DS90UB960-Q1: About CSI

Dear David,

Thank you for your question and supporting image!

Yes, so the image that you have provided shows how a video frame can use two different ways (low D-PHY and high D-PHY ways) of transferring data using the CSI-2 video transfer protocols.

Now, an advantage of using CSI-2 is that it does NOT require transmission of discrete horizontal timing information allowing flexibility for video aggregation. This is especially useful when transmitting at high D-PHY lane speeds.

The image on the left that you have provided is how video, active region, is transferred using low D-PHY lane speeds (eg. 800 megabits/sec). The image on the right is sending the same video, active region, but using high D-PHY lane speeds(eg. 1500 megabits/sec). 

As reflected in the image, increasing the lane speeds while sending the same frame with the same frame rate will increase the vertical blanking time. However, it is important to note that the the transmitter is outputting it's lower power, or LP11 state during the blanking time so that no data bytes are being sent by the transmitter. During the blanking state, the transmitter is idle and does not consume any signaling bandwidth at that time.

Very Respectfully,

William Yi

Dear David,

Thank you for your questions!

1. Horizontal timing information would have been sent in the form of the short packet. The short packets use a 32 bit frame structure and are used for signaling frame start, frame end and synchronization information. It would have been sent between the long packet states separated by LPS, low power states.
   
   
2. Horizontal Blanking is not transferred in CSI-2, but there is line time used to calculate latency.
   Reference here: e2e.ti.com/.../4114344
   
   
3. No, the picture resolutions remains the same at 1920*720. Changing the transmission speed does not change the resolution. Resolution is dependent on how much bandwidth can be sent through. Refer to the last question in the video on how Vtotal lines affects bandwidth.
   
   
4. Selecting the D-PHY speed would depend on how much LP-11 time and latency your system will be able to handle. You may expect more LP11 time with higher D-PHY speeds.
   
   
5. Yes, D-PHY lane speed is useful because it directly affects the data rate that you are sending. 

If you would care to learn more about these protocols, I would recommend researching more about CSI-2 video data transfer protocols and standards. 

Very Respectfully,

William Yi

1. Vertical and Horizontal Blanking includes the vertical and horizontal width and porch information. Below shows what data types are stored in the Short packets. 

2. During the LP11 blanking states, these are low power states and would not be transferred as data or take bandwidth. They are still in the packet.

3. Yes, that is right. 

4. LP-11 blanking time is affected by the data rate. When you pass through a higher data rate, there will be more vertical blanking time. This would affect the latency due to the extra amount it would take. 

If you have more question on CSI-2. This is a helpful data sheet on CSI-2 that may help clear up any confusions. 

CSI-2 Datasheet

Dear David, 

Thank you for your follow up questions.

1. No, that is not correct. All horizontal timing information including width and porch information is sent with the long packets. The only difference between CSI-2 and other video protocols it that horizontal timing is not discrete. Meaning that it does not take bandwidth or data to send. It is still in the packet as NULL in the blanking state, which does not take bandwidth.

2. There is still horizontal blanking. However, since it does not take bandwidth or data, it is considered not to be transferred.

Very Respectfully,

William Yi

Dear David,

Good question. Here is another picture from the datasheet:

Once the receiver reads the data type is blanking data or NULL. It will ignore the packet payload and consider it to have no data. 

Very Respectfully,

William Yi

Dear David,

Thank you for your question!

It is a separate long packet that has a variable length. 

Respectfully,

William Yi

Hello David,

Thank you for your question!

1. Lets say that in your first example on the left, your data rate (D-PHY) is 800Mbps.

Data rate = PCLK * number of bits per pixel. (In standard RGB888 video this would be 24 bits. 8 bits each for red, green, and blue)

Inversely, your PCLK = Data Rate / number of bits per pixel = 800Mbps / 24bits =   33 Mhz.

Lets say your frame rate = 60 Hz

Then, your the total pixels that you are sending = 33Mhz / 60 Hz = 0.55 Mpixels per frame. 

2. Lets say that in your second example on the right, your data rate (D-PHY) is 1.6Gbps.

Data rate = PCLK * number of bits per pixel. (In standard RGB888 video this would be 24 bits. 8 bits each for red, green, and blue)

Inversely, your PCLK = Data Rate / number of bits per pixel = 1.6Gbps / 24bits =   66 Mhz.

Lets say your frame rate = 60 Hz

Then, your the total pixels that you are sending = 66Mhz / 60 Hz = 1.1 Mpixels per frame. 

Because you are increasing your data rate, and keeping your frame rate the same. The extra pixels per frame are added to your vertical blanking that you see in the second image. 

I hope this helps clear up any confusion!

Best Regards,

William Yi