DC blocking at high frequencies (10GHz)

Hi!

I'm designing a circuit at 10GHz and have a bit of a problem with DC blocking. I can't seem to find capacitors with high enough capacitance (in 3pF would give ~5ohms reactance at 10GHz which would be fine in a 50ohm system) and SRF. Another thought was to use coupled lines as DC blocks, but with the thin substrate I'm working with I can't seem to achieve tight enough coupling based on AWR Microwave Office simulations. Does anyone have any insights?

Thanks in advance!

Have you tried interdigital capacitors? I used them at 24GHz on thin substrates. To get started, I designed dimension with circuit models, and then switched over to EM simulation because the circuit models might miss some extra resonances.

i hve used plenty of interdigital caps at high frequencies. You just have to make sure the finger length is not too long, or it will act more like a transmission line component.

Thank you!
I also heard that theoretically full coupling (0dB) should be able with a coupled transmission line pair. Is that true? (It seems impossible.)

Yes, you can design for very low insertion loss. Here is the EM simulated insertion loss for a 24GHz interdigital cap.

sure a coupled line quarter wave long should be good for maybe 30% bandwidth. you can also use it to transform impedance if it is part of a matching network

Thank you everyone for the useful insights! Meanwhile I managed to achieve 1.5dB coupling, but the width of the lines is much smaller than that of a 50ohm line and it just seems unlikely that a 50ohm line continued in a 100ohm coupled line pair (100ohm being the stand-alone impedance of one line of the pair) would provide the best answer. The next step will be to simulate it in CST or fabricate a test circuit using home methods, just to make sure. Thank you again!

Antenna94 said:

but the width of the lines is much smaller than that of a 50ohm line and it just seems unlikely that a 50ohm line continued in a 100ohm coupled line pair (100ohm being the stand-alone impedance of one line of the pair) would provide the best answer.

Click to expand...

Don't worry, that might be a very valid solution. I also used that in a 24GHz design long ago.

Now I am a bit baffled. As the characteristic impedance of the system will be 50ohms, I thought the reactance of the DC-blocking capacitor at the working frequency should be at most 5ohms, so the capacitance has to be greater than 3pF at 10GHz. I designed an interdigital capacitor based on AWR Microwave Office simulations and S21 seems to be >-2dB, however, the capacitance of the structure is only about 0.2pF (a realistic value for such a structure after some reading, but still can't wrap my head around the physical processes happening there). Another problem is probably the input impedance of the capacitor, as the exact length of the line connecting it to a transistor really affects the gain of the transistor - this is an oscillator and the capacitor has to be in the feedback loop as there is a tuning voltage in the feedback filter and that has to be decoupled from the DC voltages of the biasing network.

So the question would be: how is it that a really small capacitance provides good transmission and is there a way to match its input impedance as we would during a filter design?

The distributed ICAP is not a lumped element capacitor. It might be a resonant structure (see #5 above) and you can't reduce the electrical response to series C.

do keep in mind that SOME chip capacitors WILL work up at high frequencies. you have to try them, and when you find one that works, stick to that manufacturer and exact type. I have seen 0.1 uF caps in use at 27 GHz, for instance.

The problem with chip capacitors is that at 10GHz for good decoupling I would need at least 3pF (<5ohm reactance), and those bigger values have resonance and/or bad Q at high frequencies. volker@melhaus: thank you! It seems to provide good coupling, but I though that in MMIC design these interdigital capacitors are used for their capacitance value, that's why I was unsure.