Conference 7.286::fddi

688.0. "Twisted Pair FDDI Standard Update" by LEVERS::B_CRONIN () Tue Sep 01 1992 17:26

    Here is this months Twisted Pair FDDI update. The committee is
    developing a standard for FDDI over 100 ohm datagrade UTP (category 5) 
    and 150 ohm STP (Type 1 or 2). 
    
    The TP-PMD ad hoc working group met the week of 17 August in Bellevue 
    Washington. As you probably know, at its June meeting the committee 
    chose MLT-3 as the encoding scheme. At this meeting the big topic was
    the scrambling technique. Scrambling is needed to help pass the FCC and
    EN55022 radiated emissions tests. 
    
    Two scrambling techniques were under discussion. The first was a self 
    synchronizing scrambler, the second a stream cipher. The committee
    chose the stream cipher technique, which is described below. This was
    an important decision, because after encoding, scrambling was the most
    difficult choice for the committee to make. With this hurdle out of the 
    way the committee is much closer to creating a standard. That said, 
    its still too soon to predict the actual time at which the standard
    will be ready (sorry about that!). 
    
    In a stream cipher, a linear feedback shift register (LFSR) is used 
    as the randomizing element. In this case the polynomial used is 
    X^9 + X^11 (the output of the 9th and 11th registers are exclusive ORed, 
    and fed back to the input of the LFSR). The transmit data is then XORed 
    with this sequence, to produce the scrambled output. At the receive end 
    another X^9 + X^11 sequence is exclusive ORed with the received signal 
    stream to recover the transmit data. In Boolean terms, if A is the 
    transmit data and B is the LFSR output, then A XOR B XOR B = A. The real 
    challenge in all of this is making sure that the LFSRs at each end are 
    in the same state at the same time, relative to the bit that they are
    encoding/decoding. I won't go into the details here, but there is an
    elaborate procedure for making sure that the links use the Halt Line 
    State to get the two LFSRs into sync, and that they stay in sync once the 
    original data is being recovered. The syncronization is accomplished
    without any communication between the two LFSRs, and happens within the 
    connect state of the PCM state diagram. A block diagram looks like this: 
    
    +-----------+                     +------------+
    | LFSR      |-----+               | LFSR       |----+
    +-----------+     |               +------------+    |
                      |   +-----+                       |   +-----+
                      +---|XOR  |                       +---|XOR  |
             data---------|     |---------------------------|     |-- data
    			  +-----+                           +-----+
    
    		TRANSMITTER                        RECEIVER 
    
    
    At the next meeting the committee will discuss connectors, and the 
    gory details of the channel attenuation and crosstalk budget. After 
    that its on to the drivers and equalization details. 
    

    I usually am just a "listener", but something in the basenote startled
    me:
    
    Could someone elaborate just a little bit on how scrambling is
    connected with reduced emission levels?  Just curious!
    
    Rolf, DEC Brazil
                                        

While it's transmitting actual data, the FDDI signal can look fairly random.
But at other times it is highly correlated.  On an idle ring, most of what
you see are Idle symbols; a stream of Idle symbols corresponds to a 62.5 MHz
squarewave.

If you look at this with a spectrum analyzer, you'll see a few tall peaks.

The scrambler "randomizes" these fixed patterns, so the Idles that used to
produce a single frequency peak now look a lot more like noise, and on a
spectrum analyzer you'll see lots of little peaks.  The important point is
that the SAME total signal power is spread over more frequencies.

Given that the government RF emissions requirements set limits on radiated
(or conducted) power per unit of bandwidth -- in other words, limits on how
tall those peaks on the spectrum analyzer are allowed to be -- it helps to
cut down the tall peaks and convert them into a larger number of lower peaks.

	paul