various problems with Openpnp

[Kai N]

Hello everybody,

2 months ago I started to build a machine based on a design by Anthony Webb. The machine is finished and works, but I have some problems with the components and also some things with the machine are not clear to me.

I use as controller a smoothieboard 5x with the drivers A5984 and microstepping 1/32.

The all axes work with openpnp as they should, but I had to limit the current of the motors from the axes X and Y (Nema17 Motors) to 0.7A. because otherwise they lose steps and do strange things. But the motors are actually designed for 1.6A. With the 0.7A the motors and the machine work fine. 

I have bought a "Peters Head" as placement head. It is a nice piece of mechanics, very high quality manufactured. Also the connected Nema 8 motors get very hot (60°C-70°C), although I only set 0.3A here and the motors should be able to handle 0.6A. I have reservations to grill them. 

Now I do not understand why these motors cannot be operated with their rated current. On the smoothieboard the heatsinks are mounted on the drivers and the currently available firmware is used. As power supply I use 24V/15A. Is this normal behaviour or am I missing something here. My Config.txt is attached. I also played with the acceleration values and set the speed down, there are no changes.

To the problem with Openpnp:

I'm using the current version of Openpnp2.0. We're just playing with the bottom vision for the components. I had the bug that I installed my camera rotated by 180°. I was wondering why all components were mounted with a big offset despite the vision. I had set all directions as described in the manual and also the jogging in the camera was correct in all directions. Now it is that with a QFN and a SOT-223 component sometimes the correction angle is not correct.

The following procedure happens:

1. the part is picked from a tray. The angle is set to 0°.

2. then the part moves over the camera. A slight offset is detected, after the vision, the cross in the viewport of Openpnp changes. However, the part does not rotate to the end position.

3. the part moves to the Place position, is rotated and 

at the place position. Now it is so that sometimes (in 10% of cases) this angle is strangely wrong and the part sits diagonally on the PCB. I have attached some pictures which show the results. In the pipeline the red rectangle is drawn correctly around the part. The angle of the offset is also different every time. Sometimes less, sometimes more. I can't see any logic behind it. Also I have seen videos where the component above the bottom camera is rotated to the correct angle. This is not the case with my machine. The linear offset of the component is always correct only the angle is wrong. I have activated pre-rotate before the vision, but I have also tried without. No difference. What I have not yet understood is what the options max. linear offset and max. angular offset do. Which maximum offset is determined here? 

I have attached my maschine.xml once. Maybe I have set something wrong and don't see it.

Thanks for your appreciated help.

 

Kai

[bert shivaan]

Trying to help with the simple things here.

First, are you using the same nozzle for the PIC and regulator as for the passives? If so, my guess is the parts are moving on the nozzle.

--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/9abeece1-cede-4917-b647-51c05f24dbc6o%40googlegroups.com.

[bert shivaan]

Also, I have had issues placing parts on blank (no solder) PCB's because the parts slip around. You can put small piece of double sided tape to see if they are moving.

[ma...@makr.zone]

Hi Kay

I'm in a hurry, just a few quick links for explanations and hints what could be wrong:

https://makr.zone/extending-bottom-vision-in-openpnp/341/

Note to self: the Wiki needs improvement. --> https://github.com/openpnp/openpnp/wiki/Bottom-Vision

https://makr.zone/smoothieware-problems-with-mixed-axis/356/

Search for "part slipping on the nozzle tip" here...

https://groups.google.com/g/openpnp/c/bEVZvYoXO98/m/hl14FcspBwAJ

... to see how you can test this.

Like I said, in a hurry, sorry this is so terse.

_Mark

[Kai N]

Hello cncmaschineguy,

yes it is correct that at the moment we always use the same Nozzle. I have all the Juki nozzles in the house and have tried bigger ones too, but with the same result. Sometimes he sets right, sometimes wrong. With the moving of the part at the nozzle I have also read and tested.

I have also noticed that when the nozzle is lifted, if no solder paste is applied, the small passive components move. However, I have observed that by increasing the waiting time after the component is placed on the PCB, the component is already twisted when it is placed.  

Anyway, thanks for your tips, as my experience with the machine is still on a low level.

Kai

Am Samstag, 11. Juli 2020 02:02:20 UTC+2 schrieb cncmachineguy:

Also, I have had issues placing parts on blank (no solder) PCB's because the parts slip around. You can put small piece of double sided tape to see if they are moving.

To unsubscribe from this group and stop receiving emails from it, send an email to ope...@googlegroups.com.

[Kai N]

Hi ma...@makr.zone,

I have already looked through their website to find possible errors on my machine.  Unfortunately I didn't know the process of pre-turning the part and I still don't have a complete understanding. I assume that by setting the options max. linear offset and max. angular offset the maximum necessary correction of the two values is set. Are these values taken into account when recording the component or are they viewed above the bottom camera? It is so that sometimes 3 pictures are taken above the camera and sometimes only 1 picture. I assume that it works like this.

Case1:  The component has a required correction angle of less than 10 degrees. I have set the 10° as an option. The camera takes a picture and the pipeline measures the angle. Now it is determined that the angle is less than 10°. Only 1 picture is taken and the correction sets the component. 

Case2: The component has a needed correction angle of more than 10°. The camera takes a picture and the pipeline measures the angle. Now it is determined that the angle is more than 10°. Now 2 more measurements are made and then the component is set with the correction.

Is this correct or am I wrong? Can you please explain the procedure again? 

I have also observed and tested the following:

I place the component into the tray turned 90° clockwise. I enter the 90° rotation in the feeder settings. Now he takes the component out of the tray. When moving to the camera the component is now rotated 135° clockwise. Then the measurement is done above the camera and the component is finally placed on the PCB with a high offset. This does not happen with passive components which are programmed as stripfeeders with 90° rotation. These are rotated counter-clockwise by 90° while moving to the camera, then measured above the camera and placed correctly on the PCB. I don't have the machine with me right now, but I will test it again next week with the 90° rotation. At the moment this is not too bad because I can always put the component into the tray with 0° rotation to the PCB. 

My understanding at the moment is that the component has to be in an angle of 0° to the PCB before it comes to the camera. Then the correction angle is determined, which has to be less than 45° so that the correction can be made to 0°. 

I think I still have a lot to learn about the machine and play with it to understand what is happening.

A quick word about the smoothie board. Is the bug with the multi-axes fixed in the current firmware? What I also see on my machine is, if I turn the C-axes with a step of 1° by hand, it happens that sometimes the C-axis is not turned. When you press the second time, the axis continues to rotate. If I turn in 10° steps, then I always see the rotation in the camera. I have checked the smooth running of the axes and cannot find any errors. I will check it again next week. Is it normal that the motors C1 and C2 heat up to 60° at a current of 0.3A?

Kai

To unsubscribe from this group and stop receiving emails from it, send an email to ope...@googlegroups.com.

[Michael Anton]

There is no way to test placement accuracy of these machines, without something tacky on the surface of the PCB.  Get some light tack double sided tape for doing test placements, if you don't want to use paste.  Otherwise, the parts will move when the nozzle is lifted.   I use a product made by 3M called Scotch Permanent Double Sided Tape, which is anything but permanent, since it is just like regular Scotch tape with adhesive on both sides. 

Also, if you plan to have your machine move fast, I've learned to always use the largest nozzle that fits on the part without interfering with vision.  Otherwise, parts will slide during acceleration, and rotation.  This of course assumes you have a nozzle changer, and that it is practical to do this.  Your vacuum level would also determine how large a nozzle is required.

[ma...@makr.zone]

Hi Kai

Yes, you got the process (almost) right. Note, all this only applies to pre-rotate is enabled.

1. It takes a picture.
   
2. If one or more of the offsets are above the thresholds it will rotate/move the nozzle to bring the part to the correct position and then restarts at step 1.
   
3. Else (if all the offsets are below the thresholds or the number of retries are exhausted) it goes on.
4. It remembers the last measured offsets to be applied when the part is placed.
   

So this is iterative and it stops when it is good enough. The rationale is this: through errors by Z parallax, lighting asymmetry, and/or camera distortions it is practically impossible to measure absolute offsets accurately. But most of these errors are symmetrical, i.e. the errors cancel each other out on both sides, so by bringing the part to the center of the camera again and again, the accuracy will improve more and more.

> I have also observed and tested the following...

There should be no difference between parts in how bottom vision works, if both have pre-rotate enabled. But there are differences in how the angle is interpreted. The Strip Feeder as well as the BlindsFeeder and many other tape based feeders interpret the angle as "rotation in tape". So you must know the tape angle too, as those two add up! Other feeders don't have a tape and the angle is absolute i.e. relative to the machine. Some of this is explained in the Wiki :

https://github.com/openpnp/openpnp/wiki/ReferenceStripFeeder

https://github.com/openpnp/openpnp/wiki/ReferenceStripFeeder#rotation-in-tape

> A quick word about the smoothie board. Is the bug with the multi-axes fixed in the current firmware?

Yes it should be. But make sure you really have the latest. Or use my firmware.

> Is it normal that the motors C1 and C2 heat up to 60° at a current of 0.3A?

Yes, I believe that is normal. Some drivers allow reduction of current when no/low motion is detected.

One example where this worked so-so:

https://makr.zone/choosing-a-motion-controller-the-panucatt-azteeg-x5-gt-32bit/455/

_Mark

To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/e6c4024a-d640-4fed-993e-5b64d1c73dd2o%40googlegroups.com.