Author Topic: Pick and Place Machine TVM802A / TVM802B (Read 406408 times)

protoneer · « **Reply #150 on:** March 18, 2016, 07:13:13 am »

Haha... I was wondering about adding a second boost pump. But I take it the one is vacuum and the other is blowing.

This might be an option for a boost pump or replacement... https://www.sparkfun.com/products/10398

ServoKit · « **Reply #151 on:** March 18, 2016, 07:40:10 am »

Ok, the first pic from the far side of the moon.

From left to right:

USB I/F board (outside frame to the left)
LAN I/F & controller board, white ribbon cable goes to front panel of machine
Driver board (attached to underside of work plate)
Blower pump (black plastic cuboid)
Two vacuum pumps (blue, mfg. Hargraves)
Power supply (outside frame to the right)

The wiring looks reasonably clean, plenty of air in the box.

Addendum: Those pumps seem to be the Hargraves BTC-IIs double head; not cheap, price starts at $240 a piece.

Addendum 2: The driver board is double-sided, the actual driver ICs sit on top, using the work plate as a heat sink. The board has seven drivers: 1 x, 1 y, 2 feeder, 2 nozzle rotation, 1 nozzle up/down. Apparently no drivers in the head. The drivers seem fairly beefy, definitely not A4988-kind of stuff but not M542-level either. The board also supplies the pumps.

Addendum 3: Both cameras are connected directly to the USB I/F board, apparently the controller only switches the camera illumination.

Regards, Axel

ServoKit · « **Reply #152 on:** March 18, 2016, 09:00:17 am »

The controller.

MCU is STM32, Cortex ARM 3, smaller chip is TI DP83848, Ethernet IF.

Regards, Axel

ServoKit · « **Reply #153 on:** March 18, 2016, 09:35:09 am »

Main point of the peek inside the machine was to establish a safe area where I can drill mounting holes to fix up stuff. The power supply, the two vacuum pumps, the blow pump, and the motor driver board are attached to the underside of the work plate. That leaves the reddish area in the photo. Trays outside this area will have to be fixed with double-sided tape etc.

Regards, Axel

ServoKit · « **Reply #154 on:** March 18, 2016, 10:32:14 am »

More pics.

#1: Vacuum pumps, seen from the right side of the machine; not sure if it's one pump per nozzle or the two are in series (or whatever that's called with pumps), will have to examine head more closely.
#2: USB I/F and controller at rear of machine; white ribbon cable goes to buttons on front panel, the left white connector is camera illumination and solenoids, the right white connector is limit switches, colored ribbon cable goes to driver, not visible is Ethernet jack behind white ribbon cable.
#3: Another view of the pumps, yellow/green is power for pumps

Regards, Axel

ar__systems · « **Reply #155 on:** March 18, 2016, 01:58:35 pm »

This only reinforces my initial impression of relatively solid build. I mean they use quality parts and components, but somehow the result is still underwhelming...

The boost pump from sparkfun is even weaker than the one they use.

I intend to attach something like this:

http://www.globalindustrial.ca/p/hvac/service/chemicals/mastercool-90066-2v-110-6-cfm-vacuum-pump-two-stage?infoParam.campaignId=T9F&gclid=CIy86MapyssCFQwPaQodPCkH-w

protoneer · « **Reply #156 on:** March 18, 2016, 06:48:51 pm »

Quote from: ar__systems on March 18, 2016, 01:58:35 pm

I intend to attach something like this:

http://www.globalindustrial.ca/p/hvac/service/chemicals/mastercool-90066-2v-110-6-cfm-vacuum-pump-two-stage?infoParam.campaignId=T9F&gclid=CIy86MapyssCFQwPaQodPCkH-w

Correct me if I am wrong but those are very loud to run? (Same as a normal compressor)

How about having an inline air-tank but for negative pressure...

ar__systems · « **Reply #157 on:** March 18, 2016, 07:09:09 pm »

Yes, kind of loud, but nowhere near as loud as my run-of-the-mill compressor. Still I keep it in another room. I run a small tube to the room where I work.

In-line tank - not sure. Might work too.

ServoKit · « **Reply #158 on:** March 19, 2016, 07:54:06 am »

Update on the driver board: Someone at mikrocontroller.net remarked that it looks a lot like the one in the Neoden TM220/240.

Regards, Axel

ServoKit · « **Reply #159 on:** March 19, 2016, 01:24:51 pm »

I've added mounting holes to the base plate for attaching trays and the like.

Step-by-step: http://servokit.com/blog/2016/03/19/adding-mounting-holes-to-the-tvm802a-base-plate/

Regards, Axel

protoneer · « **Reply #160 on:** March 20, 2016, 04:48:09 am »

Axel, do you have some photos of the reel feeders and how they work?

You also mentioned that the controller board looked the same as Neoden's, what model was that? Wondering because the lower priced models seem to not have vision and ethernet.

ServoKit · « **Reply #161 on:** March 20, 2016, 05:40:12 am »

@protoneer: Motor driver board is apparently like TM220/240, not controller board. Also, the controller on the TVM doesn't handle the vision, the cameras are directly connected to the PC. Re. feeder, it's really very simple: all the motors do is peeling the cover tape, description of process is on my blog post about the head:

http://servokit.com/blog/2016/03/14/a-closer-look-at-the-pick-and-place-head-of-the-tvm802a/

Regards, Axel

protoneer · « **Reply #162 on:** March 20, 2016, 05:47:44 am »

Thanks.. I have studied all your blog posts..

But I am not sure about the motor that turns the plastic cover collector wheels. Does it have some clutch system for each reel? Tension on the wheels seem to be hard to control...

ServoKit · « **Reply #163 on:** March 20, 2016, 06:01:18 am »

There are actually two motors turning the rod with the wheels that wind up the cover tape. The wheels themselves rotate independently from each other but they don't rotate freely, you have to apply some force to do it, it's a bit like an overload clutch. Yesterday, while playing with manual mode, I accidentally engaged the stripper motors without advancing the tape and the respective wheel didn't move. How well this works when more than one reel is attached remains to be seen. Have to wait for my PCBs.

If there are problems like ar_systems mentioned they might come from the fact that the load / tension on the rod is distributed asymmetrical, e.g. when only using eight feeders on the lower end. It might help to only use every other feeder position or so in such a case, to distribute the load across the rod.

Regards, Axel

ar__systems · « **Reply #164 on:** March 20, 2016, 03:48:27 pm »

Today I finally run a test with an external vac pump. I'd like to say the result is better, however with many factors at play it is difficult to say with certainty that it was due to the better vacuum.

I first ran a base line test with the built in pump populating 320 components at 80% speed and then another 320 at 100%. 100% actually came out more decent. In the 80% test 3 were placed on the sides and a couple were rotated almost 90 degrees. 100% did not have any such defects.

So far I run only 160 parts with external vac at 80 speed. All parts were placed ok.

Now, I noticed something else. There is a systemic error in component placement. I mean, consistency is quite good, definitely good enough to place 0603 without help of a camera, but the coordinates are messed up.

I ran a total of 5 test boards, each with 160 components split into 8 sections. The attached picture shows section 1 and 2 of the board I placed today. As you can see section 1 is all shifted down. And it is like that on all 5 boards. But not all of the components are shifted down. Section 7 (not shown) is shifted up. And other sections all shifted a bit in different directions. Looks like entire board was slightly rotated.

So, just like with Neoden, looks like the software is not capable of doing basic math correctly

ServoKit · « **Reply #165 on:** March 20, 2016, 04:21:55 pm »

Could be a resonance effect. I have this at 30%, the machine shakes a bit and the motors make funny noises, at 40% it's gone. The table also plays a role here.

Best way to test coordinates: Create an array of components (right click parts list for context menu, it's in there somewhere) and place parts 10mm apart in x, with y constant. Then put .5mm divided ruler (or printout) along x and let it run.

Also: Does it make a difference which feeder bank you use?

Regards, Axel

ar__systems · « **Reply #166 on:** March 20, 2016, 04:35:41 pm »

No. Original tests with built in pump were done with the left banks, and today I used north banks. The shift is the same.

I have the table bolted to a wall, so it is quite stable.

Your test is only good to gauge consistency, and I already know it is quite good. It screws up coordinates when matching fiducials locations to the placement data.

ServoKit · « **Reply #167 on:** March 22, 2016, 10:27:54 am »

Created a 3D printed rail to attach tapes to the work area and have the nozzles pick parts directly from the tape. Preferable to taking parts out of the tape and putting them into a tray, I think.

Write-up: http://servokit.com/blog/2016/03/22/adding-a-tape-holder-to-the-tvm802a/

Regards, Axel

rwb · « **Reply #168 on:** March 22, 2016, 03:27:59 pm »

That's pretty cool Axel.

I had been thinking about how long it would take to load a custom milled CNC tray but your way allows you to pre-program the part spacing and then just pull load the part strip and pull back the clear tape which has to save a good amount of time vs hand placing parts.

I think this is a excellent way to go.

You just have to make sure the tapes are loaded properly spacing wise so the head picks the parts at the right spot if the parts are small. No so much a problem with larger parts since you have vison.

Can't wait to see how your first runs go and if any of the issues AR_Systems has experienced also happen to you.

Ichan · « **Reply #169 on:** March 22, 2016, 06:02:54 pm »

Hi Axel,

No need to lathe some brass for the reel locating pin, just use 1.5mm ferrule - mine is blue in color.

You can not use only two pin on the edge as the tape actually shrink or stretched in use, you will need several locating pins along the tape.

Also with this kind of feeder, carefully set the pick height of the nozzle - if the nozzle push the part too hard then many other neighboring parts will jumps over their pocket, like fleas

.

-ichan

ServoKit · « **Reply #170 on:** March 22, 2016, 06:32:02 pm »

Thanks, ichan, good idea - I like to work on the lathe, though...

Regards, Axel

ServoKit · « **Reply #171 on:** March 23, 2016, 06:23:23 am »

@rwb: Good point re. the spacing. The parts that are candidates for the tape fixture, like the MOSFETs in the pic and our TSSOPs, are compartmentalized in their tapes. The tapes have little grooves on the underside between the compartments which can be used as a guide, just have to add a counterpart to the rail.

Regards, Axel

ar__systems · « **Reply #172 on:** March 24, 2016, 01:13:34 am »

We are almost done with the protocol. At this point we can control everything, the only thing still left to figure out is setting the speed of the movements.

Looks like the stepper driver in the machine is the most basic one - no microstepping, unfortunately. That's the reason for horrible noises the machine produces at slow speeds - it basically comes to full stop in between steps. I really wish they at least support acceleration control. Will know next week.

ServoKit · « **Reply #173 on:** March 24, 2016, 06:54:37 am »

The motors must run with microstepping: With the belt drive (diameter of pulley c. 16mm = circumference 50mm) and the standard 200 steps/revolution we would only have a resolution of 0.25mm...

Regards, Axel

protoneer · « **Reply #174 on:** March 24, 2016, 07:16:52 am »

How well does these machines deal with Micro's that are angled 45 degrees?

I take it the vision will ensure that the micro is straight first and then it will be turned into the 45 degree angle....

Also how well are these machines at placing 0.5mm pitch micro's? I am looking at using it for 100pin 0.5mm pitch micro's if possible...

(My machine is close but still stuck with customs


EEVblog Main Site	EEVblog on Youtube	EEVblog on Twitter	EEVblog on Facebook	EEVblog on Odysee

Author Topic: Pick and Place Machine TVM802A / TVM802B (Read 406408 times)

Share me