CHMT-528P / CHMT-530P4

[nisma]

To my knowledge the machine have working nozzle holder, the nozzle base is missing. This for commercial reasons and because of missing automatic
nozzle offset correction code.

[Spikee]

The nozzle system they use can fully support the commonly used auto nozzle change method.
Heck they are even implementing it in one of their new higher spec machines. I even told them / offered them that they should charge like 300-400 USD for this option. But no response of thus yet.

Kimi has not reacted to my last effort to get in contact. I'll try again tomorrow.

[Spikee]

After talking with kimi I did some experimenting.
And with a -0.25 degrees rotation compensation and y compensation of 0.25 mm it seems to place the TQFP consistent at this moment.
The next assembly with these kind of parts will show if it is good enough or not. But it QFN should have less of an influence because of a rotation offset since these
are generally smaller chips. And the spacing is the same as the TQFP I'm testing with (0.5mm).

Regarding auto toolchange there has not been any progress. Their engineer is not sure at this moment if it can be done on this line of machines.
It seems they are fully focused on their new 6 head one.

[Spikee]

I did a run of 20 boards for a new design and they all came out fine. Including the 0.5mm pitch qfn and the big SOIC MCU.
My small batch assembly in Shenzhen China is up and running

[danielm]

Hi , is there any expected time-frame four launch of the 6-head machine?

[mgbolts]

Well, I have taken the plunge. My 530P4 ships by TNT today.

@spike - begging your understanding for some dumb questions!

[mgbolts]

Hi Spikee,

Would you mind posting some pictures of your tape installation on the yamaha feeders including the clear plastic .

Thanks

[mgbolts]

One week in and we have made a few boards so far. I have been pleasantly surprised, machine is working great.  Large learning curve based on trial and error.

A few observations:
1. Damned heavy, need a sturdy table and reasonable sized room.
2. Instructions are average but enough to get you going.
3. Menus are intuitive and you can muddle your way through most functions.
4. Panel feature works ok
5. Vacuum detection effective. 100% detect rate when I had a feeder problem (my fault)
6. Menu touch screen was better than I expected. Not that fussed no network connection.
7. PLACEMENT DELAY WAS ESSENTIAL FOR NOT THROWING THE COMPONENT AFTER PLACING.  GRRR... THAT WASTED A FEW HOURS! I am using 0.25s at this points, not sure if I can get away with less, needs more testng.

Issues So Far

Panel Array Feature:
a) Needs an offset function for edge as maths assumes corner is pcb1
b) Higher decimal resolution required for interval when array gets large, error creeps up.
c) Capped at 20x20

Panel Coordinate System
a) No way to import this.....need to enter manually

Head
a) Needed to reduce speed down from 100% as was skipping belt when moving from home position (back right)
b) should have ordered more smaller tips.  Assuming that it can pick up two components in one action....
c) Head height maths still does my 'head' in, will need to spend more time to work out what is zero!

Component Alignment Camera
a) Its a black box with no ability to help it learn. 
b) Seems to do the trick.  Still have to see how it performs with more complex parts.

PC Software Tool
a) Only imports placements based on simple column order - simple but effective
b) No ability to import placement heights, need to be manually entered in one by one for every component on the list ....not each component stack.....can get tedious.
c) No ability to store component info for future use.
d) Placement Delay - Cant import and cant enter into software.

In Summary

Works great and no regrets.  Plenty of room for improvement with the software though. 

[JPlocher]

(From a longer post over in the overview 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

[mpi]

7. PLACEMENT DELAY WAS ESSENTIAL FOR NOT THROWING THE COMPONENT AFTER PLACING.

d) Placement Delay - Cant import and cant enter into software.

Where/how did you enter Placement Delay?

[mgbolts]

Delay is on the component screen on the panel.

[theatrus]

Placement delay can also be entered in the text .dpv file.

[DinkDonk]

I am experiencing the same rotation variance as you Spikee.
I can't get a hold of Charmhigh (Kimi) to resolve the issue.
Placing large parts like QNF64 results in variable part rotation. It's like an oscillating offset error. Every 5-10th component placement zeroes out the error.
All things points to a fault in the computer vision algorithm. If the software was open source I could have fixed it in a jiffy 

After talking with kimi I did some experimenting.
And with a -0.25 degrees rotation compensation and y compensation of 0.25 mm it seems to place the TQFP consistent at this moment.

Was the -0.25° offset the only thing that had to be done on your end?
Did you input the offset in the global system setting or per component?

Attached is a sample of 5 QNF64 placements as an animated GIF.

[DinkDonk]

Update:

It seems the rotation variance only happens on head #2.
If I use the correct nozzle on head #1, the vision system corrects the rotation like it should, and placement is very stable.

This leads me to believe that the vision does not work properly with head #2, and it might be offsetting head #1 no matter what head is active.

I'm on a CHM-T48VB machine.