PWM fast stop operation

[Petro Karashchenko <pe...@gmail.com>]

Hello team,

Recently I had a project where I had to control a motor using PWM. In
general things worked pretty well, but I've met an issue when I need to
stop PWM as soon as possible. My custom board is based on SAMv7 and uses
the standard NuttX PWM driver. The issue with the SAMv7 PWM peripheral is
that when a PWM stop is issued the PWM generation is not stopped
immediately, but waits at the end of the current cycle. I'm not sure if
that is unique to SAMv7, so maybe you can give feedback about the other
platforms. Initially I tried to replace stop by keeping PWM running and
updating config to have 0% duty cycle, but that still waits till the end of
the current cycle to apply.
The SAMv7 PWM has a signal override capabilities, so I used that instead of
stopping the PWM. I've extended the "struct pwm_info_s" structure and
included an override field there, but feel that it might not be the best
approach.

I would appreciate your advices on what is the best way to integrate output
override feature into existing PWM driver.

Best regards,
Petro

Re: PWM fast stop operation

["Alan C. Assis" <ac...@gmail.com>]

Hi Petro,

Maybe instead of creating an override you can create a capability like
exists for SDIO on NuttX.

It could be used to control features/characteristics unique for some
PWM controllers.

Buf of course it will need some override/sync flag because some
applications will prefer to wait up to the next cycle.

BR,

Alan

On 11/12/22, Petro Karashchenko <pe...@gmail.com> wrote:
> Hello David,
>
> Thanks for your reply. I fully understand that. In my case I have the half
> bridge schema of four pins and PWM frequency is very low. In case I need to
> do immediate stop I need to stop pulse generation and short pin to the
> ground. It is more because the overall half bridge solution is controlled
> by 2 pins with PWM and other 2 pins are controlled as GPIOs. There where
> the sync problems come in place.
>
> Best regards,
> Petro
>
> On Sat, Nov 12, 2022, 9:34 AM David Sidrane <da...@nscdg.com>
> wrote:
>
>> Hi Petro,
>>
>> On the STM32 there are shadow registers that buffer the updated data to
>> ensure that outputs are updated synchronously with the end of the period.
>> There is a bit that can be set to force the update to the shadowed
>> registers
>> to be immediate, which will change the count of the current period in
>> that
>> period. This can be dangerous; in that it could extend a pulse and create
>> an
>> undesired position or speed. In the case of going to 0, this could create
>> a
>> runt pulse and also create an undesired position or a slow speed (not so
>> bad), Because the update is asynchronous the pulse with the output, it is
>> not deterministic. So synchronous updates are a good thing most times.
>>
>> David
>>
>>
>> -----Original Message-----
>> From: Petro Karashchenko <pe...@gmail.com>
>> Sent: Friday, November 11, 2022 6:50 PM
>> To: dev@nuttx.apache.org
>> Subject: PWM fast stop operation
>>
>> Hello team,
>>
>> Recently I had a project where I had to control a motor using PWM. In
>> general things worked pretty well, but I've met an issue when I need to
>> stop
>> PWM as soon as possible. My custom board is based on SAMv7 and uses the
>> standard NuttX PWM driver. The issue with the SAMv7 PWM peripheral is
>> that
>> when a PWM stop is issued the PWM generation is not stopped immediately,
>> but
>> waits at the end of the current cycle. I'm not sure if that is unique to
>> SAMv7, so maybe you can give feedback about the other platforms.
>> Initially
>> I
>> tried to replace stop by keeping PWM running and updating config to have
>> 0%
>> duty cycle, but that still waits till the end of the current cycle to
>> apply.
>> The SAMv7 PWM has a signal override capabilities, so I used that instead
>> of
>> stopping the PWM. I've extended the "struct pwm_info_s" structure and
>> included an override field there, but feel that it might not be the best
>> approach.
>>
>> I would appreciate your advices on what is the best way to integrate
>> output
>> override feature into existing PWM driver.
>>
>> Best regards,
>> Petro
>>
>

Re: PWM fast stop operation

[Petro Karashchenko <pe...@gmail.com>]

Hello David,

Thanks for your reply. I fully understand that. In my case I have the half
bridge schema of four pins and PWM frequency is very low. In case I need to
do immediate stop I need to stop pulse generation and short pin to the
ground. It is more because the overall half bridge solution is controlled
by 2 pins with PWM and other 2 pins are controlled as GPIOs. There where
the sync problems come in place.

Best regards,
Petro

On Sat, Nov 12, 2022, 9:34 AM David Sidrane <da...@nscdg.com> wrote:

> Hi Petro,
>
> On the STM32 there are shadow registers that buffer the updated data to
> ensure that outputs are updated synchronously with the end of the period.
> There is a bit that can be set to force the update to the shadowed
> registers
> to be immediate, which will change the count of the current period in that
> period. This can be dangerous; in that it could extend a pulse and create
> an
> undesired position or speed. In the case of going to 0, this could create a
> runt pulse and also create an undesired position or a slow speed (not so
> bad), Because the update is asynchronous the pulse with the output, it is
> not deterministic. So synchronous updates are a good thing most times.
>
> David
>
>
> -----Original Message-----
> From: Petro Karashchenko <pe...@gmail.com>
> Sent: Friday, November 11, 2022 6:50 PM
> To: dev@nuttx.apache.org
> Subject: PWM fast stop operation
>
> Hello team,
>
> Recently I had a project where I had to control a motor using PWM. In
> general things worked pretty well, but I've met an issue when I need to
> stop
> PWM as soon as possible. My custom board is based on SAMv7 and uses the
> standard NuttX PWM driver. The issue with the SAMv7 PWM peripheral is that
> when a PWM stop is issued the PWM generation is not stopped immediately,
> but
> waits at the end of the current cycle. I'm not sure if that is unique to
> SAMv7, so maybe you can give feedback about the other platforms. Initially
> I
> tried to replace stop by keeping PWM running and updating config to have 0%
> duty cycle, but that still waits till the end of the current cycle to
> apply.
> The SAMv7 PWM has a signal override capabilities, so I used that instead of
> stopping the PWM. I've extended the "struct pwm_info_s" structure and
> included an override field there, but feel that it might not be the best
> approach.
>
> I would appreciate your advices on what is the best way to integrate output
> override feature into existing PWM driver.
>
> Best regards,
> Petro
>

RE: PWM fast stop operation

[David Sidrane <da...@nscdg.com>]

Hi Petro,

On the STM32 there are shadow registers that buffer the updated data to
ensure that outputs are updated synchronously with the end of the period.
There is a bit that can be set to force the update to the shadowed registers
to be immediate, which will change the count of the current period in that
period. This can be dangerous; in that it could extend a pulse and create an
undesired position or speed. In the case of going to 0, this could create a
runt pulse and also create an undesired position or a slow speed (not so
bad), Because the update is asynchronous the pulse with the output, it is
not deterministic. So synchronous updates are a good thing most times.

David


-----Original Message-----
From: Petro Karashchenko <pe...@gmail.com>
Sent: Friday, November 11, 2022 6:50 PM
To: dev@nuttx.apache.org
Subject: PWM fast stop operation

Hello team,

Recently I had a project where I had to control a motor using PWM. In
general things worked pretty well, but I've met an issue when I need to stop
PWM as soon as possible. My custom board is based on SAMv7 and uses the
standard NuttX PWM driver. The issue with the SAMv7 PWM peripheral is that
when a PWM stop is issued the PWM generation is not stopped immediately, but
waits at the end of the current cycle. I'm not sure if that is unique to
SAMv7, so maybe you can give feedback about the other platforms. Initially I
tried to replace stop by keeping PWM running and updating config to have 0%
duty cycle, but that still waits till the end of the current cycle to apply.
The SAMv7 PWM has a signal override capabilities, so I used that instead of
stopping the PWM. I've extended the "struct pwm_info_s" structure and
included an override field there, but feel that it might not be the best
approach.

I would appreciate your advices on what is the best way to integrate output
override feature into existing PWM driver.

Best regards,
Petro