High-Power RF Multiplexer

tch.

y-

It looks like I would need to use an RF Directional Coupler type of architecture, to divert a portion of the RF power to the RF-to-DC converter. There's a nice-looking discussion of transformer-based RF directional coupler design (with free software) at

Would it work to use a "standard" lab-type directional coupler, with

Yes, you can do that as well. But all that adds cost because now you have to install yet another module. Question: Is this for a product or just a one-time thing where cost really doesn't matter much? What's the total frequency range over which it must perform?

itch.

ely-

Thanks, Joerg!

I am still having trouble getting my head around the interface between the 25W of (say 2 GHz) RF and the RF-to-DC converter circuitry. Everything else will already be 50 Ohms, with N-type connectors. Would I have to use something like a directional coupler? Or am I missing something obvious?

en

an

en

nd

le

witch.

rely-

Hi Joerg,

It's not for a commercial product and is for support of a very good cause. But if it looks like it would work well-enough we might use over a hundred of them. So unit cost would eventually be a concern.

For the operating frequency, we can make it a little easier and say that it would probably be a fixed frequency band of about 100 MHz width, somewhere between 1 and 2 GHz.

Right now, I want to try to decide how feasible it might be, and would like to try to decide how it should be done, and then try to get a couple of engineering development models designed and built.

Regards,

Tom

between

than

no

that

switch.

=20

=20

There is a way to deal with that, use a shunt mode regulator. Nearly=20 constant power draw. And a relatively big "swamping" capacitor.

entirely-

between

the

than

the

between

no

unit

and

couple

used?

that

switch.

concern.

so.

that's

be

good

windings,

regulate

frequency,

any/

get

switcher

for

can

available

entirely-

thought

ideas

A directional coupler may or may not be a good idea. In either case i=20 would expect not more than 6 db, maybe less. The advantage is less AM caused by your leach device.

between

than

between

no

and

couple

used?

that

Wasn't there a recent thread where 100 W, GHz PIN diode switches=20 were discussed?

Well, you essentially need a fast rectifier diode with very low capacitance. Maybe two in series of this type (with leakage resistors across each):

John Larkin is quite the expert on fast diodes, maybe he knows a better one off the top of his head.

You can simulate that in SPICE. The diode needs to see a DC path, meaning there should be a choke to GND on the RF or some other sort of DC path. The other side of the diode dumps into a small ceramic cap. Make it large enough so the ripple is acceptable at max load and min frequency but not much larger than needed. This allows a fast ramp-up. Again, make sure your RF source doesn't fry when the switch hasn't reached a stable state yet.

See the other post. I'd pick a diode, then simulate things on SPICE. Then build up the rectifier part. Now pick a buck converter chip, and if you use one from LTC you can simulate that as well. This leads to a better design confidence before building stuff. But buck regulators are really run-of-the-mills these days, no need to simulate until the cows come home.

But keep in mind the AM modulation of the RF signal that this causes.

You do realize of source, the switching will produce sidebands relative to the rate of switching? This would be relevant if you are feeding an antenna for communications purposes.

een

han

een

no

and

ple

d?

at

switch.

..

's

T*

s,

irely-

ht

Oh wow! It simulates brilliantly.

Is part of the idea to use the right number and type of diodes in series to get a 50 Ohm input impedance at a frequency of interest?

Just playing around, I used some zener models that were already in my LT-Spice library (maybe altogether unsuitable for this application, some big 10A MBR20100CT, Cjo=3D10p) and found that putting three in series got the impedance around 50 Ohms at 1 GHz (but only about 25 Ohms at 2 GHz). And with a .01uF to gnd downstream, and then a 1 Ohm series R and a .022uF to gnd, then a 1k to gnd to simulate a 2.33-Watt load (plus 100uH to gnd, before the diodes), and with input of 50V peak at 1 GHz, I get 48.3 Vdc out with 63 uV ripple, with a rise time to 90% of that in 70 ns. (I had no idea what value of leakage resistor to use across each diode, but tried 10k, 100k, and 1 Meg and didn't see much difference.)

If I could find a good way to make the input peak voltage a little over 16 V, instead of the 50 V from the 25W of RF in 50 Ohms, this way would quite fast in getting to the needed DC voltage, with the hope of not frying the pin diodes in the switch.

Thanks, so much, Joerg. It looks like I'm getting closer to being on the right track.

Regards,

Tom

Interesting point. Luckily it shouldn't matter much, for my application.

Thanks,

Tom

rote:

ween

than

he

ween

no

and

uple

ed?

hat

switch.

...

's

T*

be

gs,

te

le

tirely-

ght

as

switch.

But don't forget that in real life capacitances are higher, inductances creep in, and the cookie never tastes as good as the TV ad promised (but brandy beans do!).

No, no, the goal is to keep it as high as possible and the capacitive load as low as possible. Anything that you connect in parallel will disturb your RF path and mainly do two things: Add capacitance and add real load (because you are tapping off real power). So the total Z the transmit stage sees drops below 50ohms meaning there could be reflected power.

Zeners? They would not be suitable as rectifiers here. You need real RF-diodes. Low in capacitance yet with enough current rating. If you use buck converters the current won't be so bad because you can easily get

For leakage you have to consult the spec of whatever diode you select and make sure the resistors handle the worst case value, where one diode shows worst case leakage and the other none.

You'd need two, one set for 5V and one for 15V. They are hysteretic, meaning the loop compensation efforts will be zilch and the trade-off is ripple on the output. But you can either filter that our via LC or the extra ripple may not even matter.

Glad to help. Whatever you do think the following through:

a. Is it ok that the RF switch will be in some sort of limbo for the brief period after turn-on before the regulators deliver 5V and 15V?

b. Will the RF source behave itself while the switch is thrown in the other direction?

c. Any other condition that could cause the RF source to go bzzzt ...

between

the

more than

toward the

between

and no

unit

design and

couple

used?

appnotes that

to switch.

the

see

gaps

concern.

limbo"

don't

so.

that.

spec

that's

state.

should be

good

two

my

feed

use a

rectify,

windings,

regulate

frequency,

any/

here?

range

get

switcher

for

varying

So

can

can

available

smoothing

entirely-

you

RF

thought

ideas

between

AM

=46eed your rectifier with about 6 dB less signal than comes out of the = transmitter=20 (6dB directional coupler). Or if you are space or parts limited, just a =

tween

he

than

the

tween

d no

it

n and

ouple

sed?

that

o switch.

ps

rn.

o"

't

so.

at's

HUT*

be

od

ngs,

ate

y,

y/

ge

get

her

for

can

an

ble

ntirely-

ou

RF

ught

eas

Thanks again, Joerg! Sorry I couldn't reply sooner. I had to go on travel for work, unexpectedly, this week. I just got home, finally, and need sleep.

OK. I don't know why I was thinking so stupidly about the impedance, since it's in parallel.

Yeah, I know I will need real RF diodes. I was just playing around and was also still thinking wrongly about the input impedance.

I'll give it another go after I've had some rest.

Thanks again,

Tom

rote:

between

the

re than

d the

between

and no

unit

ign and

couple

used?

s that

to switch.

-78...

the

ee

aps

ern.

bo"

n't

so.

ec

hat's

PHUT*

ld be

good

two

my

e a

fy,

dings,

ulate

ncy,

any/

re?

nge

get

cher

for

ng

can

can

lable

ng

entirely-

you

RF

hought

ideas

AM

nsmitter

a 2:2:1

Thanks! But only 6 dB down seems like I will be leeching-off too much power. That would be about one-fourth of my 25 Watts, or about 6.28 Watts. Will I need to take that much, to end up with 1.4 Watts? On the other hand, if it was out of 50 Ohms that would give about 17.7 volts, which would be handy for getting 15 volts. Is that why you said 6 dB?

Tom

transmitter

Don't rely on stable voltage levels here. A wee change in the antenna (bird sits down on it, etc.) or whatever is connected can change that voltage substantially. You really need switchers here if you want to get by with a minimum of leeched power.

Is the RF on Coax, or in a waveguide? If it's on coax, why don't you insert some DC at the transmitter and pick it off at the switch? A coupling capacitor at each end and coils with a SRF below the frequency used by the transmitter to insert and remove the DC. That way you don't lose less than 1% of the RF.