Fibre Channel signal characteristics
Fibre channel electrical signals are sent over a duplex differential interface. This usually consists of twisted-pair cables with a nominal impedance of 150 ohms, in differential mode. This is a genuine differential signalling system so no ground reference is carried through the cable, except for the shield. Signalling is AC-coupled, with the series capacitors located at the transmitter end of the link.
I don't actually understand the definition of the Fibre Channel signalling voltage, since it's declared complex.
On it's db9 (copper) connector, it offers the following signals
Code: Select all
Pin Signal-name Comments
1 +OUT Fibre channel output
2 +5V Optional
3 Module Fault Detect Optional
4 * Reserved
5 +IN Fibre channel input
6 -OUT Fibre channel output
7 Output Disable Optional
8 GND Optional, return for pin 2
9 -IN Fibre channel input
Basically, it's a differential TX+/- and RX+/-, with the following characteristics
Tx PECL Signal Requirements
VI (max) 2000 mV
VI (min) 650 mV
Line Impedance ZO 150 ? ± 30 ?
Input Impedance RI 150 ? ± 10 ?
Det. Jitter (max) 0.09 UI p-p
Total Jitter (max) 0.24 UI p-p
Mark ratio 50 ± 10 %
Rise/fall time Max 350 ps Min 100 ps
Diff. Skew (max) 20 ps
Rx PECL Characteristics
VO (max) 2000 mV †
VO (min) 370 mV †
Line Impedance ZO 150 ? ± 30 ? †
Impedance | ZE |75 ? ± 25 ?
Det. Jitter (max) 0.38 UI p-p
Total Jitter (max) 0.60 UI p-p
Diff. Skew (max) 205 ps
I wonder if I can recycle two of those MIAs as "RS422-like". Of course the analogical side needs an adapter from TTL to PECL Characteristics. Can it work?
Optical fiber is better than copper against noises when you have to connect things around motors. I could use BASE1000-SX, but ... I don't need the complexity of TCP/IP.
The coding system on Fibre Channel ensures an equal number of ones and zeros are always sent. This is useful as it allows very tight automatic gain control in the receiver (esp for optical)
I wouldn't expect it to like data the does't have a balanced number of ones and zeros, or it the runs of ones and zeros are too long (upsetting any AGC loop)
What do *you* think, guys?