# Re: [PATCH] staging: iio: ad5933: fix handling of settling time cycles

From: Luis de Bethencourt
Date: Wed Jun 01 2016 - 14:15:42 EST

On 01/06/16 17:22, Lars-Peter Clausen wrote:
> On 06/01/2016 05:55 PM, Luis de Bethencourt wrote:
>> Correctly handle the settling time cycles value. The else branch was an
>> impossible condition (> 1022 in the else branch of > 511) and the handling
>> of the values was dividing by 2 and 4, with a left shift, instead of
>> multiplying.
>>
>> Based on the Table 13 at the bottom of Page 25 of the Data Sheet:
>>
>> Signed-off-by: Luis de Bethencourt <luisbg@xxxxxxxxxxxxxxx>
>> ---
>>
>> Hi,
>>
>> I decided to use the hexadecimal values instead of (1 << 10) and (1 << 9), for
>> briefness, I could resend using those instead if it is prefered.
>>
>> I also decided to use multiplications instead of right-shifts for readability.
>> I could use change that as well.
>>
>> Thanks,
>> Luis
>>
>> 1 file changed, 7 insertions(+), 4 deletions(-)
>>
>> index 9f43976..3a2cf8f3 100644
>> @@ -444,10 +444,13 @@ static ssize_t ad5933_store(struct device *dev,
>> st->settling_cycles = val;
>>
>> /* 2x, 4x handling, see datasheet */
>> - if (val > 511)
>> - val = (val >> 1) | (1 << 9);
>> - else if (val > 1022)
>> - val = (val >> 2) | (3 << 9);
>> + if (val & 0x400 && val & 0x200) {
>> + val &= 0x1ff;
>> + val *= 4;
>> + } else if (val & 0x200) {
>> + val &= 0x1ff;
>> + val *= 2;
>> + }
>
> This does not look correct. D10 and D9 select an additional multiplier of
> either 1, 2 or 4. So dividing the value before writing it to the register is
> the right approach in that case. Just flipping the order in which the
> conditions are evaluated should be sufficient.
>
>>
>> dat = cpu_to_be16(val);
>>
>

I misunderstood the register being read instead of being written. Looking at it
now, I have no idea why. Sorry.

Will resend a patch flipping the order of conditions.

Thanks for the review,
Luis