above fT, especially if you cancel out Ccb (and Cce if applicable) with tuning and neutralization.  Or if you use a common-base amplifier with more input current than DC bias (something a cascode cannot do, but a transformer coupled circuit can).

It's interesting that, although this approach works particularly well for dominant capacitive components (like vacuum tubes and MOSFETs), BJTs have the additional phenomenon of recombination, which fundamentally limits how fast they can switch.  This is a large contributor to fT, so -- although it might not be strictly true that fT is a hard stopping point, it's probably just not worth it to push those extra few MHz.  And so, typically, you see transistors with fT of GHz, being used at 100s of MHz where the gain is high and performance is predictable.

(Vacuum tubes are pretty ridiculous about this aspect.  A typical DC wideband amplifier will roll off in the MHz.  But the tube itself might be good to 300MHz or more, if you merely add tuning!  Most semiconductors can't pull off such a feat; but they wouldn't want to, since their fundamental gain is so much higher that you get the added range for free!)

Tim

[commie]

I'll make the assumption you are talking about transmitting rather than receiving?, two things to remember, 1) always match your impedance when going from stage to stage (maximum power theorem)  and 2) using small rf formers get your source impedance as low as possible.

At the input of every transistor is a series cr circuit which determines your bandwidth, there's not much you can do about c but there is something you can do about r and that's why it's necessary to get your impedance down as low as possible.

 

[commie]

I've been trying to figure out how to select transistors for radio frequency circuits but not being able to find much information on the web.

Transistors designed primarily for rf are very very expensive, so the trick is not to use them but instead use general purpose and/or switching transistors. There are techniques you can use which will make them work at hf.

Commie

[edavid]

The TO39 classics like 2N3866 and 2N5160 have gotten expensive... any good alternatives?

(I mean general purpose parts I can keep in stock, with a similar f_t, in a cheaper package.)

[commie]

The TO39 classics like 2N3866 and 2N5160 have gotten expensive... any good alternatives?

It depends at which frequency you interested in, I assume hf (dc~30MHz), you can replace the 2n3866 with 2n2219a at much lower cost and will work no problem on the hf band.Be careful of buying 2n3866 on ebay, there are alot of fakes around parading as 'Motorola'.

Commie

[KJDS]

Generally, for a given silicon topology then you can scale the drain or collector length to give higher power at the expense of higher noise figure and operating current. Siemens produced a lovely app note about it in the early 90s, or possibly the late 80's.

If you can provide a good description of what you want to achieve then I'm sure someone can suggest good cheap devices for your needs.

[rfeecs]

What is it with the hobbyist's love for 40 year old technology?  2N3904??  TO-39?? 

The OP mentioned UHF, which goes up to 1 GHz.  These days RF goes well into the microwave bands, covering wifi, cellular bands, etc.

For output power under about a watt, you can get an actual, modern RF transistor for a buck or two.  You could maybe look at Digikey:
http://www.digikey.com/product-search/en?FV=fff40015%2Cfff802eb%2C1c0002%2C1c0003%2C1c00c5%2C1c00fa&mnonly=0&newproducts=0&ColumnSort=1000011&page=1&stock=1&pbfree=0&rohs=0&quantity=&ptm=0&fid=0&pageSize=100

Yes, you can get up in the hundreds of dollars for parts that put out hundreds of watts at 1 GHz.  Because you will need something better than a 2N3055.

Another nice thing about real RF parts is they are usually specified with real RF specifications, like noise figure, gain, and output power at some target frequency.

There are also nowadays plenty of inexpensive RFICs and MMICs, even high speed differential amps and op-amps.  Discrete transistors are used less and less as time goes on.

[apis]

Thanks for all the replies!

I don't have a HAM licence so I'm only allowed to make recievers.

I have been wanting to try and make a AIS receiver for my family's boat. AIS is transmitted with GMSK-modulation at 161.975 MHz and 162.025 MHz. I'm guessing I would need a low noise pre-amplifier and then some buffer before mixing the signal down? I was planing on building an ordinary FM radio receiver first though since the demodulation is more straight forward (or maybe it isn't, but FM is more familiar to me) and it uses similar frequencies.

You can supposedly receive weather image data from both satellites and radio stations which could be fun as well. Not sure at what frequencies.

Then there is the 433 MHz ism band that all kind of cheap sensors use, would be nice to be able to log some of that data to a micro controller. I know there are cheap 433 MHz modules, but that would be cheating.

It would also be cool to calibrate my OCXO with the DCF77 signal from Braunschweig, but that is broadcast on 77.5 khz so in that case it's more about the antenna I suspect.

I have also been planning on building this counter:
http://lea.hamradio.si/~s53mv/counter/history.html
more details:
http://www.vhfcomm.co.uk/Counter.pdf
But I haven't been able to find all the parts for the amplifier stages so I need to figure out what replacement parts to use, especially for the "ERA-2 MMIC" That part is way over my head unfortunately. The digital and the low frequency stuff I am familiar with though. The design is from 1998 so maybe there are better chips to choose from now?

A few common ones to get you started (some are quite old but you'll come across them in old circuits).
<...>
These walkthroughs of simple transceiver circuits show where some of the above devices are used.

Thanks for the list. The "Beach 40" radio looks very nice! I might try and build the receiver part. I'm not entirely sure about the function of the bifilar toroid after the 2n2222 stage though, the other parts I think I understand.

For output power under about a watt, you can get an actual, modern RF transistor for a buck or two.
<...>
There are also nowadays plenty of inexpensive RFICs and MMICs, even high speed differential amps and op-amps.  Discrete transistors are used less and less as time goes on.

They were actually cheaper than I remembered, but the problem is there are thousands to choose from and if you don't know which parameters to look for it's difficult (and eventually gets expensive to buy new ones). RFICs are nice if you wan't to get something done quickly, but not as much fun, it's like buying a complete radio.

[commie]

What is it with the hobbyist's love for 40 year old technology?  2N3904??  TO-39?? 

Well, 2n3904 have a complement 2n3906 and both cost pennies to buy in quantity. The to-39 are useful because they can dissipate 3~5 watts of power with a heatsink, and they are all easy to work with.

For output power under about a watt, you can get an actual, modern RF transistor for a buck or two.

The only reason multi GHz smd transistors are available at low cost is because of mobile phone manufacture which has significantly reduced the price of these transistor devices.Another important point of these transistors is that they are characterized for the GHz band only limiting there use other than what they are supposed to be used for.

[AF6LJ]

Another good transistor for both HF and low to mid VHF
The 2N5109 this is good for linear service although it may be a bit pricy..
www.mouser.com/ds/2/68/2n5109-16404.pdf

[commie]

Another good transistor for both HF and low to mid VHF .The 2N5109 this is good for linear service

Yes, I have few of these in stock, however I find them a bit 'harsh' because of ft=1200MHz and Vce_max.=20V where as the 2n3866 has a Vce_max.=30V and ft=400MHz so it's... nice and 'softer'.

[rfeecs]

I have also been planning on building this counter:
http://lea.hamradio.si/~s53mv/counter/history.html
more details:
http://www.vhfcomm.co.uk/Counter.pdf
But I haven't been able to find all the parts for the amplifier stages so I need to figure out what replacement parts to use, especially for the "ERA-2 MMIC"

A likely replacement to the ERA-2 MMIC is the ERA2+ from Mini-Circuits for less than $2 each: http://www.minicircuits.com/pdfs/ERA-2+.pdf.  It is an HBT Darlington gain block.  There are tons of equivalent parts available from distributors.  You can use their parametric search tools to try to narrow it down.  You will find you can often buy an MMIC amplifier for less than you could put together one from discretes.

Well, 2n3904 have a complement 2n3906 and both cost pennies to buy in quantity. The to-39 are useful because they can dissipate 3~5 watts of power with a heatsink, and they are all easy to work with.

...

The only reason multi GHz smd transistors are available at low cost is because of mobile phone manufacture which has significantly reduced the price of these transistor devices.Another important point of these transistors is that they are characterized for the GHz band only limiting there use other than what they are supposed to be used for.

Most of the applications the OP mentions are above 100 MHz.  A part like the 2N3904 doesn't apply at all.  But I understand it may be great for some applications.  IMHO, the TO-39 is obsolete.  It's only advantage is that it is hermetic, and unless you are doing a high rel space application you don't need it.

For RF, surface mount is your friend and becomes a necessity at higher frequencies.

None of these transistors are used in mobile phones.  Mobile phones all use MMICs for the RF front end.  Yes, mobile phones have spurred the technology development that has led to lower prices.  But the semiconductor die is often the cheapest part of a discrete transistor.  A wafer has maybe 100,000 die, all batch processed at once.  So divide the cost of the wafer and process by 100,000.  The die costs pennies, maybe less than one cent.  The cost is in the packaging and testing and handling.  Of course for power devices the price goes up because die are bigger and packages are more expensive.

The fact that RF transistors may be spec'ed at a different frequency than you are using doesn't mean it won't work at a lower frequency.

I'm just saying use the right part for the job.  It may not cost more and it might be easier and better.

[AF6LJ]

Another good transistor for both HF and low to mid VHF .The 2N5109 this is good for linear service

Yes, I have few of these in stock, however I find them a bit 'harsh' because of ft=1200MHz and Vce_max.=20V where as the 2n3866 has a Vce_max.=30V and ft=400MHz so it's... nice and 'softer'.

Not sure what you mean by harsh, I have found them easy to work with.

[commie]

Not sure what you mean by harsh, I have found them easy to work with.

Parasitic oscillation hazard.

[AF6LJ]

Not sure what you mean by harsh, I have found them easy to work with.

Parasitic oscillation hazard.

Oh I get it...
That usually can be avoided by careful attention to layout.
They really are not good parts for the beginner in that regard.

[KJDS]

For non-critical stuff at VHF, I quite like the BFS17, Ft is 1GHz

For UHF, the BFR93 has an Ft of 6GHz.

If I want low noise, then it's probably an MGF series pHEMT, but these will oscillate unless you spend at least a day getting the design right, but for sub .3dB noise figure at 1GHz it's worth it.

[T3sl4co1l]

The fact that RF transistors may be spec'ed at a different frequency than you are using doesn't mean it won't work at a lower frequency.

Although come to think of it, there are "pretuned" parts out there, that are designed to solder into a given fixture (usually narrow band) and give reasonable impedances (like, around 50 ohms nominal) in that frequency range, and.. shit all elsewhere.  I think these are usually for common bands, like 2.45GHz.  Where you kind of don't want to have to muck around with tuning and matching anyway, so they're useful.

But yeah, the datasheet will tell you that, at least.

On a related subject, being able to read the scattering (s-) parameters is pretty helpful for working with RF devices, which are rarely specified in normal means (like capacitance and resistance).  Specifically: a reflective and capacitive input (s11) or output (s22) coefficient means capacitance is dominant on that port; typical examples being FETs and most outputs (i.e., pretty much anything besides a vacuum tube triode -- which might instead have a fairly reasonable match (s22 ~ 0) due to its low plate resistance).  s21 is forward gain, and s12 is reverse transfer (predominantly D-G or C-B capacitance, but often exacerbated by lead parasitics).

Tim

[rfeecs]

The fact that RF transistors may be spec'ed at a different frequency than you are using doesn't mean it won't work at a lower frequency.

Although come to think of it, there are "pretuned" parts out there, that are designed to solder into a given fixture (usually narrow band) and give reasonable impedances (like, around 50 ohms nominal) in that frequency range, and.. shit all elsewhere.  I think these are usually for common bands, like 2.45GHz.  Where you kind of don't want to have to muck around with tuning and matching anyway, so they're useful.

But yeah, the datasheet will tell you that, at least.

That's a good point.  Often large power devices have internal matching because it is difficult or impossible to match over a decent bandwidth without putting matching elements right at the device inside the package.  If you look inside these, they have tons of bond wires and a few capacitors for matching (L-C-L), using the bondwires for inductance.  You can see why these things cost a fortune:

[rfeecs]

On a related subject, being able to read the scattering (s-) parameters is pretty helpful for working with RF devices...

The classic app note on S-parameters:  http://literature.cdn.keysight.com/litweb/pdf/5952-0918.pdf?id=922621

[apis]

Thanks again everyone! I will try to get some of the 2N5109 and BFS17 as well, and read up more on s-parameters.

For jFets they don't always seem to specify a transition frequency, how do you figure out what frequency range they are intended for?
Example: http://www.digikey.se/product-detail/en/2N3819/2N3819CS-ND/4806873

And this BJT sounds too good: Ft=85GHz, less than 1db NF and it cost less than a dollar, is there a catch? (Aside from the higher risk of parasitic oscillation that commie mentioned).
http://www.infineon.com/dgdl/Infineon_Robust_WiFi_SiGe_LNA_BFP840FESD_Rev1_1.pdf?folderId=db3a30431f848401011fcbf2ab4c04c4&fileId=db3a304336ca04c90136e88f755435bf

A likely replacement to the ERA-2 MMIC is the ERA2+ from Mini-Circuits for less than $2 each: http://www.minicircuits.com/pdfs/ERA-2+.pdf.  It is an HBT Darlington gain block.  There are tons of equivalent parts available from distributors.  You can use their parametric search tools to try to narrow it down.  You will find you can often buy an MMIC amplifier for less than you could put together one from discretes.

Thanks, that sounds like it's the right one!
I can't access their website right now though, will look on it later. It looks very easy to use?

If I want low noise, then it's probably an MGF series pHEMT, but these will oscillate unless you spend at least a day getting the design right, but for sub .3dB noise figure at 1GHz it's worth it.

Are these the ones (could only find them on ebay): http://www.ebay.com/itm/5pcs-Original-MGF4919G-SUPER-LOW-NOISE-/280897340734?pt=LH_DefaultDomain_0&hash=item4166c9453e
Would it make sense to use one of those for vhf?

[KJDS]

And this BJT sounds too good: Ft=85GHz, less than 1db NF and it cost less than a dollar, is there a catch? (Aside from the higher risk of parasitic oscillation that commie mentioned).
http://www.infineon.com/dgdl/Infineon_Robust_WiFi_SiGe_LNA_BFP840FESD_Rev1_1.pdf?folderId=db3a30431f848401011fcbf2ab4c04c4&fileId=db3a304336ca04c90136e88f755435bf

It's got an absolute maximum supply voltage of 2.2V, and looks internally matched to 5.8GHz, though not in a very narrowband way. Don't try using it unless you're confident with the subtleties of RF design or need a 5.8GHz discrete amp.

If I want low noise, then it's probably an MGF series pHEMT, but these will oscillate unless you spend at least a day getting the design right, but for sub .3dB noise figure at 1GHz it's worth it.

Are these the ones (could only find them on ebay): http://www.ebay.com/itm/5pcs-Original-MGF4919G-SUPER-LOW-NOISE-/280897340734?pt=LH_DefaultDomain_0&hash=item4166c9453e
Would it make sense to use one of those for vhf?
[/quote]

It doesn't make sense to use one at VHF unless you're looking for sub 0.2dB noise figure, so unless you're doing EME work, or have some other very specialized need, then don't bother. I spent a few weekends when working abroad with a need for something to stop me getting bored doing a series of designs using the MGFs. The S11 best match and best NF are a long way apart, so there's lots of tricks needed to get a reasonable match and reasonable noise figure, and a whole heap of work to have some confidence that once built you'll be able to find a way to ensure it won't oscillate. It's a bit like trying to use a dragster to go to the supermarket, possible, but dumb.

[apis]

Ah, maybe not the best part to begin with then.