Skip to contentSkip to conversation

IBM HDAPS, I need a tip.

DISCUSSION Alejandro Bonilla Beeche 9 participants 15 messages Aug 1, 2005

lore
Share & tools
lore
#0 lore raw

Hi Guys,

I hope you all aren't sick about the topic. I have a quick question...

Thanks to development of the driver from some nice guys, we are able to get data from the accelerometer. There is one problem. How do we calibrate the values that are outputed from the userspace? All PC's get a different value, and we can't really find the best solution. What is the scientific and smartest way to do this?

i.e. of the driver output from the userspace.
abonilla@debian:~/hdaps/hdaps-dave-0.02

$ ./ibm_hdaps_userspace /dev/ibm_hdaps 
x_accel: 409
y_accel: 528
   temp: 47
  temp2: 47
unknown: 7

If I move the PC 45 deg right.(Looking from front the left side is higher)

km_activity (keybd) = 0
km_activity (mouse) = 0
x_accel: 396
y_accel: 579
temp: 47
temp2: 47
unknown: 7

The thing is, people have different values, and I think they are also different depending on where they are.

Another question for this kernel inclusion (heh) Should we use Sysfs or should we use the userspace that outputs this data, else what is recomended?

Thanks in advance,

.Alejandro

#1 lore raw

Second Try... ;-)

Anyone?

.Alejandro

Quoted 41 lines · Alejandro Bonilla Beeche

Hi Guys,

I hope you all aren't sick about the topic. I have a quick question...

Thanks to development of the driver from some nice guys, we are able to get data from the accelerometer. There is one problem. How do we calibrate the values that are outputed from the userspace? All PC's get a different value, and we can't really find the best solution. What is the scientific and smartest way to do this?

i.e. of the driver output from the userspace.
abonilla@debian:~/hdaps/hdaps-dave-0.02

$ ./ibm_hdaps_userspace /dev/ibm_hdaps 
x_accel: 409
y_accel: 528
   temp: 47
  temp2: 47
unknown: 7

If I move the PC 45 deg right.(Looking from front the left side is higher)

km_activity (keybd) = 0
km_activity (mouse) = 0
x_accel: 396
y_accel: 579
temp: 47
temp2: 47
unknown: 7

The thing is, people have different values, and I think they are also different depending on where they are.

Another question for this kernel inclusion (heh) Should we use Sysfs or should we use the userspace that outputs this data, else what is recomended?

Thanks in advance,

.Alejandro

#2 lore raw
Quoted 3 lines · Alejandro Bonilla Beeche

Second Try... ;-)

Anyone?

You're obviously getting different values because "moving the laptop left to right" will produce a different acceleration every time.

So in order to calibrate it you need a readily available source of constant acceleration, preferably with a known value.

Hint: -9.8 m/sec^2.

Lee

#3 lore raw
Quoted 5 lines

So in order to calibrate it you need a readily available source of constant acceleration, preferably with a known value.

Hint: -9.8 m/sec^2.

Drop it out of the window? :)

Jan Engelhardt

#4 lore raw

Well don't forget there is a bios 'calibration' routine that you will see on start up (especially if you are on a moving vehicle/train). What is this calibration used for, and does it provide calibration information to the windows driver? Could we use it somehow to help solve this problem?

Yani

Quoted 13 lines · Jan Engelhardt
Quoted 5 lines

So in order to calibrate it you need a readily available source of constant acceleration, preferably with a known value.

Hint: -9.8 m/sec^2.

Drop it out of the window? :)

Jan Engelhardt
--

#5 lore raw
Quoted 2 lines · Yani Ioannou

Well don't forget there is a bios 'calibration' routine that you will see on start up (especially if you are on a moving vehicle/train).

I've never seen that. Could you please elaborate on what you see, and when?

-- Dave

#6 lore raw
Quoted 7 lines · Dave Hansen
Quoted 2 lines · Yani Ioannou

Well don't forget there is a bios 'calibration' routine that you will see on start up (especially if you are on a moving vehicle/train).

I've never seen that. Could you please elaborate on what you see, and when?

Sorry it says diagnostics, not calibration, but the way it behaves leads me to believe its doing some sort of calibration. If you boot up your thinkpad, get rid of your bios splash screen (esc), and tilt the thinkpad back and forth you will see something about IBM Active Protection diagnostics running and a spinner. If you keep tilting the notebook back and forth you can actually prevent the machine from booting because this will fail after a while. Normally you won't see it, I mainly notice it while in a train or car.

Yani

#7 lore raw
In reply to Jan Engelhardt
Quoted 6 lines · Jan Engelhardt
Quoted 4 lines

So in order to calibrate it you need a readily available source of constant acceleration, preferably with a known value.

Hint: -9.8 m/sec^2.

Drop it out of the window? :)

No, no. Constant gravity (like having the laptop sitting on the desk) "feels like" constant acceleration.

Dropping it out of the window should measure 0 m/sec^2, because the accelerometer is not working on an inertial referential (I hope this is the correct term in english...). For the accelerometer, this is just like the feeling of free falling inside an elevator: no gravity :)

Signature · 5 lines
Paulo Marques - www.grupopie.com

It is a mistake to think you can solve any major problems
just with potatoes.
Douglas Adams
#8 lore raw
Quoted 11 lines · Paulo Marques
Quoted 6 lines · Jan Engelhardt
Quoted 4 lines

So in order to calibrate it you need a readily available source of constant acceleration, preferably with a known value.

Hint: -9.8 m/sec^2.

Drop it out of the window? :)

No, no. Constant gravity (like having the laptop sitting on the desk) "feels like" constant acceleration.

Quoted 7 lines

Dropping it out of the window should measure 0 m/sec^2, because the accelerometer is not working on an inertial referential (I hope this is the correct term in english...). For the accelerometer, this is just like the feeling of free falling inside an elevator: no gravity :)

--
Paulo Marques - www.grupopie.com

You need a centrifuge or something that works like one. You can make one and you can calibrate it using simple techniques. For instance, you can swing the mass (laptop) over your head, round and round with your arm. If the rate is such that the mass "just" falls over the top, the 'G's at that point are zero. At the bottom of the swing, the 'G's will be maximum. If you used a fish-scale (spring-scale) to swing the whole thing, you could measure its static weight, then observe the apparent weight everywhere along the swing arc. If you don't have to go through 0 Gs, you can just swing the whole thing around horizontally and observe the fish-scale. The 'G' load is exactly the indicated weight / static weight.

As a relative measure, a 5.5 lb weight, swung through a 5ft diameter, (2.5 ft radius) arc at such a speed that it "just" falls through the top, exerts about 4 Gs (22 lb) at the bottom. Don't accidentally hit something. Impact Gs may make your experiment expensive.

Cheers,
Dick Johnson
Penguin : Linux version 2.6.12 on an i686 machine (5537.79 BogoMips). Warning : 98.36% of all statistics are fiction.
.
I apologize for the following. I tried to kill it with the above dot :

****************************************************************
The information transmitted in this message is confidential and may be privileged. Any review, retransmission, dissemination, or other use of this information by persons or entities other than the intended recipient is prohibited. If you are not the intended recipient, please notify Analogic Corporation immediately - by replying to this message or by sending an email to [email protected] - and destroy all copies of this information, including any attachments, without reading or disclosing them.

Thank you.

#9 lore raw
Quoted 16 lines · linux-os (Dick Johnson)
Quoted 16 lines
Quoted 6 lines · Jan Engelhardt
Quoted 4 lines

So in order to calibrate it you need a readily available source of constant acceleration, preferably with a known value.

Hint: -9.8 m/sec^2.

Drop it out of the window? :)

No, no. Constant gravity (like having the laptop sitting on the desk) "feels like" constant acceleration.

Dropping it out of the window should measure 0 m/sec^2, because the accelerometer is not working on an inertial referential (I hope this is the correct term in english...). For the accelerometer, this is just like the feeling of free falling inside an elevator: no gravity :)

Quoted 5 lines
Quoted 2 lines

--
Paulo Marques - www.grupopie.com

You need a centrifuge or something that works like one. You can make one and you can calibrate it using simple techniques.

Not at all. It's enough to let the laptop lie on the table for [0,0]G calibration, then put sequentially it on all the four sides for [-1,0]G, [1,0]G, [0,1]G, [0,-1]G calibration.

Quoted 1 line

From these five measurements you have both the zero point and the

slopes, including a good error estimate.

I've done that before when toying with IMUs.

Signature · 2 lines
Vojtech Pavlik
SuSE Labs, SuSE CR
#10 lore raw
Quoted 21 lines · Vojtech Pavlik
Quoted 19 lines · linux-os (Dick Johnson)
Quoted 18 lines
Quoted 6 lines · Jan Engelhardt
Quoted 4 lines

So in order to calibrate it you need a readily available source of constant acceleration, preferably with a known value.

Hint: -9.8 m/sec^2.

Drop it out of the window? :)

No, no. Constant gravity (like having the laptop sitting on the desk) "feels like" constant acceleration.

Dropping it out of the window should measure 0 m/sec^2, because the accelerometer is not working on an inertial referential (I hope this is the correct term in english...). For the accelerometer, this is just like the feeling of free falling inside an elevator: no gravity :)

--
Paulo Marques - www.grupopie.com

Quoted 7 lines
Quoted 2 lines

You need a centrifuge or something that works like one. You can make one and you can calibrate it using simple techniques.

Not at all. It's enough to let the laptop lie on the table for [0,0]G calibration, then put sequentially it on all the four sides for [-1,0]G, [1,0]G, [0,1]G, [0,-1]G calibration.

Huh? A laptop on the table is subjected to 1G on earth. If turned over, is still subjected to 1G, although the sensor may show -1G. Unless the correct k (calibration factor) is known, for each load direction, it cannot be derived from +/- 1, the only readings you have, because of the 1G bias of the load-cell itself.
A pair of load-cells mounted orthagonally, will show the accelleration in 4 axis. However, you can't assume that if you held the mass at a 45 degree angle, for instance, that both sensors would read 0.707 (cos 45) or that if at 90, one pair would read 0.0

You need to calibrate using real values. With a sping-scale, and some room to swing the device, you can readily obtain some accurate load values.

Quoted 9 lines
Quoted 1 line

From these five measurements you have both the zero point and the

slopes, including a good error estimate.

I've done that before when toying with IMUs.

--
Vojtech Pavlik
SuSE Labs, SuSE CR

Cheers,
Dick Johnson
Penguin : Linux version 2.6.12 on an i686 machine (5537.79 BogoMips). Warning : 98.36% of all statistics are fiction.
.
I apologize for the following. I tried to kill it with the above dot :

****************************************************************
The information transmitted in this message is confidential and may be privileged. Any review, retransmission, dissemination, or other use of this information by persons or entities other than the intended recipient is prohibited. If you are not the intended recipient, please notify Analogic Corporation immediately - by replying to this message or by sending an email to [email protected] - and destroy all copies of this information, including any attachments, without reading or disclosing them.

Thank you.

#11 lore raw
Quoted 5 lines · linux-os (Dick Johnson)

Huh? A laptop on the table is subjected to 1G on earth. If turned over, is still subjected to 1G, although the sensor may show -1G. Unless the correct k (calibration factor) is known, for each load direction, it cannot be derived from +/- 1, the only readings you have, because of the 1G bias of the load-cell itself.

The MEMS accelerometer measures the projection of the acceleration/gravity vector in two orthogonal directions.

If you place the notebook on a horizontal table, the projection of gravity to both the directions measured will be zero, because the gravity vector is orthogonal to both of the directions.

Thus for each direction you get the value at -1, 0, and +1 G, and that's perfectly enough to get a calibration, as long as the accelerometer is linear, which MEMSes to a reasonable degree are.

Quoted 4 lines

A pair of load-cells mounted orthagonally, will show the accelleration in 4 axis. However, you can't assume that if you held the mass at a 45 degree angle, for instance, that both sensors would read 0.707 (cos 45) or that if at 90, one pair would read 0.0

Well, accelerometers are not load cells with a weight mounted to them, that's the thing.

Quoted 3 lines

You need to calibrate using real values. With a sping-scale, and some room to swing the device, you can readily obtain some accurate load values.

The Earth's gravity is a very well known real value.

Signature · 2 lines
Vojtech Pavlik
SuSE Labs, SuSE CR
#12 lore raw
In reply to Alejandro Bonilla Beeche
Quoted 8 lines
Quoted 8 lines

Thanks to development of the driver from some nice guys, we are able to get data from the accelerometer. There is one problem. How do we calibrate the values that are outputed from the userspace? All PC's get a different value, and we can't really find the best solution. What is the scientific and smartest way to do this?

I'm not convinced we need to get so hung up on the calibration. Sure, each laptop has somewhat different resting values - but surely what we're looking for is any rate of change in either the X or Y values thats over a predefined 'safe' threshold? (I would imagine that we're only going to find that safe threshold from some imaginative testing once we've got the head parking sorted....)

Just my 2p worth,
Jon.

______________________________________________________________
Email via Mailtraq4Free from Enstar (www.mailtraqdirect.co.uk)

#13 lore raw
Quoted 18 lines
Quoted 1 line
Quoted 1 line

Thanks to development of the driver from some nice guys, we

are able to

Quoted 2 lines
Quoted 2 lines

get data from the accelerometer. There is one problem. How do we calibrate the values that are outputed from the userspace?

All PC's get

Quoted 1 line
Quoted 1 line

a different value, and we can't really find the best

solution. What is

Quoted 3 lines
Quoted 3 lines

the scientific and smartest way to do this?

I'm not convinced we need to get so hung up on the calibration. Sure, each laptop has somewhat different resting values - but surely what we're looking for is any rate of change in either the X or Y values thats over a predefined 'safe' threshold? (I would imagine that we're only going to find that safe threshold from some imaginative testing once we've got the head parking sorted....)

Quoted 2 lines

Just my 2p worth,
Jon.

hard drive parking was already sorted out. We have a script that does this and works great parking the heads.

The problem here is that we have 10 different models.

One will have 20 as X and the others will have 500 as x. Some will increment in 20 when you move them 45Deg, and some will increment 50.

How can you determine from an shake, to a fall?

.Alejandro

#14 lore raw
Quoted 3 lines

hard drive parking was already sorted out. We have a script that does this and works great parking the heads.

Would have to disagree with you here. Jens' ATA7 parking code is a great start but we still have a couple of issues to address before it's usable for this purpose. (1) it only works on a subset of the devices that the Windows driver works for, and (2) we need a way to freeze the device for a short period so that the next I/O request doesn't wake it up before it hits the floor...

Quoted 6 lines

The problem here is that we have 10 different models.

One will have 20 as X and the others will have 500 as x. Some will increment in 20 when you move them 45Deg, and some will increment 50.

How can you determine from an shake, to a fall?

If there really are devices with resting values of 20 then I would agree we have a problem. I've tried on a few different Thinkpads and have always seen resting values of several hundreds. If this is true for all, I would like to think we could estimate a sensible value that works for everyone...

Regards,
Jon.

______________________________________________________________
Email via Mailtraq4Free from Enstar (www.mailtraqdirect.co.uk)