Pat, There are a couple of reasons for 8b10b encoding. DC wander is one of them. The encoding servos the DC level of the signal around. That way, if the signal is AC coupled (air, cap) then the baseline of the signal won't drift too far away from center and it makes it easier to design the comparators on the receive end. Also, the 8b10b encoding ensure a run length of no greater than 5 before getting a transition. If not using encoding, then it is possible to get long strings of ones or zeros. The pll would basically have to ignore all of these long strings which in turn means a very low bandwidth for the pll. This means increased die area and less jitter tolerance. A raw binary data stream has a PSD (Power Spectral density) that follows a sin(x)/x curve. It has a spectral null at the bit rate. 8b10b has a very good attenuation at DC and also the spectral null at the bit period. The clock recovery is usually handled by first doing frequency aquisition, and then when frequency locked the pll is switched into phase tracking mode. The phase detectors I have seen for this (phase mode) are the Hogge detector. Basically, it will not pass a pulse when data is not present. If you are familiar with the classic PFD, it passes pulses from both the reference and the VCO. The Hogge detector won't allow the charge pump to dump charge if no edges are present. So, this gets us back to defining the pll filter. With long strings of ones and zeros, the filter must be large so that it doesn't drift appreciably. With encoding, the clock recovery and pll are simplified by forcing edges at a certain rate. Hope that helps Tim "Zabinski, Patrick J." wrote: > Jeffrey, > > Thanks for the feedback. > > Looking at the spectrum of an un-encoded/raw data stream versus > an 8b/10b-encoded data stream, I can see how the power > spectral density will be increased at a frequency > equal to the data rate/2, which would provide more > information for a PLL to synch onto. > > If this is the case, is there a way of analyzing exactly how > much better a PLL can synch when using 8b/10b vs when > not using 8b/10b? If there is, can the analysis be > generalized to allowing me to determine how often > transitions need to be in order for a PLL to lock? > > Thanks, > Pat > > > > > # What's the purpose of 8b/10b encoding? > > > > Generally as I understand, it is to be able to recover the clock by > > ensuring some number of transitions per period (8 per 10?). > > Clocks change > > period frequently when working at high speeds, due to > > temperature, etc... not > > to meantion the fact that otherwise the sending clock would > > inevitably be > > out of phase, plus period, of your receiver's clock. > > > > > ------------------------------------------------------------------ > To unsubscribe from si-list: > si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field > > or to administer your membership from a web page, go to: > //www.freelists.org/webpage/si-list > > For help: > si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field > > List archives are viewable at: > //www.freelists.org/archives/si-list > or at our remote archives: > http://groups.yahoo.com/group/si-list/messages > Old (prior to June 6, 2001) list archives are viewable at: > http://www.qsl.net/wb6tpu > -- ------------------------------------------- Analog Devices "Enjoy yourself - 7910 Triad Center Dr it is later than you think" Greensboro, NC 336-605-4317 tim.stroud@xxxxxxxxxx ------------------------------------------- ------------------------------------------------------------------ To unsubscribe from si-list: si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field or to administer your membership from a web page, go to: //www.freelists.org/webpage/si-list For help: si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field List archives are viewable at: //www.freelists.org/archives/si-list or at our remote archives: http://groups.yahoo.com/group/si-list/messages Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu