In comp.arch.embedded,
rickman <g...@gmail.com> wrote:
>> second-source OR roll my own.
>
> I don't think you will find that I2C is intended for hot-swap.  I am
> not saying it can't be done, I am saying that it has no built-in
> support.  The one-wire interface uses a CRC to verify communications
> and in general is intended for noisy or intermittent communications
> and supports hot-swap.

We did a schematic once with external I2C through an LTC4303 buffer
which is designed for hot-swap I2C applications. The external I2C
was dropped befor the actual board was made, so no real experience
with this buffer. If we go for I2C, I suspect I will have another
look at it.

> The one-wire spec is a bit tricky to read for timing, but the info is
> all there.  The only real issue in rolling your own is the timing.  An
> MCU will likely be run on an internal RC oscillator and will need to
> have some margin of stability and accuracy.  The timing margins are
> pretty wide to handle this, so the RC oscillators on many MCUs will be
> good enough, but you need to pay attention to this detail.  Otherwise
> the protocol is not difficult at all.
>
> I bet you can even beat the Dallas price on many one-wire devices and
> end up with a smaller footprint as well (other than the cap and the
> diode).

If we go with the 1-wire, I think well go with the standard parts at
first, but having the option of emulating a part is nice. That leaves
our option to bigger memories open and solves the single source issue.

-- 
Stef    (remove caps, dashes and .invalid from e-mail address to reply by mail)