Pick and Place Machine TVM802A / TVM802B

[protoneer]

Also, don't run the s/w on W10. the docs says XP is not supported but we actually ran it on XP and it worked fine. On W10 it has some connectivity issues resulting in some commands not executed properly. I first attributed it to trying the machine over wifi, but no. I ended up with a broken nozzle and bent pick up needle as a result of trying w10.

Thanks for the tips... I was heading the W10 way.   

How are you finding the nozzles? I see there are adapters for Juki nozzles that are much easier to find...

How are you finding the Vision system?

Sorry for all the questions.... 

[ar__systems]

Regarding the nozzles, mine actually came already equipped with Juki nozzles (well... probably not original ). I'm surprised to see here photos of the machines with some yellow metal nozzles. What brand are those?

I recommend buying some spares from the start.

Vision system is a must for PNP. Although s/w overall looks like it was written by a student over a weekend, the visual system actually works consistently reliably. I don't remember a single issue with it. Ok there is one - looking for fiducials it can get confused by the presence of other things in the field of view, like silk screen or traces or holes. So it is better to keep area around fiducials free of anything else.

[protoneer]

Regarding the nozzles, mine actually came already equipped with Juki nozzles (well... probably not original ). I'm surprised to see here photos of the machines with some yellow metal nozzles. What brand are those?

The metal nozzles are the same as what Neoden uses. Its hard to tell where things originated when products are being developed side by side...  They are fitted on the TVM802A before the B was developed. Adapters are available but might be fun to press-fit.

 

[protoneer]

I am very interested in the network protocol. 

It was one of the main reasons I picked this machine. Possibility of writing custom software for it opens lots of doors even if its just used for special movements.

AR_Systems, do you have any thoughts on solutions for the feeder issue?

[ar__systems]

Again, it's part of the design. Rather than using a super sensitive sensor with no calibration, they use something they can actually measure changes on. When the thing is in the "up and empty" situation they can (but may not always) calibrate the sensor. When they change heads from giant to next to no diameter at all, they can calibrate for that head. They might store that value or measure it each change. Storing it lest them detect the "oops missing head / wrong head" problem. Yes it's a bit more firmware. Yes it works better than the alternative. There are likely as many variations on how to do it as people doing it.

It still is not perfect, but no solution ever is. The smallest head does still set a limit on the leaks. You can't have more air through the leak than through the head. A screw down gasket fitting has a *much* lower leak rate than the flow through that small head. You can have a lot more than a 1x10^-8 cc/atm/sec leak and still detect a missing part just fine.

Bob

Nice theory and I wish things were done that way, but the reality is that for TVM802A/TVM802B the pressure sensor is a dumb mechanical pressure switch that can't be adjusted (I mean not without disassembly). The pressure status transmitted to the PC is ***ONE BIT***. It works accordingly poorly. Often times it does not feel the part is on the nozzle and throws it out, even though it looks absolutely fine. It especially hates ceramic caps. But also some huge parts that were sitting absolutely solidly on the nozzle. So I'm guessing the switch is not very consistent. I had to disable pressure for several components because it was just discarding parts over and over and over.

[uncle_bob]

Again, it's part of the design. Rather than using a super sensitive sensor with no calibration, they use something they can actually measure changes on. When the thing is in the "up and empty" situation they can (but may not always) calibrate the sensor. When they change heads from giant to next to no diameter at all, they can calibrate for that head. They might store that value or measure it each change. Storing it lest them detect the "oops missing head / wrong head" problem. Yes it's a bit more firmware. Yes it works better than the alternative. There are likely as many variations on how to do it as people doing it.

It still is not perfect, but no solution ever is. The smallest head does still set a limit on the leaks. You can't have more air through the leak than through the head. A screw down gasket fitting has a *much* lower leak rate than the flow through that small head. You can have a lot more than a 1x10^-8 cc/atm/sec leak and still detect a missing part just fine.

Bob

Nice theory and I wish things were done that way, but the reality is that for TVM802A/TVM802B the pressure sensor is a dumb mechanical pressure switch that can't be adjusted (I mean not without disassembly). The pressure status transmitted to the PC is ***ONE BIT***. It works accordingly poorly. Often times it does not feel the part is on the nozzle and throws it out, even though it looks absolutely fine. It especially hates ceramic caps. But also some huge parts that were sitting absolutely solidly on the nozzle. So I'm guessing the switch is not very consistent. I had to disable pressure for several components because it was just discarding parts over and over and over.

Hi

That's the whole point, If you do a more exotic machine, a *lot* of things change. It's not just a "put this on and you are done" mod. The whole thing has to work together. The price goes up in big steps *not* just because of the accessory you add. It goes up at least as much for all the other stuff you have to mess with to get it to work right.

Bob

[ar__systems]

Not sure what you mean.... I've read entire Neoden-4 thread over the last few days. Many comments are like "oh you bought cheap machine, what do you expect". Here is what I think.

When I buy cheap car, I don't expect it to be on par with 200K jaguar. It might not be as fast, it might not have 4WD, it might have crappy interior. But I sure as hell expect NOT to have to get out of the car every 100 meters to do some tweaking under the hood.

Yes we buy cheap machine and we already compromise on a lot of things. It is slow, it does not have conveyor, it does not have nozzle changer, it can't mount 0201 etc etc. But I'm kind of expecting it to be more or less reliable. it is not.

Well ok, at least the amount of money I paid for mine is about how much a DYI kit to build a PNP would've cost me.

And yes on this machine several things can be changed that will make significant improvement even with stock s/w.

[ar__systems]

do you have any thoughts on solutions for the feeder issue?

I don't have any ideas for quick and easy fix I'm planning completely replacing feeders together with the peeling mechanism. I plan to use one dc motor per reel to peel off the cover tape. I don't believe the existing single shaft system can be made to work. Although I spoke to an engineer today and he said that if we replace rubber rings in those spools with bronze it might be made to produce more consistent friction. But I'm not holding my breath.

[protoneer]

In the end it all needs to play nice.

Not sure if this is possible with the TVM802X’s s/w but at work I have written proxy app’s that basically sits between hardware and 3rd party software. The proxy app would just listen on the machine port and forward the data to the 3rd party software.

This opens the door to injecting messages to and from the machine and even transforming them.

An example would be to add better pressure sensors and get the proxy app to spit out the appropriate comms to the 3rd party s/w to place or dump a part.

[protoneer]

do you have any thoughts on solutions for the feeder issue?

I don't have any ideas for quick and easy fix I'm planning completely replacing feeders together with the peeling mechanism. I plan to use one dc motor per reel to peel off the cover tape. I don't believe the existing single shaft system can be made to work. Although I spoke to an engineer today and he said that if we replace rubber rings in those spools with bronze it might be made to produce more consistent friction. But I'm not holding my breath.

How about commercial reel dispensers? Like the Juki, Panasonic and so on…. I know they will be much more expensive but in my case I could add a few at a time. That would increase productivity over time.

[ar__systems]

Not sure if this is possible with the TVM802X’s s/w but at work I have written proxy app’s that basically sits between hardware and 3rd party software. The proxy app would just listen on the machine port and forward the data to the 3rd party software.

Should be possible. You can direct the S/W to talk to a specific port on specific IP, which can operated by your application. Not sure if it is a good approach. You need to build in a lot of intelligence into this application. It needs to know what the original s/w is intending to do, just from observing it's actions. I think it is just easier to write an entire application and scrap the original one altogether.

[ar__systems]

How about commercial reel dispensers? Like the Juki, Panasonic and so on…. I know they will be much more expensive but in my case I could add a few at a time. That would increase productivity over time.

I did not look into those. Might be impossible to fit mechanically. The pickup location on the TVM802A/B is about 10-15cm inside of the perimeter of the machine, and the vertical clearance is only about 20mm. I only briefly looked at various big machine feeders and I don't think they would fit without major modifications to the TVM.

[ServoKit]

@rwb, Bob:

I think the 10mm capability is there but really just theoretically as long as you can't tell the software to navigate around obstacles. I'm mostly interested in caps of the 5.8mm variety and those shouldn't be a problem. Re. the weight of parts: The #3 nozzle does lift reliably an ESP8266-12F. That's a fairly large part (24x16mm, 2-3grams).

Re. W10: I'm running the machine on Windows 7 Home, 64 Bit with 8GB. So far no issues. This is a dedicated computer that does nothing else.
Re. nozzles: It seems that the 802B is different in this regard, photos on Ali show them with Jukis but they don't seem have the quick change. Maybe ar_systems can clarify this.

Vision system is a must for PNP. Although s/w overall looks like it was written by a student over a weekend, the visual system actually works consistently reliably. I don't remember a single issue with it. Ok there is one - looking for fiducials it can get confused by the presence of other things in the field of view, like silk screen or traces or holes. So it is better to keep area around fiducials free of anything else.

My exact sentiment and experience.

I'm looking forward to our PCBs so that I can do some real tests.

Regards, Axel

[ar__systems]

Nozzles are easy to change just by pulling on them down. They are only held in place by the friction in o-ring between the nozzle and the yellow tube they are insterted into.

[uncle_bob]

Not sure what you mean.... I've read entire Neoden-4 thread over the last few days. Many comments are like "oh you bought cheap machine, what do you expect". Here is what I think.

When I buy cheap car, I don't expect it to be on par with 200K jaguar. It might not be as fast, it might not have 4WD, it might have crappy interior. But I sure as hell expect NOT to have to get out of the car every 100 meters to do some tweaking under the hood.

Yes we buy cheap machine and we already compromise on a lot of things. It is slow, it does not have conveyor, it does not have nozzle changer, it can't mount 0201 etc etc. But I'm kind of expecting it to be more or less reliable. it is not.

Well ok, at least the amount of money I paid for mine is about how much a DYI kit to build a PNP would've cost me.

And yes on this machine several things can be changed that will make significant improvement even with stock s/w.

Hi

If you go back through the entire sub-thread that I was commenting on, the discussion was related to an auto changer for multiple nozzles for this machine. That generally involves a wide range of nozzles as well as the changer it's self. The comments were related to the various other parts of the system's hardware (and software) that might be impacted by simply bolting on a changer. The original comments started on a "that should only be a couple hundred dollars" sort of theme.

Bob

[uncle_bob]

@rwb, Bob:

I think the 10mm capability is there but really just theoretically as long as you can't tell the software to navigate around obstacles. I'm mostly interested in caps of the 5.8mm variety and those shouldn't be a problem. Re. the weight of parts: The #3 nozzle does lift reliably an ESP8266-12F. That's a fairly large part (24x16mm, 2-3grams).

Hi

The other item that gets big are power inductors / transformers. Those tend to get heavy when they get big.

Bob

[ar__systems]

With a big nozzle it has no problem holding inductors or anything else.

[uncle_bob]

With a big nozzle it has no problem holding inductors or anything else.

Hi

... then I guess all those parts I've dropped over the years must have been from another issue...

Bob

[ServoKit]

I tried it with the largest nozzle (#4) and little pieces of brass I have lying around. Pictured is the highest weight that will be picked reliably, it's about 4grams (1/8oz.). The larger one in the pic drops off. The Juki nozzles (even knock-offs) are perhaps better, i.e. air tighter.

Regards, Axel

[uncle_bob]

I tried it with the largest nozzle (#4) and little pieces of brass I have lying around. Pictured is the highest weight that will be picked reliably, it's about 4grams (1/8oz.). The larger one in the pic drops off. The Juki nozzles (even knock-offs) are perhaps better, i.e. air tighter.

Regards, Axel

Hi

Which is not terribly surprising. You only have a finite suction area and a max of 15 psi over that area. Once the weight of the part exceeds the area times pressure ... the part falls off. It's just physics. The standard drill here is to always use at least one nozzle bigger than you think you need (go up one from the smallest one that will pick up and hold the part).

Bob

[ar__systems]

Larger Juki nozzles have rubber tips so they will hold even better.

Unfortunately due to a poor pump in the TVM we do not have 15 PSI differential. I measured the pressure of 560mmHg, and that's with no direct path for leaking.... This is why I want to replace the vacuum pump with a much bigger one. Not only this one creates only a weak vacuum, it takes a couple of seconds for it to reach even this level. But in a couple of seconds the part is already delivered to the destination so it is kind of late....

[ar__systems]

Servo, keep in mind just lifting is not enough. It needs to withstand lateral acceleration without shifting or being dropped.

[ServoKit]

Yup, I tried with our heaviest and largest part, 16x24mm, 1.45g precisely. No problems.

Regards, Axel

[protoneer]

How does the Vacuum pump on this unit look?

[ServoKit]

I've opened the machine (will post pics later); from a quick glance there are two pumps marked "Hargraves", this seems to be the mfg: https://www.hargravesfluidics.com/

Regards, Axel