Calibration Solutions are provided by Issues and Solutions when targeting the Calibration milestone. This page as a whole provides the background information for it. Individual sections are linked from the corresponding solutions.

The Calibration Solutions are part of an overall holistic (whole-machine) approach to calibrate both the cameras and various mechanics of the machine by playing the two against each other. On one hand, the known metrics of machine motion are used to calibrate the cameras. On the other hand, the camera is used to make the motion more precise. The inherent chicken and egg situation between the two is circumvented by exploiting various forms of symmetry and by approximation through iteration, things that would be hard or tedious to do by hand.

To get some impressions, watch a video of these and other calibration solutions. The video is not complete, so come back to this page.

Backlash compensation is used to avoid the effects of any looseness or play in the mechanical linkages of machine axes. Using the camera, looking at the calibration primary fiducial, a special calibration process can automatically determine the backlash of the machine X and Y axes. It can therefore configure Backlash Compensation with the correct method, speed and offsets to compensate the errors.

NOTE: on mechanically superior machines, backlash compensation will automatically be switched off if it is not needed. So even if you are convinced that your machine does not need it, it makes sense to perform this calibration and confirm everything works as expected. You'll also get a ton of diagnostic information in the process (see below).

Tolerance ± can be used to set the wanted tolerance for the axis. The calibration routine may be able to select a more performant method if the granted tolerance is greater. More performant methods avoid direction changes for a more fluid and faster overall machine performance. The calibration routine will do its best to aim for the wanted tolerance, but no guarantee can be given.

More information about the rationale and the methods used etc. can be found on the Backlash Compensation page.

The driver must first be properly configured in terms of Motion Control Type. One of the advanced motion control types must be selected, i.e. ConstantAcceleration or better, where OpenPnP can effectively control the speed factor, i.e. not just the feed-rate, but also the acceleration (and optionally jerk). The calibration needs to perform elaborate measurements using different speed factors, otherwise the routine is useless or even misguiding.

You also need to make sure your Kinematic Axis Rate Limits are properly set and tuned to the maximum. Backlash will typically change when you make an axis accelerate (or jerk) harder. Therefore, you need to tune the axes' rate limits first, and/or repeat the backlash calibration again, after changing them.

Failure to prepare properly will be reported with this Error:

After having prepared everything, press the Accept button to perform the calibration. NOTE: the routine will test many aspects of the axis behavior, this will take many minutes.

Press the Machine Disable button, if you want to interrupt it.

After the calibration has completed, go to Machine Setup / Axis to see the diagnostics.

The calibration function is also available on the Axis.

Press the Calibrate now button to perform an axis backlash calibration at any time.

The settings proposed by the routine can be edited as you like. Refer to the Backlash Compensation page for more information.

Elaborate graphical diagnostics are present after each calibration (both from Issues & Solutions and from the button). It contains a rich set of information about how the axis performs, that actually goes beyond the mere backlash behavior (explained below the screenshot):

The first graph shows you the positional precision of the axis, stepping over one Millimeter.

• The Relative Step Errors (blue dots) should remain inside the dark blue tolerance limits.
• The Absolute Error (red line) should not show any sagging and no trend that goes outside the dark blue limits.
• The Random Move Error (brown dots) should also mostly remain inside the dark blue tolerance limits. These dots represent random distance moves made with the newly calibrated backlash compensation settings applied, i.e. this is a success test. The same brown dots are also plotted in the second graph, this time over move distance.

Failure to meet these goals may indicate that the motor currents are too weak and/or that the axis resolution/micro-stepping is too fine. The transmission ratio may be too large, e.g. the pulleys too large for the given motor power. Also check any dynamic motor current control such as Trinamic CoolStep™. You may have to switch it off.

The second graph shows the measured Backlash Offset (blue) and Overshoot (red) at different approach distances. These are measured at minimum (usually 25%) speed and full speed, respectively. Both are performed after a straight and after a direction reversing move. The routine tries to find a minimum approach distance, where the backlash remains consistent within the given Tolerance ± window. This minimum approach distance with consistent backlash is called Sneak-up Distance.

There is a dark blue letter-F-shaped indicator superimposed on the graph:

• The determined Sneak-up Distance is indicated by the vertical line of the letter "F".
• The determined consistent Backlash Offsets with Tolerance ± window is bracketed by the horizontal lines of the letter "F".

Based on the observed performance, a heuristic will select the appropriate Backlash Compensation Method automatically:

• If the Backlash Offset and Sneak-up Distance are within Tolerance ±, the backlash compensation is switched off. Bravo for a superb machine!
• If the Sneak-up Distance is smaller than the average Backlash Offset, the most efficient DirectionalCompensation method is chosen. This method is as fast and fluid as with no compensation.
• If the Sneak-up Distance is larger than the average Backlash Offset but still acceptable, the still quite efficient DirectionalSneakUp method is chosen.
• If the Sneak-up Distance is not acceptable, the classical OneSidedPositioning is chosen. It incurs extra slow approach moves and direction changes, i.e. motion is significantly slower. This is a sign that backlash is large and rather inconsistent. Check the brown dots in the first graph for how consistently accurate the machine really is (no guarantees to uphold the wanted tolerance at this point).

The green line shows you the Move Time over distance. Note the logarithmic scale, both on distance and move time. This is a good indicator of your axis performance.

In this graph, the calibration routine measures Backlash Offset consistency at different approach speeds. If possible the speed is increased, up from the initial 25%.

The graph also shows you how effective the speed control is (green line). Ideally, the line is perfectly diagonal, i.e. effective speed factor and nominal speed factor (on the horizontal axis) are the same at each dot.

To calibrate precision camera ↔ nozzle offsets, we let the nozzle pick, rotate and place a small round test object and then measure the result using the camera. By exploiting the effects of rotational symmetry and averaging, the precise nozzle axis can be determined.

Watch the video sequence to understand how it work.

The test object should be circular and flat on the top. The nozzle must be able to pick and place it precisely. There must be good contrast between test objects and background. As you are likely to lose or damage these little test objects, best make several. They can be as simple as punched out "confetti" from a hole punch, used on matte card stock. You can also create heavier ones made from metal sheet etc. Press them flat. Don't worry if they are not perfect, this will all be cancelled-out through circular symmetry.

The calibration is executed using the regular pick and place cycle, so all the vacuum control settings etc. apply. If you need to control the properties of the part to be picked, like the part height or package blow-off level, etc. you can create a part with the ID "TEST-OBJECT". It will be used automatically.

You already performed the preliminary camera calibration earlier, but now it is time for the advanced calibration that includes compensating lens distortion and camera mounting tilt. A more profound and precise 3D Units per Pixel calibration is also included.

The calibration must be performed with the same calibration rig, that you used for the preliminary calibration, make sure it is ready, and locations (including Z heights) are still valid. If not, please revisit the Calibration Primary Fiducial and Calibration Secondary Fiducial steps first: enable the Include Solved? checkbox on the Issues & Solutions tab, to see revisitable solutions.

See the Advanced Camera Calibration page for more information. Note, that Issues & Solutions will perform an automated calibration, with the calibration fiducials predefined.