We started discussing about that on this thread http://arduino.cc/forum/index.php/topic,97347.msg917036.html#new
but we went OT. So I decided to move it here.
This is the target and assumptions:
- A set of boards, Atmel 328 based, targeted to domotics
- All boards connected through a single twisted pair where either communication data, either DC is conveyed
- Bus topology: star or multi-star
- Bus sharing: multi-master with collision avoidance
- Data transfer: enough to provide command and parameters transfer for remote actuations (i.e. no realtime audio and/or video). A simple 9600bps should be enough for the most actuations
- Cheap and with a small PCB footprint
Discussions on this forum revealed that the bus should supply between 24V to 30V at, more or less, 1A. This is enough for a medium/small installation with up to 50 nodes on a single bus.
These are some already evaluated (and already discarded) solutions. More details on the mentioned thread.
- RS485 transceivers de-coupled with capacitors
- Tx based on open collector and/or open drain decoupled with capacitors
- Tx push-pull based decoupled with capacitors
- Each power supply on each board decoupled by mean of inductors
- Main power supply decoupled by inductors
We already discussed about the modulation that could be used to transmit data over the DC. An FSK solution was proposed because seems something doable taking advantage of some hardware peripheral present in the Atmel. The proposed two symbols where located at, more or less, 4KHz and 7KHz.
The solution to use inductors and capacitors, after simulations, ended up with the idea that it's not a reasonably cheap solution when running at that frequencies. So we need to think to a different hardware and, probably, a different modulation.
The main problem we faced are:
The proposed Tx behave like a pulsed generator. Either if implemented by a driver, either if implemented through an open collector or open drain, it was always powered by an onboard 5V regulated. It was always decoupled to the DC bus through a capacitor thus producing a no more than a maximum of 5V injected signal (when no load at the bus!). The internal resistance of the stage was not able to properly drive the bus when more than two or three units were connected to it.
Idea: what if replacing the proposed Tx stadium with something working directly with the DC bus? That means no decoupling capacitors and (theoretically) infinite current...
The proposed onboard power supply was connected to the bus through an inductor. This prevents the FSK signal to be shorted by the big capacitor in front of the voltage regulator. Without the inductor the capacitor is like a few ohms resistor for the FSK signal. This provides a "short way" for the ground... In order to provide a high impedance for the FSK signal, a too large inductor should be used. The best would be to have NOT a capacitor... but this is not possible for the voltage regulator (right?).
Idea: is there something we can use to replace the inductor? At school I remember I studied that an emitter follower presents an input equivalent impedance that was related to the beta of the transistor used.... Is that something valid even in this situation?
But I think we can start focusing on the Tx part than move to the other problems...
Who is interested on that!? This could poses the basis for a new set of open based protocol, shields and products.