Hi Ulrich, After rereading your original email some more, I am wondering if I correctly understood the trouble you are having. Is the following understanding correct? Most of the time things are good. Every once in a while (every few seconds?) you get one (many?) frames where the textured and untextured channels are swapped. Is that accurate? If not, could you possibly post a bag with the offending image_raw camera output so that I can be sure of my understanding of the problem. Keep it short as these bags get big very fast. A single camera is probably sufficient, no need to send both, unless something extra is illustrated. Could you try killing the following code from pr2_camera_synchronizer/synchronizer_classes.py? {{{          # In case the controllers got restarted, refresh their state.          controller_update_count += 1          if controller_update_count >= 10:              controller_update_count = 0              for controller in self.controllers:                  controller.update(); }}} It causes a problem similar to my understanding of your problem in the tests I have been doing. On Wed, Sep 22, 2010 at 3:13 AM, Ulrich Friedrich Klank wrote: > Hi Blaise, > > sorry for posting the last email twice. > > Thanks for the quick response. Unfortunately, neither the trunk version > nor "projector tweak" could solve the problem. > > Here is the dynamic-reconfigure-dump which gives the leasts corrupted > images in the no-texture topic (but still a significant amount of empty > point clouds): > > {camera_reset: false, forearm_l_rate: 30.0, forearm_l_trig_mode: 1, > forearm_r_rate: 30.0, >   forearm_r_trig_mode: 1, narrow_stereo_trig_mode: 5, projector_mode: > 2, projector_pulse_length: 0.001, >   projector_pulse_shift: 0.0, projector_rate: 40.0, projector_tweak: > 0.0, prosilica_projector_inhibit: false, >   stereo_rate: 20.0, wide_stereo_trig_mode: 2} > > Regarding projector tweak i tried "rosparam set projector_tweak -0.001". > Is this what you meant? I could not see any effect. > > Best regards, > Ulrich Klank > > On 09/22/10 07:09, Blaise Gassend wrote: >> One thing I forgot to mention is that you can relatively easily set >> your own waveforms for the camera and projector by directly calling >> the set_waveform services and dynamically reconfiguring the cameras. >> If you want one particular waveform, that is a good approach. For >> something more generic, the right thing is to get the bugs out of >> pr2_camera_synchronizer. >> >> Have you tried the projector_tweak? From the symptoms you describe, >> that should get you going. >> >> There is an updated version pr2_camera_synchronizer in trunk >> (https://code.ros.org/svn/wg-ros-pkg/stacks/pr2_robot/trunk/pr2_camera_synchronizer) >> that eliminates some problems with low frame rates. If you were >> experimenting with low frame rates, this fix and projector_tweak >> should get you going. >> >> On Tue, Sep 21, 2010 at 12:13 PM, Blaise Gassend >>  wrote: >> >>> I see that my previous reply to Dejan didn't go to the whole list, so >>> here it is. >>> >>> There aren't currently any calibration knobs other than >>> projector_tweak. Have you tried setting projector_tweak to -0.001? >>> (Described here >>> http://www.ros.org/wiki/pr2_camera_synchronizer/Troubleshooting). This >>> should get you going when grabbing images when the camera frame rate >>> is half the projector frame rate, but apparently not when the camera >>> frame rate is less than that. >>> >>> Since these problems seem to be coming to a head the past couple of >>> weeks, I think it is time for me to finally stop and fix them. >>> >>> Can you send me the output of dynparam dump on the synchronizer node? >>> >>> On Tue, Sep 21, 2010 at 6:36 AM, Ulrich Friedrich Klank  wrote: >>> >>>> Hi pr2-users, >>>> >>>> On our PR2, we are experiencing currently a temporal offset between >>>> emitted textured light and labeled data in the narrow+wide point cloud >>>> and images. We get images with texture on it in the wrong topic and some >>>> empty point clouds. >>>> >>>> While filtering out empty point clouds is easy, it is  difficult to >>>> detect in non-colored images if the pattern is visible or not.  (Which >>>> makes image processing based on edges impossible) >>>> >>>> So I wanted to ask, if and how it is possible to influence the time when >>>> the textured light is emitted. Or even better, can it be calibrated with >>>> the exposure of the cameras in "AlternatingProjector" mode? >>>> >>>> Are there any settings beyond the parameters available in >>>> camera_synchronizer_node? (Those parameter did not help) >>>> >>>> Thanks in advance, >>>> >>>> Ulrich Klank >>>> >>>> >>>> -- >>>> Ulrich Klank                        klank@in.tum.de >>>> Technische Universität München | Boltzmannstr 3 | >>>> 85748 Garching bei München  | Germany >>>> ias.in.tum.de/people/klank | Tel: +49 89 289 - 26909 >>>> >>>> _______________________________________________ >>>> PR2-users mailing list >>>> PR2-users@lists.willowgarage.com >>>> http://lists.willowgarage.com/cgi-bin/mailman/listinfo/pr2-users >>>> >>>> >>> > > > -- > Ulrich Klank                        klank@in.tum.de > Technische Universität München | Boltzmannstr 3 | > 85748 Garching bei München  | Germany > ias.in.tum.de/people/klank | Tel: +49 89 289 - 26909 > > _______________________________________________ > PR2-users mailing list > PR2-users@lists.willowgarage.com > http://lists.willowgarage.com/cgi-bin/mailman/listinfo/pr2-users >