Pick and Place Machine TVM802A / TVM802B

[mikeselectricstuff]

Surely a belt would be faster and cheaper than a leadscrew?

[protoneer]

What would be the best way to calculate the needle-to-nozzle offset?

Camera to nozzles can be done with a peace of blue-tac. Press the nozzle into the blue-tac(Note XY), center  camera over the blue-tac(Note XY). Deduct two coordinates to get off set.

[protoneer]

Interesting!!! OpenPnP might be an option in the future...

https://groups.google.com/forum/#!topic/openpnp/ZW4fu7Gy5vs

If only we can decipher the protocol... 

[ServoKit]

@mike: I try to avoid belt drives when possible, components *may* be cheaper but overall the setup means more parts and more work (mounting the motor, regulating belt tension etc.) Also, for 250mm travel, speed is not an issue.

In the meantime I began working on the dispenser; the drive train of  the PCB transport can be exploited for Y direction, adding support for X is easy.

Regards, Axel
 

[ar__systems]

Interesting!!! OpenPnP might be an option in the future...

https://groups.google.com/forum/#!topic/openpnp/ZW4fu7Gy5vs

If only we can decipher the protocol... 

It is done. I'm on vacation now, I'll post it when I get back. Next week.

[ServoKit]

Doing fiducial recognition with Raspberry PI and $25 camera.

Write-up w/ Python code: http://servokit.com/blog/2016/05/09/an-automatic-smt-solder-paste-dispenser-iii/

Regards, Axel

[harry4516]

Interesting!!! OpenPnP might be an option in the future...

https://groups.google.com/forum/#!topic/openpnp/ZW4fu7Gy5vs

If only we can decipher the protocol... 

It is done. I'm on vacation now, I'll post it when I get back. Next week.

this would enable us to save all the machine settings in a file. There are so many numbers without a backup. Currently I save them as screenshots.

[protoneer]

I did another run with the machine today. 600+ parts with 10 panel changes.

I had one attempt at picking a part not succeeding. Made a small adjustment to the tape peeler and all was good.

With the current settings I can start the job and do other things while it works. Saving me plenty of time.

Working without fid's is also not so bad. With my current set of panels I have used via's on each sub board to do the alignment. I used manual alignment for now and its amazing how each board differs.

I am really pleased with the machine and recommend it as a  machine to start with. Cant wait to see what the TVM9XX offers. The removable reel holders sounds very tempting.

[harry4516]

protoneer, I can confirm your experience 100%.

One question,
If a part is not pick successfully then the empty nozzle is dunked into the paste on the board.
This makes problems with further picking.
Does your machine detect empty nozzles ?

How do you clean the nozzles in that case ?

[protoneer]

Mmmm.... there is a machine setting to check for picking issues.(uses a pressure sensor)

In my case the plastic cover did not remove in time for the nozzle picking the part. It then went straight from the reel to the reject bin. It also has a try limit of 5 attempts if I recall correctly.

Does anyone know of a vibration feeder that might work well with this machine?

[uncle_bob]

Mmmm.... there is a machine setting to check for picking issues.(uses a pressure sensor)

In my case the plastic cover did not remove in time for the nozzle picking the part. It then went straight from the reel to the reject bin. It also has a try limit of 5 attempts if I recall correctly.

Does anyone know of a vibration feeder that might work well with this machine?

Hi

I think you could easily shorten that to "Does anyone know of a vibration feeder that works well".

Vibratory bowl feeders are tricky beasts. They tend to be a bit sensitive to the exact parts being feed. That makes them a pretty expensive item unless you are doing a *lot* of the same parts. When you go out to custom tool one (so it works well with your parts) you find that there are relatively few people who understand how to do one right. Those that do know generally are pretty fully booked ....Thus not just pricy, but expensive *and* long(er) lead time.

Of course you may be looking for a different sort of feeder ....

Bob

[harry4516]

Mmmm.... there is a machine setting to check for picking issues.(uses a pressure sensor)

In my case the plastic cover did not remove in time for the nozzle picking the part. It then went straight from the reel to the reject bin. It also has a try limit of 5 attempts if I recall correctly.

...

I had the issue with the plastic cover too, but where can I find and activate this setting ?

[protoneer]

I had the issue with the plastic cover too, but where can I find and activate this setting ?

Its on the Other tab of the machine settings. Can't recall the wording they used... (Was not very obvious  )

[ServoKit]

@Harry, you have to activate it in the Machine tab settings ("Use pressure sensor") and for the individual part in the part list. The machine will try 5x to get the part right, then a dialog window comes up with an error msg.

Also, if the nozzle gets into the paste, that shouldn't happen; check the settings for distance from nozzle to PCB.

Regards, Axel

[sergeil]

With the pressure sensor and smaller tips, a bit of blockage (e.g. a 0402 part that half-slips in or some paste that happens to get in the nozzle) can easily confuse pick-up sensor. Then it'll think every pickup is successful.

You can quickly check if that's working correctly by going to "manual" mode, turning on the vacuum + appropriate nozzle vacuum, and checking that the pressure indicator is off usually and on when you block the nozzle.

[harry4516]

thanks 
that works great It just picks the next part.
I had this problem when the cover tape was not peeled off correctly (my fault, an adhesive tape was on the wrong side and glued on one of the rods).

Here is a new board, just to show the people what this fantasic machine is able to do.
The SMB diodes adjusted in Quick mode, the ICs in Accurate mode and all other parts without camera:

[harry4516]

Interesting!!! OpenPnP might be an option in the future...

https://groups.google.com/forum/#!topic/openpnp/ZW4fu7Gy5vs

If only we can decipher the protocol... 

It is done. I'm on vacation now, I'll post it when I get back. Next week.

any news ? can't wait for this valueable information 

[dermeister]

I have a TVM802B on the way right now and would also be quite interested in seeing the reverse engineered control protocol.

[harry4516]

looks like we do not get it this way, so why don't we do it by ourself ?

Lets build a team of TVM802 users who are able to work with  wireshark.
Wireshark has a nice function "follow TCP stream" which may be useful.
We have to capture the stream, make a little change in the TVM802 settings and then look for changes in the TCP stream.

I am ready to start this project. But the stream is quite long, so more people should work on it to get the information in reasonable time.
We could exchange our results here.

Harry

[uncle_bob]

looks like we do not get it this way, so why don't we do it by ourself ?

Lets build a team of TVM802 users who are able to work with  wireshark.
Wireshark has a nice function "follow TCP stream" which may be useful.
We have to capture the stream, make a little change in the TVM802 settings and then look for changes in the TCP stream.

I am ready to start this project. But the stream is quite long, so more people should work on it to get the information in reasonable time.
We could exchange our results here.

Harry

Hi

I have a copy of Wireshark if anybody wants to send me a TVM802

===

Having been down this road once or twice:

Wireshark is free, so that's the easy part. The best way to watch things is with a second computer. That way you don't get in the way of the control process on the "main machine". The gotcha there is that you may need an ethernet switch that does port forwarding to see the whole process. Fortunately they are pretty cheap these days. They are also fairly easy to set up.

The objective is to interpret the transactions. That can be done in a number of ways. One is by writing code is incrementally updated as you figure out what is what. Ultimately you have a monitor program that can tell you "send: move head X = 332 Y= 211", "rec: head at location" or some such thing. To do that kind of code with a group, you need agreement on what you can or can't run on your machines. Visual Studio is one (free) option. The nice thing about a program is that it's quite happy to sit there for hours watching the dialog. It can tag the stuff it does not understand so you can come back and work it out later.

Lots of details .... MANY different ways to get the job done.

Bob

[ar__systems]

looks like we do not get it this way, so why don't we do it by ourself ?

sorry guys, I've been busy with other stuff that makes me money

pm me your email and I'll forward what I've got so far.

[ar__systems]

Unfortunately the code is very.... bad. My student wrote it and I did not have time to work on it since. That said, it does everything.

[vonnieda]

Unfortunately the code is very.... bad. My student wrote it and I did not have time to work on it since. That said, it does everything.

Could you just chuck it up on a Github repo or something? If you don't want to take the time, and wouldn't mind it being done, feel free to send me what you've got and I will post it for you.

Thanks,
Jason

[protoneer]

looks like we do not get it this way, so why don't we do it by ourself ?

sorry guys, I've been busy with other stuff that makes me money

pm me your email and I'll forward what I've got so far.

Hi @Ar__systems, I have sent you a quick PM. I can setup the code on Github and then people can start working on this. We don't mind about the state of the code, we are use to cleaning up code...

By the way, thanks for willing to share this. Once this has been integrated into OpenPNP everyone is going benefit from an even better running machine.

By the way, I have been playing with 0402's the last few weeks and I am very surprised at how well the machine handles it. 0603 and 0402 are almost half the price of the 0805's. Nice cost saving over the long-run...

[ar__systems]

I sent the code to protoneer. He will set it up on Github or wherever, as he suggested