Tuesday, November 10, 2015

Computer Science Notes

exam time again ! :|

7-state process Model - note that a process can go also from  the running state directly to the Ready/Suspended state


Some encoding schemes applied at the physical layer in a network protocol model - either the OSI model or a '5 layer model' omitting the session and Presentation layers ( Please Do Not Touch Steve's Pet Alligator, PDNTSPA, Phy, DataLink, Network , Transport, Session, Presentation, Application)


Encoding (the primary function of the PHY layer - turning bits into signals)

Services are provided to the layer above

NRZ Non-Return to Zero

a high voltage indicates a one - a low voltage indicates a 0
This exhibits a problem when there is a large stream of consecutive ones or zeros - the scheme is prone to base-line shift, where source and destination host clocks become out of synch

NRZI Non Return to Zero Inverted
To counter the previous problem, a one is indicated with a phase shift and a zero is indicated by no shift, but this only solves the problem for ones, the same succeptability to baseline shift is present with zeros. ie only half the problem is solved

Manchester

there is a phase shift every clock-tick as the signal is XOR'D with the clock. This solves the baseline shift problem, but costs bandwidth though - bandwidth is reduced by 50%. A one is encoded with a Descending voltage and a zero with a Ascending Voltage.

This is possibly counter intuitive - intuition may assume that a 1 is associated with an ascending voltage, and zero with a descending voltage - but this is opposite of the
case.


4B/5B
This is the strongest solution to baseline shift of all previously mentioned encoding schemes. The 16 possible combinations of every 4 bits are mapped to 16 of the 32 possible combinations of using 5 bits instead, and the other 16 combinations ignored or used for frame detection. The 16 look up codes chosen are those with phase shift in at least 2 in the Block (B) of bits - so there is no risk of baseline shift.

More bits are thus needed to encode the same message compared to NRZ and there is the overhead of the look to dictionary process.


No comments: