The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> High-Speed I/O Design >> Why partial response is so called/named?
https://designers-guide.org/forum/YaBB.pl?num=1315933802

Message started by neoflash on Sep 13^th, 2011, 10:10am

Title: Why partial response is so called/named?
Post by neoflash on Sep 13^th, 2011, 10:10am

Like duo-binary coding (ak + ak-1).

Why it is called partial response? I understand it is a kind of linear Tx FIR filter, but I don't know what the "partial response" stands for?

Anyone can help understanding this in intuitive way?

Title: Re: Why partial response is so called/named?
Post by rfcooltools.com on Sep 13^th, 2011, 1:45pm

neoflash,

I believe its a method of optimizing data for a known communication channel so when it is received through that channel the least amount of information is lost.

A little history from the older hard disk drive "PRML" Partial Response Maximum Likelihood method of storing and retrieving information through an imperfect channel (actually this was my first IC design job). Is a form of error correction which uses the Viterbi algorithm and works by correcting errors based provided that the data is formatted and stored in a manner that when read the algorithm can correct it. The encoding on the storage of the data is expanded by inserting ones and zeros to essentially shape the power spectrum to maximize the spectrum of the constraining channel. For example imagine you had two data sequences.
a. 0101010101
b. 0011001100

if both a and b are passed through a low pass filter b has a better chance of not loosing information due to its lower harmonic content, even though both data sets represent 10 bits of information.

If the entire channel is charectorized from write to read-back then you can encode the data in such a way that its spectral content is a partial response of that channel. Unfortunately there is no panacea so the physical world data rate must increase beyond the bit rate to get an overall optimum system.

http://rfcooltools.com

Title: Re: Why partial response is so called/named?
Post by rfcooltools.com on Sep 13^th, 2011, 1:52pm

neoflash,

I also found this paper that describes it better than I did.
http://www.lintech.org/comp-per/09PRML.pdf
look at page 10, its probably what your looking for.

http://rfcooltools.com

Title: Re: Why partial response is so called/named?
Post by neoflash on Sep 13^th, 2011, 2:51pm

Very helpful explanation.

What did you mean by "partial response of the channel"?

You mean mimic the channel frequency response with the Tx PR filter?

Title: Re: Why partial response is so called/named?
Post by loose-electron on Sep 26^th, 2011, 6:46pm

Ah yes, read channels for disk drives, now applicable to SerDes.

PRML = Partial Response and Maximum Likelihood.

They are two separate things to understand.

Partial response is all about low pass filtering such that the higher frequency harmonics are removed and a better SNR and better BER are the result.

However, Partial Response means with higher harmonics removed the signal will have some "ringing" because the higher frequencies have been removed.

If you sample the signal at the null points of the ringing the signal sampled still represents the original signal.

Those samples can be used to represent the signal in a time domain sampled system, where the samples are compared to the expected ideal sampled signal. The difference between observed and expected are recorded and used to make a decision with respect to what was the actual signal. A Viterbi detector is used to do the math for this.

Best short form explanation I can toss out there.

BTW, back when PRML was a new thing to disk drives (circa 1992) some people used to call it:

PRML = Partially Right Mostly Lucky!

:D ;D