[lm-sensors] Further documentation for w83627hf module

[lm-sensors] Further documentation for w83627hf module

[Julian Brunner]

Hello,

I'm trying to write an lm-sensors configuration for the mainboard
"Kontron 986LCD-M/mITX", which uses the "Winbond W83627THF" sensor
chip.

I contacted the manufacturer to request information about how the
sensors/fans/voltages are connected to the chip and how to convert the
raw values to sensible information. They sent me a table listing which
sensor/fan/voltage is connected to which pin on the chip, together
with the used resistors and conversion formulas. They table also has a
column saying on which lm-sensors output the sensor/fan/voltage should
appear (inX, fanX, tempX).

Unfortunately, the values that are reported don't match up (for
example, +5V is supposed to appear on in2, although values near 3.26V
are reported). Now, I guess it's possible that they used a different
module or a different version of the same module (to drive the sensor
chip) when they created the table.

So I started looking for information on the "w83627hf" module that I'm
running, to find out which input pins of the sensor chip are mapped to
which lm-sensors outputs. Unfortunately, the documentation I found for
this module (for example, the one at
http://www.lm-sensors.org/browser/lm-sensors/trunk/doc/chips/w83627hf?rev222
(I also checked the one of "w83781d" mentioned there)) doesn't seem to
contain any information on how the sensor chip inputs translate to
lm-sensor outputs.

So, I'd be very thankful for any advice on where to look for
information regarding this mapping from sensor chip inputs to
lm-sensor outputs.

Thanks in advance,
 Julian Brunner

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Further documentation for w83627hf module

[Jean Delvare]

Hi Julian,

On Sun, 14 Nov 2010 22:40:13 +0100, Julian Brunner wrote:
> I'm trying to write an lm-sensors configuration for the mainboard
> "Kontron 986LCD-M/mITX", which uses the "Winbond W83627THF" sensor
> chip.
> 
> I contacted the manufacturer to request information about how the
> sensors/fans/voltages are connected to the chip and how to convert the
> raw values to sensible information. They sent me a table listing which
> sensor/fan/voltage is connected to which pin on the chip, together
> with the used resistors and conversion formulas. They table also has a
> column saying on which lm-sensors output the sensor/fan/voltage should
> appear (inX, fanX, tempX).

This is very nice to them. Most manufacturers don't bother providing
that kind of technical details.

> Unfortunately, the values that are reported don't match up (for
> example, +5V is supposed to appear on in2, although values near 3.26V
> are reported).

Are you sure you declared the proper compute formula based on the
resistor values they provided? +3.26V looks like an unscaled value.

> Now, I guess it's possible that they used a different
> module or a different version of the same module (to drive the sensor
> chip) when they created the table.

"module"? Do you mean kernel driver? There has only ever been one Linux
driver supporting the W83627THF device, and that driver is the w83627hf
driver. This driver has been mostly unchanged for the past several
years, so I doubt that even an older version of that driver could
explain the difference.

> So I started looking for information on the "w83627hf" module that I'm
> running, to find out which input pins of the sensor chip are mapped to
> which lm-sensors outputs. Unfortunately, the documentation I found for
> this module (for example, the one at
> http://www.lm-sensors.org/browser/lm-sensors/trunk/doc/chips/w83627hf?rev222
> (I also checked the one of "w83781d" mentioned there)) doesn't seem to
> contain any information on how the sensor chip inputs translate to
> lm-sensor outputs.

The document above is outdated. It refers to the legacy version of the
driver, that was aimed at Linux kernel 2.4. For an up-to-date document,
please check the 2.6 kernel source tree:
  http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=Documentation/hwmon/w83627hf

But anyway, none of our driver documentation files lists the pin
mapping of sysfs attribute files. It would be redundant with the
datasheets. These datasheets were public, so even though I don't think
they are available on the manufacturer's site any longer, you should
have no difficulties getting your hands on the W83627THF datasheet.

> So, I'd be very thankful for any advice on where to look for
> information regarding this mapping from sensor chip inputs to
> lm-sensor outputs.

Google for "W83627THF datasheet". On the datasheet, there's one page
with the chip drawn and pin numbers with functions.

That being said, the information provided by Kontron should be
sufficient to write a libsensors configuration file. If you could share
the information with us, together with the configuration file you came
up with, we should be able to help you fix it.

Please also include the raw output of sensors (sensors -c /dev/null),
we will use it as a basis to validate the compute statements.

-- 
Jean Delvare
http://khali.linux-fr.org/wishlist.html

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Further documentation for w83627hf module

[Julian Brunner]

Hello Jean,

On Mon, Nov 22, 2010 at 10:49 PM, Jean Delvare <khali@linux-fr.org> wrote:
> Hi Julian,
>
> On Sun, 14 Nov 2010 22:40:13 +0100, Julian Brunner wrote:
>> I'm trying to write an lm-sensors configuration for the mainboard
>> "Kontron 986LCD-M/mITX", which uses the "Winbond W83627THF" sensor
>> chip.
>>
>> I contacted the manufacturer to request information about how the
>> sensors/fans/voltages are connected to the chip and how to convert the
>> raw values to sensible information. They sent me a table listing which
>> sensor/fan/voltage is connected to which pin on the chip, together
>> with the used resistors and conversion formulas. They table also has a
>> column saying on which lm-sensors output the sensor/fan/voltage should
>> appear (inX, fanX, tempX).
>
> This is very nice to them. Most manufacturers don't bother providing
> that kind of technical details.

Yes, I've done (or rather, tried doing) the configuration for two
mainboards from other manufacturers and had to learn the hard way that
the sensors conversion formulas are some of the best kept business
secrets, so I'm really grateful for this one.

>> Unfortunately, the values that are reported don't match up (for
>> example, +5V is supposed to appear on in2, although values near 3.26V
>> are reported).
>
> Are you sure you declared the proper compute formula based on the
> resistor values they provided? +3.26V looks like an unscaled value.

Well, the document doesn't say anything about resistors for that
particular pin (AVCC, Pin 114), and the "Multiply factor" is supposed
to be 1.

>> Now, I guess it's possible that they used a different
>> module or a different version of the same module (to drive the sensor
>> chip) when they created the table.
>
> "module"? Do you mean kernel driver? There has only ever been one Linux
> driver supporting the W83627THF device, and that driver is the w83627hf
> driver. This driver has been mostly unchanged for the past several
> years, so I doubt that even an older version of that driver could
> explain the difference.
>
>> So I started looking for information on the "w83627hf" module that I'm
>> running, to find out which input pins of the sensor chip are mapped to
>> which lm-sensors outputs. Unfortunately, the documentation I found for
>> this module (for example, the one at
>> http://www.lm-sensors.org/browser/lm-sensors/trunk/doc/chips/w83627hf?rev222
>> (I also checked the one of "w83781d" mentioned there)) doesn't seem to
>> contain any information on how the sensor chip inputs translate to
>> lm-sensor outputs.
>
> The document above is outdated. It refers to the legacy version of the
> driver, that was aimed at Linux kernel 2.4. For an up-to-date document,
> please check the 2.6 kernel source tree:
>  http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=Documentation/hwmon/w83627hf
>
> But anyway, none of our driver documentation files lists the pin
> mapping of sysfs attribute files. It would be redundant with the
> datasheets. These datasheets were public, so even though I don't think
> they are available on the manufacturer's site any longer, you should
> have no difficulties getting your hands on the W83627THF datasheet.
>
>> So, I'd be very thankful for any advice on where to look for
>> information regarding this mapping from sensor chip inputs to
>> lm-sensor outputs.
>
> Google for "W83627THF datasheet". On the datasheet, there's one page
> with the chip drawn and pin numbers with functions.

I've had had a look at this datasheet a few weeks ago already, I've
also seen the page with the pin layout and description. Still, I can't
figure out how this tells me on which sysfs file a particular pin's
value would appear (for example, the description of VIN0 only says "0V
to 4.096V FSR Analog Inputs"). And how would the authors of the
datasheet know about this mapping in the first place? I feel like I'm
getting something completely wrong here, it'd be great if you could
clarify how I would get the information I need out of this document.

> That being said, the information provided by Kontron should be
> sufficient to write a libsensors configuration file. If you could share
> the information with us, together with the configuration file you came
> up with, we should be able to help you fix it.
>
> Please also include the raw output of sensors (sensors -c /dev/null),
> we will use it as a basis to validate the compute statements.
>
> --
> Jean Delvare
> http://khali.linux-fr.org/wishlist.html
>

I've uploaded the document Kontron sent me at
http://multiserver.ath.cx/LM%20Sensor%20correction%20tables.pdf
(unfortunately, the connection is somewhat unstable, if it doesn't
work, try a few hours later or the next day, I think I can't attach
pdf files when sending stuff to the mailing list).

The configuration for lm-sensors I'm currently using is this:

chip "w83627thf-*"

    label in0 "VCore A"
    label in1 "+12 V"
    label in2 "+5 V"
    label in3 "+3 V"
    label in4 "+3.3 V"
    label in5 "VCore B"
    label in7 "+5 VSB"
    label in8 "V Battery"

    label temp1 "System Temperature"
    label temp2 "CPU Temperature"
    label temp3 "External Temperature"

    label fan1 "System Fan"
    label fan2 "CPU Fan"
    label fan3 "External Fan"

    compute in0 1.000 * @, @ / 1.000
    compute in1 3.801 * @, @ / 3.801
    compute in2 1.000 * @, @ / 1.000
    compute in3 1.000 * @, @ / 1.000
    compute in4 1.000 * @, @ / 1.000
    compute in5 1.000 * @, @ / 1.000
    compute in7 1.000 * @, @ / 1.000
    compute in8 1.000 * @, @ / 1.000

The output from sensors -c /dev/null is:

w83627thf-isa-0a00
Adapter: ISA adapter
in0:         +0.97 V  (min =  +0.70 V, max =  +1.87 V)
in1:         +3.17 V  (min =  +2.77 V, max =  +0.27 V)   ALARM
in2:         +3.26 V  (min =  +1.54 V, max =  +0.54 V)   ALARM
in3:         +2.96 V  (min =  +3.07 V, max =  +0.29 V)   ALARM
in4:         +1.52 V  (min =  +0.10 V, max =  +1.18 V)   ALARM
in7:         +2.96 V  (min =  +0.00 V, max =  +2.05 V)   ALARM
in8:         +3.12 V  (min =  +1.02 V, max =  +0.05 V)   ALARM
fan1:          0 RPM  (min = 15340 RPM, div = 2)  ALARM
fan2:       2789 RPM  (min = 84375 RPM, div = 2)  ALARM
fan3:          0 RPM  (min = 5075 RPM, div = 2)  ALARM
temp1:       +77.0°C  (high =  +0.0°C, hyst =  +4.0°C)  ALARM  sensor
= thermistor
temp2:       +67.5°C  (high = +80.0°C, hyst = +75.0°C)  sensor = thermistor
temp3:       +60.5°C  (high = +80.0°C, hyst = +75.0°C)  sensor = thermistor

So 12 V seems to be alright, VCore, too, I'll have to check with the
BIOS values later to confirm. 3 V could be right too, but seeing how
almost everything else is messed up, I'm not too sure that's not just
a coincidence, too. The temperatures are kinda off, too, seeing how
coretemp reports 47 ºC.

I'm not really sure what to do at this point, and whether I got
something backwards or the information from Kontron is incorrect. So
any pointers on what might be going on here are greatly appreciated.

Thanks in advance,
 Julian Brunner

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Further documentation for w83627hf module

[Jean Delvare]

Hi Julian,

On Tue, 23 Nov 2010 21:25:48 +0100, Julian Brunner wrote:
> Hello Jean,
> 
> On Mon, Nov 22, 2010 at 10:49 PM, Jean Delvare <khali@linux-fr.org> wrote:
> > Are you sure you declared the proper compute formula based on the
> > resistor values they provided? +3.26V looks like an unscaled value.
> 
> Well, the document doesn't say anything about resistors for that
> particular pin (AVCC, Pin 114), and the "Multiply factor" is supposed
> to be 1.

This isn't possible. The W83627THF has a 4.08 V ADC, so any voltage
above 4 V has to be scaled down using resistors.

> > (...)
> > Google for "W83627THF datasheet". On the datasheet, there's one page
> > with the chip drawn and pin numbers with functions.
> 
> I've had had a look at this datasheet a few weeks ago already, I've
> also seen the page with the pin layout and description. Still, I can't
> figure out how this tells me on which sysfs file a particular pin's
> value would appear (for example, the description of VIN0 only says "0V
> to 4.096V FSR Analog Inputs"). And how would the authors of the
> datasheet know about this mapping in the first place? I feel like I'm
> getting something completely wrong here, it'd be great if you could
> clarify how I would get the information I need out of this document.

The datasheet tells you the mapping between pins and registers. The
driver's source code tells you the mapping between registers and sysfs
attributes. I admit the information isn't immediately available, you
have to work out your way through the datasheet and the source code,
but the information is definitely there.

I admit that it would be easier if our documentation listed the pin <->
register <-> sysfs attribute mappings, however this would have to be
updated with every new chip we support, because manufacturers try to
make register-compatible devices (to some extent) but most often don't
care about pin compatibility (not for hardware monitoring features at
least.) Also, cases where you need to deal with the pin numbers are
rare, so this seems to be quite an investment for little return.

> > That being said, the information provided by Kontron should be
> > sufficient to write a libsensors configuration file. If you could share
> > the information with us, together with the configuration file you came
> > up with, we should be able to help you fix it.
> >
> > Please also include the raw output of sensors (sensors -c /dev/null),
> > we will use it as a basis to validate the compute statements.
> 
> I've uploaded the document Kontron sent me at
> http://multiserver.ath.cx/LM%20Sensor%20correction%20tables.pdf
> (unfortunately, the connection is somewhat unstable, if it doesn't
> work, try a few hours later or the next day, I think I can't attach
> pdf files when sending stuff to the mailing list).

I managed to download it :)

Assuming the labels are OK, there are clearly resistor values missing.
+5V and 5VSB definitely can't be measured without scaling resistors.

> The configuration for lm-sensors I'm currently using is this:
> 
> chip "w83627thf-*"
> 
>     label in0 "VCore A"
>     label in1 "+12 V"
>     label in2 "+5 V"
>     label in3 "+3 V"

There is no 3V in computers except for Vbat. So this would be  3.3V as
well (but see below.)

>     label in4 "+3.3 V"
>     label in5 "VCore B"

Better labelled "+1.5 V". Modern CPUs don't have a second Vcore. This
line is more likely monitoring the AGP voltage, or chipset voltage, or
memory (DDR3) voltage.

>     label in7 "+5 VSB"
>     label in8 "V Battery"
> 
>     label temp1 "System Temperature"
>     label temp2 "CPU Temperature"
>     label temp3 "External Temperature"
> 
>     label fan1 "System Fan"
>     label fan2 "CPU Fan"
>     label fan3 "External Fan"
> 
>     compute in0 1.000 * @, @ / 1.000

Note that you can simply omit such compute statements. Omitting them
makes libsensors faster.

>     compute in1 3.801 * @, @ / 3.801
>     compute in2 1.000 * @, @ / 1.000
>     compute in3 1.000 * @, @ / 1.000
>     compute in4 1.000 * @, @ / 1.000
>     compute in5 1.000 * @, @ / 1.000
>     compute in7 1.000 * @, @ / 1.000
>     compute in8 1.000 * @, @ / 1.000
> 
> The output from sensors -c /dev/null is:
> 
> w83627thf-isa-0a00
> Adapter: ISA adapter
> in0:         +0.97 V  (min =  +0.70 V, max =  +1.87 V)
> in1:         +3.17 V  (min =  +2.77 V, max =  +0.27 V)   ALARM
> in2:         +3.26 V  (min =  +1.54 V, max =  +0.54 V)   ALARM
> in3:         +2.96 V  (min =  +3.07 V, max =  +0.29 V)   ALARM
> in4:         +1.52 V  (min =  +0.10 V, max =  +1.18 V)   ALARM
> in7:         +2.96 V  (min =  +0.00 V, max =  +2.05 V)   ALARM
> in8:         +3.12 V  (min =  +1.02 V, max =  +0.05 V)   ALARM
> fan1:          0 RPM  (min = 15340 RPM, div = 2)  ALARM
> fan2:       2789 RPM  (min = 84375 RPM, div = 2)  ALARM
> fan3:          0 RPM  (min = 5075 RPM, div = 2)  ALARM
> temp1:       +77.0°C  (high =  +0.0°C, hyst =  +4.0°C)  ALARM  sensor = thermistor
> temp2:       +67.5°C  (high = +80.0°C, hyst = +75.0°C)  sensor = thermistor
> temp3:       +60.5°C  (high = +80.0°C, hyst = +75.0°C)  sensor = thermistor
> 
> So 12 V seems to be alright, VCore, too, I'll have to check with the

Yes.

> BIOS values later to confirm. 3 V could be right too, but seeing how
> almost everything else is messed up, I'm not too sure that's not just
> a coincidence, too.

From the above, I am almost certain you have the following:

in0 -> Vcore
in1 -> +12V
in2 -> +3.3V
in3 -> +5V (scaling needed)
in4 -> +1.5V
in7 -> 5VSB (scaling needed)
in8 -> Vbat

See how in5 doesn't exist, contrary to what the Kontron documentation
claims. So their documentation is definitely not correct. I suggest
that you recontact Kontron with that information and ask them for a
corrected document. Feel free to include me in the e-mail discussion.

Could you please check in the BIOS, which voltages are displayed and in
which order? And also the values of +5V and 5VSB.

> The temperatures are kinda off, too, seeing how
> coretemp reports 47 ºC.

Coretemp doesn't report accurate values, so I wouldn't use it as a
reference.

Do you really have a single fan in the system?

> I'm not really sure what to do at this point, and whether I got
> something backwards or the information from Kontron is incorrect. So
> any pointers on what might be going on here are greatly appreciated.

The information provided by Kontron is inconsistent, so I'm certain it
is incorrect. Hopefully they can provide fixed data.

-- 
Jean Delvare
http://khali.linux-fr.org/wishlist.html

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors