Electronics > Projects, Designs, and Technical Stuff
SIGSALY 1942 ADC recreation with tubes
<< < (2/2)
johnwa:
Oh bugger, you beat me to it!

Some time ago, I was looking at some of GK's epic creations, and thought it would be good to start a large-scale analogue project of my own. I thought a recreation of SIGSALY would be a suitable choice.

I wasn't aiming for complete authenticity, choosing to use BJTs instead of glassfets, as complete construction of even a simplex link would require a huge number of active devices. So far, the only bits I have designed are the bandpass filter, envelope detector, quantisers, modulo summer, and frequency modulator. I am currently laying out a PCB for these modules, but there is still a lot of design work to be done. I really don't know when/if it will be finished!

But enough about my project. Congratulations on your achievement, Jon! I haven't had time to study the circuit in detail, but you have definitely done a nice job on the mechanical construction.  :-+ Were you working directly from an original circuit?

I don't know if a link has already been posted, but the Crypto Museum has an excellent description of how SIGSALY worked (and lots of other interesting things as well...)

jonpaul:
Hello all, many thanks for the comments.

I have researched the history of digital technology since 1990s and SIGSALY since 2003, and presented papers at SMPTE, AES, NAB, SBE, Bletchley Park, Invalides and ENST. Besides the first ADC and DAC, SIGSALY had the first unbreakable speech scrambler, first spread spectrum an was the first true digital signal processor, all this in 6 months from Bell Labs in NYC, 1942. Claude Shannon, Harry Nyquist, Homer Dudley and Alan Turing were involved to varying degrees.

My SIGSALY Quantizer goal was to make a dramatic demonstration of analog to 5 level digital conversion with voice input, that even non-technical and young viewers can understand.

After I reverse engineered  the topology of the SIGSALY"Stepper" or quantizer, and its use of five argon tetrodes 2051 thyratrons as  comparators, the rebuild took ~ 3 years overall.  I wanted to use vintage WWII era parts as close as possible to the original 1942 SIGSALY Quantizer and microphone preamp.  The design included some modern parts in the utility and support  circuits, like the power supply,  meter, buffer.

There are 2 articles on the IEEE Spectrum site, one for the general project and the second for the design and debugging details.

Finally here are private links to a 3 min video demos in English and French.

http://crypto-museum.org/QUANT/VIDEO/QUANTEN.mp4

http://crypto-museum.org/QUANT/VIDEO/QUANTFR2.mp4

Finally a complete SIGSALY reconstruction needs 72 quantizers and 384 thyratrons.

A DSP real-time simulator in software is planned.

Your comments and feedback appreciated.


Jon




johnwa:

--- Quote from: jonpaul on February 14, 2019, 02:11:48 pm ---
I have researched the history of digital technology since 1990s and SIGSALY since 2003, and presented papers at SMPTE, AES, NAB, SBE, Bletchley Park, Invalides and ENST. Besides the first ADC and DAC, SIGSALY had the first unbreakable speech scrambler, first spread spectrum an was the first true digital signal processor, all this in 6 months from Bell Labs in NYC, 1942. Claude Shannon, Harry Nyquist, Homer Dudley and Alan Turing were involved to varying degrees.

--- End quote ---

Yes, and the coding used is probably the first example of data compression being applied to speech. By my calculations, the effective bit rate was only ~ 1600bps, which is quite impressive, even if the result wasn't exactly "toll grade". The designers obviously thought about this quite a bit, even down to the nonlinear transfer function of the quantiser.


--- Quote from: jonpaul on February 14, 2019, 02:11:48 pm ---
My SIGSALY Quantizer goal was to make a dramatic demonstration of analog to 5 level digital conversion with voice input, that even non-technical and young viewers can understand.

After I reverse engineered  the topology of the SIGSALY"Stepper" or quantizer, and its use of five argon tetrodes 2051 thyratrons as  comparators, the rebuild took ~ 3 years overall.  I wanted to use vintage WWII era parts as close as possible to the original 1942 SIGSALY Quantizer and microphone preamp.  The design included some modern parts in the utility and support  circuits, like the power supply,  meter, buffer.


--- End quote ---

Well, you seem to have implemented at least the quantiser section with authentic 1940s parts! I can imagine how frustrating getting all the threshold voltages would have been.  :-/O I guess it makes you realise, although an ADC is fairly mundane these days, it was quite a precision piece of circuitry back then. And I suppose that all of the circuitry after the envelope detector would have to be DC-coupled, which would only add to the difficulties. Did you experience much trouble with drift over temperature and line voltage with your circuit Jon?

And I noticed the URL of your video, are you associated at all with http://www.cryptomuseum.com ?
jonpaul:
Hello all, I greatly appreciate your comments and feedback! More notes:

1/ There were 33 patents granted to BTL overall, some by very well-known engineers like Homer Dudley, inventor of the VOCODER. Many were classified secret at the time and only declassified in 1976.

Claude Shannon's fundamental paper Communication Theory of Secrecy Systems was inspired by his work at BTL on the algorithms of SIGSALY and classified during the war. It was finally declassified in October 1949 and published in BSTJ.

http://pages.cs.wisc.edu/~rist/642-spring-2014/shannon-secrecy.pdf
https://archive.org/details/bstj28-4-656

2/ RE The transmission rate and BW:  voice was analyzed into 12 parameters, 10 sub-band levels and 2 pitch parameters, also one bit for voiced/unvoiced. Each of the 12 analog parameters had 25 Hz bandwidth. 12 X 25 = 300 Hz, or 10X compression of voice information. Twelve quantizers converted the 12 vocoder parameters into six level digital signals. Bit rate: 50 Hz sample rate, 12 channels 2.5 bits (6 levels) per ch, thus total 30 bits at 50 Hz or 1500 BPS.

See for info on vocoder theory https://en.wikipedia.org/wiki/Vocoder

3/ The coding of the 12 quantized parameters employed MOD-6 addition. Each 6 level parameter was added to a 6 level band limited noise, and indeed recorded on 16" phonograph records.

4/ The final step before SW transmission was spread spectrum with SSB and FM, which expanded the bandwidth back to 3 kHz. SIGSALY had the first working spread spectrum, used not for secrecy but to reduce transoceanic SW radio fading by 20 dB.

5/ I know of no original SIGSALY recordings. But NSA and NCM have recordings of post war versions of SIGSALY using the same techniques. Very few if any are on the Internet.

See my interview and a short sound clip on PRI 99% invisible program
https://99percentinvisible.org/episode/vox-ex-machina/

6/ To probe further, see the IEEE and NSA papers, and my AES, NAB and SMPTE papers on the Origins of Digital Media and Origins of Compression.

This link has a precise technical description, which should answer many of your questions: https://www.cryptomuseum.com/crypto/usa/sigsaly/index.htm

7/ My 3 minute demo of the quantizer in action (pardon the amateur video quality, it was my very first video, done with limited equipment)

https://youtu.be/4TLMnFyapjA

8/ Connection from SIGSALY to a local phone eg at Cabinet War rooms in London, used secure copper circuits up to a few miles. The 4 wire SIGSALY secure circuits used noise and phantom circuits. They connected a SIGSALY terminal to a remote telephone e.g. Churchill War Rooms to SIGSALY miles away in the basement of Selfridges Department store. ( "phantom" is a circuit that uses common mode currents, for a second channel e.g. transmitted thru the center tap of a balanced line transformer).

https://en.wikipedia.org/wiki/Phantom_circuit
Its origin was in the 1800s to add a circuit on a telegraph line.

Again thanks to all for your fine comments!

Jon Paul

jonpaul:
Forgot: RE Drift: See the RCA Industrial tube manuals.

The 2051 thyratons show a 2:1 ratio in trigger voltage depending on filament voltage, anode V etc. In the articles I mention selecting and serializing the tubes, and that the BTL engineers and techs encountered the same issues.  SIGSALY had regulated power supplies and was maintained in an air conditioned room at 25 deg C +/- 0.5 deg. C. I had to readjust the R ladder several times to get the log levels correct and all 5 thyratrons to fire.

In the next versions, I will add trim-pots at each ladder tap point, just as BTL did in 1942.
After serializing and adjustment, the machine has performed perfectly at many venues and on various mains 120V, 240V.
I must test it on a variac at high/nom/low line!

Finally, I am now certain the quantizer was the first ADC (Reeves in UK had theory and papers on PCM in 1930s but could not build one!), first flash converter and first log law convereter!

SIGSALY was the  first unbreakable speech scrambler, first working spread spectrum and first digital signal processor, all designed at BTL in 6 months in 1942.

Enjoy,

Jon

Navigation
Message Index
Previous page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod