Pick and Place Machine TVM802A / TVM802B

[ManuFerHi]

I have put many of this type, generally small components go well, the problem with vision is when the components are large.
Even so it is essential to make a visual review before reflow.

[diyengineering]

Sharing this version (v3.06) of the software that was provided with my 802A

fyi, seems most versions of this software are detected as malware, use at own risk... I cannot verify it as I have not received my machine yet. but since this is newer than other versions of the software that I have seen, I wanted to share it.

https://1drv.ms/u/s!AjSyWl01MM23jogmKNNmTgaelR9pcQ

[Conjure]

Hey all,

I just got mine yesterday and calibrated it pretty easily, actually. Thought it would be a nightmare.

One question: With the test alignment board, I'm trying to use the bottom side with multiple fiducials and split between PCB1 and PCB2 (in essence, get the machine to place the same thing on "two" mapped PCBs). I map out the fiducial coordinates for PCB1 Mark 1 & 2 and PCB2 Mark 1 & 2. When it goes to "PCB 2" it aligns the fiducials correctly, however it starts placing components on top of PCB1 (where it already placed components).

Any ideas?

Thanks!

[Conjure]

https://www.facebook.com/groups/213033402766980/

We've also made a Facebook group for owners of TVM802B. Discussions, questions, sharing scripts for pick and place export (Diptrace, etc.).

[mcsmarte]

I have been playing around with this TVM802B machine and made several successful builds with it. However, I am still not very satisfied with it. As a couple of other people mentioned here, the vacuum pump system is not really well designed. It puts my components in the trash area on my second nozzle about 50% of the time, while they are perfectly in place. If I disable the use of the pressure sensor, I can place components just fine, so I conclude that the vacuum is actually working

I seem to have an older version of the TVM802B, I can't run the 3.x software that a couple of guys shared, only the 2.xx versions. Apparently the guys in China made a new FW for the machine, and I guess a few new HW parts too! There are a lot of features in the 3.x that I don’t have access to, like the vacuum configuration for each nozzle for example, or aiming at components when setting up the IC pickup place for trays or strip holders.

A lot of people have big plans, making a new software, writing a new FW as well. That's all very good, but humanity will be on Mars before that project is complete!
At this point I have not had my machine taken apart yet, but I am getting there. I read some other guy said it was a mechanical pressure sensor that was either high or low. Is there anyway to "configure" this little HW thing, or maybe replace it with something a little more refined? Maybe somehow be able to adjust at what pressure it thinks the component is not there? The new FW/SW has an upper and lower threshold, that would be nice.

[Darius]

In my TVM802 machine the pressure sensors like this https://www.ebay.com/itm/Sensore-Pressione-MD-PS002-Vacuum-Sensor-Absolute-Pressure-Sensor-For-Arduino/122884435359?epid=1031456898&hash=item1c9c7bad9f:g:4AkAAOSwYwJaRM3M

+op Amp + Comparator. I think adjusting comparator level you can change pressure sensitivity.

[mcsmarte]

I also started taking the machine apart today, but didn't get as far as you to dissect the actual assembly. But yeah, that's what it must look like underneath the board. Did yours break down after looking at enough pressure sensors I concluded that it is indeed the one you posted the link to.
I followed the hoses back to the pump and it seems like one of the pumps actually isn't performing very well, they are both Parker D861-22-01.

I took a few pictures of the board, trying to figure out what is going on, but maybe you have better pictures now that you have the board off the mount? I only measured a few voltages as well and saw that when there a component is on the nozzle pin 7 on both the OP amps go from ~0V to around 1.4V-2V using nozzle 503, and U1 seems to be for Nozzle 1.
The PC SW does not show anything else than if there is pressure or not, but since the new version of the SW has thresholds for the nozzle pickup, I doubt that the OP amps are actually working as a comparator, but instead measuring the sensor, and amplifying the signal to be somewhere between 0-5V so that micro can deal with it? I could be wrong, would need to measure all the pins

But whether it is a comparator sending a digital signal back or the microcontroller is actually doing an ADC measurement or the sensor reading, I guess I can get around this by adding a potentiometer to one of the resistors so I can control the voltage leaving the board.

If you have some good pictures of the little board on both sides, please do post them

[ar__systems]

But whether it is a comparator sending a digital signal back or the microcontroller is actually doing an ADC measurement or the sensor reading, I guess I can get around this by adding a potentiometer to one of the resistors so I can control the voltage leaving the board.

On 802 the comparator sends out a binary signal. Adding POT is possible, but then you need to keep in mind that for different components you might require different thresholds. Say, ceramic caps are usually more leaky, and a pressure level of a well seating cap might the same as that of a mispicked resistor.

[nisma]

Why not use openPnp, there exist a driver for the standard firmware , however i don`t know if it works with latest one.
Further there are some people that have converted it using arduino mega as alternative controller using a protoshield for plugging in the existing connectors and
remaining necessary circuits.

[ar__systems]

Do you use this setup?

[nisma]

No, i had for some time a cmht28 that i have converted to openpnp adding top and bottom camera.
Had written the openpnp driver at that time you'r reverse engeenered the protocol, and helped some people to convert the tvm802 to openpnp using
6axis grbl firmware. OpenPnp is slower, assembly is more accurate and further it is possible to mount nozzle changer (approx 200$).
   

[luiHS]

 
Hi.

Does anyone have experience placing TQFP100 and TQFP144 chips with this machine?

I also have doubts about the limitations in relation to the height of the components, such as aluminum capacitors. For example, if we can place aluminum capacitors with a height of 12 mm.

I recently bought this machine, but after seeing the Neoden4, I think I will replace the TVM802 with a Neoden4, it is more expensive but I can safely place TQFP100/144, although there are limits with the height of the components that can not exceed 10 mm .

regards

[ManuFerHi]

No, i had for some time a cmht28 that i have converted to openpnp adding top and bottom camera.
Had written the openpnp driver at that time you'r reverse engeenered the protocol, and helped some people to convert the tvm802 to openpnp using
6axis grbl firmware. OpenPnp is slower, assembly is more accurate and further it is possible to mount nozzle changer (approx 200$).
 

I want to try openpnp with TVM802B although it gives me a bit of fear, afraid of making mistakes and having the heads or prick hit with something.
Is it possible to control the prick with openpnp?

[ManuFerHi]

Hi.

Does anyone have experience placing TQFP100 and TQFP144 chips with this machine?

I also have doubts about the limitations in relation to the height of the components, such as aluminum capacitors. For example, if we can place aluminum capacitors with a height of 12 mm.

I recently bought this machine, but after seeing the Neoden4, I think I will replace the TVM802 with a Neoden4, it is more expensive but I can safely place TQFP100/144, although there are limits with the height of the components that can not exceed 10 mm .

regards

Hola Luis.
QTF100 yes it fits in the vision of the camera, but TQFP144 no, I at least I have not got it you can only place it without vision. Anyway, even if you use vision, my experience is that all CIs with 0.5mm pitch must visually inspect them and correct their position before the reflow. Sometimes the machine hits but others do not, the small integrated ones place them much better than the big ones.

That is one of the reasons why I want to try openPNP, and a bug that has the original software, when it places two components simultaneously with both nozzles, the first nozzle does not stop sucking until the second does so it stays stuck in the nozzle until the second nozzle places its component. It only happens sometimes but this bug is very annoying.

[luiHS]

.

Thanks Manu.

Even before buying the machine, I was warned by the chinese that the TVM802 did not accurately place these chips with such a fine pitch.

My idea was to position with the machine, and then review and adjust with the microscope.

But shortly after I bought the TVM802, I saw the Neoden4, and I liked it a lot. My idea is to sell the TVM802 to a customer of mine, and buy the Neoden4.

The TVM802 I still have it in the unopened box, I think that as it is, it's going to go for my client, brand new. I do not know if it's worth it, take it out, install it and try it, because if it does not position well the TQFP144, I prefer the Neoden4.

Do you know what limitations the TVM802 has, in terms of the height of the components? This is with respect to the electrolytic capacitors, I have some of 13mm height.

.

[ManuFerHi]

I have been using the machine for 9 months, I have done a lot of assembly, I think I know it well enough.
If the component you will use in feeder left and back, max height will be about 4mm, if you are going to use it on a tray you can reach 6mm forcing a lot, I have tried to put 10mm but can not even get it out of the tape anymore from 6mm you can touch with other components of the PCB.
The software allows you to work at any height, but you encounter physical limitations.

[luiHS]

 
About the height of the component, in Neoden they tell me that they will make me some customized nozzles for the Neoden4, to place components with a height of up to 10 mm. I think they are shorter nozzles.

I recently asked about the nozzles for the TVM802, but these seem to be all standard, do not customize them or make shorter models to solve this limitation.

I do not understand why these machines have these limitations, so that the Z axis could move more, let's say about 2cm, it would be enough to be able to easily place high components such as electrolytic capacitors and some power inductors.

 

[harry4516]

you could place the PCB lower, there is enough space to the body of the machine.
Of course, this only makes sense if you have a lot of big caps or inductors since you will
also need a big nozzle and usually a better VAC pump to hold heavy parts.

I'm using an external VAC pump for a year now, and it works MUCH better then the tiny build in pumps,
even for very small components.

[mcsmarte]

Hey Harry,

It was your videos on Youtube, your reviews on this blog that caused me to put money into this machine too. Your Eagle to TVM converter got me going much faster than I would have been otherwise. Thanks!

I have the pumps hanging out of my machine at the moment, wondering if adding another small pump would make me happy or if I should follow in your footsteps once again. I have a couple of other vacuum pumps that I could put in serial with the existing dual chamber pumps. But I have a feeling you might already have been down this road and concluded that an external pump is the way to go?
One feature I guess you can't really use then, is the pressure? Or are you controlling the vacuum in the external system to be in the "correct" range?

I spend some time trying to ohm through the little board that sits on top of the nozzles and measures the vacuum. The attached schematic is what I found. It might be wrong, but at least this makes sense.
The vacuum itself is setup up as a wheatstone bridge. One side of the bridge is fed into the opamp (blue) which just works as a buffer as the amplification is close to 1, then in the green box the signal gets amplified/compared, the output driver will drive the pulled up 3.3V signal to ground once the opamp signal is high enough. At least something like that.
I don't know exactly how the sensor works, if one side of the bridge is ambient pressure serving as a voltage reference through one opamp, and the other the sensed pressure? But I am guessing that the 10k resistors in the diagram can be made smaller, to amplify more, giving a signal earlier with the weaker pump? Will try this over the weekend

I guess the feeders mechanism could be better, but as a whole the machine works pretty good. I have serious issues with the SW/FW though, it could be so much better, that's really a shame. Therefore I am leaning a bit towards trying to do a conversion to the OpenPNP now after someone commented about the Smoothie boards and swapping out the control boards.
This guy did the hard work it seems, has pictures and wiring of most things. https://github.com/johngrabner/tvm802B_2_smoothie_and_OpenPnP

Reverse engineering the FW as some people started is of course possible, but making a couple of adapter boards with the correct connectors which connected easily to the existing TVM802 connectors and to the smoothie-boards is probably much easier.
Watching a few videos from OpenPNP, I especially like the option to teach the machine about strips with components, using the camera to find the exact center of very small components, that looked very neat. It is very cumbersome to setup to the machine when having multiple small strips of components.

[SMdude]

Hey Harry,

It was your videos on Youtube, your reviews on this blog that caused me to put money into this machine too. Your Eagle to TVM converter got me going much faster than I would have been otherwise. Thanks!

I have the pumps hanging out of my machine at the moment, wondering if adding another small pump would make me happy or if I should follow in your footsteps once again. I have a couple of other vacuum pumps that I could put in serial with the existing dual chamber pumps. But I have a feeling you might already have been down this road and concluded that an external pump is the way to go?
One feature I guess you can't really use then, is the pressure? Or are you controlling the vacuum in the external system to be in the "correct" range?

I spend some time trying to ohm through the little board that sits on top of the nozzles and measures the vacuum. The attached schematic is what I found. It might be wrong, but at least this makes sense.
The vacuum itself is setup up as a wheatstone bridge. One side of the bridge is fed into the opamp (blue) which just works as a buffer as the amplification is close to 1, then in the green box the signal gets amplified/compared, the output driver will drive the pulled up 3.3V signal to ground once the opamp signal is high enough. At least something like that.
I don't know exactly how the sensor works, if one side of the bridge is ambient pressure serving as a voltage reference through one opamp, and the other the sensed pressure? But I am guessing that the 10k resistors in the diagram can be made smaller, to amplify more, giving a signal earlier with the weaker pump? Will try this over the weekend

I guess the feeders mechanism could be better, but as a whole the machine works pretty good. I have serious issues with the SW/FW though, it could be so much better, that's really a shame. Therefore I am leaning a bit towards trying to do a conversion to the OpenPNP now after someone commented about the Smoothie boards and swapping out the control boards.
This guy did the hard work it seems, has pictures and wiring of most things. https://github.com/johngrabner/tvm802B_2_smoothie_and_OpenPnP

Reverse engineering the FW as some people started is of course possible, but making a couple of adapter boards with the correct connectors which connected easily to the existing TVM802 connectors and to the smoothie-boards is probably much easier.
Watching a few videos from OpenPNP, I especially like the option to teach the machine about strips with components, using the camera to find the exact center of very small components, that looked very neat. It is very cumbersome to setup to the machine when having multiple small strips of components.

There was a member here and on the openpnp group that made a driver that sat between openpnp and the tvm. That way you didn't need to change out the control system. I don't know if he ended up making it opensource. Glen English I think was his handle.

Openpnp works pretty good too and depending how proficiant you are with Java can be customised to your needs unlike the provided software.

[mcsmarte]

The other weekend I made 4 boards with about 192 components and I am guessing there were something like 250 discarded 0603 components, mostly capacitors and inductors, but only a few resistors. I kept trying to adjust the pickup point for the stacks, adjusting the prick correction. I eventually started running only with nozzle 1 as the other nozzle would discard almost any components, which were sitting on the nozzle just fine. I figured it was time to spend a couple of hours to fix some of these items:

1. Nozzle 2 pickup place
I also messed around with the nozzle/camera calibration built in functions. Big surprise: it doesn't work very well. Tried many different 503 nozzles, mostly the same offset calculated. When the pickup place for nozzle 1 is perfect, nozzle 2 always picked up the component too far to the right. But since it was mostly a problem on the left stack, the back stack was a bit better. So I tried adjusting the "calibrated" values manually, and what do you know, it started to work much better!

2. Vacuum pressure signal
The nozzle was picking up much more accurate now, but still no vacuum pressure signal. I'm still not sure if the 2nd pump is actually worse, or it is the vacuum sensor that has a problem, but.... I already mapped the vacuum sensor board components and so I tried to change the R10 (earlier attached schematic) from 10k to 8k, thinking that this was the way to go. Now I measured even less voltage on the transistor basis pin, so went with a 13k instead and suddenly it started working a lot better.

3. Vision aligned components off a bit
With the learning from the nozzle 2, started playing around with the camera aligned components. I often had the problem that bigger TQFP parts ended up falling in between the solder pasted pads. Turned out to be the same "calibration" that was the problem. Corrected the location of the nozzle 1 compared to the camera and now it stated putting components so very accurate.

Later that evening I ran 8 of the same boards and had a total of only 6 discarded components. There are more components lost due to problems with the reel tape not being pulled off properly all the time, but it's a big step forward


When I run the machine at very high speed, the thick aluminum plate where the nozzles, camera and prick sit, will randomly hit the screws on the back stack, around B10-B16, when not using the camera correction. I guess the acceleration is so fast and abrupt that the head actually twists a bit and when it hits, the stepper engine loses a few steps. I guess these are open loop-steppers? Took the metal shell of the head piece, to reduce the weight, have not seen it since. May have bent the prick a bit because whenever this happened, the machine got stuck when trying to pick up the next component, but the location was now off.
There is a play in the linear bearing on the right side, one end of it can move somewhat. Maybe that causes the head to hit those screw now and then.
I have seen it try to go all the way too the left after picking up a component. The alarm sounds and the machine stops. Sometimes it totally misplaces a component, but the next ones are ok. That's kind of annoying because I have to visually inspect every single component before the oven. The USB connectors on the machine are very crappy, causes the cameras to fail quite often.

On the positive side, I don't think I have even seen the PC SW crash. V3.x has some nice features that I would like to use, they could even be ported back to V2.x, but China Jim seems to have stopped responding - the 1 year warranty period is over. Stuck at v2.37

[hugeone]

Hi,

I plan to buy one next week but I read only about the problems with this machine.
My question:
Is it worth buying? Should I go ahead or start looking for another one?

Thank you
Peter

[HHaase]

Hi,

I plan to buy one next week but I read only about the problems with this machine.
My question:
Is it worth buying? Should I go ahead or start looking for another one?

Thank you
Peter

The truth is, these are bottom-dollar pick and place machines, so you need to have appropriate expectations. You also need to give yourself time to learn the personality of the machine and work out any issues that affect your needs.   Different boards, different parts, different designs, they all have varying needs.   So it all really depends on what YOU need the machine to do.

But even full-on industrial machines that cost hundreds of thousands of dollars have issues.  Any pick and place machine is going to need engineering time to program and tune, repair, and debug.   Doesn't matter the price, source, or age of the machine.  It's the price of admission to surface mount manufacturing.  If you ever worked on a Fuji CP series machine you'd know exactly how incredibly complex and maintenance intensive a machine can get.  Compared to the old CP's,  which were industry standard workhorses for a decade or better,  this itty bitty Qihe on my desk is a dream to keep going. 

I've done a few hundred boards on mine now,  mostly just rows and rows of good sized passives.  Some QFP32's and such. Probably 15,000 placements so far I'd guess. Working within the limits of the machine I've been pretty happy with it.  It's not without its quirks but for my needs it's a very good fit.

The only area that's caused me any actual issues are in the prick needle.  It was getting hung up just slightly during motion, causing parts to jump from the pocket, particularly when advancing two components at a time.    I took a file and tapered the point, and the problem went away instantly.   Now the only real lingering issue I have right now is that the opto sensor which detects the needle being extended sometimes doesn't work.  The software compensates but it does dramatically slow down the unit.   I don't know if it's overheating or getting dirty just yet.   I haven't been to the point yet where I am ready to pull the electronics off the bottom of the head. 

[hugeone]

The truth is, these are bottom-dollar pick and place machines, so you need to have appropriate expectations.

I actually have. I need to place some tqfp64 parts and many passives. For me the speed is not important at all. I have to prototype boards I design and I hope that this machine will help me with it. Probably even the slowest machine will be faster than me. My eyes are not as good as they used to.

[ar__systems]

I did not have much success with TFQP. The machine will routinely place it at an angle. They are very easy to bridge and hard to clean up the bridges.