Calibration Error (response -7)
Having difficulty calibrating the 435 inside the RealSense Viewer
Invalid Value in rs2_run_on_chip_calibration_cpp
hwmon command 0x80(8 3 0 /) railed (response -7)
- Multiple distances
- Different setups (hallway, on back looking at ceiling, office door)
- Adjusting to "High Accuracy"
- Setting 1280x720 (resolution)
I cannot seem to find the error code in the documentation anywhere.
What seems to be the problem?
Hi Stevew If you are using Windows, could you try downloading and running the 2.49.0 version of the standalone RealSense Viewer - Intel.RealSense.Viewer.exe - and activating On-Chip calibration in that Viewer to see whether it resolves your problem, please.
I am using 2.49 already.
I was hoping that there was some documentation that would explain the error (response -7)
As it appears an intentional error "code".
Playing with this more...it seems to be an issue with some state of software and/or hardware, as it can sometimes work, albeit without any obvious changes (which can be frustrating).
This error has been reported by a small number of RealSense users at the link below.
They were controlling on-chip calibration from program scripting instead of the Viewer and were informed by their 'response - 7' error message that it related to HW not ready. Most of the users who reported the problem on that discussion said that performing repeated retries of the calibration process enabled it to work.
Circling back to this.
Checking our deliveries, we're finding 435's that are arriving outside the acceptable margin of error.
I wanted to see what I could do to help our assembly team correct this error, so I wanted to do some calibration tests.
(I'm not a programmer or engineer)
Of the 3 types of calibration, the "On-Chip" and "Tare" would seem the appropriate "first try" when trying to improve calibration.
Because the camera doesn't have to be moved, we presumably could test and calibrate in the same jig (set for 2M calibration tests)
I continue to get this error (reported in this ticket) --- starting and restarting in the HOPE it resolves doesn't seem a good solution.
Is there anything else I can do?
Taking a depth coordinate reading in the RealSense Viewer from the calibration image attached to the ceiling may not work well. This is because it is a general physics principle (not specific to RealSense) that dark grey or black absorbs light and so makes it more difficult for depth cameras to read depth information from such surfaces. The darker the color shade, the more light that is absorbed and so the less depth detail that the camera can obtain.
Taking depth readings from a tiled ceiling may also have a chance of increasing depth error due to the camera becoming confused by scenes with visual patterns that repeat horizontally or vertically, such as a sequence of similar looking tiles. Other examples of repeating patterns are a horizontal row of fence posts or tree-tops, or a vertical stack of window blind slats.
The Intel guides at the links below provide advice about reducing the impact of the repeating pattern phenomenon.
I printed out a "Ground-Truth" Target (8.5x11")
From File in documentation here:
This is what I get---the target is square (I opened in different applications--including adobe reader)
This doesn't match the measurements provided
("The included targets are designed to be 175 x 100 mm")
And doesn't look like the sample images, that are clearly not square.
You guys don't make this very easy :)
- Camera has 3M Depth
- Accuracy is measured at 2M
- TARE Calibration needs to be done at less than 2M
The most important factor will not be the dimensions of the print-out (as it is is going to vary depending on whether A4 or US Letter format paper is used) but instead that the yellow box surrounds the edges of the paper sheet during the calibration process and does not overlap the sheet if possible.
Thanks for all this help.
I've got the cameras calibrating.
- I reconfigured my testing environment (more ambient light and TARE calibration (significantly) under 2 Meters)
I was able to do tests with a control camera, and a known failing camera, and have now been able to calibrate both within spec (deviation not exceeding 2% at 2 Meters) GREAT!
I want to do a little more testing.
In the "TARE" Calibration I have a question
- I understand the "Ground Truth" measurement
- But what should I put under "Get"---is this the Horz/Vert distance between the dots in the Calibration Image (seen here)
May I ask?
What is the value of user input fields then? --why does it allow overwrite?
And if the calibration is already off...then wouldn't the measurement likely be wrong?
I assume you calculate the length by first knowing the distance of two legs of the triangle.
(Distance of the camera to the 2 target dots)
The default Get value of 1200 is an arbitrary value chosen by the tool's developer. It is easiest to use the Get button to have the tool automatically calculate an appropriate value, but you can equally choose to provide your own value if you prefer to do so. A member of the Intel RealSense team who is a calibration specialist provides detailed advice about this tool at the link below.
Please sign in to leave a comment.