0.5 microfarad will not work.  It will resonate somewhere far below 18 GHz and give spikes of high insertion loss (dielectric constant is too high, and the LxWxH is too big).

Someone does make a composite DC block capacitor comprised of two individual ceramic caps in parallel.  It is someone like AVX, ATC, or Johanson, don't remember.  

yes, atc & dielectric labs both make "hybrid" parts; 2 caps in parallel; a large one for the low freq's, & a small cap for the high frequencies. the drawback is that the small caps have too low a voltage rating.

i prototyped with the dielectric labs co8 with 2400pf; its only rated to 50 volts also; worked pretty good esxcept down at the low frequencies.

people like ma, inmet, & midwest build broadband coaxial dc blocks, so there must be a suitable part in the market somewhere

Well, you could try to make your own pair of capacitors that do the same thing, but with 200V parts.  The layout will be tricky at the high frequency end.  Also, depending on how the big cap works, you may run into a brick wall with Foster's reactance theory, which might make the two caps resonate as a series open circuit at some narrow frequency band.

If you really need that wide of frequency range, it will take at least two, maybe three capacitors in parallel, the lower frequency larger caps will need multiple RF coils in their path to stop propagation of energy from 18 ghz down to ?Xghz.
So layout a straight line with a high frequency cap in a straight line, then add a U shaped second line, or actually an O shaped second line that contains a larger low frequency cap in the center, and inductors on either side of the high frequency capacitor where the O shaped line meets the thru line. Do some circuit simulation with C and L values to optimize the transmission loss, especially at the crossover point where the straight thru C and O loog L C L have similar S21 losses, those components will have to be optimized for that area of frequency.
You might improve things by adding some quarter wave opens in the O loop to help out the loss characteristics.
I haven't done this, so it's just an idea, take this with a grain of salt please.

kch

thanks!

one of my concerns is with the voltage rating of the smaller caps used for the high frequencies. high freq caps tend to be low voltage, like 50 volts; but i need 200 to be safe

Will this blocking capacitor being in series actually have 200 volts across it? If the cap is very low impedance and your device on the other end is higher impedance, then your cap won't have the full voltage that you apply across it. Only if your item after the cap is very low impedance could you have a 200 volt problem. The cap is small too, so you can't have too much differential across it even if it's output is open circuited at 18 ghz.
How much loss can you tolerate in your DC block design?
kch

thanks!

200volts is a design spec. there may be 150-160v bias on one end, though this seems unlikely to me. it also seems unlikely the other port will be grounded.  but 200volts its what i have to design to

Is the lower limit .01 Hz or 0 MHz ? What is in input
impedance of the following circuit and what is the
max. attenuation?

<nbucska@pcperipherals DOT com> subj: eng-tips
read FAQ240-1032

10mhz is the lower limit

rf feeds into a 50 ohm circuit, though there may possible be a dc return to ground; but its 50ohms from 10mhz-18ghz

max loss: 0.5db

thanks!

bob

I suggest use parallel/series combination of capacitors
so that one path has about 50 times larger C and their
sum gives .5 dB attenuation at 10MHz with the 50 ohm.

If only some discrete frequencies are required, you may
shift the resonance freqency into the not used band.

If continuous, you may need three parallel paths...
You will have to experiment, wiring (wire L ) is critical, too.

<nbucska@pcperipherals DOT com> subj: eng-tips
read FAQ240-1032

Higgler,

you have confused the RF impedance with the DC blocking capability. The DC bias means that the capacitor will definitely have a large voltage across it. The AC impedance of the capacitor is irrelevant to the argument.

Bhawkins,
What I would point out is this, the DC bias can be a problem when switching on and off. Suppose the DC bias voltage rises quickly. A large capacitor will couple this through to the (sensitive) GHz amplifier. Thus the switch-on (or switch-off) transient may take out the amplifying device which is being "protected" by the DC blocking capacitor. You therefore may need to consider slew rate limiting the bias voltage and/or clamping the amplifier input to the transient.

True, I was thinking RF energy.
I sit corrected and feel good about it.
Now for a cup of coffee.

kch