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 >> High-Pass Filter bandwidth Consideration for 8b/10b coding
https://designers-guide.org/forum/YaBB.pl?num=1242271987

Message started by neoflash on May 13^th, 2009, 8:33pm

Title: High-Pass Filter bandwidth Consideration for 8b/10b coding
Post by neoflash on May 13^th, 2009, 8:33pm

How we calculate the necessary HPF bandwidth for a 1Gb/s 8b10b coded NRZ bit stream?

I think in some spec it is defined as 75nF and 50ohm. What's the theory behind this number?

Title: Re: High-Pass Filter bandwidth Consideration for 8b/10b coding
Post by loose-electron on May 17^th, 2009, 11:07am

whats the spectral content of 8b/10b at a particular data rate?

Might want to look at that in the frequency domain and my guess is that the answer will fall out of that.

Title: Re: High-Pass Filter bandwidth Consideration for 8b/10b coding
Post by wave on May 21^st, 2009, 9:06pm

Is there a pre-able or some training waveform?

The data rate x # of training bits will set your HPF settling requirement, relative to your Comparator / ADC resolution.

:)
Wave

Title: Re: High-Pass Filter bandwidth Consideration for 8b/10b coding
Post by ywguo on Jun 29^th, 2009, 2:41am

Hi neoflash?

I guess what you need is AC couple between the transmitter and receiver. What Jerry said was correct. You must look at the spectral of the bit stream.

8b10b coding is used to scramble the bit stream in order to avoid consecutive 1's or 0's longer than 5 bit. I don't remember very clear, perhaps 4 bit. But the detailed length of 0's or 1's is not important. As well known, long length of consecutive 0's or 1's deteriorate the inter symbol interference (ISI). It is one kind of deterministic jitter in serial data communication. So the 8b10 coded bit stream does not have low-frequency spectral. Probable that is the theory what you want to know.

Yawei

Title: Re: High-Pass Filter bandwidth Consideration for 8b/10b coding
Post by rf-design on Jul 11^th, 2009, 1:01am

I did not remember it exactly but I guess the lower corner is 1/8 of the brutto baudrate. The design approach is to limit the DC-wander. So each of the 256 codes have a specific DC content. You can interprete the high pass filter as a integrator as feedback element. So a long sequence of symbols which have the highest number of ones or zeros change the DC-value of the feedback integrator. The DC-value reduce the useable eye-size. So independend of the integrator value a long sequence of the critical symbol will set the limit.

Title: Re: High-Pass Filter bandwidth Consideration for 8b/10b coding
Post by neoflash on Jul 12^th, 2009, 2:23pm

DC wander is something I'm worrying about.

One question is that 8B10B coding is defined as DC balanced. However, it is hard to define the low end of spectrual content.

By the way, what is the definition of brutto baud rate?

rf-design wrote on Jul 11^th, 2009, 1:01am: