Rf switch using pin diodes

Hello Jamie,

Up there stuff like trace impedance does begin to matter.

Nice. Just a suggestion: If you will be doing lots of RF stuff in the future get used to rounded trace turns. At least muffle the 90 degree bends such as that right before the SMA to a couple 45 degree bends like you did with most of the other traces.

Supposedly those are almost as easy as a TV dinner. You know, the stuff you throw in the microwave, press the buttons that the package text advises and it's cooked ;-)

If you go to not-so-standard thicknesses you'll price yourself out of the bargain fabs. Also, you might want to look for a trace calculator that takes into account adjoining copper pour.

Looks cool. The pin diode pair may be a bit tough to solder since there are no thermal reliefs for its pads. That could become a concern should this circuit go into series production.

Typically that's the usual trade-off. Ready-to-go units can be a lot more expensive than taking a nice transistor from NEC or another company and rolling your own.

Regards, Joerg

Hi, Jamie -

I recommend you get a c "...epoxy fiberglass is useful up to about 1.5 GHz before its nonisotropic characteristics and losses become excessive."

Teflon (registered) is good to 10 GHz.

Just thought you'd like to know.

Cheers, John

I might have to use a bigger amplifier to compensate for the extra losses then? :) Not very efficient but I think teflon pcb's are too expensive, not sure though. I guess some people make their own teflon pcb's?

cheers, Jamie

I downloaded most of the NEC website, and will look through the datasheets later, but this one looks like it is a decent candidate:

NE651R479A

Mouser stocks it for $5.10US in small quantities.

cheers, Jamie

Hello John,

Well, yes, but when you open a 2.4GHz phone it's usually all FR4 in there. Even one 5GHz phone had a PCB that looked like some better grade of FR4, certainly not Teflon. Same for the WLAN stuff around here which all runs at 2.4GHz.

Regards, Joerg

I'd like to make the amplifier to take in 1mW at 2.4GHz and outputs

1Watt. None of the NEC transistors I found can do both this high gain and high power output but the more expensive TC3141 does but it is expensive. I guess I will have to use two stages of amplification, from 1mW to 100mW and then from 100mW to ~1W using two of the NEC transistors? Would it be harder to get a working amplifier using two transistors rather than a single TC3141? The evalboard circuitry for the TC3141 looks just as complex as that for the transistors from NEC so I guess the only reason it is so much more expensive is that it has high gain and high power combined? Any suggestions on good amplifiers to use for 1mW to ~1W? :)

cheers, Jamie

Hello Jamie,

At 2.4GHz the transistor that could do 30dB at 1W out ain't invented yet ;-)

With a tube I'd have no problem doing single stage ...

It's going to be at least two stages. These amps aren't rocket science to build but I strongly suggest to pick up a hands-on book about that. Can't recommend the one that I have here because it's in German but I bet the ARRL will have a good book on microwave design. The good old ARRL Handbook might already be all you need, can't say because I do not have the latest edition. ==> John, do you have one?

Regards, Joerg

This is a nice 3 stage amp:

I noticed that NEC/CEL datasheets are pretty short. There is more info on this chip from a press release than in the datasheet! :)

Mouser has it for $5.17US

some more info on it from: "

"

"..and an output stage in a 4-mm-sq, leadless package and provides a typical gain of 44 dB at 1.8 GHz. The

UPG2118K from NEC, available through California Eastern Laboratories, is useful in the 800- to 2500-MHz range;

absolute output power is at least 32 dBm at 1.5W.

Minimum power efficiency for this 2.8 to 5.5V-dc, single-supply device is 42%, and its on-chip active bias .."

Now if I could just find an example circuit for it at 2.4GHz..

cheers, Jamie

Hello Jamie,

But it still wants almost +10dBm input to produce the desired +30dBm out.

It should be feasible. But quite frankly, while at $5.17 it might be ok for a one-up solution I would still do it discrete for any serious production run. There is also less of a chance that a crucial part falls from grace and becomes obsolete. Plus you'd need another preamp anyway to get from your 0dBm level to a watt with this chip. So why not just add a couple more transistor stages to get there?

Regards, Joerg

Is that from the datasheet? I didn't see the gain specified there (or more likely I couldn't understand it) but the EDN link said it has 44dB gain (at 1.8 GHz though) "

"

I've checked the prices on the discretes from CEL/NEC at mouser and they seem to be around the $5 mark too, at least these ones are: "

"

Do you have a link to any cheaper discretes that you would recommend?

I think using discretes is like using 74xx logic gates instead of an CPLD? :)

For a multistage amp I think the max2240 would be a good first stage: "

" It can amplify the 1mW output of the nRF2401A to 100mW and it has

4 levels of gain. Fun to solder too, 0.5mm pitch 9pin BGA - perfect for my toaster oven :)

cheers, Jamie

Hello Jamie,

Yes.

It's been a while but you should be able to find them via the major semi companies. To me bipolar was always the ticket because they are cheaper. The Philips BFG135 comes to mind, IIRC about 50c to 60c at Arrow. Don't know if one would do in the final, you might need two. The lower power versions for the younger stages are much cheaper. Don't go with what was in recent press releases, it's mostly going to cost through the nose. Go through the parameter lists on the mfg sites.

However, you may not be able to buy stuff like this single qty. And with most EU companies, forget samples. Customer service in my experience is... well, I'm not going to use foul language now. Let's say I use most of them only if I absolutely have to.

One thing I have to say though is that the specsmanship of EU companies (especially Philips) is generally excellent. The data sheets are detailed and for RF parts they often give you the complete layout. Kinda takes the fun out of it, doesn't it? Pull the data sheet through Arrow though, when going through the Philips web site you could make a whole pot of coffee and it may still be loading. Or trying to find it. With other companies I have found that their "web servers" couldn't even find their own parts at all at time. Pathetic. Ok, rant mode off now.

Logic isn't truly discrete. It's transistors, diodes and stuff. Most recent grads wouldn't even know what to do with a 2N2907, it is sad.

Well, 100mW would be a piece of cake with most any cheap RF transistor.

With all this keep in mind the FCC and other legal and safety considerations. A mishap with a design could seriously mess with other vital communications. I am not sure 1W is permitted, probably not.

Regards, Joerg

The ARRL UHF/Microwave Experimenter's Manual. It, along with other excellent VHF/UHF publications, can be found on this page:

John

Hello John,

Thanks. That ought to do it and the price is probably a bargain just like the other ARRL publications.

I'd buy it as well but I already got the fat edition (all five books in one) Karl Weiner:"UHF Unterlage". It's about the heaviest book on my shelf. IIRC it was translated into English as "The UHF Compendium". Don't know about today but even my old 1987 copy has GHz stuff in there so UHF may be a bit of a misnomer here. Lots of photos which helps in seeing how to build it all.

Jamie, check ham radio sites and links as well. Radio amateurs are known for their sensitive noses when it comes to finding the best microwave transistor deals out there.

Regards, Joerg

(snip)

I think you should move C4 and 5 close to the chip to reduce the chance of standing waves in the trace between those parts.

There is RF energy near by, fringing off the signal lines and through the switch inter lead and device capacitance. I would want my control signals uncontaminated with it. Since the control current is down in the micro amperes, adding a resistor in series with the control lines near the cap would also damp any ringing induced in the long end of the control trace (which should be thin to reduce its Q if it happens to be a resonant length) without reducing the control voltage much. I'm just trying to imagine anything that could go wrong.

That's what I heard.

The trouble with pessimists is that are right so damn often.

Thanks guys for the links!

I found an alternative to using pin diodes for the RF switch that seemed better in all ways so I made a new board using it.

It uses the upg2009tb SPDT RF switch from NEC: "

"

It is easier to drive than pin diodes, and is only $0.99US at mouser.com. The circuit is simpler too.

Here are some pics of the new circuit/board and the eagle files:

schematic: "

"

layout: "

"

boardfiles: "

"""

I used "coplanar waveguide" for the RF trace impedance calculations from Agilent Appcad with the top ground layer 20mils away from the RF traces and 45mil trace width on 1/32 FR4. It works out to about 50ohms, but I am not sure about how the vias will effect the impedance, so I kept them a bit away from the RF traces, I think they would lower the impedance of the traces the closer they are but I don't know how to calculate that!?

cheers, Jamie

Why would there would be standing waves in those traces? I take your word for it though:

"

"

cheers, Jamie

Hello John,

Doesn't the first law of Murphy say "If something can go wrong, it will"?

Regards, Joerg