Pick and Place Machine TVM802A / TVM802B

[ar__systems]

It is useless. The s/w is crap and should be written from scratch.

[thommo]

Guys,

I look at it this way.

You spend peanuts and you'll certainly get monkeys - there's no two ways about it. These are cheap-arse machines, and when you see the level of work that's gone into the 920 [can't speak for other models], it's hard to imagine how these companies even survive.

I did my research and the 802 doesn't suit my business case. I've invested a fair amount of time [with a translator] speaking with these guys, and I believe they really want to help, and when they've understood the issues I've presented to them, so far they've responded brilliantly - no complaints. I believe the 920 has a lot of potential and is the only one of the bunch I'd consider investing in for this use-space. The main reason being that they have opted to design around industry standard 3rd party Yamaha component feeders.

But another thing that has also struck me now that I read this complaints - and that is the dependency on 3rd party PC and OS [which the 802 models appear to have]. That's very hard work for anyone to maintain - just ask those that have tried.

Again, the TVM920 is different, and although it doesn't have a high-end commercial MOBO, it does have it's own MiniITX hidden away inside the machine, so what a user experiences is 'more likely' what was experienced in the factory by QiHe R&D guys [so that's another variable out the window].

It sounds to me that many of you are not really concentrating on the 'direct to QiHe' path, instead relying on others [from this forum and elsewhere] to solve the various issues. So I'd really encourage you to be patient and contact them 'directly' - otherwise they cannot learn or experience the issues that are being discovered.

I am happy to share my experiences in regard to the 920 as they occur, but I do not have the time or interest to extend that further - say into the 802 space. I hope you can understand the practicality of what I am saying here. It's not that I don't want to help, rather I need to 'stick to my knitting'. I have provided my links and contacts at QiHe in posts on the TVM920 forum for those wishing to delve.

The answer may just be that, for many of you, the 802 provided sufficient 'visibility' into the potential of handling PnP requirements 'in-house', and now it's time to move up to the 920 - something I'll know more about personally, and soon I hope!

Good luck guys - Peter

[tmf]

So I am live online with them now trying to figure it out, china never likes letting secrets out :-)

RichardS

[SPY GUY]

My TVM802B came with Sucface V2.23Beta2 on the disc, which was odd, as the pics in the basic printed user guide that came with it showed V2.25.
That and the name typo set the expected standard, which has been maintained 

I had missing line of manual move mode buttons that included the Pump button, which was actually there just but not shown due to a lack of my screens resolution.

I went through a V2.27 which addded X and Y scaling that was initially set to -0001 causing the No home tmf is probably suffering with, just a quick jerk in the wrong direction, easily fixed by setting both to 1.

I have since had two V2.28's, the latest to add fields to change the discard co ordinates, which does that but has added other errors.
I will drop them on one of our servers if anyone needs this.

re thommos (peters) Peanuts and monkeys post
I fully accept and appreciate the low cost, built to a budget, limited capability and  entry level aspect of ths equipment.

But at it's low price, it is good starting point, that attracts DIY buyers, most able and willing to contribute to a forum and thread such as this, which is a huge benefit all round, and I certainly did not purchase this without any expectation of not having to do some work to customise it and make suitible upgrades, based on this thread and it's many inspirational contributors alone, to who I give huge thanks 

I just tried out my tube feeder option this weekend and sure enough it does not work right out of the box. So whats new   

With a seemingly endless chuckle, I discovered that the parts are too low to be picked up by the nozzles at even their lowest travel point.

Just another interesting challenge, and an opportunity to get inventive and creative.
And also another opportunity to reciprocally post and inpire and help other owners out.

[tmf]

So far 2.27 is working well for me in trials right now..... the scaling helped tonnes on the 300x300mm pcb..... just placed a SSOP24 part in a far coordinate and its almost dead on.....

RichardS

[harry4516]

I have sent an email to QiHE,
they resonded immediately and sent me 2.29

I will test it tomorrow and keep you informed.

Harry

[huebsch]

Folks!
I have got access to TVM 802A and B SW version 2.27. The SW seems to be the same, the only difference is that the B offers more component sources which the A does not have physically. I did about 50 boards with the "B" SW on my TVM802A. It used the upward camera correctly, as you might know the position of the camera is different on both models. So the SW recognices which HW is attached.

This morning I got 2.29 bits per mail directly from Gera. My mail scanners tagged it as "dangerous" but all virus scanners did not complain so I stared it. I just started the SW and played around a few minutes. The x/x Axis Scale was set to 0,001 so the machine did not even home. Already in the 2.27 SW that option was available and I have configured it before. We had that reported on a previous posting already. As I was aware that this was added and I had to readjust it already on the 2.27 SW. I immediately checked the configuration screen (last tab on the SysConfig). I have set both values to 1 and will check calibration. I expect if I have to adjust those values to anything different of 1 I need to recalibrate all other machine values. For now I positioned the head to the camera and strip positions which I have stored on my PC from previous SW revisions. With 1/1 values on the scaling I find the same positions again as before. But I need to check if the stepping is precise on large x/y values.

I have attached a screen shot of the new configuration option.

Finally I asked Gera if I'm allowed to share the bits publicly, up to now I have no response. Might be a time zone issue, I'm sitting in Europe, so it might be too late in China. I plan to offer it on one of my servers of if we get up a WIKI or something I will post it there.
rgds
Arnold

[ar__systems]

Technically the s/w does not need XY scaling option to provide perfect placement. It only complicates setup and adjustment process. All of the positions of the machine (like feeder's positions) are set in absolute coordinates even if they are slightly of. So that's fine. Once the PCB is configured, the exact scaling within the space of the PCB is established by means of fiudicials. 

[huebsch]

Technically the s/w does not need XY scaling option to provide perfect placement. It only complicates setup and adjustment process. All of the positions of the machine (like feeder's positions) are set in absolute coordinates even if they are slightly of. So that's fine. Once the PCB is configured, the exact scaling within the space of the PCB is established by means of fiudicials.

Hello!
You are completely right BUT:
on my machine when I run the head to 0/0 and next to 300/0 I find the head misaligned, less than a millimeter. So If I had a small component far at the right It may be an advantage to change the stuff. I'm only doing short boards under 10cm so I decided to ignore that little error. So the possibility to calibrate the movement itself might help in special situations.
-AH-

[ar__systems]

The math in the software is a bit screwed up I admit.... For example, on a long board (20cm in X direction) I've found that without vision the further right I go, the more off the components get in X axis. So I modified my export script to fix the "scaling error" and got a perfect placement. WITHOUT VISION. Now, as soon I enabled vision for these components, I got X errors again. I can't understand it...

All I'm saying they introduced XY scaling to let people workaround some other problems in the sw instead of fixing them

[huebsch]

Hi ar__systems!
I agree with you, I wonder why the errors happen. Driving steppers correctly does not allow to have those misalignments. Nevertheless it's still a fine machine and saves me a lot of time!
-AH-

[tmf]

All I'm saying they introduced XY scaling to let people workaround some other problems in the sw instead of fixing them

Bingo my thoughts exactly, however since I used this "feature" all my PCB's are on track now ....

However read my next post :-)

RichardS

[tmf]

OK so I just ran my first real pcb, its 300x300mm (12x12") with 144 leds, of which 140 were perfectly placed, the other 3 were 25mm or 1: off at least and one was rotated.

I used vision... and when these error occured the image on the screen left was a part that was not centered..... so I think there is some form of error from the camera, not sure why yet.... it was board #1, but I will watch it much more closely next time and have my finger on the pause button!

RichardS

[sergeil]

Update from the TVM802B machine in Zurich:

Been using it extensively. 0402, QFN, etc components, you name it. Placed a total of around 40-60 PCBs by now, component count 200-300 per pcb (we use the machine to make test adapters for QCing and calibrating the Fotokite Phi. This lets us have a 3-day revision time on these adapters including 2-day PCB external manufacturing)

The machine has been working great except recently it has developed 2 ailments:
 1) one of the nozzles places randomly doesn't enable the vacuum anymore. It somehow fixed itself. Not sure yet if it's mechanical, software, or electrical problem
 2) The machine has a certain corner where it makes a horrible grinding "step" sound and I've noticed completely messes up pick-up. It seems like a mechanical problem but I haven't found anything loose or any obvious signs of wear. The nozzle head seems to become "non-rigid", and dips down, causing the feeder pin to get stuck consistently in the leftmost back pickup position. It can be "manually" made to work by applying corrective force to the head, but something is definitely not right.

Can anyone comment on experiencing something similar? A bit puzzled here since again, no obvious wear&tear marks or loose fasteners -- though we haven't yet fully disassembled the machine to investigate.

Another fun note: I noticed that if you get PCB's with HASL/immersion tin surface finish, we've had some so reflective as to confuse the fiducial camera (it reflects itself and the fiducial no longer stands out). Turns out there is a simple fix -- apply a bit of clear tape over the fiduacial to make it less reflective!

The software is shit but workable once you know the bugs. Haven't had a chance to use the more recent versions.

[SPY GUY]

sergeil:

Sure sounds like you have some physical issues with your particular TVM machine.
It seems they all have some individual personality 

Thre is a contrast adjustment on the down facing camera that may solve your HASL problem.

tmf:

so I think there is some form of error from the camera, not sure why yet

Harry:

During "accurate" adjustment I can see the red line jumping. Looks like the contrast with these type of cases is too bad.
I never had this problem with TQFP where the pins (with good contrast) define the edges.

Having quickly gone through much of the step learning curve required of any owner, I am now also at this camera error / red line jumping point too, in my case on a small BGA device. It seems like the Illumination amount, or angle, Camera or software lacks sufficient contrast provision or detection between the pads and case edges.

I live in hope that I can simply and physically improve this to the required degree somehow ;-)

[uncle_bob]

......
 2) The machine has a certain corner where it makes a horrible grinding "step" sound and I've noticed completely messes up pick-up. It seems like a mechanical problem but I haven't found anything loose or any obvious signs of wear. The nozzle head seems to become "non-rigid", and dips down, causing the feeder pin to get stuck consistently in the leftmost back pickup position. It can be "manually" made to work by applying corrective force to the head, but something is definitely not right.

Can anyone comment on experiencing something similar? A bit puzzled here since again, no obvious wear&tear marks or loose fasteners -- though we haven't yet fully disassembled the machine to investigate.

........

Hi

More from the global belt driven 3D world rather than specific to the 802, general troubleshooting:

1) Binding could be the issue. Everything needs to slide easily on the rails with the motors powered off. Misalignment of the rail sets (one rail is out) can cause things to bind at one end or the other.

2) Dragging of cables can be the problem. It's the same basic thing, only at the extreme of the cable distance. One link of the drag chain gets messed up and the cable no longer extends nicely.

3) Never rule out good old dirt. A bit of lube (not much) on a rag will sometimes pull a lot of gunk off of an apparently clean rail.

4) Loose end stops switches will allow the travel to go to far. Rather than stoping before it gets to the end of the rail, it crashes and makes noise.

That all assumes the machine is working correctly otherwise. If there are multiple issues, that adds a few items to the short list of things to check ....

Have Fun

Bob

[harry4516]

after 6 months and hundreds of boards with the TVM802B I came to a conclusion:
the 802 is as it is, it helps me a lot and saves a lot of time.
It has some issues, like the inefficient camera, but I can live with it.
Manual correcting the placement of a few parts before soldering is not a problem.
I will never use it for BGAs (by the way, I would not trust my soldering oven or my soldering knowledge in case of BGAs).

I will continue working with the 802 and in parallel I will watch the progress
of the new smallsmt machine. If this machine is available I will possibly make a new decision.
Until then I am happy with the TVM802.

Harry

[mrpackethead]

You spend peanuts and you'll certainly get monkeys - there's no two ways about it. These are cheap-arse machines, and when you see the level of work that's gone into the 920 [can't speak for other models], it's hard to imagine how these companies even survive.

Hope you are right with that 920.   Its still ( reletively ) a cheap machine..

[thommo]

Yeah - it's definitely still in the 'cheap' machine space Mr Packethead, maybe not at the bottom of the list, but still 'relatively inexpensive' for sure.

So, following my own statement, I don't expect to have it deliver full SMT Production Line capacity or longevity. I understand what I'm buying, and that's perfectly OK with me.

I believe that they've sold a reasonable qty of them now, so looks like they are 'taking off' which will be a nice position and comfort to me too.

You spend peanuts and you'll certainly get monkeys - there's no two ways about it. These are cheap-arse machines, and when you see the level of work that's gone into the 920 [can't speak for other models], it's hard to imagine how these companies even survive.

Hope you are right with that 920.   Its still ( reletively ) a cheap machine..

[tmf]

One word "BOUNTY"....

There is very nice software being made from the Open PNP guys but no support yet for 820A/B

I have a TVM802A right now and the software and vision is garbage, do you think a bunch of us with 802A/B could put some money into a "bounty" so we might find someone to complete this code for us?? We could use https://www.bountysource.com/

Richard.

[vonnieda]

One word "BOUNTY"....

There is very nice software being made from the Open PNP guys but no support yet for 820A/B

I have a TVM802A right now and the software and vision is garbage, do you think a bunch of us with 802A/B could put some money into a "bounty" so we might find someone to complete this code for us?? We could use https://www.bountysource.com/

Richard.

Jason from OpenPnP here. To be honest, I don't know if you'll be happy with OpenPnP either. It lacks some of the things that I see folks on here talking about being important for production use. What I would recommend is downloading it and taking it for a test drive. It comes with a fully functional simulator / demo. There are instructions for running through a job at https://github.com/openpnp/openpnp/wiki/Quick-Start

If you check it out and you feel like OpenPnP would be an improvement on what you have now, let me know, and we can discuss prioritizing support for the TVM. The big issue will be getting me access to a machine. I can knock out the basic driver pretty easily without a machine but making everything work correctly will really require some face to face time with it. I live in Kansas City now, so if there is anyone within an hour or two of there maybe we could work something out?

Jason

[tmf]

Nice fast response... always a good sign

When you say missing production features, what might some of these be, I think for my case use its not important....

I have run the demo, looks good.

Wish I was closer to you... I am Ottawa Ontario Canada... but if a developer were close I would be more than willing to work with them.

From what I can see the machine mechanically is good, its the software that is very suspect, especially when they need a XY scale factor fudge! mine are like 0.9985 0.9995, ouch!

Richard.

[vonnieda]

Nice fast response... always a good sign

When you say missing production features, what might some of these be, I think for my case use its not important....

I have run the demo, looks good.

Wish I was closer to you... I am Ottawa Ontario Canada... but if a developer were close I would be more than willing to work with them.

From what I can see the machine mechanically is good, its the software that is very suspect, especially when they need a XY scale factor fudge! mine are like 0.9985 0.9995, ouch!

Richard.

A few things off the top of my head:

* Not very easy to run a partial job, or continue an aborted job.
* Doesn't show status of each placement during the job run. This is tied to the above and is a long running TODO. Just haven't quite gotten to it.
* Does not use fiducial board scaling. I have not really had people report this being an issue, but it might be for people who are running a lot of boards. By scaling I mean that OpenPnP uses the fiducial to find the position and rotation of the board, but it does not adjust the placements based on the distance between fiducials. That would help handle boards that are not made perfectly, so that could be a concern for some users.
* Bottom vision is pretty new and may be somewhat confusing to new users. UI needs work.

In general, OpenPnP is pick and place software imagined by someone who has never touched a real pick and place. That's me I saw the need for the software, no one else was writing it and now here we are. It does work, and there are a lot of people using it for production work, but I feel like if you are coming from a commercial pick and place expectation there may be some disappointment. I guess the one major benefit, though, is that it's open source, so if there's something that really bugs you there is at least the possibility that it can be fixed.

Anyway, I will monitor this thread and the other TVM one and gauge interest. If it seems viable, I suppose the bounty funds could be used to fly me to someone who has a machine

Also, as for working with a developer nearby - I am completely open to that. If we can find someone who knows Java and is interested in the task I'd be more than happy to guide them remotely.

Jason

[protoneer]

One word "BOUNTY"....

There is very nice software being made from the Open PNP guys but no support yet for 820A/B

I have a TVM802A right now and the software and vision is garbage, do you think a bunch of us with 802A/B could put some money into a "bounty" so we might find someone to complete this code for us?? We could use https://www.bountysource.com/

Richard.

Jason from OpenPnP here. To be honest, I don't know if you'll be happy with OpenPnP either. It lacks some of the things that I see folks on here talking about being important for production use. What I would recommend is downloading it and taking it for a test drive. It comes with a fully functional simulator / demo. There are instructions for running through a job at https://github.com/openpnp/openpnp/wiki/Quick-Start

If you check it out and you feel like OpenPnP would be an improvement on what you have now, let me know, and we can discuss prioritizing support for the TVM. The big issue will be getting me access to a machine. I can knock out the basic driver pretty easily without a machine but making everything work correctly will really require some face to face time with it. I live in Kansas City now, so if there is anyone within an hour or two of there maybe we could work something out?

Jason

Very keen on putting money into a bounty but my backup plan is to have a look at this over Christmas when I have some time off.

I agree that access to the machine will be critical.

@vonnieda , I think having openPNP as an option will still add value. The current software for the machine is working but there will always be more features requested.

Jason, do you have an Interface document for people wanting to write drivers for openPNP?   

[vonnieda]

One word "BOUNTY"....

There is very nice software being made from the Open PNP guys but no support yet for 820A/B

I have a TVM802A right now and the software and vision is garbage, do you think a bunch of us with 802A/B could put some money into a "bounty" so we might find someone to complete this code for us?? We could use https://www.bountysource.com/

Richard.

Jason from OpenPnP here. To be honest, I don't know if you'll be happy with OpenPnP either. It lacks some of the things that I see folks on here talking about being important for production use. What I would recommend is downloading it and taking it for a test drive. It comes with a fully functional simulator / demo. There are instructions for running through a job at https://github.com/openpnp/openpnp/wiki/Quick-Start

If you check it out and you feel like OpenPnP would be an improvement on what you have now, let me know, and we can discuss prioritizing support for the TVM. The big issue will be getting me access to a machine. I can knock out the basic driver pretty easily without a machine but making everything work correctly will really require some face to face time with it. I live in Kansas City now, so if there is anyone within an hour or two of there maybe we could work something out?

Jason

Very keen on putting money into a bounty but my back plan is to have a look at this over Christmas when I have some time off.

I agree that access to the machine will be critical.

@vonnieda , I think having openPNP as an option will still add value. The current software for the machine is working but there will always be more features requested.

Jason, do you have an Interface document for people wanting to write drivers for openPNP?   

No specific document, but there are a few resources that will help:

The API documentation for the driver interface: http://openpnp.github.io/openpnp/develop/org/openpnp/machine/reference/ReferenceDriver.html

The source code for the simplest working driver: https://github.com/openpnp/openpnp/blob/develop/src/main/java/org/openpnp/machine/reference/driver/GrblDriver.java

Source code for all the drivers: https://github.com/openpnp/openpnp/tree/develop/src/main/java/org/openpnp/machine/reference/driver

I *think* someone has taken a stab at writing a driver. I'll need to go digging, though, as I'm coming up empty right now. There are some comments from Cri S in this thread on the OpenPnP mailing list: https://groups.google.com/forum/?utm_medium=email&utm_source=footer#!msg/openpnp/ZW4fu7Gy5vs/gxpi9WrJFwAJ

Another thing to note, since the machine is entirely TCP/IP based (correct me if I'm wrong there) I could potentially work with the machine remotely with someone's on site assistance. We could open a port to the machine and I can just connect to it from here.

Jason