CSC Function on ADV7610

Hi,

  Please find the below comments,

1. Kindly let us know, Are you converting from RGB color space to YPbPr OUT then the register setting from "address 0x02 -> 0xF2" to "address 0x02 -> 0xF0" rare correct and also crosscheck your configuration with reference one some of the recommended settings are missing.

2. Regarding 24bit SDR422 mode, Please refer Pixel pin assignment and there we can find the details of how the pixel port pins are assigned for YPbPr output of 24bit SDR422 Mode and also we came to know the pins that need to be not connected.

Thanks,

Poornima

Hi,

  Please find the below comments,

1. Kindly let us know, Are you converting from RGB color space to YPbPr OUT then the register setting from "address 0x02 -> 0xF2" to "address 0x02 -> 0xF0" rare correct and also crosscheck your configuration with reference one some of the recommended settings are missing.

2. Regarding 24bit SDR422 mode, Please refer Pixel pin assignment and there we can find the details of how the pixel port pins are assigned for YPbPr output of 24bit SDR422 Mode and also we came to know the pins that need to be not connected.

Thanks,

Poornima

Item 1 : following settings are all of the registers they use currently. As you commented "some of the recommended settings are missing", does it still have any missing? Please add the registers in below list if there're any required register.

Item 2 : The customer is prepareing the additional comments. Please wait for the updates.

Adding my comments for the details of Item 2. Could you clarify following each question?

[ Case of "YPbPr input YPbPr output" ]

As for the red letter part in below image, lower 4 bit may be 0 padding on both Y PbPr, Is it correct?

[ Case of "RGB input YPbPr output" ]

As for the red letter part in below image, In the CSC calculation is the bit accuracy inside far more than12bit? And is output bit width is it rounded or truncated to 8bit according to the input 8bit? or fully 12bit output for keeping calculation bit accuracy?

Please also keep comment for Item1

Hi,

  Please let us know, why we need to go for lower 4 bit zero Padding/Truncating Since ADV7610 have an option to support 24 bit output mode ?

   If we have the respective output bit support, then why should we Rounding / Truncating the output bits.

Thanks,

Poornima

Hi,

This is a question of CSC calculation accuracy, not for pin assignment. As I mentioned above, they're using 24bit SDR 422 mode for the pin assign. Since the format is 422, they're asking for the CSC calculation spec when it working in each case such as Rounding / Truncating with the outputs of lower bits. It's needed for defining the customer product specification.

Please let us know the answer for Item1 and 2.

Hi,

 We don't have any CSC calculation spec other than what provided in the manuals so we need to check with expert about how Rounding / Truncating can affect the CSC calculation.

 CSC calculation been little elaborately provided in ADV7401 Manual. Please refer section 6.5.1 of the ADV7403 Datasheet Manual (ADV7403_Manual_RevB.pdf) that provides more details about CSC coefficient calculation, which is available at ADV7403 Design Support Files.

 Also Please refer this thread understanding ADV7513 CSC behavior - Q&A - Video - EngineerZone (analog.com)

Thanks,

Poornima

Hi,

I understand the comments for item2. How about the question of Item1? Do you have any comments whether it's enough with the uploaded list?

Item 1 : following settings are all of the registers they use currently. As you commented "some of the recommended settings are missing", does it still have any missing? Please add the registers in below list if there're any required register.

Hi,

 Regarding Item 1,  We have script for YPrPb input to YPrPb output So Please follow our exact script register configuration.

 Difference from reference one been highlighted in yellow color.

Please note that, If those ADI Recommended writes are not included, the part would not work reliably (unless those values are default anyway).Those registers are generally internal settings that let us handle process variation which is why they are determined during characterization. The settings that show up as "ADI Recommended Settings" are the values that work reliably across voltage and temperature.

Thanks,

Poornima

Sorry for jumping in, I also have similar questions on ADV7610, and I have several following questions.

Item3 

There are 2 points for CSC function, CP and Backend ColorSpace.

There are many settings for CP part CSC, but I can't find setting for Backend ColorSpace Conversion part.
Is there a setting for Backend ColorSpace Conversion part?

And also I'd like to know detail function on Backend ColorSpace.
What is the purpose of this Backend ColorSpace conversion and why it is located here?

Item4 (similar to Item1)

As for the "ADI recommended" setting,

I found the same setting on ADV7611, and there are comment of
"for non-fast switching application".

What is this "non-fast switching" ? A frequent format change?
I can't link this "for non-fast switching application" to what your mentioned
"the values that work reliably across voltage and temperature". Is the situation of ADV7610 different from that of ADV7611?

And also I'd like to know the "risk" that these "ADI recommended settings" are not set.
What is the risk to be considered?

Item5

As for the IO address 0x02h, the setting of 0xF5h means the Auto CSC and YPbPr out as the output of CP block.
This "Auto CSC" means that checking the input AVI info and by the combination of input color space detected by the AVI info
and RGB_OUT setting, color space conversion is determined, is it correct?

If so by this 0x98h, 0x02h 0xF5h setting both HDMI YPbPr input and RGB input can be converted YPbPr output as ADV7610 colorspace output automatically, am I right?

Item6

I use the setting of IO map address 0xBF and set 0x01, it means bypass CP core.

If this setting is done and if the "24bit SDR ITU-656 422 Mode2 " is chosen as the output assignment

and if HDMI input is RGB, what would happen?

Hi Katsuya San,

        Please find the below comments,

What is the purpose of this Backend ColoSpace conversion and why it is located here?

    "Back End CSC" is nothing, but the CSC in the CP and it is not an extra one.

Sorry for my delayed reply.

I actually checked Auto CSC function using ADV7610.

Input is RGB 444 (limitedb range) 1920x1080 59.94p 75% color bar

and expected output is YCbCr 422, BT709 color bar.

The setting of address 0x02h of IO block is 0xFDh, and I found at the color change, for example, 

cyan to green position, Cb/Cr data has slight overshoot and undershoot.

And when the setting of 0xE0h of IO block is 0x80h (DS_WITHOUT_FILTER to be no filtering)

this overshiit/undershoot disappears.

But according to the downsampling theory, filtering is needed.

Is it possible to know the chraracteristics of this downsampling filter?

I'd like to know this overshoot/undershoot is inevitable or not.