adoug
steve_mI've implemented bit-banged 100 Mbit/s Ethernet UDP transmission on the RP2040/RP2350. That is sufficient to transfer the data of the internal ADC overclocked to 5 MSPS and digitize some IF signal...
Here's the source: https://github.com/steve-m/Pico-100BASE-TX
private may, girlsoldier 
@steve_m what does this mean 
Klaus FrankThey're toggling a GPIO pin fast enough to simulate Ethernet without an Ethernet chip. Basically doing:
set pin HIGH
sleep x cycles, ms, or ns
set pin LOW
....
To form the network packages. It is cursed but it is also genius.
(And you're basically doing 100% CPU load most of the time to get the timing right for this)
Edit: Imagine you'd be flipping a light switch to send network packages. It's basically that.
@agowa338 @cyb3rm0shp1t Although the output is performed by a PIO state machine, which is fed with a 32 bit DMA. So the second CPU core, which I reserved for all the encoding, is not used 100% of the time, and the first core is available for the main 'user' application.
Yea, I kinda glanced over the PIO part that makes the Raspberry PI more suited for this as I haven't worked with it myself yet.
But you'd have 100% CPU load without it on "regular" GPIO pins.
Kevin Karhan 
@kkarhan What do you mean? The MCU is running at 250 MHz (double the Ethernet bit rate), with normal core voltage. It doesn't even get warm.
@steve_m really?
Huh...
Next thing you gonna tell me you're OpenSourcing it as a.general-pirpose Ethernet adaptor to DIY with pinout and firmware to fladh and add a socket to...
@kkarhan Yeah, I'll release it as OSS in the next weeks after ironing out some things and doing some proper host-side receive code.
@steve_m Cool, cuz that'll certainly make it a nifty choice for many projects...
MadMike77@MadMike77 @steve_m Eeyupp!
jiub 
@steve_m wow that's amazing
i thought you'd need to level shift to do ethernet, but i don't know much about the physical layer of ethernet
i thought you'd need to level shift to do ethernet, but i don't know much about the physical layer of ethernet
@jiub Fast Ethernet (100BASE-TX) is MLT-3 encoded, so there are 3 signal levels: -1, 0, +1. If you take the voltage between two digital GPIOs, you can emulate this by using 0b10, 0b00, 0b01.
@steve_m TIL! that's really clever
the 2040/2350 are such cool chips, it kind of boggles my mind that they can bitbang protocols that operate at hundreds of megahertz
the 2040/2350 are such cool chips, it kind of boggles my mind that they can bitbang protocols that operate at hundreds of megahertz
lydia! was at eh23@steve_m minecraft server when :D (/s)
Z̈oé@steve_m I’m impressed that this works and at the same time scared for signal integrity
like, what if someone deploys this en masse
„my sensor only works with this specific cable“
It's not that bad actually. The signal integrity isn't the issue. The issue here is that ANY interrupt occurring will mess up your timing and you also basically can't do anything else on that CPU core as it needs to "busy-wait" to toggle the pin at the correct time.
a RP2040 is technically even better suited for this than e.g. ESP8266 that I've seen others do this with in the past as it has special hardware for bit-banging but even though less severe still same issues apply
@agowa338 @uint8_t So far it works with all devices and switches I've tested. Of course your could add a matching network and proper magnetics to get close to a 'real' 100Base-TX transmitter, but unless you're using tens of meters of cable it's not required. EMI might be a different issue, though. As for the interrupts, this is not a problem - the PIO does clock-accurate output of the bits supplied by the DMA.
Just adding a properly* twisted and shielded cable should already allow you to get longer distances to work. After that also moving RX and TX to more "distant" pins could also help to reduce the near end interferences.
But besides that there shouldn't really be much more required to get a clean (but limited by the PIO/CPU of the microcontroller's speed) signal across.
* don't undo the twisting and shielding until the very end, like with patch panels
Zack Weinberg@steve_m I never dreamed I would live to see a day it was possible to bit-bang 100Mhz anything.
Luke Wren@steve_m I appreciate the artisanal hand-crafted twisted pair
Should you be able to get energy harvesting via the Ethernet port working as well then I'd be interested in c'n'p-ing your project 😅 (you basically beat me to it then. Wanted to build something like that to read data from the HAN-Port of the smartmeter and send it over wifi/lora/bluetooth/zigbee/... towards a server for charting)
Tofu 
​
HP van Braam@steve_m wow! That's super impressive! I wouldn't have thought it would be possible to bit-bang ethernet!
Super cool hack!
truh
__luna (headpats appreciated)how do you write gpio that fast?? isnt that ~2cycles/symbol?
unnick
juno doe âšµ@steve_m this is one of the coolest things I've seen recently
itchydog/SN-1572@steve_m I'm a huge fan of bit banging protocols and this is AMAZING
Baggypants@steve_m ah wait, was that 10Base-T?