Pick & Place MachineTVM920

[thommo]

Hi,

We (fortunately) haven't seen this behaviour (yet).

You say it's always the same direction, but I notice that there appears to be 2 random (opposite direction) rotations during that placement. One, when it is above the camera - anticlockwise, and two, the clockwise rotation at the board placement. It's very odd.

Are you actually 'speaking' with the QiHe guys?

What software version are you running - 1.34?

Pete

[jedas]

After upgrading software, rotation during transportation happens not that often. Qihe responded and gave an instructions to adjust "Down to PCB" height in the "Nozzle" section.

http://svn.fpv.lt/~jedas/tvm920.jpg

I'm running 1.34 version, but I'm not exactly sure I have this window. These options looks unfamiliar to me, even though quite useful. We were not able to configure different heights for back and front back ago. I'll play with it on Monday.

[jedas]

It looks like, that these options are available only on v1.35 beta. Support guy has sent me this version today. I'll test it couple of days. I hope it will solve at least some of the problems.

[thommo]

Hi Jedas,

The link you've attached is, I believe, from a much earlier version of the App.

Since v1.34 [the version you are running] this GUI has been available.

You will find it in v1.35beta under SYS CONFIG > [tab] OTHER. Also check out the [tab] ADVANCE tab.

It doesn't explain that unusual behaviour in v1.34 nonetheless. I have come to realise that, pending on where you are in the App, and how you got there [navigated via ..., then ..., set ..., then - you get the picture], the result has a tendency to 'change'.

In the STRONGEST POSSIBLE terms I employ you to heavily 'request' of the QiHe SW guys [address everything also to Mr Kim - owner] that they provide proper and full documentation for each release of their App, as would any professional developer would.

This documentation must, at a minimum, include the Item was was reported as a Bug, the Fix that was applied, and where to observe the outcome, and any new Features included in the release. New Release docs must also include the Intended Functionality, how to access it, and if if replaces any previous functionality or GUI mods.

A few of us have been 'at' QiHe about this for some time, and I feel only if we maintain this demand are we ever likely to obtain a result.

Without this information, any new release is a Mystery Box at best. Things are sometimes observed to have changed but, are they actually functioning as designed and intended, these and other questions which it raises only make things next to impossible to provide any meaningful feedback to other and to QiHe.

Sorry about the rant - but I feel it's very important - both for them and for us, and I'd really appreciate your effort and support in this request to them.

Hope the 'pointer' helped - Pete

[jedas]

Thanks for the notes. We are testing 1.35 right now. Either way, I think I'll try openpnp too. Looks promising. I've tried to compile Glens' driver for TVM920, it went kind of ok. When machine will be less busy, I'll try to switch over at least for one project.

[thommo]

OpenPnP has great potential however, for the moment, we have given up on it as it is just too many things for too many people. It's the 'please everyone' solution and as a result it becomes FAR more complicated to deal with and comprehend [assuming you are working on mods] than just dealing with what you already have IMO.

It would be sensational of it could be 'filtered' for a particular machine right up front to exclude all irrelevant functions and code - but that's a huge job! and not expected of anyone to undertake.

No doubt Michael's software in Germany/China would resolve a lot of things IF it could be properly directed to a machine like the TVM920 - and tested before release. Hardware in the TVM920 is great [for the money], but their insistence in continuing to protect a terrible piece of software and not opening up the source is holding back this companies sales tremendously IMO. It's difficult to comprehend how they can be so 'blind' to this issue, and at the same time attempt to save pennies by not commissioning a native English speaking person to write a manual and instruction video.

Good luck - I'll be interested to learn how you go with the OpenPnP.

BTW - what's the nature of the component-mix/range you're generally placing with the TVM920, and what's your experience with regard to its accuracy?

Peter

[mikeselectricstuff]

It's difficult to comprehend how they can be so 'blind' to this issue, and at the same time attempt to save pennies by not commissioning a native English speaking person to write a manual and instruction video.

I think it's a cultural thing about not "losing face" by admitting they need external help. It's holding a lot of Chinese businesses back from selling in the West.
Or maybe they sell enough domestically that they don't think it's worth it, though that's somewhat of a chicken-egg situation

[jedas]

So far we made ~300 boards. Usually it's 80-150 parts per board. Placement error rate is quite high. Sometimes 10 boards in a row are fine, but sometimes you need to fix at least one part per board.

Components which works ok:
- Passive 0603.
- Small diodes, transistors
- Mid range packages, like so8, so16.

Problems:
- Sometimes part is lifted at weird angle and vision is unable to help. Bad aligned feeder, or something stuck.
- Bigger LFQP64-100 packages sometimes are placed by some offset, sometimes rotated a little bit, just enough that it would cause short on 2 pins or even more.
- Big parts, like RF modules, which are >40mm or so wont fit into vision angle of view, and it's difficult to place them reliably.  Usually we need to do some calibration every morning to get those parts placed.

The thing we are missing is repeatablity. Someone in the forum has noticed, that nozzles a little bit wobbly. Maybe part rotation servos aren't precise enough. Maybe controlling software is buggy. I guess we are to find out in a months while using it. My attempt at openpnp, should rule out PC software problems, assuming software in the controller is ok.

I've attached picture of the board, which is most complicated we assembled yet. Those big caps are placed by hand. Even they are 10 mm height, we were unable to use p&p on them.

[thommo]

Thanks for the feedback.

I don't think QiHe take their SW development very seriously. From their point of view I reckon they believe the software is 'done' now, and that's it.

There doesn't appear to be any attempt at a managed, development path, let alone for the improvement of existing functionality.

So there is definitely still room for a 3rd party solution, but for us that solution is not going to be OpenPnP at this stage.

With regard to your component placement, are all of your part loaded via their VISION system, or is it dependent on the component? Do you get those erratic results even when you use the EXACT option?

I ask this because we are having an issue at present which is leading us to believe that there are irregularities when using their Vision system. We believe that some of the 'compensation' values [like the offsets, or distance matching values] are not being applied after the 'pick' stage.

It would be interesting to know what you observed when placing the same component in different Vision options - None, Quick, Exact.

Is this with V1.35 beta?

Are you also finding the the ORIGIN wanders from reboot to reboot?
There is definitely a Bug in V1.35 which can cause a crash, after which the Origin needs to be completely reset for all values, FIDs, Feeders, Offsets - the lot! because it selects new 'X&Y' values.

Peter

[mikeselectricstuff]

- Big parts, like RF modules, which are >40mm or so wont fit into vision angle of view, and it's difficult to place them reliably.  Usually we need to do some calibration every morning to get those parts placed.

A machine should be able  to vision oversize parts by moving it around to see each corner.
I'm totally not surprised that a Chinese manufacturer didn't think of this

[mikeselectricstuff]

Thanks for the feedback.

I don't think QiHe take their SW development very seriously. From their point of view I reckon they believe the software is 'done' now, and that's it.

There doesn't appear to be any attempt at a managed, development path, let alone for the improvement of existing functionality.

This is pretty much par for the course for any Chinese product. I have no doubt that if you could see the source it would be a total unstructured mess.
The Chinese simply haven't figured out how to design software yet.

[jedas]

With regard to your component placement, are all of your part loaded via their VISION system, or is it dependent on the component? Do you get those erratic results even when you use the EXACT option?

We try to use vision on every component, which is allowed by size. Also we try to use 'quick' vision, instead of 'exact' to save time. Generally IC's are set for exact, passives for quick. For some components we had to adjust "Vision factor" (edge sensitivity) to get it work. I can't say that we get erratic results with 'exact' vision setting, but sometimes they are erratic for sure. Few posts ago, I've posted a problem, with random component rotation during nozzle down operation. Qikeh admited that some of their customers had this problem and they advised to adjust 'down to pcb' height. We've tried that, but I still have no conclusive results about that, because that problem appeared quite randomly.

By the way vision system was working poorly with the cheap plastic window above the camera. When we've removed that plastic, we've got way better performance. For some small components without contrast edges it changed from totally unworkable to normal.

It looks like that exact setting for vision tries to rotate part, but in quite visible incremental steps. Maybe it fails during that process. Maybe part misses alignment on the nozzle during movement. Even we tried slow transport. This won't matter for most components, but big LQFP packages are really sensitive to this.

We believe that some of the 'compensation' values [like the offsets, or distance matching values] are not being applied after the 'pick' stage.

In my opinion, the only setting is applied after vision is "Offset with down camera center to nozzle". This needs to be calibrated quite often. Especially after changing the nozzle, or different moon phase.

Is this with V1.35 beta?
Are you also finding the the ORIGIN wanders from reboot to reboot?
There is definitely a Bug in V1.35 which can cause a crash, after which the Origin needs to be completely reset for all values, FIDs, Feeders, Offsets - the lot! because it selects new 'X&Y' values.

We are testing v1.35, but only for a couple of days, so it's early to say something. We do "home" calibration after each reboot. I'm not sure if we loose settings after program crashes, but maybe it's easy to backup .par config files?

[qihe_kim]

hi
sorry for puzzle of pirces.
we will updata the price and unify the price.

[m72]

About offsets.

For the best result, when using two fiducials, it is necessary to recalibrate the movement coefficients along the X and Y axes. It is not so important how much they correspond to the real, the main thing is that the "machine mm" along the X axis is equal to the Y axis. From this there is an error along the edges of the PCB outside the axis between fiducials.

The main idea: in the settings you need to specify the axis of rotation of the nozzle, not the tip. The axis position is more stable even you change the nozzle.
With parallel transfer, the displacement of the tip of the nozzle does not matter. The main thing is that the part has been captured.
When rotating, the center of the part must be on the axis of rotation of the nozzle.

To configure nozzles from "down camera".
I put a thick layer of plasticine (about 1-2 mm) on a black block at the height of the PCB. Then from the empty feeder I put the virtual part 8 times changing the angle from 0 to 315 degrees through 45. The center of imprint is the axis of rotation of the nozzle. I use it in the settings.

To configure nozzles from "up camera".
The center of the nozzle does not correspond to its axis of rotation. To check this, just place the nozzle above the "up camera" and in the manual mode rotate 180 degrees. I use the geometric center for the nozzle rotated by 0 and 180 degrees in the settings.

[JPlocher]

(From a longer post over in the overview thread, also in the CHMT thread...)

At the risk of being overly simplified, in the $10k range, the choice seems between the CHMT530P4 and the Qihe TVM920, with the price difference mostly attributable to number of feeders supported by the 920 (56 -vs-30)

Is there anything in the EEVBlog collective mind's experience that would swing the pointer one way or the other, if feeder quantity wasn't a factor?

  John

[thommo]

Hi JP,
We own the 920 and I haven't looked at Charmhigh's product for some time now.

If they are still not doing 56+ feeders, then the 920 is a no-brainer. Those feeders fill up at an alarming rate, and because they are so inexpensive, we now have a dedicated feeder for each of our components.

Initially, the reason we didn't consider the Charmhigh was because you were limited to a machine-mounted screen for the GUI. Having to stand/sit right next to the machine and use that tiny display would have driven me mad!

Also, it appeared that the 920's X&Y axis were more rugged that the Charmhigh model.

Either way, you won't know yourself once you bring this in-house.

One very important factor that seems to be given little consideration is what to do once the board is loaded - eg the Reflow oven.

Give this some serious consideration because it will have a major effect of your in-house PnP 'pleasure rating' I reckon.

Thommo

[dtf]

is anyone here in touch with QiHE? have they closed shop?

[thommo]

yeah - we are

why do you say that?

[dtf]

Website is down, several products on aliexpress are listed as unavailable, none of their Skype contacts are ever online or have responded in several months...

Looking for software updates on specific vision issues, any news?

[thommo]

Well, I know some people have left the company recently, so that may explain the Skype situation.

QiHe are not really a 'service or after-sales' driven company unfortunately, and we too are finding it very difficult to get reasonable responses to reasonable questions, in a reasonable time frame of late.

Interesting that a number of alternative, very promising looking machines have been made available lately. We are considering another machine ourselves and are questioning if QiHe is still worth the pain.

This machine in particular caught our eye.

http://www.smthw.com/Product-detail-id-546151.html

6 Heads
Ball Screw drive on X axis
64 Slots
7 Cameras - 6 simultaneous, and a separate one for large components
PCB Conveyor
Stand/Cabinet
USD$11,400

Website is down, several products on aliexpress are listed as unavailable, none of their Skype contacts are ever online or have responded in several months...

Looking for software updates on specific vision issues, any news?

[dpenev]

Hi Thommo,

http://www.smthw.com/Product-detail-id-546151.html

6 Heads
Ball Screw drive on X axis
64 Slots
7 Cameras - 6 simultaneous, and a separate one for large components
PCB Conveyor
Stand/Cabinet
USD$11,400

Wow, this machine looks like being in a different league.
Is the ball screw version they call HW-T6-64F ?
Do you have some trusted source of information about the machine
outside of the manufacturer? 
Are the feeders and the compressor included in the price you have mentioned?
Are they shipping it out of China?

[thommo]

No further info at this stage - sorry, but if you find anything please post it for everyone's benefit

Feeders & Factory Air - almost never included in machine cost

Yes - it is definitely a Step Up and all in the right direction too

Not confirmed, but can just about guarantee they are shipping out of CN, as that is where this machine is manufactured

[thommo]

DTF,

It seems that I'm also getting the 'cold [QiHe] shoulder' treatment at the moment too!

It's such a shame that Chinese companies are so ignorant of the basics of business, and fundamental expectations, of the Rest of the World [at least the Western bit, and goodness help you if you're European].

It wouldn't be so bad if there wasn't such an underlying arrogance about it, but i guess that's what comes from decades of undervaluing the Human element - life's still a cheap commodity - at least that's how it seems to be in CN.

It's just a shame that it's not SO CHEAP that they can be bothered to identify an English-educated Tech Support Team, and provide them with sufficient resources to work properly with!

[sam512bb]

Well, I know some people have left the company recently, so that may explain the Skype situation.

QiHe are not really a 'service or after-sales' driven company unfortunately, and we too are finding it very difficult to get reasonable responses to reasonable questions, in a reasonable time frame of late.

Interesting that a number of alternative, very promising looking machines have been made available lately. We are considering another machine ourselves and are questioning if QiHe is still worth the pain.

This machine in particular caught our eye.

http://www.smthw.com/Product-detail-id-546151.html

6 Heads
Ball Screw drive on X axis
64 Slots
7 Cameras - 6 simultaneous, and a separate one for large components
PCB Conveyor
Stand/Cabinet
USD$11,400

I found this one on Aliexpress ($11200 USD including shipping)... and it sure looks similar and/or could be the same one that you referenced with better pictures and English info:

https://www.aliexpress.com/item/High-Stability-Visual-Position-Placement-Machine-Pick-Place-Machine-with-64-Feeders-and-Servo-Motor-for/32821254406.html?spm=2114.search0104.3.150.5I2wKD&ws_ab_test=searchweb0_0,searchweb201602_4_10152_10065_5000015_10151_10344_10068_10130_10345_10342_10547_10343_10340_10341_10548_5130015_10541_10084_10083_10139_10307_10539_5080015_10312_10059_10313_10314_10534_100031_10604_10603_10103_10605_5060015_10596_10142_10107,searchweb201603_14,ppcSwitch_7&algo_expid=37202cac-c903-4f8b-927d-d77ce772b8b3-21&algo_pvid=37202cac-c903-4f8b-927d-d77ce772b8b3&rmStoreLevelAB=3

Also a smaller version too ($7400 USD including shipping):

https://www.aliexpress.com/item/High-Technology-Pick-Place-Machine-SMT460-With-Conveyor-0201-BGA-Led-SMD-T-Industrial-Equipments/32832988812.html?spm=2114.search0104.3.163.5I2wKD&ws_ab_test=searchweb0_0,searchweb201602_4_10152_10065_5000015_10151_10344_10068_10130_10345_10342_10547_10343_10340_10341_10548_5130015_10541_10084_10083_10139_10307_10539_5080015_10312_10059_10313_10314_10534_100031_10604_10603_10103_10605_5060015_10596_10142_10107,searchweb201603_14,ppcSwitch_7&algo_expid=37202cac-c903-4f8b-927d-d77ce772b8b3-23&algo_pvid=37202cac-c903-4f8b-927d-d77ce772b8b3&rmStoreLevelAB=3

Cheers,

Sam

[dpenev]

Hi Sam,

Note that the SMT660 from your first link is the machine with the belt drive on the X axis.
There seems to be also version with ball screw on X axis which is supposed to be more accurate.