[David Spencer]

David Spencer asked the Naked Scientists:
Hello Naked Scienstists!
   
I discovered your show recently and have spent considerable time listening to your backlog of podcasts, thanks for keeping my me awake and interested on the morning train to work!
   
One I was pondering a while ago is, with countrywide data services such as 3G, Wi-Fi, bluetooth, radio, television broadcasting, and radio communications, are we going to eventually run out of useable frequencies?  Or will we keep refining our technology to use more frequencies in between the old ones?

Thanks for you time in advance,

Spencer in Sale, Manchester
What do you think?

[another_someone]

There is a limit to how much information can be carried across the whole radio spectrum, so in theory we could run into physical limits (eventually); but at present it is more about finding technical solutions to using parts of the spectrum that are not presently used, reallocation old frequencies to new purposes, and compressing data to remove redundancy.

[graham.d]

The spectrum is generally being expanded upwards too. Mobile phones now operate in bands around 1.8GHz and with more efficient use of the allocated bands. In professional radio bandwidths for speech and data are coming down so that the band spacings are going to be just 6.25kHz. Voice is vocoded to a digital bitstream of just 2.4kb/s with a further 1.2kb/s of data for error correction in poor reception conditions. The use of controlled range for transmission (as in cell phones for example) also means that frequencies can be used simultaneously providing the transmitters are spaced apart far enough.

Spread spectrum techniques are also widely used now which is another way to make efficient use of the available bandwidth.

There is a finite resource here but technology is advancing so as to make better use of it. Nonetheless, the demand will keep growing so eventually we could run out of bandwidth. It is already expensive to buy bands from the regulatory authorities and I expect this will continue. It is a way of forcing the technology to get better though.

[LeeE]

Using amplitude modulation instead of frequency modulation would theoretically allow infinite channels because each channel would require just a single frequency instead of a frequency band.  On top of that, it would be possible to sub-divide the amplitude into a series of different bands, giving multiple channels per frequency (each channel then being bandwidth limited).  However, there are many practical problems with this, not least being the effect of noise on amplitude.  At the same time, I believe there are some severe problems associated with the very steep filters that would be needed to separate the discrete frequencies.  I'm less sure about the filters aspect of this but for some reason phase change problems spring to mind in connection with steep filters.

[another_someone]

Using amplitude modulation instead of frequency modulation would theoretically allow infinite channels because each channel would require just a single frequency instead of a frequency band.

This is not actually correct.

The nature of amplitude modulation is that the moment you modulate a signal you create sidebands (in fact, conventional amplitude modulation creates two sidebands, one above the carrier frequency, and one below the carrier frequency - but a lot of shortwave people use the more efficient single sideband amplitude modulation, but that was never used for commercial broadcasting).  If you modulate a 10Mhz carrier with a 10KHz signal, then you will create a sideband that is 10KHz above and below your 10MHz carrier.  Since audio sound has a whole spectrum of frequencies, depending on quality, from about 100Hz (lower if you include very low frequency sounds) to anything from about 8KHz to 20KHz, so you will be using a bandwidth of anything up to 40KHz spread around the carrier frequency.  This is generally still more efficient than frequency modulation (but the greater bandwidth usage of FM also gives better noise immunity).

[Pumblechook]

Of the useful 0 - 1000 MHz range.........

Analogue TV is the greedy one.....390MHz is used in the UK.  Some will be released (112+ MHz) after analogue switch off.  AM (including Short Wave) and FM radio only takes up about 25 MHz.  Mobile phones take up less than 50 MHz.   The military use 175 MHz.  Air traffic control takes about 30 MHz.    If you listen across the range 0-1000 MHz or look on a spectrum analyser there are actually lots of quiet frequencies.  I would have thought the space could be better organised.       

[Pumblechook]

-----------Using amplitude modulation instead of frequency modulation would theoretically allow infinite channels because. -------------

It is more complex than that.

You have to distinguish between AM as used on the present AM bands which has audio limited to a max of 5 KHz and hi-fi AM which might be used on VHF (88 - 108 MHz or whatever) ...  So each AM station is only 10 KHz wide. 

Audio is limited to 15 KHz on FM so a similar hi-fi AM transmission would be 30 kHz wide. 

FM transmissions are much wide because FM has multiple sidebands... 150 kHz mono.. 240 kHz Stereo.   

Stereo with AM would either use two separate transmission so 60 kHz total or a system similar to that as used on FM (Zenith-GE) and be 100+ kHz wide.

So in theory with AM you could fit in far more stations but FM allows re-use of frequencies over much shorter distances...  Distant stations cause far less interference on FM.. Capture effect. And as somebody said FM has better noise or interference immunity.   Stereo FM needs a heck of lot more signal though to give noise (hiss) free reception..  23 dB more in fact (200 times in power terms).   

[lyner]

No one will ever use an AM broadcast system again. They probably won't use FM either OR SSB. They are all hopelessly inefficient compared with digital coding and modern modulation systems.
AM is the worst value of all, using twice the  audio bandwidth and expensive /sweaty transmitting equipment. The only good thing is that you can use a crystal set to receive it on - hardly relevant.
FM, depending on the actual deviation you choose, can either perform about the same as AM, or, as you use more and more RF bandwidth, improve your signal to noise/interference performance. Per audio channel, it is spectrum hungry, though. The way it performs with a stereo multiplex signal makes it 'die gracefully' as the signal level gets less but there is quite a noise penalty for going stereo.
Apparently, when they were making the choice of systems for VHF broadcasting, broadband AM was considered as a serious contender; using the same sort of channel spacing as is used by FM transmissions, a good AM demodulator ('dynamic limiting'?) deals very well with impulsive noise.
Stereo, using analogue systems just HAS to use some multiplex system - not separate channels - or the image shifts all over the place as relative levels change  even a little bit.
DAB, in all its possible modern forms, is great value, using much less bandwidth per audio channel bandwidth. The problem is that the broadcasters take the mickey and use lunatic bit rates to get lots of channels in but then nothing sounds good quality. A properly engineered system has no propagation problems - multipath propagation and multiple source reception are dealt with by the coding / modulation process. The receivers would have cost £1,000,000  each, not long ago but now they're as cheap as chips.

[lyner]

Lee:

Using amplitude modulation instead of frequency modulation would theoretically allow infinite channels because each channel would require just a single frequency instead of a frequency band.

That 'theory' is only true if you wanted to  transmit zero information along each AM channel.Don't forget that there are sidebands on either side of an AM carrier, extending +/- the audio bandwidth.

[LeeE]

Yes re AM - I hadn't gone into sidebands:)

I was thinking more in terms of pulsed data transmission, more like the way that data is sent across a network or down a fibre, as opposed to modulating a carrier as in audio AM, but in the end it amounts to the same thing - you're just modulating with a square-wave:)

It then occurred to me that if, for example, you could send the pulses at four different levels of amplitude you'd be able to run two independent binary streams.

Amplitude	Stream1	Stream2
0	0	0
1	1	0
2	0	1
3	1	1

Hmm... you'd probably need to include calibration pulses in the stream too, to establish a reference amplitude.

Actually, I think there may be a way to get an extra half-speed channel using just three amplitude levels and parity techniques - can't remember it clearly enough though.

[lyner]

No data is sent as 'square wave pulses', in real life. The spectrum of a signal like this covers many times the data signalling rate. The resulting AM sidebands would be really embarrassing. In practice, data tends to be filtered in such a way that the effective bandwidth is round about the same as the data rate, for binary systems and less, pro-rata, as you go to multi level. But if you have more than two levels the error rate is worse for a given level of channel noise.
Carrier keying (the AM you are talking about) is not good value, if you are talking in terms of a modern system because of the fact that you are sending the same information twice, effectively (carried by both sidebands at once). There are a lot of alternative systems, using phase shift keying or combinations of phase and amplitude shift keying. The more complex they are the better they can be made to perform.
 There has been a lot of work done on coding and modulation systems for radio channels, optical fibres and magnetic recording etc.. Each medium  has a different 'best buy' because of the different characteristics of channel noise and interference. The system used for DAB in the UK is fiendishly complicated- multiple data channels, multiple carriers, multiple transmitters (see http://www.bbc.co.uk/rd/pubs/papers/paper_15/paper_15.shtml).
It is not a simple problem to get the best performance out of a channel. 
The first time your system (carrier keying) was used was round about the time of the Titanic sinking. In those days, the channel occupancy wasn't too much of a worry because so few people were using the airwaves.

[another_someone]

Lee,

If you want a true square wave, with nice sharp corners to the square, you will have sidebands going out to infinity.  In the real world, the corners of the square wave will be rounded, and you will have ripple noise (ringing) which will occur as the channel is squeezed into less than infinte bandwidth.

Try calculating the Fourier transform for a square wave and you will see what I mean about infinite sidebands.

[lyner]

In a normal data stream you get pulses of all different widths, corresponding to  00000100000  or 1010101010 type patterns, where the pulses are narrow and 11111100001111000 where the pulses are wide. The filtering will end up making the 1010101 pulses look more or less like a sinewave and the 1111100000 look like rounded square waves. All the information gets through and saves oodles of bandwidth. (Basically)

[another_someone]

In a normal data stream you get pulses of all different widths, corresponding to  00000100000  or 1010101010 type patterns, where the pulses are narrow and 11111100001111000 where the pulses are wide. The filtering will end up making the 1010101 pulses look more or less like a sinewave and the 1111100000 look like rounded square waves. All the information gets through and saves oodles of bandwidth. (Basically)

In terms of rounding of the wave, fair enough; but in terms of ringing, it can be more complex (particularly if you are looking for transitions rather than absolute levels).  Making sure you properly shape the wave before transmission will avoid the problems with ringing, but then we are no longer dealing with true square waves (and it was only the issue of square waves I was referring to, not more complex shaped waves).

Incidentally, dealing with signals such as 00000010000000 can be more of a problem if your circuitry does not handle very low frequencies.  Some coding systems ensure that there is at least one signal transition within every so many bits in order to ensure you don't get the ultra low frequency problems.

[lyner]

Incidentally, dealing with signals such as 00000010000000 can be more of a problem if your circuitry does not handle very low frequencies.  Some coding systems ensure that there is at least one signal transition within every so many bits in order to ensure you don't get the ultra low frequency problems.

Yes. Biphase coding turns every symbol into, either a 01 or a 10. and eliminates the problems of AC coupling.
But 'ringing' need not be a problem - it just represents inter symbol interference and is often built into the  signalling channel. It can be rectified with the appropriate filtering at the receiver.
It can be a problem where there are non linearities and overshoots can exceed the 'headroom' but it's all part of the overall engineering of the channel.
A  properly engineered system will match the transmitter filtering, the transmission channel characteristics and the receiver filter to produce optimal performance.

[LeeE]

Thanks for those comments - they point to some interesting stuff that I'll have to look up.

[graham.d]

Modern radio transmissions will all be digitally encoded. The form of this digital encoding and how this modulates the signal is complex (in both senses for those mathematically inclined). Early digital transmissions may have used Frequency Shift Keying so that each of two frequencies represents 1 or 0 or Phase Shift Keying where it is a single frequency but the phase changes to indicate a 1 or 0 transition. These can be optimised so that the spread in bandwith can allow a maximal number of bits/second for a particular frequency deviation. There are regulatory requirements to prevent transmissions overlapping on to other allocated bands, and packing more Bits/second into your allocated band is what it's all about.

It is too difficult to go into detail here regarding various modulation schemes but the most advanced scemes use both phase, frequecny and amplitude to pack the data in. A QAM (Quadrature Amplitude Modulation) system is a simple example of phase and amplitude modulation but the latest schemes use up to 64QAM superimposed on multiple subcarriers (a number of frequencies at small regular spaced offsets either side of the main carrier). The protocol management has to cope with the effects of multipath reception that can result in errors and Viterbi decoding (or more often turbo codes) used to mitigate the effects of the resulting errors. The effect of this is to make very efficient use of the allocated narrow band without flowing over into an adjacent band. The filtering required is tricky as this can also introduce errors as it causes data patterns to interfere with adjacent data patterns. This is also minimised by using digital phase linear filtering and by the relationship between the positions of the subcarriers. This can give a data rate of about 350kb/s in a 100kHz wide channel. The other tradeoff that has to be made is that usually the more bits/sec in a particular band, the lower is the range of transmission because the more sensitive is the signal to being disrupted by interfering noise that has a theoretical floor.

[techmind]

David Spencer asked the Naked Scientists:
One I was pondering a while ago is, with countrywide data services such as 3G, Wi-Fi, bluetooth, radio, television broadcasting, and radio communications, are we going to eventually run out of useable frequencies?  Or will we keep refining our technology to use more frequencies in between the old ones?

Modern digital communication modes allow you to get far more information transmitted per section of radio spectrum.

In particular Coded Orthogonal Frequency Division Multiplexing (CODFM) allows you to get practically the theoretical maximum amount of information through a given communications channel. When coupled also with lossy/perceptual audio and video coders, this gives an extremely efficient communications link - far more efficient than the old AM or FM or (AM) analog television transmissions.

(CODFM is used in DAB radio, digital terrestrial TV, ADSL and ordinary "56k" dialup modems (down a cable), possibly for WiFi...)

CODFM is essentially a super-suped up version of AM. For one communications channel it uses thousands of carriers all modulated with discrete amplitudes and phases. In "64QAM" each carrier can take on one of 64 discrete amplitude/phase permutations. Given then 2000 or 8000 carriers, you are transmitting so much data "in parallel" that you only need to change or update the phases relatively slowly (this is called the symbol rate, and may be as much as 100 microseconds). This slow symbol rate makes the system robust to multipath reflections. Very cleverly, this same property means that, provided you synchronise transmitters, you can transmit the exact same programs from many transmitters across the country all on the same frequency - even with overlapping coverage areas - without interference. This is known as a single-frequency network (SFN). I believe some parts of Europe use SFN for digital TV (we don't, yet). I believe the main national DAB radio multiplexes in the UK use an SFN configuration too. For broadcast applications this results in a significant saving in spectrum requirements.

The "cellular" based communications systems, mobile phones etc, reuses the same frequencies in different (non-adjacent) cells. When demand causes them to "run out" of frequencies, they just build more base stations and reuse the frequencies across shorter distances. This is a big part of the reason why there are a huge number of phone base-stations in the major cities (propagation is also poor through the buildings, so more local base stations help there too).

Similarly WiFi is relatively short-range and uses only a very few channels (about 11 in this country at 2.4GHz of which only 3 are supposed not to interfere with each other) - which keep getting reused. I'm quite surprised WiFi hasn't logjammed in suburbia!

It is certainly argueable that in any one place (and time) in the country and awful lot of spectrum is not being used. There is pressure on creating systems which (for short-range communications - a few metres) use all "unused" frequencies at that moment. This is completely contrary to the regulatory frameworks which were developed 60-odd years ago, and will probably take some time to change.

Spectrum is a finite resource, but there's plenty of room for "efficiency savings" by rolling out newer technologies and regulatory frameworks.