For CSI bandwidth > 1.5 Gbps, the skew calibration should be enabled. The L4T documentation mentions:
An initiation deskew signal should be sent by sensor or deserializer to perform the skew calibration. If the deskew signals is not sent, the receiver will stall, and the capture will time out.
This is indeed matching the experiments I did.
The chipset I’m using can additionally send a skew calibration sequence at the end of every frame. Is the Orin Nano able to make use of these calibration sequences, or only the initialization sequence is used?
deskew calibration is a must if data-rate > 1.5 Gbps, else the camera firmware will continue to wait for deskew signal from the sensor side. it’ll enable pixel parser when deskew calibration has completed.
please see-also Topic 268833 to apply the bug fixes regarding to deskew calibration.
Thanks for your fast reply. I will try to deploy the FW linked on the other post to experiment.
Maybe my initial post was not fully clear:
My current setup uses 1.6 Gbps/lane
The chip (a V3Link deserializer) is sending the skew calibration sequence prior to the video data (validated with the scope)
The Orin Nano dts (pixel clock value) is configured so that the skew calibration is required; I also checked this point: if the deserializer chip is not sending the skew calibration sequence, the acquisition is stalled. With it it works. So, exactly as expected.
With the current configuration, everything works fine: the Jetson needs the skew calibration sequence, the deserializer sends it, the Jetson receives it and the acquisition runs fine.
My question was about extra skew calibration sequences that the chip I’m using over CSI can send (after each “end of frame” packet). The Jetson seems to be fine receiving these extra sequences.
But I was wondering if periodic skew calibration sequences are of any help for the Orin Nano (or could cause issues), or if only the initial skew calibration sequence is used. I assume this is part of the RCE firmware.
Could you please share some hints about how the skew calibration is implemented on Jetson side?
To answer your question, the deserializer chipset I’m using is a TI TDES954. Here are some quotes from the datasheet:
The initial skew calibration (as required by the RCE) is enabled with this setting:
CSI_CAL_EN: Enable initial CSI Skew-Calibration sequence When the initial skew-calibration sequence is enabled, the CSI Transmitter will send the sequence at initialization, prior to sending any HS data. This bit should be set when operating at 1.6 Gbps CSI speed
The extra skew calibration sequences can be enabled though:
CSI_CAL_PERIODIC: Enable periodic CSI Skew-Calibration sequence When the periodic skew-calibration sequence is enabled, the CSI Transmitter will send the periodic skew-calibration sequence following the sending of Frame End packets.
Please find attached annotated traces showing the details about the skew calibration sequences: skew_calibration_details.pdf (31.0 KB)
RCE firmware it doesn’t support periodic CSI Skew-Calibration sequence at the moment.
anyways, is it camera use-case with data-rate higher than 2.5Gbps to enable CSI_CAL_PERIODIC?
Thanks for the confirmation.
The max speed supported by the deserializer is 1.6Gbbps/lane. There is no particular use case for the CSI_CAL_PERIODIC setting mentioned in the datasheet, and higher speeds are not possible.
I will go with the initialization sequence only, which is the mandatory one for bandwidth > 1.5 Gbps.
From my side, your replies answered my question. Thank you very much.