Convert Bits/sec to Hz

Without knowing more, one can't. In general, for a given modulation scheme, increasing the data rate involves scaling everything down equally in time, which scales everything up equally in frequency.

But there are so many different ways to encode data onto a signal that the bandwidth vs. data rate spread is pretty wide. There are practical systems that have 16-point constellations (four bits/symbol) that can pack in four bits/Hz; there may be practical (or at least fitfully practical) systems that can do six bits/Hz. On the other hand, there are systems that are vastly spectrally inefficient; with those the sky's the limit as far as wasted bandwidth.

Hi, Tim is quite knowledgeable on these subjects and gave some usefull info but maybe your question is not quite so deep. Hertz is used to describe cycles/second (cycles per second). If you allow it to mean events per second then a bit can be an event. So 1Gb/sec is 10^9bits / sec or 1 Ghz. This is for mental clarity and the actual units may vary.

Tom

Yes, Tim touches on the old baud rate (symbol rate) vs frequency that you're suggesting as the simple answer.

Both are not incorrect (double negative only to show grey area).

For example 100Mbps network link bangs four symbols on each of 1/25MHz to get the speed, and Gbps copper network links don't run at 1GHz, they run slower speed, using a couple pairs in each direction.

So, in this case, context matters.

Where I said symbols, Tim said constellation, Tim is more accurate :)

Grant.

Not really. Symbols are the stars in a constellation of stars. The symbol is the basic unit of information transferred. A constellation is a representation of the possible physical values (voltage, frequency, phase, whatever).

A little more specifically, a symbol may be a specific value in the constellation or a transition from the current point to another.

Thanks to all for your feedback. I had thought about the baud rate idea that Tim elaborated on, but the context in which I asked this question was Ethernet bits going from the MAC layer to the PHY layer at say 1 Gb/s or more (for the newer standards).

MIME encoded XML is one serious contender for wasted bandwidth.

Again, you need to know how many bits per symbol are used. IIRC GBE is four bits (4-pairs in use).

I think 100Mbps symbol rate is 25Mhz, 4 bits/symbol, 1 pair each direction, but no idea what magic's done for 1Gbps over the 4 pairs.

Grant.

According to the Shannon-Hartley theorem, the highest possible data rate is:

C = B*log2(1 + S/N)

Where C is the channel capacity in bits per second

B is the channel bandwidth in Hz

Log2 is base 2 logarithm

and S/N is the signal to noise power ratio (as a linear ratio not in dB)

In practice you will always get less than this, though some modulation schemes can get quite close.

depends how you do the data framing too, how do you tell if a bit is a one or zero?

Grant.

What exactly are you trying to do here? a bit of background may help answer your question. E.g. are you trying to send data down a cable? if so will it be synchronous or asynchronous? Is it just one way communication between two devices?

Gareth.

PS top posting, that's posting your reply above the relevant text, can make threads difficult to follow. it is usually clearer to put your reply below the relevant text in the post you are replying to.

On 08/06/2010 03:42 AM, Daku wrote: (top posting fixed)

GB Ethernet uses 5-level signaling on four pairs of wires. I can't find a reference that gives the symbol rate; without forward error correction (FEC) it could be as low as 110MBaud. I don't believe this at all-- I'm sure there's at least some coding to keep the average voltage at 0 for transformer coupling, for error detection, and maybe even for error correction.

But surfing on Wikipedia for a few moments didn't tell me -- you'll have to dig up the appropriate standard and take a look for yourself!

OOps --

And data framing? Bit and byte/word separation? A square wave conveys no information until you agree on how to interpret it, frequency? So you need to modulate the pulse train in some fashion, agreed beforehand at both ends before you can convey information.

Basic information theory: sender and receiver must first agree on a dictionary :)

Grant.

it

is

Rather magical, part of the training sequence (on copper) determines cable length ;)

Grant.

I am creating Verilog models of high-speed Ethernet protocols (e.g.,

an integrated circuit. However, some of the IEEE specification documents that I have read do not mention any of the coding schemes etc.