About change the DM8127 stream resolution.

Hi Gomo,

Some questions:

1) What is the frame rate you want to capture with? 5M @ how many fps?

2) Will you be using the resizer / scalar operations to reduce resolution?

3) What resolutions will be used for encode? Please note we do not support 5M for H.264 in our current software framework.

The capture drivers can be used for implementing 5M capture but the frame-rate and rest of the system features will need some analysis.

Regards

Rajat

Hello,

The current set of ISS capture driver and codec can support only upto 1080p resolution.

However, in furture version we will support encoding and capture upto 8M (hardware does not have any limitation)

This will be available end of April

Regards,

Raghu

Thanks your reply at first.

Rajat Sagar said:

Hi Gomo,

Some questions:

1) What is the frame rate you want to capture with? 5M @ how many fps?

2) Will you be using the resizer / scalar operations to reduce resolution?

3) What resolutions will be used for encode? Please note we do not support 5M for H.264 in our current software framework.

The capture drivers can be used for implementing 5M capture but the frame-rate and rest of the system features will need some analysis.

Regards

Rajat

1) IPNC seneor full frame(15fps) is enough to us.

2)Using RszA to reduce the resolution. Our design data flow (using DM8127 IPNC to modify ) is: Sensor(3664 x 2748, 15 fps) -->FPGA(3664 x 2748, 15 fps)-->RSZA(3432x2048, 15fps) -->Encode(H.264 or MJPEG)-->Live555(RTSP)-->PC client;

3)   About 5M H.264, TI empolyee  Venugopala have answered that DM8127 can encode 5M H.264 , see the issue http://e2e.ti.com/support/dsp/davinci_digital_media_processors/f/716/t/164479.aspx, he said: "One in the User Guide is correct. Codec can encode 4Kx4K. We would correct the documentation."

Please give us more advices about the 5M capture. Thanks again.

Hi Raghu, Thanks your reply.

Raghu Kudva said:

Hello,

The current set of ISS capture driver and codec can support only upto 1080p resolution.

However, in furture version we will support encoding and capture upto 8M (hardware does not have any limitation)

This will be available end of April

Regards,

Raghu

I know the limitation of current ISS capture driver and codec.  The purpose of post this issue are:

1) Make sure the feasibility of capture 5M stream;

2)Get more advices to help me to modify by myself. The time , at the end of April, it's too late to me.

BTW, Upto 8M capture and encode, do you means to updata DM8127 IPNC software to new version which support more big resolution?

Thanks  a lot.

Hi Gomo,

5M capture is feasable from the hardware. But currently the capture driver is not tested for this resolution.So if you go about implementing this - you would need to change the capture driver yourself. 5M @ 15fps seems is feasable from ISP performance and DDR load point of view.

Q1) Do you want to do MJPEG encode or H.264 encode? Please note that MJPEG will work as Venu has pointed out - but the H.264 encode link will require update to support megapixel resolutions.

Q2) Will the FPGA output be YUV format or RAW format?

If its RAW format you can leverage the current driver flow to implement the capture - though you will need to change the input and output resolution settings.

If it is YUV then the driver needs more changes - I can talk more about it if this is the case.

Regards

Rajat

hi Rajat,

       Thanks for your information 5M capture.

Rajat Sagar said:

Q1) Do you want to do MJPEG encode or H.264 encode? Please note that MJPEG will work as Venu has pointed out - but the H.264 encode link will require update to support megapixel resolutions.

I want do MJPEG and H.264 respectively which it can switch through web. I had read the document of HDVICP2 MJPEG and H.264 datasheet, the all can encode 4Kx4K resolution. You said "H.264 encode link will require update", is it the encLink_h.264.c? Can you give more detail information?

Rajat Sagar said:

Q2) Will the FPGA output be YUV format or RAW format?

If its RAW format you can leverage the current driver flow to implement the capture - though you will need to change the input and output resolution settings.

If it is YUV then the driver needs more changes - I can talk more about it if this is the case.

The first step, using DM8127 IPNC to realize our function: 5M H.264 and MJPEG encode output; And then to build our owner camera with CCD sensor.

Raw format data is enough to our first step. As for the YUV capture,  if the FPGA have no limitation, I am very interest with it.

Thanks a lot.

Hi Gomo,

The codec support for H.264 & MJPEG is available - however we have not tested the mcfw framework with greater than 1080p resolutions. Thus, minor code change may be required for implementing 5M encode.

I would agree that for 1st pass you go with RAW output from FPGA, this way you can leverage the current drivers without change i.e. do on-the-fly RAW to YUV processing from the ISP and generate the required streams.

May I ask what do you use the FPGA for? Is it generating the data or for your custom processing requirements?

Regards

Rajat

Please mark this post as answered via the Verify Answer button below if you think it answers your question.  Thanks!

Hi Rajat,

  Thanks foryour reply.

Rajat Sagar said:

May I ask what do you use the FPGA for? Is it generating the data or for your custom processing requirements?

Regards

Rajat

We use FPGA to convert serial data to parallel data, do some statistic, reduce noise and do edge processing etc.