signal pickoff

You have passive choices: Reduce the 2pF to 0.5pF by making a 1:4 divider, you can make up for the loss later if required. Add resistance in series.

I guess I could make the loop into an official 3-resistor tee attenuator, but that could have more hazards than a simple pickoff resistor.

I'll think about a capacitive divider too, but I also want to run low,

You may have an easier time achieving a wide bandwidth for the pass-through, than the pickoff. SMT components can be trouble makers. See how the LEE-19 incorporates transmission-line signal paths within its structure.

Huh it's 50 ohms all around? You could make it some sort of attenuator, add some series C if 2 pF is too much. George H.

That much is a given. You can't pick off some energy *and* leave it all to pass through.

Which is what Win suggested as a 4:1 capacitive divider. Please do keep up :)

JL should first accept that he's building a splitter, then decide how ornate to make it. I'd be looking at resistive attenuators, boosted with extra C to balance the buffer/amp input... depending on the importance of flat input and output impedances, and source/load impedance isolation.

Make one of these:

Should just be something like so:

An entirely passive system is not so unattractive. No power supply requuired. Perfectly fine into a scope, etc., or into an SMA amp module, provided by the user.

Joe Dunsmore' coupler is 16dB (with 1.6dB through loss), but it's trivial to modify the ratios. It's easy and cheap to build from

The trade-off for broadband performance and directionality (port isolation) is that it's resistive; up to half the input energy is simply dissipated. If you don't need it to be directional, you can reduce the losses, but if you don't want to either measure or pass reflected power, you have to dissipate it.

Example through/coupled dB ratios:

The 6dB version is limited at the botton end by the ferrite behaviour, bit that can be improved using VK3DIP's technique with binocular cores, as ripped off in:

Note that this is a faulty Chinese copy of a Ukrainian ripoff of VK3DIP's design, but fix the errors and fit nicely-trimmed semi-rigid hardline and the performance is remarkable from 100KHz to 3GHz. VNA test results on request.

Better geometry and smaller resistors than 0805 would push that to 10GHz.

Clifford Heath.

What happens if you mount a very low C fet gate direct onto the co-ax between the connectors via an SM R? Then make the co-ax a planar pcb thing, as I expect it is.

A decent compromise is a pickoff that uses a 100 ohm to 50 ohm divider, with a small inductor in series with the 50 to AC ground. That gives me 1/3 of the customer voltage for my IC (a comparator) and a bit of high frequency peaking where I probably want it. The damage to the thru signal is small, even less at high frequencies.

It's an ADCMP572, and the 50 ohm resistor is actually inside the chip.

The phemt source follower is still interesting. I might try one for fun.

You can compensate for some capacitive loading by narrowing the microstrip track for a few mm on either side of the lumped capacitance.

John

That still becomes as an LCL T-filter at about 2x what the roll-off would have been. That's fine as long as you know it.

I Spiced that but it doesn't work for my geometry and speed. 2 pF of pickoff capacitance is a huge hit on such a small short trace. The 50 ohm trace is about 3 pf/inch, and my whole run between connectors is about 1".

Your signal level is nearly 4dBm, which in the RF world is HUUUUGE, so HUUU GE in fact many small signal monolithic amps introduce serious harmonic + s pur content hitting their intercept point with that kind of *input* level. You knock that thing down to -16dBm where you can get reasonably harmonic a nd spur free amplification. But if you want to pay $100 for a high IIP3 amp to do a $0.99 job, have at it, your call. You want that 20dB coupling to i solate your through channel from the pickoff channel, the more attenuation the better. Attenuators are deliberately introduced in all kinds of circuit s to kill non-linearities.

Did I mention 4 dBm?

We have the LEE-19 in stock, an 8 GHz 12 dB MMIC. Costs $1.19. A resistor from the microstrip trace, hundreds of ohms, would get me about net unity voltage gain at the MMIC output.

Too much gain doesn't worry me. I can sell the users attenuators.

The signal is a clock, into a comparator, so RF type distortion specs don't matter.

It will be a top layer trace, microstrip or CPW. I'd love to use a source follower, probably a PHEMT or even a fast NPN... if it doesn't oscillate. I could breadboard that maybe.

The SKY65050 makes a semi-decent follower, about like your average JFET. The Avago ones (ATF38143 etc.) have such low drain impedances that their gains as a follower are down around 0.6 or something horrible like that. Small pHEMTs are surprisingly stable due to their very low interelectrode capacitances--you have to really work to make one oscillate.

I haven't tried the Mini Circuits ones yet, though I have a hundred or so in stock. The new Renesas 12-GHz class pHEMTs look pretty interesting too.

Bootstrapping a pHEMT with a SiGe:C transistor (BFP640) makes a nearly perfect gain stage, except that you have to de-Q the base of the BJT or it'll oscillate at 14 GHz or so. I usually use a BLM18BB05 (5 ohms at

Cheers

Phil Hobbs