I2C 2.1 support for High Speed I2C, 3.4 Mb/s

Woodruff, Richard r-woodruff2 at ti.com
Sun Apr 24 02:52:01 CEST 2005

Hello Mark,
> I don't think anybody's ever asked us about it, much less worked on
> Looks like HS mode was defined in I2C V2.0 in 1998.

That's what I thought.

> Would you be working on a bit-banging driver or is here an I2C bus
> out there supporting HS mode?

Bus master.

> If you are going to write a driver please keep in touch and let us
> how it's going and what API extensions you propose.

I didn't think much on it.  Off the top of my head, I though client chip
drivers could advertise an attribute which indicates if they are hs-mode
capable.  If the bus master supports hs-mode and the client advertises
itself capable, then the master would just do it for transactions larger
than some number X.  After the lower speed setup phase is done, data can
be sent at max speed between hs targets even on a bus with slow devices.

> If a bit-banger big desision is to mod and/or start from i2c-algo-bit
> i2c-algo-biths.
> algo-bit is 33% duty cycle, biths is 50% duty cycle and in theory can
> go faster. However i2c spec says HS mode is 33% duty cycle. Go figure.
> See TODO for more info.

I wonder what rate a bit banging interface can achieve...

Richard W.

More information about the lm-sensors mailing list