Talk:SIGSALY

Stimulus

James H. Ellis' non-secret encryption for which SIGSALY or something simliar seems to have been the stimulus

Do we have any evidence that Ellis was motivated by SIGSALY? — Matt 12:17, 22 Jun 2004 (UTC)

Duplicate article

Blast! I've just spent a day writing an article on the "X System", as my reference book (see 'Further reading' I just added) calls it, only to discover that we already have an article on this topic, just one which wasn't listed under all its names.

Which raises the question of what we ought to call the article. Both the Bell book, and the Turing bio by Hodges, which also covers it in some detail, call it the "X System" (or some typographical variant thereof, such as "X-system"). Do we want to leave it here (which is, I gather, the name the NSA museum has it under), or use the one in all the references? Noel (talk) 19:25, 6 Mar 2005 (UTC)

Sorry to hear about your reduplication of effort, although you've appear to have merged in a lot good stuff into this article regardless — nice work! Regarding the name: I don't have strong feelings either way, but a few minutes with Google suggests that SIGSALY is the more common name on the Web. — Matt Crypto 01:09, 8 Mar 2005 (UTC)

Spread spectrum

I don't see how this system was a spread-spectrum system. To me, SS implies that the signal is spread out across a wide frequency range so as to be imprevious to narrow-band jamming. It's true that the transmission technique used by this system was relatively good against atmospheric distortion (natural "jamming", as it were), but I would describe it more as a "frequency modulation" system, rather than SS. They used FM because an amplitude modulated signal wouldn't give the required S/N (the vocoded signals needed to be reproduced +- 10% in amplitude for the unvocoded speech to be intelligible), and there was too much fading on long hops to meet this target if they used AM. After all, I don't think anyone would describe ordinary FM commercial radio as SS! Noel (talk) 23:02, 6 Mar 2005 (UTC)

I'm out of my depth technically, but perhaps we could directly attribute the claim to Bennett's IEEE article? That is, something like "Bennett (1983) suggests that SIGSALY can be thought of as being one of the very first successful applications of spread spectrum technology".— Matt Crypto 01:09, 8 Mar 2005 (UTC)

A spread spectrum system uses a larger range of frequencies than is necessary to transmit the signal. Different modulation schemes do this to different degrees. FM uses more bandwidth ($\Delta f_{FM}$ ≈ $2(f_{Modulation}+\Delta f_{max})$, where $\Delta f_{max}$ is the maximum frequency shift) to achieve the higher resistance than AM (Δ$f_{AM}$ ≈ $2f_{Modulation}$). In this sense, FM is a spread-spectrum method.

SIGSALY transmits a form of noise (because the one-time-pad is noise) and probably uses the entire frequency bandwidth of the transmission medium to transmit that "noise". Even for the transmission of a payload signal consisting of just a single tone, the entire bandwidth is used. In this sense, SIGSALY uses a spread-spectrum technique. It is comparable to CDMA, which also uses "noise" for modulation.

It would be interesting to see if it would work to transmit two (or several) SIGSALY payloads over the same transmission medium (at the same time), encrypted using two (or several) independent SIGSALY systems each using a different "key record". My guess is that only one noise bit is used for each payload bit. In this case it would not work. --RainerBlome 20:30, 24 October 2005 (UTC)

Bit rate

According to Diffie and Landau, Privacy on the Line, SIGSALY's vocoder ran at 2400 bps. It must have sounded terrible. Phr (talk) 09:40, 24 July 2006 (UTC)

Accurate to call it PCM?

The article claims that SIGSALY transmitted speech using PCM, but don't think that is a fair claim. SIGSALY transmitted the quantized parameters of an entirely analog vocoder rather than digitizing the the speech waveform directly as in a true PCM system. The vocoder parameters were not even transmitted via PCM: instead a subranging quantizer was used and its output was sent directly rather than in PCM form (see US patent 3,912,868). Although the output of a subranging quantizer can be losslessly converted to PCM, the distinction there is important because SIGSALY was in production use before Alec Reeves's patent on PCM was issued in the US. As far as I can tell, SIGSALY never adopted true PCM principles, which is not shocking since it mostly predated that information theoretic basis which allows us to understand the fundamental mathematical equivalence of multi-level and binary encoding. --Gmaxwell 19:35, 19 March 2007 (UTC)

emulation/simulation on modern computers?

Does such thing exist? It would be nice to "hear" how this analogue vocoder sounds (for people having the right decoding key and for people who don't. ;-)) --RokerHRO (talk) 21:07, 22 April 2010 (UTC)

I second this wish. Or, are there any samples of what the original system sounded like in existence? 88.195.53.186 (talk) 10:32, 13 September 2011 (UTC)

Why just "effectively" a One Time Pad?

This could probably take a 'minor' flag, but I'm new to this, and I know at least one Wikipedia editor will disagree with me, so I'm make a proposal rather than editing the text first.

The text reads: '"The records were effectively the SIGSALY "one-time pad"'

I'd like to change it to

'"The records were the SIGSALY "one-time pad"'

I don't see why the qualification is required. The records contained random signal which was quantized in the same way as the amplitude signal, then added to the data on transmission, using modular arithmetic, and then removed at the far end, by using an identical record. The records lasted as long as the data, and were used just once, then destroyed. The information was concealed in very much the same way as an old fashioned letter-based One Time Pad.

I can't see any way this is not a 'one time pad', and so I'd like to remove the qualification 'effectively'.

If I don't see major objections in a few days, I'll make the edit.

thanks

Petertrei (talk) 02:57, 18 May 2010 (UTC)

Math confusion

The article says "ten narrow-band (less than 25 Hz) signals, giving the amplitude in ten separate frequency bands, which together covered the telephone passband (250 Hz - 2,950 Hz);" but when I do the math I get 270 channels of 25 Hz to cover the specified band. Please explain and please provide a reliable source for the statement in the article. Edison (talk) 23:44, 7 December 2014 (UTC)

It's flat-out wrong. I linked in the patent. What it actually means is 'it samples 10 bands and then measures their amplitudes at 25hz' - the bands width is not 25hz. Corrected. --SpeedEvil (talk) 13:08, 25 March 2015 (UTC)

I agree with you both that that sentence is self-contradictory. I see one reference say that the vocoder parameters were measured every 20 milliseconds (50 times a second). (The patent mentions "zero to 25 cycles per second"). So I tweaked that sentence a little more. Better now? --DavidCary (talk) 23:44, 14 November 2019 (UTC)