OMAP3 Multiple camera support

OMAP3 Multiple camera support

[Sriram V]

Hi

Does OMAP3 camera driver support multiple cameras at the same time.

As i understand - You can have simultaneous YUV422 (Directly to memory)
and another one passing through camera controller ISP?

I Also, wanted to check if anyone has tried having multiple cameras on omap3
with the existing driver.

-- 
Regards,
Sriram

Re: OMAP3 Multiple camera support

[Laurent Pinchart]

Hi Sriram,

On Wednesday 17 September 2014 23:06:42 Sriram V wrote:
> Hi
> 
> Does OMAP3 camera driver support multiple cameras at the same time.
> 
> As i understand - You can have simultaneous YUV422 (Directly to memory)
> and another one passing through camera controller ISP?
> 
> I Also, wanted to check if anyone has tried having multiple cameras on omap3
> with the existing driver.

The driver does support capturing from multiple cameras at the same time, 
provided one of them is connected to the CSI2A receiver. You can then capture 
raw frames from the CSI2A receiver output while processing frames from the 
other camera (connected to CSI1/CCP2, CSI2C or parallel interface) using the 
whole ISP pipeline.

Please note that the consumer OMAP3 variants are documented by TI as not 
including the CSI receivers. However, several developers have reported that 
the receivers are present and usable at least in some of the chips.

-- 
Regards,

Laurent Pinchart

Re: OMAP3 Multiple camera support

[Sriram V]

Hi Laurent,

What you are essentially saying is

CSI2A => Smart Sensor (YUV422)
CSI2C => Sensor that gives out RAW Bayer.

I guess this is a driver limitation? Am i correct?

Also, Can i have something like this?

SMART Sensor => CSI2A => H3A => MEM (Can i have this)
CSI2C => ISP => H3A => MEM

Can't i have H3A for both the pipelines?
Or

Can i enable H3A on the fly for both the sensors? One After the other?

Regards,
Sriram


On Thu, Sep 18, 2014 at 1:55 PM, Laurent Pinchart
<laurent.pinchart@ideasonboard.com> wrote:
> Hi Sriram,
>
> On Wednesday 17 September 2014 23:06:42 Sriram V wrote:
>> Hi
>>
>> Does OMAP3 camera driver support multiple cameras at the same time.
>>
>> As i understand - You can have simultaneous YUV422 (Directly to memory)
>> and another one passing through camera controller ISP?
>>
>> I Also, wanted to check if anyone has tried having multiple cameras on omap3
>> with the existing driver.
>
> The driver does support capturing from multiple cameras at the same time,
> provided one of them is connected to the CSI2A receiver. You can then capture
> raw frames from the CSI2A receiver output while processing frames from the
> other camera (connected to CSI1/CCP2, CSI2C or parallel interface) using the
> whole ISP pipeline.
>
> Please note that the consumer OMAP3 variants are documented by TI as not
> including the CSI receivers. However, several developers have reported that
> the receivers are present and usable at least in some of the chips.
>
> --
> Regards,
>
> Laurent Pinchart
>



-- 
Regards,
Sriram

Re: OMAP3 Multiple camera support

[Laurent Pinchart]

Hi Sriram,

On Thursday 18 September 2014 15:10:13 Sriram V wrote:
> Hi Laurent,
> 
> What you are essentially saying is
> 
> CSI2A => Smart Sensor (YUV422)
> CSI2C => Sensor that gives out RAW Bayer.

Not quite. You can have the following pipelines running in parallel.

Sensor A -> CSI2A -> memory
Sensor B -> (CSI1/CCP2/CSI2C ->) CCDC -> preview -> resizer -> memory

The format captured at the CSI2A output will be set by the sensor. If you use 
a smart sensor that outputs YUV, you will capture YUV images. If you use a 
Bayer sensor, you will capture raw Bayer images. Which sensor to use will 
depend on your use case.

The second pipeline can use the whole ISP. I've showed the full pipeline, but 
you can capture frames at the output of the CCDC, preview engine or resizer, 
depending on your use case.

The H3A blocks are connected to the CCDC output, so they're only available to 
the second pipeline.

Another option would be to run three pipelines.

Sensor A -> CSI2A -> memory
Sensor B -> CSI2C -> memory
memory -> CSI1/CCP2 -> CCDC -> preview -> resizer -> memory

(The second pipeline isn't supported by the driver now, but that's a software 
limitation only, it shouldn't be difficult to fix)

The third pipeline would then be multiplexed between the two sensors, your 
userspace application would push frames captured from sensors A and B 
alternatively through the memory-to-memory processing pipeline.

This will consume more memory bandwidth than the first option, whether it 
would work depends on the frame sizes and rates. Another disadvantage of this 
solution is that reconfiguring the processing pipeline for each frame will 
take time and thus lower the maximum possible frame rate (I don't have numbers 
though).

On the upside you will be able to use H3A for both sensors.

> I guess this is a driver limitation? Am i correct?
> 
> Also, Can i have something like this?
> 
> SMART Sensor => CSI2A => H3A => MEM (Can i have this)
> CSI2C => ISP => H3A => MEM
> 
> Can't i have H3A for both the pipelines?
> Or
> 
> Can i enable H3A on the fly for both the sensors? One After the other?

-- 
Regards,

Laurent Pinchart