Skip to contentSkip to conversation

sharpsl_pm: using milivolts instead of custom units?

DISCUSSION Pavel Machek 2 participants 3 messages Nov 11, 2005

lore
Share & tools
lore
#0 lore raw

Hi!

It seems to me that sharpsl.c is using some very custom units:

#define SHARPSL_CHARGE_ON_VOLT         0x99  /* 2.9V */
#define SHARPSL_CHARGE_ON_TEMP         0xe0  /* 2.9V */
#define SHARPSL_CHARGE_ON_ACIN_HIGH   0x9b  /* 6V */
#define SHARPSL_CHARGE_ON_ACIN_LOW    0x34  /* 2V */
#define SHARPSL_FATAL_ACIN_VOLT        182   /* 3.45V */
#define SHARPSL_FATAL_NOACIN_VOLT      170   /* 3.40V */

...what are they? Unfortunately collie uses very different units:

#ifdef CONFIG_SA1100_COLLIE

struct battery_thresh spitz_battery_levels_noac[] = {
        { 368, 100},
        { 358,  25},
        { 356,   5},
        {   0,   0},

...

...so it could get very confusing. Would it be feasible to convert to mV as soon as possible and have all constants in milivolts? I realize they may be slightly different for different models, but they should at least be comparable.
Pavel

Signature · 1 line
Thanks, Sharp!
#1 lore raw
Quoted 4 lines · Pavel Machek

Hi!

It seems to me that sharpsl.c is using some very custom units:

Quoted 8 lines
#define SHARPSL_CHARGE_ON_VOLT         0x99  /* 2.9V */
#define SHARPSL_CHARGE_ON_TEMP         0xe0  /* 2.9V */
#define SHARPSL_CHARGE_ON_ACIN_HIGH   0x9b  /* 6V */
#define SHARPSL_CHARGE_ON_ACIN_LOW    0x34  /* 2V */
#define SHARPSL_FATAL_ACIN_VOLT        182   /* 3.45V */
#define SHARPSL_FATAL_NOACIN_VOLT      170   /* 3.40V */

...what are they? Unfortunately collie uses very different units:

They're raw values as provided by the ADC. Collie will have a different ADC and hence different raw values. There's no real need to convert them to mV and back again all the time.

Quoted 6 lines

#ifdef CONFIG_SA1100_COLLIE

struct battery_thresh spitz_battery_levels_noac[] = {
        { 368, 100},
        { 358,  25},
        { 356,   5},
        {   0,   0},
Quoted 6 lines

...

...so it could get very confusing. Would it be feasible to convert to mV as soon as possible and have all constants in milivolts? I realize they may be slightly different for different models, but they should at least be comparable.

The battery levels are totally different between the models and each is going to need a levels translation table as its an extremely none linear decay curve. Using ADC values here makes a lot of sense as that's as granular as the information ever gets and you don't start to lose accuracy anywhere with rounding.

I still think you've much bigger problems to worry about as this code is heavily PXA biased and is going to need a lot of changes to work on the SA1100. Its why I've put it in mach-pxa rather than common and a separate driver for collie based on this code might be easier.

Richard

#2 lore raw

Hi!

Quoted 13 lines
Quoted 6 lines

#ifdef CONFIG_SA1100_COLLIE

struct battery_thresh spitz_battery_levels_noac[] = {
        { 368, 100},
        { 358,  25},
        { 356,   5},
        {   0,   0},
Quoted 6 lines

...

...so it could get very confusing. Would it be feasible to convert to mV as soon as possible and have all constants in milivolts? I realize they may be slightly different for different models, but they should at least be comparable.

Quoted 5 lines

The battery levels are totally different between the models and each is going to need a levels translation table as its an extremely none linear decay curve. Using ADC values here makes a lot of sense as that's as granular as the information ever gets and you don't start to lose accuracy anywhere with rounding.

Well, all the models have li-ion batteries, so levels (in milivolts) for stuff like "battery empty" should be very similar. If conversion from ADC to voltage is non-trivial, that's a show stopper I guess.

Quoted 4 lines

I still think you've much bigger problems to worry about as this code is heavily PXA biased and is going to need a lot of changes to work on the SA1100. Its why I've put it in mach-pxa rather than common and a separate driver for collie based on this code might be easier.

Actually I think I can refactor in a way that a common core can be share. I'd hate to copy 1000 lines...
Pavel

Signature · 1 line
Thanks, Sharp!