RE: AT91 ADC support

[Sergei Gavrikov <sergei.gavrikov@gmail.com>]

On Thu, 27 May 2010, Christophe Coutand wrote:
> Hi Sergei,
>
> I have attached an updated var_io.h that includes ADC support for the
> AT91M55800A. I took the opportunity to define the number of channels
> per ADC in this file. It can be overwritten per platform basis since
> AT91SAM7L64 for instance has 4 channels only.

Hi Christophe,

Thanks for your time. I will take a look this morning.

> The definition of AT91_MAX_ADC_CHAN must be removed from
> devs/adc/arm/at91/current/src/adc_at91.c. I have not joined any patch
> for it since I will hopefully submit a new driver that support both
> ADCs.

Excellent! I did want to ask you about and it seems fetchamil is
retrieving it just now.

Sergei

> Christophe
>
> -----Original Message-----
> From: Sergei Gavrikov [mailto:sergei.gavrikov@gmail.com]
> Sent: 27. mai 2010 18:38
> To: Christophe Coutand
> Cc: John Dallaway; eCos Patches
> Subject: RE: AT91 ADC support
>
> On Thu, 27 May 2010, Christophe Coutand wrote:
>> Hi John,
>>
>> It's pretty easy to add the required definition for the AT91M55800A
>> targets. The only thing I see now is that this device contains 2 ADCs
>> which I have not considered before. I guess there are several ways out
>> of this:
>>
>> 1- Update the actual AT91 ADC driver to make full use of the
>> AT91M55800A targets. I guess should be done by loading a second ADC
>> instance (one for each ADC. I have not been through all the thinking
>> here...).
>>
>> 2- or limit the AT91 driver to use only ADC0 of the AT91M55800A target
>> for the time being.
>>
>> 3- or exclude AT91M55800A targets for the time being.
>>
>> IMO #1 is best but I cannot give any time frame for completing it.
>
> Hi Christophe, John
>
> I would prefer #1, if you want I can add that in AT91/ADC CDL/sources.
> One thing: I won't be able to test it on real hardware (only build
> process can be tested me).
>
> John, what do you think: Can we add the second ADC instance without a
> testing?
>
>> One additional weakness of the driver is that it is made for up to 8
>> channels. It is defined nowhere what the targeted CPU can actually
>> handle, this is left to the user when configuring eCos. I believe this
>> is pretty fine since the user must anyway know which signal he wants
> to
>> sample but you might disagree on that one.
>
> I know only one target which would use all 8 channels :-) That's eCos
> i386/Linux synthetic target.
>
> Thanks for collaboration.
>
> Sergei
>
>> Regards,
>> Christophe
>