Forums

non-integer fractionally spaced equalizer?

Started by Muzaffer Kal April 18, 2008
Hi everyone,
I'm considering implementing a fractionally spaced equalizer and one
question which has bugged me in the past is whether a non-integer,
smaller than 2 fractional spacing would have any benefits (ie better
equalization). Suppose the signal rate is 100 M bauds and the signal
bandwidth is around 70 Mhz. A T/2 fractional equalizer would need a
200 MHz ADC but would 140 MHz sampling have any benefits?

Any and all comments, ridicules, opinions welcome.

Muzaffer Kal wrote:

> Hi everyone, > I'm considering implementing a fractionally spaced equalizer and one > question which has bugged me in the past is whether a non-integer, > smaller than 2 fractional spacing would have any benefits (ie better > equalization). Suppose the signal rate is 100 M bauds and the signal > bandwidth is around 70 Mhz. A T/2 fractional equalizer would need a > 200 MHz ADC but would 140 MHz sampling have any benefits? > > Any and all comments, ridicules, opinions welcome.
Sampling at T/2 is the minimum sufficient for dealing with the non-integer delays. Sampling faster then T but slower then T/2 would be beneficial in some special cases. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
On Fri, 18 Apr 2008 14:24:25 GMT, Vladimir Vassilevsky
<antispam_bogus@hotmail.com> wrote:
>Muzaffer Kal wrote: > >> Hi everyone, >> I'm considering implementing a fractionally spaced equalizer and one >> question which has bugged me in the past is whether a non-integer, >> smaller than 2 fractional spacing would have any benefits (ie better >> equalization). Suppose the signal rate is 100 M bauds and the signal >> bandwidth is around 70 Mhz. A T/2 fractional equalizer would need a >> 200 MHz ADC but would 140 MHz sampling have any benefits? >> >> Any and all comments, ridicules, opinions welcome. > >Sampling at T/2 is the minimum sufficient for dealing with the >non-integer delays. Sampling faster then T but slower then T/2 would be >beneficial in some special cases.
Let's assume for the time being that I am doing phase optimization in my timing recovery loop (which is not decision dependent). What other benefits, under what cases would a faster than T, slower than T/2 sampling buy me?

Muzaffer Kal wrote:


>>>I'm considering implementing a fractionally spaced equalizer and one >>>question which has bugged me in the past is whether a non-integer, >>>smaller than 2 fractional spacing would have any benefits (ie better >>>equalization). Suppose the signal rate is 100 M bauds and the signal >>>bandwidth is around 70 Mhz. A T/2 fractional equalizer would need a >>>200 MHz ADC but would 140 MHz sampling have any benefits? >>> >>>Any and all comments, ridicules, opinions welcome. >> >>Sampling at T/2 is the minimum sufficient for dealing with the >>non-integer delays. Sampling faster then T but slower then T/2 would be >>beneficial in some special cases. > > > Let's assume for the time being that I am doing phase optimization in > my timing recovery loop (which is not decision dependent). What other > benefits, under what cases would a faster than T, slower than T/2 > sampling buy me?
The equalizer deals not only with ISI but with the noise also. Regardless of the data rate, the higher is the EQ sample rate, the less are the problems with the aliased noise components. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com