Is Sensing in Cognitive Radio history?

Doglas Sicker, FCCSome interesting talks from top regulators where given at the QUASAR regulator workshop hosted by BT in London. Among the most interesting developments is that the two “pioneer” countries in TV White Space usage now both abandon what has been perceived as THE key feature of Cognitive Radio – spectrum sensing. Both the FCC and OFCOM now issue regulation that requires geolocation and database lookup (see previous blog post) to get transmission permission whereas “sensing is no longer required” as Douglas Sicker, FCC Chief Scientist expresses it(left). So if its not required, why would any equipment vendor include it? Is the Cognitive Radio concept dead – when one of its corner stones is set a side?

Well, I think declaring the demise of Cognitive Radio is somewhat premature. As has been demonstrated several times, TV White Space is not the place where sensing makes sense ;-). The key problem is that sensing the TV transmitters provides very little information of the whereabouts and pathloss to the (silent) TV receivers, the victims of interference. Knowing that a certain TV channel is used in the area is about as much information we can get – and that information we can get more reliably through the geolocation database. This is a fundamental problem and has nothing to do with imperfections of sensing – we can improve the sensing schemes as much as we like – even “perfect” sensing can add anything significant to the database approach. The problem is that sensing detects signals, and not what we really want to detect – spectrum opportunities (= if we can avoid TV-receiver interference).

So, is there anywhere sensing works? Well, sensing should for example work well whenever the transmitter and receiver are closely co-located, i.e. for short range primary systems. One such example is Wireless Microphones. If we from afar detect such a transmitter, that reveals significant information regarding the interference we may cause on the primary receiver. Any two-way system like WLANs, should also work great since there is no silent receiver. Another interesting example is radar systems, where in fact the transmitter and receiver are co-located. If we know the transmitter power of the radar we can calculate the path loss to the radar and exactly estimate the interference at the receiver.

So Spectrum Sensing is not (quite) dead – but not as vigorous tool as we believed it to be! But what about Cognitive Radio – is it blindfolded without effective sensing?

About Jens Zander

Professor Jens Zander is professor in Radio Communication Systems at the Royal Institute of Technology, Stockholm, Sweden. He has been among the few in Swedens Ny Teknik magazine's annual list of influential people in ICT that have been given the epithet “Mobile Guru”. He is one of the leading researchers in mobile communication and is the Scientific director of the industry/academia collaboration center Wireless@KTH. His research group focuses on three main areas – the efficient and scalable use of the radio frequency spectrum, economic aspects of mobile systems and application and energy efficiency in future wireless infrastructures.
This entry was posted in Spectrum. Bookmark the permalink.

10 Responses to Is Sensing in Cognitive Radio history?

  1. Ki Won Sung says:

    I think secondary devices should be connected to a database with or without the feasibility of spectrum sensing. Detection is just a beginning of the secondary spectrum access. What’s equally (and perhaps even more) important is how to access the spectrum.

    With the number of primary users and secondary users increasing and their service requirements gradually changing, the detection threshold and the transmission decision criteria should be adpated accordingly. Any “legacy” secondary device with an obsolete rule will be a headache.

  2. Johnny Cognitive says:

    Spectrum Sensing is dead. With xMax Cognitive Radio there is absolutely no need for any spectrum sensing…thats just the way it is.

    • Jens Zander says:

      So let’s for a moment assume that xMax does sensing in a “smarter way” than all the rest. Unfortunately in TVWS, this does not help us very much. My point is, that this is an inherent problem of broadcast systems – no sensing scheme (regardless of fancy acronyms even the theoretically optimal ones) is capable of detecting what is not there (e.g. silent receivers)

      • Johnny Cognitive says:

        hey Jens glad to see you have become more civil…in TVWS, xMax is best used because it can signal up and down the spectrum, which ultimately results in more/better services to the end consumer. Also, there is no chance of interfering with any signals when using xMax.

      • Johnny Cognitive says:

        In other words Jens, it is all wide open spectrum to xMax.

    • sibi says:

      hi guys, i could not find any technical papers on xMax technology. if you have any link please share.


  3. Peter_in_stockholm_82 says:

    Not having high hopes that anyone would be reading a comment on such an old post but i’ll write what i have in mind. Please bear in mind that my knowledge in cognitive radios etc. is around 4-5 days old :)
    If sensing helps obtaining path loss information, is that enough to know the amount of interference generated at the unintended receiver(s)? Path loss is a function of distance and shadow fading, meaning that fast fading will not be captured with the sensing. Wouldn’t this result in either large inaccuracies, or in the need of being very conservative when allowing a secondary user to transmit on the primary spectrum?
    Wouldn’t it make more sense to have transmissions by secondary users in scenarios that are similar to the uplink of cellular systems, i.e. we have one receiver (assuming one cell just for the sake of the example and simplicity) instead of the downlink where we have many receivers? If done in uplink, the secondary transmitters need to determine how much interference they generate to a very specific unintended receiver (i.e. the base station in the example i have) instead of trying to guess the interference to cellular users in their neighborhood?

  4. Anonymous says:

    Have a look a this paper
    Here is a brief expose on the uncertainties even if we have perfect sensing in the sense (!) that we can perfectly estimate the pathloss to the primary transmitter. If the correlation between the paths is low and we do not know were the primary receiver is the fade margins we need to include is in the order of 40-50 dB.  We are likely to have just as good (or even better results) by looking up the use of the spectrum in a data base.

  5. Peter_in_stockholm_82 says:

    Thanks for the reply and the article Jens.
    I completely agree that a database look-up is the most straightforward approach, and the safest best for cognitive radio to gain momentum. However, such an approach has the inherent limitation of being too static from a temporal perspective. For instance, the database will tell you that frequency band A is reserved in region R meaning that no secondary users are allowed to use band A. However, band A might be used intermittently while being empty of transmissions X% of the time.
    Using sensing, or some better technique, secondary users could be allowed to exploit the time dimension to further increase spectrum utilization.
    I personally do not think we have reached the point where the spectrum is that crowded, but…

Leave a Reply

Your email address will not be published. Required fields are marked *