Both approaches depend upon frequency separation of the Tx and Rx signals. Both provide (it's actually the whole point of either approach) 'full duplex'. The advantage of using a single antenna (and feedline) is often far more valuable than eliminating the diplexer; the diplexer is just a harmless box that should be reliable and maintenance free (as compares to an antenna left outside in the wind and rain).


Other than some niche applications (e.g. certain low power Doppler radar applications), I can't think of any communications systems that use full duplex on a single (Tx and Rx) frequency. In theory, one could null out any one frequency, but the sidebands wouldn't be precisely nulled.


Some full duplex systems are sensitive even to Passive Intermodulation (PIM) in the cables, connectors, and any mounting hardware within the beam.

It appears that full-duplex is achievable without frequency separation:
http://www.ehow.com/list_7551272_13-ghz-microwave-...
http://www.commconnect.com/wp-content/uploads/2011...


This could possibly be done with polarization separation, i.e., H/V or right/left. http://ieeexplore.ieee.org/xpl/login.jsp?tp=&a...

Seems like there are analogies to conventional wired comm duplexing, ala RS-232.

TTFN
FAQ731-376: Eng-Tips.com Forum Policies

oops, that should have been RS-422

TTFN
FAQ731-376: Eng-Tips.com Forum Policies

Full duplex on almost-adjacent channels within a given band is (of course) absolutely standard.

I've followed all the links and I didn't see anything that indicated that anyone had cracked the nut of achieving full duplex using only a single frequency (Tx and Rx full duplex within one channel). It would be very exciting if it's been made practical. Did I miss a detail on those references?

Oh, OK, I see what you mean.

Although, some people claim to have done it:
http://sing.stanford.edu/fullduplex/

TTFN
FAQ731-376: Eng-Tips.com Forum Policies

Yep. Perfectly applicable link.

They mention 45dB and 73dB cancellation.

The 45dB is not at all impressive; that could be achieved with the twigs and a bit of moss , or something similarly low tech.

The 73dB is achieved "in the digital domain". Of course, these are the easy 73dB (at the fat end). If a practical system needs (for example) 146dB isolation to match the system path loss (so that the Rx signal can stand above the local Tx signal), then they've got a lot more work to do at the more difficult tiny end.

They should look into the echo cancellation techniques used by long distance telephone carriers. They should feed forward a clean sample of the Tx signal to the DSP so that it knows what to look for. Correlate and kill. Repeat because there will be multiple echoes. The practical implementation may involve significant buffering and thus latency. Latency is the enemy of high speed bi-directional data. It should be integrated with the demodulating and data decoding.

Years ago I invented an application for this sort of technology to make crypto voice systems more usable. It got buried by the company. Oh well, ideas are a dime a dozen.

I was interested in the echo cancellation topics, because either the double reflection of the feed line or the multipath reflection can be viewed as echo sources.
I want to explain the system with more detail. For Eband(71~86G), the pathloss is 92.4+20*log80Ghz+20*log2km=136.4dB, and each subarray's gain is 40dB, so the receive singnal will attenuate 136.4-40*2=56.4dB, if we plus addtional rain and oxygen effect, it will be larger.And without additional bandpass filter, the isolation of locate end is 75~90dB, which is not enough for same channel working of TX and RX.
But what I concerned is the double reflection of feed line can also cause a -30dB weaker echo. And the multipath reflection also can cause a -25dB echo. So what's the difference of this echos with the locate end leakage? Further more, if the leakage obvious tiny than the RSL from far end, do this decreased the correlation and cancellation work load? Of course,if same channel working is impossible, the adjacent channel working of TX and RX also make it a superior virtual diplexer with high isolation, low group delay, and reconfigutable.
BTW, the antenna is fully integrated planar subarrays.

This is why we use different radio channels for Rx and Tx when using full duplex. It's not yet a solved problem to use the exact same frequency for both - for the reasons you're outlined.

Isolation is not too difficult if the antennas couple in the H plane and no common radomes are used and there is a little separation.
If you E plane couple and they are real close to each other, it's a 30 dB worse problem.

Isolation needs to be defined too. Input to output single connector is different than individual antenna Tx to Rx antenna. You can swamp a Rx input amp and still pretend isolation out the Rx connector is ok.

Hi, Higgler. What's no common radome meaning? Would you reveal littile more.

Oh, Ubiquiti also ready realized the same thing in parabolic antenna, but they call it 2X2 MIMO without diplexer. It is base on HDD(hybird domain diplexer), including both FDD and TDD. And it can obtain a unbalanced air port capacity.
http://www.ubnt.com/airfiber

Assuming it's the same "HDD", then it's described here:
http://etrij.etri.re.kr/Cyber/Download/PublishedPa...

If I understand it (?), then it's simply a hybrid combination of FDD (frequency division) and TDD (time division). Thus it still avoids the present-day-impossibility of full duplex on a single radio channel.