Re: TI-H: Re: IR Link Status (was: expensive remote control)


[Prev][Next][Index][Thread]

Re: TI-H: Re: IR Link Status (was: expensive remote control)




Jon Olson wrote:

> But it's getting calcs to recognise their own string, and to make sure it
> got sent to the RIGHT person, etc.

This is exactly what I2C (the MBus) does. It just can't be done with the TI
protocol, which is designed strictly for two calcs. 

Let me take back what I just said. Perhaps it could be made so, that the
first byte would be an address byte, just like in I2C, and it would be sent
with 100 baud or whatever is the maximum rate for a multiple calc transfer.
The slaves (receivers) which wouldn't recognize the address as their own
would then switch off, leaving only the calc that initiated the transfer
and, possibly, the one that recognized the address, active on the bus.
These two could then send the data with the TI protocol at any rate they
can deal with, without others interfering.

However, this type of protocol would combine the disadvantages of the TI
protocol and I2C/MBus. It would require its own driver, because the TI
protocol routines in the ROM only deal with high-speed (point-to-point)
transfers and not ~100 baud ones. The variable transfers in ROM in
particular wouldn't work. These are also the main disadvantages of I2C.
In addition, there wouldn't be any collision detection, and no support for
external devices, as there is when using I2C. The first byte would take
about one tenth a second, compared with under 10 milliseconds (one
hundredth a second) for I2C.

The only advantage I see for this kind of a link protocol is that it would
work with the present IR links. However, I am currently helping PEA to
develop a PIC-controlled IR link with native I2C support (in addition to TI
protocol and line-state support). I think this, or a similar IR link, will
put an end to these discussions, and in the future, we will only have two
link protocols: The TI protocol for point-to-point transfers, and MBus/I2C
for multicalc transfers.

-Ozone

-- 
*** Osma Suominen *** ozone@clinet.fi *** http://www.clinet.fi/~ozone ***


References: