Can you really put the entire RF spectrum on a couple of millimeters of fiber?

November 6, 2006 3:59 AM

RF-to-IF over fiber. I'm not much of an RF geek lately but i'm learning a bit here and there. One of the things i've found some interest in is these RF-to-IF converters, which supposedly allow you to take chunks, or even the entire RF spectrum, convert it to a data signal, relay it over fiber, and then move it back into RF spectrum at the other end allowing traditional receivers to listen to it.

It sounds to me like this is most useful for signals intelligence, e.g. if I was able to put an antenna in place X and maybe another box but not hundreds of thousands of dollars in people and receivers. Can someone provide some more information on how this would work in practice?

posted by arimathea to technology (13 comments total)

I've spent five minutes sitting here thinking about this, and the more I look at it the more farfetched it sounds. I can't think of any way to do what you're talking about. Do you have any hints at all as to just what the encoding on the fiber might be?
posted by Steven C. Den Beste at 6:05 AM on November 6, 2006

i'm no communications engineer, but i don't see what's so farfetched about it. the full "RF" spectrum spans ~DC - 300 GHz, which is something like 0.1% of the bandwidth of the visible spectrum. you modulate a CW laser with some broadband RF signal and demodulate it at the end of the fiber with a diode or something.

arimathea - sorry if i'm not understanding your question, but a google search has all kinds of details - is there something else you're looking for?

i don't know if you can access this pdf, but it provides a pretty good introduction to the concept.
posted by sergeant sandwich at 6:34 AM on November 6, 2006

No. The way I have interpreted it, I-Q data is converted with an ADC at one end, transmitted over the fiber, then converted with a DAC at the other (or perhaps the data stays digital). I don't question that the fiber certainly has the bandwidth to cover the spectrum (especially if you use multiple lambdas). A google on 'radio-on-fiber' seems to return some info, but i'm looking for some more detailed views.
posted by arimathea at 6:41 AM on November 6, 2006

Thanks sgt. sandwich - yes, i've Googled, I was just wondering if someone had more information on it from a pure implementation perspective, e.g. someone who has done it before.
posted by arimathea at 6:41 AM on November 6, 2006

The usual "RF spectrum" is the frequency range from about 1 Hz (ultra longwave) to 5 GHz (microwave); in theory, it is possible to encode and transmit all this spectrum on a greater bandwidth fibre channel of sufficient capacity, and to do so in real time, one would not use an "intermediate frequency" (IF) in the way a standard superhet circuit uses IF, but would directly transmit the spectrum slice after A/D conversion. From a rough "rule of thumb" interpretation of Shannon's law, we'd find that to do this with any fidelity, we'd need a minimum fibre bandwidth of about 25 GHz, which is a lot for single circuit multi-mode fibre over long distance, but not at all impossible. At the other end, we'd need D/A conversion and UW spectrum transmitters of appropriate design to recreate the full range and types of modulation of the original RF slice; again, this is possible, but really more practical for simplified signal sets. In fact, if the RF spectrum at the other end becomes as wide as sergeant sandwich postulates, and the modulation it encodes as subtle and difficult as that created by the latest generations of military UWS gear can create, the Shannon limits get pretty nasty, and probably quickly exceed the parallel channel capacities of even the experimental transmission systems I've linked here.

In general, it would be much more effective if only selected signals could be sent, stripped in transit of the modulation and carrier components used to make them RF signals, and then have those broadcast components re-created at the far end of the point-to-point fibre link. In this kind of system, an IF stage like that you describe, could be useful in creating a transmission signal within Shannon limits of a much narrower band. This is useful for reducing the costs of transmission by maximizing signal-to-noise ratio of the intelligence transmitted, so it is the usual and preferred mechanism in practical communications systems.
posted by paulsc at 6:51 AM on November 6, 2006

ah, sorry. thought you wanted to do it all analog.

paulsc's answer is (as always) fantastic, but i thought i'd link to this (enormous) document from IBM, which is awesome in its depth and coverage of pretty much every aspect of fiber communictaions. in particular you might want to look at chapter 7 which covers some of the encoding methods.
posted by sergeant sandwich at 6:58 AM on November 6, 2006

In the digitizing scenario, the challenging parts are making the signal digital and developing the algorithms. The digitizing still has to occur at the remote end, and the IF part is really just called The Internets.

If you were to use the fiber optic as an analogue communication channel, as sergeant sandwich suggests.. Well, I think that is the question you intend to ask.
posted by Chuckles at 7:01 AM on November 6, 2006

I didn't think anyone could make a 25 GHz analog-to-digital converter. Has the state-of-the-art gone that far now? And consider how many bits of resolution you'd need.

That was the part that stopped me. I still don't believe it.
posted by Steven C. Den Beste at 9:48 AM on November 6, 2006

I'm on the fence about that.. Here is a 25 giga-sample per second 'scope. Poor resolution might not be too troublesome, if you get close to the signal source.
posted by Chuckles at 9:57 AM on November 6, 2006

paulsc, thanks for the detailed answer. Now we're talking. So in your second paragraph, can you offer suggestions of what that pipeline would look like from a practical perspective? The first one seems pretty obvious. I guess you'd use an SDR to rip the data out and just encode it and call it a day, or what?
posted by arimathea at 11:24 AM on November 6, 2006

Not really the sort of "how" information you are after, but I use a bit of gear that does this sort of thing every day. The PPM Sentinel II takes in an analogue signal up to 1GHz in bandwidth, transmits it over fibre, then converts it back to an electrical signal.

I use it to make EMC measurements. It allows you to connect a remote antenna to your spectrum analyser or EMC receiver without having to run long lengths of coax through the field you are trying to measure.
posted by markr at 4:23 AM on November 7, 2006

If you can digitize the signal at the remote end, why convert it back up in the RF again, instead of remaining in the digital domain? To me it doesn't make sense unless the analog IF is what's passed through the fibre.
posted by springload at 3:20 PM on November 7, 2006

Well, obviously I was thinking the same thing.. Actually though, the processing power required to work on so much data in the digital domain must be enormous. It isn't hard to dump immense amounts of digital data down a pipe, but to process it in a timely manner? I bet it won't be too many years before it does become practical, but it might not be practical today. Even if it is, if you have a lot of legacy equipment already..
posted by Chuckles at 7:39 AM on November 8, 2006