[Ronja] 100Mbit over LEDs Gregory Maxwell gmaxwell at gmail.com Sun Nov 18 20:59:40 GMT 2007
* Previous message: [Ronja] e-mail of Ronja author * Next message: [Ronja] 100Mbit over LEDs * Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
How much thought has been given to the possibility of achieving 100mbit using LEDs?
In many countries laser emitters are subject to much more regulation than other visible light sources. For example, in the United States any laser with more than 5mw output is at risk of complications from regulations, even if the aperture is so large that only a tiny amount of light could enter the eye. Non-laser sources are not subject to the same scrutiny. Also, for a given output power non-laser sources are much cheaper (red-orange Luxeon 3W outputs >50mw of light and costs $10 .. I'm not aware of any laser diode with price performance near that)
Many years ago I measured the bandwidths of some leds that I had and found that they had a decent response to about 20MHz. I do not know what the ronja hardware delivers, but I would assume that it is around that, since about that much is required for 10mbit performance. This is not anywhere near enough bandwidth to do 100mbit using the binary NRZ encoding that ronja uses. It is also not enough to do 100mbit using the three level encoding used on 100TX (which needs about 35MHz).
It should, however, be possible to encode 100mbit into that bandpass by using more levels of amplitude modulation. Using multi-level amplitude modulation will require better signal to noise ratios and good driver and photo-cell linearity. I would not expect SNR to be a problem for ronja, but the use of multi-level encoding would add a lot of complexity to the electronics.
Quite aside from NRZ, surely the 4B5B encoding used by 100BaseTX can readily be converted to light? or am I missing something?
>How much thought has been given to the possibility of achieving >100mbit using LEDs?
The HPWT (and other) leds can be persuaded to signal att 100 Mbps, just using "brute force driving". However, an equalised led suffers from a much lower level, so distances possible are greatly reduced.
Multi level encoding is the way to go I think, encoding is easy, decoding involves linearity and AGC, so rx will be more complicated. But the phillips tza3033 TransimpedenceAmp does AGC etc. in one chip.
The AGC will be solved using TZA3033 transimpedance amplifiers from phillips.