Pick and Place Machine TVM802A / TVM802B

[ar__systems]

random thought but 1.6 is a pretty common pcb thickness...?

Now that sounds like a totally sensible guess! You are probably correct.

[protoneer]

random thought but 1.6 is a pretty common pcb thickness...?

Looking at it again last night, PCB settings gets saved to the export file but not the machines settings as ar__systems pointed out. 1.6mm is the PCB thickness and gets saved,yes.

Focus on both my camera's seems to be good. The cameras seems to have adjustable lenses so that might be something to play with. Exposure set at 7 works well for red boards on the down camera.

Last night's test went well and I was able to align my old boards without fid's. The software has the option to manually align boards with the down camera or to run it as it is. (Used TH pins to align the 6 boards on my panel)

The reel feeders are the trickiest bit to setup. "Nozzle 2 Cam" , "Nozzle 2 Advancement Needle" needs to be calibrated properly.

At the moment I have two small issues.

• The nozzle does not pick up the part in the center
• The advancement needle keeps getting stuck with one particular reel.

[ServoKit]

Focus on my down camera is very good, up camera seems to be a little out of focus. At least on my machine the focus is not adjustable, lenses are glued / loctited.

The off-center picking happened on one of my reels, too (0805 caps in the tape).

Regards, Axel

[ar__systems]

Mine were glued too, but I was able to remove the glue from the down camera. Now I can adjust focus.

Initially I was testing down camera only with gold plated boards. And Fiducial detection worked well. But now I tried with HASL boards, green mask - and none of the "exposure" settings are working. Either the the fiducial and unmasked PCB around it are underexposed, or overexposed.

I wonder why they decided to use that stupid red backlight?

[ServoKit]

With HASL boards, reflection of incident light seems to be the biggest problem, darken the room or put a cardboard shade around the camera. The ENIG boards with a matte finish from PCB Pool worked best in my tests (too bad they cost so much more that the Seeed boards, at least for small volume.)

Regards, Axel
 

[ServoKit]

Alpha version of automatic loading and unloading of PCBs into the machine

Write-up about progress so far:

http://servokit.com/blog/2016/04/08/automated-pcb-handling-with-the-tvm802a-pick-and-place-machine-i/

Video follows.

Regards, Axel

[l0wside]

Axel, the Chinese manufacturers do ENIG as well, although the cost-up is a bit steep.
My latest boards were from MultiCB, at first look same quality as PBB-Pool, but about1/3 cheaper.

Max

[protoneer]

Alpha version of automatic loading and unloading of PCBs into the machine

Write-up about progress so far:

http://servokit.com/blog/2016/04/08/automated-pcb-handling-with-the-tvm802a-pick-and-place-machine-i/

Video follows.

Regards, Axel

     I did 40 similar size boards,hand feeding it. The boards did not have Fids and I only had one that I stopped after one part to realign.

I am very happy with this little machine. I load the CSV configuration,insert the board and hit the play button on the machine. Once the board is done, I load another empty board and press play button to start the next run.

Next set of boards will be panelized and will have 6 sub boards and a 6 in one SS stencil. I once again found that solder paste done properly helps a lot. If the paste is misaligned it makes it even harder to see misaligned parts.

With the 250+ parts I placed I had 4 stops where the advancement needle got stuck. All on the same reel with the first pick. Luckily these machines are easy to diagnose.

The needle only gets stuck for two alignment reasons.

• Y direction, the reel position needs adjustment(Easy app setting).   
• X direction, its the cover stripping that is moving the tape instead of the advancement needle. To fix it put some slack in the cover tape collector and if need be adjust the screws on it to make it pull with less force.

[ar__systems]

Yeah, I got about the same results, but I don't find them quite satisfactory. I've had boards that have 220 parts per one board, and that would mean 4 stops every board. Not quite good enough for me.

[ServoKit]

Another user of the 802A in Germany has placed .5mm QFN successfully:

https://www.mikrocontroller.net/topic/371594#4539279

Regards, Axel

[ServoKit]

Some open heart surgery... - I soldered wires to the Run/Stop LED and the start/stop button so that I can see the machine state and control the machine from an Arduino.

Regards, Axel

[harry4516]

Another user of the 802A in Germany has placed .5mm QFN successfully:

https://www.mikrocontroller.net/topic/371594#4539279

Regards, Axel

yes, thats me.
I purchased the TVM802B last week.
My boards contain mainly 0603 parts and QFN pitch 0.5mm microcontrollers.

It took a few days and many hours to setup the machine correctly. The threshold of Cam-1 was very important for accuracy.
After some trial and error things worked well and it was time to test the first QFN, an ATXmega128A4U-MU with 44 pins and a thermal pad.

Vision set to Accuracy and speed to 50%.
The QFN was positioned very precise, I could not see any significant offset with the microscope. It is much better compared to my previous manual placement, so I do not expect any soldering problems.

Currently I am working on an easy integration into Eagle, that is almost done. Then I will CNC a few trays for various ICs. I hope to do the first real boards this week and will post microscope-photos when done.

Regards, Harry

[ar__systems]

script for eagle. This is modified script I took from mathias' blog. parts named as FIDx will be included in fiducials.

[uncle_bob]

Another user of the 802A in Germany has placed .5mm QFN successfully:

https://www.mikrocontroller.net/topic/371594#4539279

Regards, Axel

yes, thats me.
I purchased the TVM802B last week.
My boards contain mainly 0603 parts and QFN pitch 0.5mm microcontrollers.

It took a few days and many hours to setup the machine correctly. The threshold of Cam-1 was very important for accuracy.
After some trial and error things worked well and it was time to test the first QFN, an ATXmega128A4U-MU with 44 pins and a thermal pad.

Vision set to Accuracy and speed to 50%.
The QFN was positioned very precise, I could not see any significant offset with the microscope. It is much better compared to my previous manual placement, so I do not expect any soldering problems.

Currently I am working on an easy integration into Eagle, that is almost done. Then I will CNC a few trays for various ICs. I hope to do the first real boards this week and will post microscope-photos when done.

Regards, Harry

Hi

There are a number of people *very* much interested in how the machine does with 0.5 mm parts. In my case, I could care less how slow it needs to go (50% is fine) when placing fine pitch. I also could care less how much one time calibration is required when setting up the machine (up to a point). If the machine takes four hours to re-calibrate every day .... that's probably worth knowing about.

Please keep us all informed about your progress !!!

Bob

[harry4516]

here is my first picture:

I put a tiny piece of double sided adhesive tape on the PCB and then placed this microcontroller  QFN 0.5mm pitch.
As we can see the precision is very well.

It is extremely important to set the fiducials as precise as possible. Just a little mistake with the measurements and the QFN cannot be placed accurately.
The TVM802 works best with little round fiducials. Rectangles do not work as well.

Harry

[uncle_bob]

here is my first picture:

I put a tiny piece of double sided adhesive tape on the PCB and then placed this microcontroller  QFN 0.5mm pitch.
As we can see the precision is very well.

It is extremely important to set the fiducials as precise as possible. Just a little mistake with the measurements and the QFN cannot be placed accurately.
The TVM802 works best with little round fiducials. Rectangles do not work as well.

Harry

Hi

That looks quite good. The PC board it is on is out by roughly as much as the placement. That has always been one of the arguments for down facing cameras that "spot" the pads on the board. The counter arguments have generally revolved around the (very) low speed of that process and the difficulty of recognizing pads in the clutter of a board layout plus solder paste.

Bob

[mikeselectricstuff]

here is my first picture:

I put a tiny piece of double sided adhesive tape on the PCB and then placed this microcontroller  QFN 0.5mm pitch.
As we can see the precision is very well.

It is extremely important to set the fiducials as precise as possible. Just a little mistake with the measurements and the QFN cannot be placed accurately.
The TVM802 works best with little round fiducials. Rectangles do not work as well.

Harry

Hi

That looks quite good. The PC board it is on is out by roughly as much as the placement. That has always been one of the arguments for down facing cameras that "spot" the pads on the board. The counter arguments have generally revolved around the (very) low speed of that process and the difficulty of recognizing pads in the clutter of a board layout plus solder paste.

Bob

Downward vision to locate pads is hopeless due to paste.
If you need higher local accuracy you use local fiducials. But you really want the basic accuracy of the machine to be enough to only need corner fids. A machine that can't do 0.5mm pitch without local help is pretty poor.   

[Max.Bz]

Hi all, purchased the same TVM802B, now I am in the process of calibration, there is a small problem with the position of the components on the board, reference points exhibit normal. What  it is necessary to slightly shift the position. maybe change Mark PCB position few degrees, or Real mark? or something else?

[ServoKit]

If all components are off, check the nozzle calibration (Sys Config > Nozzle > Test buttons). If just one is off, might be feeder issue, see if the needle comes down at the right place, correct if necessary in Sys Config > Left Stack / Back Stack.

Regards, Axel

[Max.Bz]

If all components are off, check the nozzle calibration (Sys Config > Nozzle > Test buttons). If just one is off, might be feeder issue, see if the needle comes down at the right place, correct if necessary in Sys Config > Left Stack / Back Stack.

Regards, Axel

All component are off, nozzles test parameters attached, camera define nozzles like on video from CD, before that had brought down the settings to stand again(left stack,back stack, prick, camera position)

[ServoKit]

The values look ok to me, much like on my machine. Some things you might want to check:

1. Make sure, the fiducial coordinates are correct relative to the origin of the board. The edges of the different layers might not be exactly aligned (see Neoden thread for an example: https://www.eevblog.com/forum/reviews/neoden-4-pick-and-place/msg897138/#msg897138)

2. Make sure PCB is fixed and can not move (stupid but happened to me).

3. Nozzle size can be an issue, look closely if parts are drifting or rotating when the nozzle comes down. Also there's a "place delay" in the config, I found "0" here working best.

Perhaps you could post a photo of a placement error.

Regards, Axel

[harry4516]

my first board is finished, here are some pictures.

The QFN was positioned very well.

The 0603 Rs and Cs are not so precise because it does not use the camera.
This has to do with the pick-up process, I'm not sure if nozzle 1 or 2 is the best choice.

The glass diodes did not work at all, I corrected them manually. For any reason the nozzle cannot hold them strong enough and they move around on the way from the stack to the PCB.
I will not use glass diodes any more. Other diodes in SOD case are no problem.

The FPC connector was placed manually because I currently don't know how to define it's offset.
The glass diodes are the only parts which I corrected manually before soldering in the oven.
Even if some R and Cs looked not so good, they corrected themself during the soldering process.

Harry

[harry4516]

All component are off, nozzles test parameters attached, camera define nozzles like on video from CD, before that had brought down the settings to stand again(left stack,back stack, prick, camera position)

after measuring the ficudials the board is synchronized with the upper camera. This works quite well.
If all components are off then I would play with the Camera2 - Nozzle1-Offset settings which should move all components.

I will test that tomorrow.

Harry

[uncle_bob]

my first board is finished, here are some pictures.

The QFN was positioned very well.

The 0603 Rs and Cs are not so precise because it does not use the camera.
This has to do with the pick-up process, I'm not sure if nozzle 1 or 2 is the best choice.

The glass diodes did not work at all, I corrected them manually. For any reason the nozzle cannot hold them strong enough and they move around on the way from the stack to the PCB.
I will not use glass diodes any more. Other diodes in SOD case are no problem.

The FPC connector was placed manually because I currently don't know how to define it's offset.
The glass diodes are the only parts which I corrected manually before soldering in the oven.
Even if some R and Cs looked not so good, they corrected themself during the soldering process.

Harry

Hi

That all looks quite good.

Even with gear much fancier than a TVM802, we have given up on round body parts. MELF's, diodes, and all the rest have been off our "products we use" list for a very long time. I can't think of any time I've been unable to do a design because of that restriction.

Bob

[rwb]

It's nice to see this machine accuratley placing .5mm pitch QFN parts 

Keep them pictures and feedback coming guys!