Naked Science Forum
Non Life Sciences => Technology => Topic started by: syhprum on 08/02/2015 10:42:41
-
WWII showed the need for secure voice communication between the UK and the USA to this end the 55 ton 30kW SIGSALY system was developed to scramble speech via a radio link.
Was a consideration given to using the telegraph cables between the UK and USA ? The better cables had a bandwidth of 100Hz and speech compressor's (Vocoders) had been developed in 1939 so it at first sight it would have seemed to be possible.
http://en.wikipedia.org/wiki/SIGSALY.
-
The description in Wikipedia suggests that each speech sample was over 30 bits, and sampled 50 times per second, for a total data rate over 1500 bits per second (plus timing/synchronisation overheads). So it far exceeded the 100Hz bandwidth mentioned above for a transatlantic cable.
Another problem with the cable is that it is strongly dispersive, ie different frequencies travel at very different velocities. This would have scrambled the scrambler.
Computer memory and processing power was very expensive in those days, so today's matched filters would have been impractically complex in the second world war.
So wireless/radio was a better medium, despite the fact that it was more likely to be intercepted, and suffered from lightning crackles and fading due to atmospheric effects.
-
Dial up modems can thru put 56Kb/s on a nominal 4kHz telephone line so 1500b/s is not so way out but as you say dispersion may be the problem, I don't think there would be any great problem doing it with modern electronics.
Modern Vocoders use far less than 1500b/s
-
Another problem with the cable is that it is strongly dispersive, ie different frequencies travel at very different velocities. This would have scrambled the scrambler.
Wouldn't that be a reason to convert from analog to digital.
With a digital signal, one can use essentially a single frequency carrier which should minimize the dispersion, although there would still be signal degradation.
How "real time" was this system?
So, did Roosevelt utter a phrase which gets sent to an encoder, transmitted, then sent to a decoder, and 30 seconds later Churchill hears the phrase, then responds... and 30 seconds more before Roosevelt hears the response????? Such a system can be extremely annoying for communication. However, it certainly would be 10 steps above written telegraph messages.
-
For security the SIGSALY system used a frequency jumping system the one time pad key for which was stored in audio disks that had to be physically distributed before a conversation could take place and synchronised with the aid of the WWV standard frequency transmissions this would not have been needed if a cable had been used meaning the system would be quickly available.
-
Dial up modems can thru put 56Kb/s on a nominal 4kHz telephone line so 1500b/s is not so way out but as you say dispersion may be the problem, I don't think there would be any great problem doing it with modern electronics.
Modern Vocoders use far less than 1500b/s
Is your question about whether this was considered in WW2 or feasible today?
When I worked on modems in the '70s the max speed offered by UK PTT was 48k. This was over a special wideband, 4 wire cable 30mm dia, so expensive that only a few were installed. Max dial up speed was 600bps (bps was still called baud as a left over from telegraph days) and used FSK. Max practical speed was 2400bps over 4wire private ccts - separate Tx Rx and used phase quadrature modulation.
Not sure what was feasible during the war.
-
Dial up modems can thru put 56Kb/s on a nominal 4kHz telephone line
Telephone lines are like a "low pass filter": low frequencies get through fairly well, but higher frequencies are more strongly attenuated.
Normal telephone lines achieve their 3 or 4kHz bandwidth because they are limited to about 5km in length.
You can get a bit further by using loading coils - this makes the loss slightly better at 3kHz, but a lot worse at 6kHz.
But a transatlantic cable is thousands of kilometers long - when you talk about a bandwidth of 100Hz, that means that frequencies above 100Hz or 200Hz are strongly attenuated (and suffer large phase shifts), and are effectively unusable. Such a cable is only useful for Morse Code.
It was only with the introduction of amplifiers in the cable that single voice calls could be made - and later multiple voice calls, by modulating radio frequencies.
so 1500b/s is not so way out
It takes some pretty fancy modulation and error correction to fit 1500b/s into a 100Hz bandwidth - especially if you are using the sea and earth as the electrical return path...
Wouldn't that be a reason to convert from analog to digital?
This effectively was digital transmission - they digitised the outputs of bandpass filters, then encrypted the digital values.
With a digital signal, one can use essentially a single frequency carrier which should minimize the dispersion, although there would still be signal degradation.
Unfortunately, when you modulate a single frequency, it "spreads" the spectrum.
I still remember 1200bps dial-up modems - they started with a single tone, but the modulated tone filled half of the 0.3 - 3.3kHz telephone band.
This type of modulation would exceed the 100Hz bandwidth of the cable mentioned above, making it effectively impossible to extract the original signal.
How "real time" was this system?
The idea of sampling every 20ms is so that the modulation process added little delay.
I imagine that they would have used analogue bandpass filters, which don't add too much delay.
Mouth-to-ear delays of up to 150ms each way are quite tolerable, but geosynchronous satellite delays of close to 300ms each way are quite a nuisance (until both parties get used to it).
-
I was considered a dab hand with modems thirty years ago I remember that if you wanted to send pictures back and forth to the USA you had to add 73 to the dialled number to make sure you got a cable connection and not a satellite which was hopeless.
Dial up modems are little valued today but they incorporate some really clever electronics for less that £10.00
-
Apparently the Idea had been around since 1928
"Computer Speech: Recognition, Compression, Synthesis
Speech compression, once an esoteric preoccupation of a few speech enthusiasts, has taken on a practical significance of singular proportion. As mentioned before. it all began in 1928 when Homer Dudley, an engineer at Bell Laboratories, had a brilliant idea for compressing a speech signal with a bandwidth of over 3000 Hz into the 100 Hz bandwidth of a new transatlantic telegraph cable. Instead or sending the speech signal itself"
If they were prepared to employ 55 tons of equipment and 30kW a quite complex even a mini ENIAC like device could be built and not needing the one time pads a superior result