[lm-sensors] Intel DH55HC Sensors

[lm-sensors] Intel DH55HC Sensors

[Christopher Head]

Hi,
I'm trying to get my DH55HC's sensors working. Detection indicates
w83627dhg-isa-0a10 as the device. According to the wiki at
<http://lm-sensors.org/wiki/VoltageLabelsAndScaling>, this chip is
capable of reading 8 mV/LSb. Here's the problem: the division ratio
doesn't make sense! In the BIOS, I observed the following values on the
12 V rail:

11.828
11.844
11.859

The difference between these values is 15 or 16 mV. But 12 V / 0.016
V/LSb = 750 LSb! Since these sensor chips don't read raw values above
255, how in the world am I getting 16 mV resolution on a 12 V input?
I'm seeing similar results on my 5 V rail, which returns values just 2
mV apart in the BIOS, and on my so-called "Memory V_SM" rail (2 mV) and
"+Vccp" rail (differences as small as 1 mV being displayed).

Any ideas?

Chris

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

Re: [lm-sensors] Intel DH55HC Sensors

[Guenter Roeck]

On Wed, Mar 14, 2012 at 05:36:59PM -0400, Christopher Head wrote:
> Hi,
> I'm trying to get my DH55HC's sensors working. Detection indicates
> w83627dhg-isa-0a10 as the device. According to the wiki at
> <http://lm-sensors.org/wiki/VoltageLabelsAndScaling>, this chip is
> capable of reading 8 mV/LSb. Here's the problem: the division ratio
> doesn't make sense! In the BIOS, I observed the following values on the
> 12 V rail:
> 
> 11.828
> 11.844
> 11.859
> 
> The difference between these values is 15 or 16 mV. But 12 V / 0.016
> V/LSb = 750 LSb! Since these sensor chips don't read raw values above
> 255, how in the world am I getting 16 mV resolution on a 12 V input?
> I'm seeing similar results on my 5 V rail, which returns values just 2
> mV apart in the BIOS, and on my so-called "Memory V_SM" rail (2 mV) and
> "+Vccp" rail (differences as small as 1 mV being displayed).
> 
> Any ideas?
> 
The BIOS may get its voltages through HECI, which uses a different set of ADCs
built into the H55 PCH.

Guenter

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

Re: [lm-sensors] Intel DH55HC Sensors

[Christopher Head]

On Wed, 14 Mar 2012 20:36:55 -0700
Guenter Roeck <guenter.roeck@ericsson.com> wrote:

> On Wed, Mar 14, 2012 at 05:36:59PM -0400, Christopher Head wrote:
> > Hi,
> > I'm trying to get my DH55HC's sensors working. Detection indicates
> > w83627dhg-isa-0a10 as the device. According to the wiki at
> > <http://lm-sensors.org/wiki/VoltageLabelsAndScaling>, this chip is
> > capable of reading 8 mV/LSb. Here's the problem: the division ratio
> > doesn't make sense! In the BIOS, I observed the following values on
> > the 12 V rail:
> > 
> > 11.828
> > 11.844
> > 11.859
> > 
> > The difference between these values is 15 or 16 mV. But 12 V / 0.016
> > V/LSb = 750 LSb! Since these sensor chips don't read raw values
> > above 255, how in the world am I getting 16 mV resolution on a 12 V
> > input? I'm seeing similar results on my 5 V rail, which returns
> > values just 2 mV apart in the BIOS, and on my so-called "Memory
> > V_SM" rail (2 mV) and "+Vccp" rail (differences as small as 1 mV
> > being displayed).
> > 
> > Any ideas?
> > 
> The BIOS may get its voltages through HECI, which uses a different
> set of ADCs built into the H55 PCH.
> 
> Guenter

It's possible; I did see a mailing list message (on this list, in the
archives) explaining that the PCH is used for fan speed reporting on
this board. Is there any more recent news on PCH support in lm_sensors?
I don't have a lot of time but if it's a programming problem I may be
able to give some assistance.

Chris

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

Re: [lm-sensors] Intel DH55HC Sensors

[Guenter Roeck]

On Mon, 2012-03-19 at 17:01 -0400, Christopher Head wrote:
> On Wed, 14 Mar 2012 20:36:55 -0700
> Guenter Roeck <guenter.roeck@ericsson.com> wrote:
> 
> > On Wed, Mar 14, 2012 at 05:36:59PM -0400, Christopher Head wrote:
> > > Hi,
> > > I'm trying to get my DH55HC's sensors working. Detection indicates
> > > w83627dhg-isa-0a10 as the device. According to the wiki at
> > > <http://lm-sensors.org/wiki/VoltageLabelsAndScaling>, this chip is
> > > capable of reading 8 mV/LSb. Here's the problem: the division ratio
> > > doesn't make sense! In the BIOS, I observed the following values on
> > > the 12 V rail:
> > > 
> > > 11.828
> > > 11.844
> > > 11.859
> > > 
> > > The difference between these values is 15 or 16 mV. But 12 V / 0.016
> > > V/LSb = 750 LSb! Since these sensor chips don't read raw values
> > > above 255, how in the world am I getting 16 mV resolution on a 12 V
> > > input? I'm seeing similar results on my 5 V rail, which returns
> > > values just 2 mV apart in the BIOS, and on my so-called "Memory
> > > V_SM" rail (2 mV) and "+Vccp" rail (differences as small as 1 mV
> > > being displayed).
> > > 
> > > Any ideas?
> > > 
> > The BIOS may get its voltages through HECI, which uses a different
> > set of ADCs built into the H55 PCH.
> > 
> > Guenter
> 
> It's possible; I did see a mailing list message (on this list, in the
> archives) explaining that the PCH is used for fan speed reporting on
> this board. Is there any more recent news on PCH support in lm_sensors?
> I don't have a lot of time but if it's a programming problem I may be
> able to give some assistance.
> 
Chris,

I briefly looked into this a while ago, after the HECI (MEI) driver was
put back into the kernel. Problem is that it is not well documented, and
it is built around the assumption that it would only be used from user
space applications. This means one would need to understand it enough to
1) port the necessary user space libraries into the kernel, 2) modify
the MEI driver to support those in-kernel libraries, and 3) use the
libraries to write a hwmon driver.

Writing the hwmon driver would be straightforward, but the rest is just
too much for my spare time.

Guenter



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

Re: [lm-sensors] Intel DH55HC Sensors

[Christopher Head]

On Mon, 19 Mar 2012 14:46:03 -0700
Guenter Roeck <guenter.roeck@ericsson.com> wrote:

> Chris,
> 
> I briefly looked into this a while ago, after the HECI (MEI) driver
> was put back into the kernel. Problem is that it is not well
> documented, and it is built around the assumption that it would only
> be used from user space applications. This means one would need to
> understand it enough to 1) port the necessary user space libraries
> into the kernel, 2) modify the MEI driver to support those in-kernel
> libraries, and 3) use the libraries to write a hwmon driver.
> 
> Writing the hwmon driver would be straightforward, but the rest is
> just too much for my spare time.
> 
> Guenter

Alright, thanks for the information. One of these days I might take a
peek and perhaps try to hack up a patch, but it won't be any time soon.
Or maybe someone else will first.

Chris

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

Re: [lm-sensors] Intel DH55HC Sensors

[Guenter Roeck]

On Mon, Mar 19, 2012 at 07:18:07PM -0400, Christopher Head wrote:
> On Mon, 19 Mar 2012 14:46:03 -0700
> Guenter Roeck <guenter.roeck@ericsson.com> wrote:
> 
> > Chris,
> > 
> > I briefly looked into this a while ago, after the HECI (MEI) driver
> > was put back into the kernel. Problem is that it is not well
> > documented, and it is built around the assumption that it would only
> > be used from user space applications. This means one would need to
> > understand it enough to 1) port the necessary user space libraries
> > into the kernel, 2) modify the MEI driver to support those in-kernel
> > libraries, and 3) use the libraries to write a hwmon driver.
> > 
> > Writing the hwmon driver would be straightforward, but the rest is
> > just too much for my spare time.
> > 
> > Guenter
> 
> Alright, thanks for the information. One of these days I might take a
> peek and perhaps try to hack up a patch, but it won't be any time soon.
> Or maybe someone else will first.
> 
I had another quick look.

Replace 1) and 2) with "re-write the kernel driver to support non-file accesses".
The libraries are not really needed, but the kernel mei driver has file accesses
written all over it, so it would require a major rewrite to enable in-kernel access.

Guenter

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