🚀 go-pugleaf

urbina@novax.llnl.gov writes:
>1.) My design right now incorporates a 14 bit serial output ADC. I sample the
>speech at 7khz which is double the bandwidth of telcom speech. The output from
>the ADC is XOR'ed with a 31 stage linear feedback shift register that generates
>several hours of pseudorandom "noise" before it repeats itself. (It's clocked
>at 7khz to give me that long turn-around time.) This encrypted data stream gets
>sent over the telco line (in theory) and the receiving party whose PN generator
>is synced with mine, decodes the encrypted data stream and the remaining voice
>bits are routed to his/her DAC. My question is 7khz too fast for telcom ?
>I've heard of speeds of 9600 baud with modems. Appreciate any help with this.

       7kHz times 14 bits = 98,000 bits/second.  This is well beyond
existing modem technology.  The usual telephone standard is 8 bits, and the
D/A isn't linear, it has a psuedo-logarithmic response curve to improve
the dynamic range.  There are standard IC for this, called "codecs".

       This still gets you 56Kb, which, in fact, is the rate used in the
digital portions of the telephone system.  But you can't push that through
a modem without some compression.  Most modern cryptosystems compress the
audio down to 9600 or 4800 baud, using linear predictive compression (LPC).
There are ICs for this, too.  Of course, an encrypted ISDN phone would
be much easier; no compression required.  That would be a nice
project.

       Then you need to encrypt.  A 31-bit recirculating shift register
is a very weak encryption algorithm.  Ask in sci.crypt for a better one.

       You might be able to do the whole job in a DSP chip, including the
modem.  That might be the lowest-cost solution; DSP chips are getting cheaper.

>2.) The quick and dirty way I've synced the pseudonoise generators of the
>calling and receiving parties is that I've 2 LM567 tone decoders that monitor
>the telco line for say...the digit 7 (852hz + 1209hz.) When it detects that..

        Using a tone decoder to indicate you want to turn the system on
is fine, but you probably won't get precise enough timing from them to
sync to the 100uS required at even 9600 baud.  Better to send some sync
pattern in the digital data stream and sync up on that.

        AT&T is now selling a real encrypted phone for around $1300,
based on the STU-III technology.  The problem still seems to requires enough
parts to keep the cost high.

        If you're more of an analog person, you might look into Turing's
original analog voice cryptosystem, described in "Turing - the Enigma".
1940s technology, but considered sound by Turing, who furfils the
criterion set down by Friedman: "No new cryptosystem is worth looking at
unless it was invented by someone who has already broken a very hard one."

					John Nagle