Sensors chip driver for MAXIM MAX6650 / MAX6651 fan controllers
===============================================================

Chip notes
----------

The MAX6551 is a fan controller. It has two distinct functions:

1) Speed control, with feedback, for a single fan. The code can write
   to a register requesting that the fan run at a certain speed, and the
   chip will adjust the fan voltage until it reaches that speed. This
   is called "closed loop" mode in the datasheet.

2) Speed measurement (tachometers) for up to four fans, including the 
   controlled one.

The MAX6550 is similar but only has one measurement tachometer.

Both chips have a few general purpose i/o pins, some of which can be 
configured to have special functions. At the time of writing an overview 
of the chip, with links to the datasheet, can be seen at:

     http://www.maxim-ic.com/quick_view2.cfm/qv_pk/2296/ln/en

If that has moved searching for MAX6550 from Maxim's home page should 
work.

Driver state
------------

Currently the driver provides the following entries in /proc/sys/
dev/sensors/max6650-xxx/:

fan1 through fan4 (read only): the current measured fan speeds, in 
rpm.

speed (read/write): when written, sets the chip to closed loop mode and
sets the configured speed, in rpm, to the value written. When read returns
the configured speed, again in rpm.

xdump (read only): returns all of the chip register values, except the
tachometers. Used for debug, and should probably go away or be conditionally
compiled.

The MAX665x uses byte wide registers with various scaling factors and
some more or less complex equations (in the datasheet) relating register
values to real world values. The driver takes care of all of this.

Note that the measured speed ("fan1") never exactly matches the set speed
("speed") because of the coarse resolution of the underlying registers.

Assumptions
-----------

The driver assumes that:

 o The MAX665x is running from its internal 254 kHz clock, not an
   external clock.

 o The monitored fans produce two pulses per rotation (This same 
   assumption is made by Maxim throughout the datasheet.)

 o The tachometer count time (register 0x16) has been configured suitably
   for the fans in use, perhaps by some boot code. The driver reads this 
   register to compute rpm, but never writes to it.

 o The speed controller prescale (low three bits of register 0x04) has 
   been configured suitably for the fans in use, perhaps by some boot 
   code. The driver reads these bits to compute rpm, but never changes
   them.

The driver currently never reads or writes any other configuration 
(gpio config, alarms, 5V/12V setting).

Bugs
----

 o The driver assumes too much: it should be possible to set more of the
   configuration dynamically instead of assuming that some other code has 
   done it.

 o It does not differentiate between a MAX6550 and a MAX6551, as I can
   see no obvious way to do this. It has only been tested with a MAX6551 
   so if anyone gets a board with a MAX6550 to test with maybe something 
   will be revealed, such as the three missing tachometer registers
   returning all ones or something. (OK, so maybe I should have called
   this driver max6651, and ignored the max6650 totally...)

 o The /proc names don't match the standards in doc/developers/proc,
   as this chip doesn't map well to other fan controllers,

Detection notes 
---------------

The module will autodetect the presence of a max6650 or max6651 at any 
of the four possible I2C addresses for these devices. There is no ID 
register, but there are several registers with all zero bits at the 
high end. The driver checks that all of these are in fact zero before
declaring the chip to be a MAX6550/1.

sensors-detect
--------------

Support for MAX6650/1 has been added to sensors-detect. It has been
tested as correctly detecting a MAX6551 at I2C address 0x1B. Like the
driver, it cannot yet distinguish a MAX6650 and MAX6651.

prog/sensors
------------

Initial read-only support has been added to the "sensors" program.

Testing
-------

The driver and associated programs have been tested under Linux 2.4.18 
running on an embedded Intel XScale processor (actually an IXP425 running
in big-endian mode) with a MAX6651 chip.

Author
------

John Morris (john.morris@spirentcom.com)