Yes, the standard bias-T looks something like this:
DC power source >>---L1--+ +--L2->> "unregulated" DC power to regulator
bidirectional RF ----C1--+- mixed RF + DC -+--C2---- bidirectional RF
where L1 and L2 are equal-size power inductors and C1 and C2 are equal-size data-coupling capacitors.
I would avoid connecting anything else to the "mixed RF + DC" line other than those two inductors and two capacitors.
(Or perhaps four capacitors, if I had a separate "transmit capacitor" and a "receive capacitor" at both ends).
Since you likely have some sort of connectors between the two devices that typically have 0.1 Ohm of resistance each,
coupling capacitors that give an impedance of less than 0.1 Ohm (across the entire bandwidth) will be more than adequate (and perhaps overkill).
So a capacitor with capacitance at least 1/(2*pi*90 kHz * 0.1 Ohm) =~= 18 uF is more than adequate (and perhaps overkill).
You'll want a cap with low parasitic resistance (ESR), so mica, film, or ceramic -- rather than tantalum or electrolytic.
You'll want a cap with low parasitic inductance, so surface-mount -- rather than through-hole.
Standard off-the-shelf capacitors and inductors are more than adequate up to 10 MHz or so.
People that work with higher frequencies use striplines and resonating stubs that may appear to be black magic.
Although there are a few people who claim it isn't.
a b c d e
Inevitably, not all the energy sent out by the transmitter will make it to the receiver.
If I cut the cable between the transmitter and the receiver and add "a few more connectors" in the path between them, a little energy is lost each time the signal crosses a connector.
Practically all digital communication systems can tolerate a lot more loss than that caused by "a few more connectors".
So keeping the distortion to something less than the loss of "a few more connectors" is overkill.
(I prefer to get my first prototype working with oversized components, rather than pick stuff that is right on the verge of not working).
Alas, I don't have a rule of thumb for figuring out how much inductance to specify.
Perhaps whatever you have that is generating or receiving your G3-PLC data might have some sort of datasheet with some recommendations?
Ed Mullins and Anass Mrabet in "Analog Front-End Design for a Narrowband Power-Line Communications Modem Using the AFE031" have many tips you might find useful. In particular, their figure 27 seems to indicate that, with
PRIME or G3-PLC,
a standard off-the-shelf voltage regulator will work,
one where the only inductance between the power line and the large bulk power storage capacitor that powers all the electronics is a standard EMI filter.
(via http://www.ti.com/solution/power_line_communication_modem ,
via http://www.ti.com/plc ).
The datasheet for one particular such voltage regulator has a detailed list of materials; its EMI filter consists of a
- 1 L1 Inductor, AC line, common choke, 27 mH, 54P512-276 Vitec Electronics Corp.
- 1 L2 Inductor, high current choke, 3.3 μH, HCP0703-3R3-R Coiltronics/Cooper
(nearby datasheets also mention a "54PR515-146" and a "AF5169-146" AC-Line Common Mode Choke).
It is getting easier to search popular electronic suppliers for "common mode chokes". Alas, while I see many with "at least" 27 mH, and many "at least" 3 A, finding one that meets both specs is difficult.
Perhaps the Bourns Inc. 7122-RC (4 A, 25 mH) might be adequate?