Fun puzzle: can you make a pure analog (bonus: unpowered) ethernet throttling knob?
-
-
The problem then is how to throttle from 10/100 Mbit down to kbit rates. Basic idea: simulate collisions.
-
Collision detection is off if full duplex is negotiated
-
Then force half-duplex by cutting 2 (ie 1 tx, 1rx) of the pairs. Specifically, the outer 2. An ethernet NIC should detect the change and fallback to 100MbE.
-
100Mbit still does full duplex
- 1 more reply
New conversation -
-
-
I think an analog packet dropper for 10/100 Ethernet is possible. Hey
@azonenberg, up to the challenge? -
You'd lose LFSR sync if you corrupted too many bits in a row. Best option would be to randomly introduce glitches to the differential zero state (short P/n together) at regular intervals, causing single bit CRC errors/invalid symbols
-
With MLT-3 your glitch has a 50% chance of happening and a 50% chance of being ignored (signal was already 0). So drop rate is half of glitch rate. Drop rate is measured in packets/sec, not bits/sec, so smaller packets less affected
-
Bigger question (might be fun to try fuzzing this): How much corrupted data can a PHY handle before dropping the link and resyncing? A single bad MLT-3 symbol triggers a cascade of malformed data before recovering some time later. You'd get 1-2 bad 4b/5b symbols in a row
-
As it stands now TRAGICLASER doesnt do a great job of resyncing but real PHYs probably have a wider recovery tolerance. Unsure if real PHYs can resync the LFSR without resetting the link entirely (autoneg is slow vs re-LFSR)
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.