Analog Bandwidth Required for 125Mbps 8b10b Serial Link Over Copper Cable?

I have a custom serial data link that is electrically equivalent to Fiber Channel(copper), the only difference being the Bitrate of the link is only 125Mbs. The link spec is DC balanced 8b10b, 2Vp-p launch amplitude, transformer coupled.....
My question is what would the 'analog' bandwidth be considered? Is it 62.5 Mhz??? I am trying to specify coax and stp cabling for this link over distances approaching 300' or more. What frequencies in the cable spec(insertion loss, return loss etc....) should I be concerned with. Considering 125Mbs, what is the fundamental analog frequency and should I be concerned with harmonics?
Putting this link over glass would solve everything, however I need to spec copper over significant distances.
Any insight would be greatly appreciated. Thanks.

On a first pass let's clarify the digital bandwidth issue and compare it to the digital throughput:

Due to the fact it is DC balanced in an 8b10b manner, the throughput only accounts for 80% of the link bandwidth utilized. In other words, for every byte of data we are also transmitting two extra bits. So, right off the bat the link must support a minimum of 156.25Mbs.

Does this part make sense?

Thank you for your interest and responding to my questions.

Yes, indeed this makes sense. To clarify my specs, the pre-encoded data rate (throughput) is 100Mbs and the resultant link rate after 8b10b encoding is 125Mbs (line rate). If this seems right to you, what kind of 'analog' bandwidth will this signal need on a copper media such as coax or utp/stp?

Thanks again for your time.

Cable impedance would seem to be an obvious consideration.
Attenuation for long cable runs.
Capacitance for high speed signaling over long distance can create dispersion problems at your receivers.
Crosstalk performance on for long cable runs if you have multiple circuits in the same cable including two simplex circuits like a dedicated RX and TX path.
Any shielding requirements?

I would call an Apps engineer at Belden and see what they recommend.

Thanks for your thoughts on my inquiry......

I ultimately need to know at what frequency or Bandwidth in Hz (Mhz) this link will operate at in order for me to spec the copper media. When my bandwidth (in Hz, not Mbs) is known I can follow through with looking at insertion losses, return loss, NEXT, ATSUM, FEXT etc......

I am leaning towards coax and UTP at various lengths.....length being determined by the above characteristics @ the operating bandwidth.

Based on my link specs (8b10b encoding, DC balanced) I think the 'analog' bandwidth or fundamental frequency would be 62.5Mhz for a line rate of 125Mbs. If I'm wrong about this, please advise.

If I were to contact any cable manufacturer, probably the first question they would ask is 'what are your bandwidth requirements?'. Since media bandwidth and bitrate are not synonymous, I need to know the 'bandwidth' in Hz.

I have tested this link on Belden 7927a (CAT6a), bidirectional link using 2 pairs @ 300', and it works like a charm. Since the 7927a is a high end industrial 10Gb Ethernet construction, I would like to test lower end cables (CAT5e solid horizontal, & stranded patch) to determine my cable length limitations.

Just would like to know what my characteristic 'analog bandwidth' is for the link.

Your questions of equivalent analog bandwidth not so straightforward. Do you mean -3dB bandwidth? Do you mean the 'knee' frequency based on the pulse risetime? Do mean the bandwidth required to implement relative error-free communication over a long cable? It sounds like the latter and I beleive attenuation and dispersion, both length dependant, are going to be your major foes. The details of your receiver play a big part in the answer.

It sounds like you will not be prepared to call the experts until someone gives you a number.
A source like this http://pdfserv.maxim-ic.com/en/an/AN3455.pdf shows the bandwdth to the first null in the spectrum is inversely proportional to the bit period, so 125Mbps NRZ data would be 125MHz. (However, you may need some of those high order harmonics to avoid dispersion over a long link.) Now take that number and call the experts at Belden!

Yes, all I want to quantify is the bandwidth of the first harmonic (fundamental frequency) of the 8b10b 125Mbs signal so I can establish a starting point for specifying cable.

If the fundamental/first harmonic is 125Mhz, this is 2x 62.5Mhz, when factoring harmonics this is significant when looking @ lossy, low pass cable with finite bandwidth.

The app note from Maxim is very good info. I have attached a partial pdf rip from an article that attempts to explain what I'm looking for. The formula on page A14 (pg3) is what I'm trying to validate. Data Rate = 2 x Bandwidth.

Also, this link: http://en.wikipedia.org/wiki/Fibre_Channel_electrical_interface ,first paragraph describes an example Fiber Channel signal of 2Gbs approximates a sine wave with a fundamental frequency of 1Ghz.....1/2 the line data rate.

Bottom line this for me, and forget about the cable characteristics. Is the fundamental frequency of the 8b10b 125 Mbs (Fiber Channel like) link 125Mhz, 62.5Mhz, or some other value not mentioned????

Per the Maxim app note I linked, the spectral lines are related to your datarate (125Mbps) and your pattern length. You say your data is FC-like so lets use 127 possible patterns. Your spectral spacing will be at 125Mbps/127 = 984KHz.

I think you are barking up the wrong tree just looking for 'fundamental frequency '. With a pseudo-random data sequence you will get an envelope of energy because your data pattern is so varied. Good luck. Please post back here what you find works.

Eric,

Sorry for my delay, the holidays are always busy for me.

Ok, without getting pedantic, there is a general rule of thumb for determining the analog bandwidth necessary for a digital signal:

BW = 0.35 / Tr.

Where BW is the Bandwidth
Tr is the rise time
and 0.35 is a constant which was derived through a rather lengthy set of general conditions and then rounded up for some insurance.

You only need to know your desired rise time. I would suggest finding out what the requirement is at the receiver... or possibly matching the transmitter's capabilities.

I assume you are familiar with reflections, impedance discontinuities and so on.

You should be good on using that equation to establish your minimum bandwidth required.

Make sense?

By the way... I just read some of the posts below...

Eric, the harmonics are only relevant for purposes of signal integrity (i.e. rise time) and the above formula completely takes this into account.