Placement of integrated circuits

[mrandt]

I just made an order for the Liteplacer and it would be great if this got implemented.
There is a lot of fragmentation in the software. Does any of the unsupported software handle IC placement?

Hi Martin,

welcome to the community

AFAIK, none of the publicly available versions currently supports this - but some of the "private" forks got pretty far.

I truly believe it is just a matter of time.

You are right about the fragmentation in the software - some of us are trying to change this; see latest posts in this long thread:
http://liteplacer.com/phpBB/viewtopic.p ... &start=120

Regards
Malte

[WayOutWest]

As a LitePlacer user, I believe the mechanics are accurate enough to place at least 0.4mm pitch IC; maybe even smaller - many BGAs should work as well.

I'm placing 0.8mm-pitch BGAs on my Liteplacer with no problems.

This would not only be relevant for ICs or BGAs but also other larger parts which need to be well aligned and accurately placed.

This is one area where I'm starting to slightly regret moving to Juki nozzles and a hollow-shaft stepper. You can't get 0.9-degree-step hollow shaft steppers, they only come in 1.8-degree-step, and since it's direct-drive there's no gearing. So even with 1/8th microstepping (which introduces its own inaccuracies, the 8 subdivisions are not equally sized) your rotational accuracy will never be better than around 0.45 degrees. That's just barely acceptable for some BGAs.

Juha really had the right idea with using an off-axis stepper to get 0.9 degrees per step, then adding gearing to reduce the effective degrees-per-step even further.

for example detection of a part's body based on template matching. This works well for QFN (no leads ICs) and a few others but uses the downfacing / flying vision camera. But be warned, there is quite a bit of diversity (to put it mildly) and of course Juha does not support those software versions.

Try MSER, don't waste your time on template matching. Once you see what MSER can do and how unbelievably knock-down robust it is you'll never go back. I even replaced the Hough circle detector with MSER followed by a circumcircle-perimiter-to-area-filter, it's so much faster and less sensitive to JPEG noise.

[s_sergiu]

Hello Adam,

Can you please post some links with current software version you are using?

Sergiu

[WayOutWest]

Hello Adam,

Can you please post some links with current software version you are using?

Sergiu

http://liteplacer.com/phpBB/viewtopic.php?f=11&t=281

[jarekk]

About 0.9deg hollow shaft steppers - you can get, actually quite easy.
I have ordered 25pcs from Alibaba. With my custom modifications ( long shafts, each side 40mm) as I need more then ordinary 25mm ( in my case linear bearing goes directly over the shaft).

So if you need 1..2pcs I can sell you and send by post

[WayOutWest]

About 0.9deg hollow shaft steppers - you can get, actually quite easy.

If it's so easy, post a link to where they are advertised for sale.

I'd be happy to be wrong here. But I spent a long long time googling for them with no success...

So if you need 1..2pcs I can sell you and send by post

I appreciate your offer, but I don't put parts in my machine unless they're stocked by at least one vendor who can be counted on to supply replacements if I need them.

[Picky]

Let me chime in on the A-axis microstepping topic. It is true that precision, accuracy and repeatability of the A-axis is paramount for the large IC placement. I know it from my own experience trying to teach my machines to place QFP-206 packages as an exercise.

Turns out that the myth about high order microstepping is not a myth at all. It really depends on the capability of the driver, not the motor to enable stable and robust sub-step positioning. A few years ago I tested a dozen or so different stepper motor drivers beginning from super cheap Allegro IC's and all the way to Parker's high-end drives with 1/256 microstepping.

The best bang for the buck at the time was Leadshine DM422, which allowed for 1/256 microstepping and actually delivered it. The A-axis motor was not just holding commanded intermediate position. It was also able to return to that position after I manually tried to move the shaft using pliers. As far as I remember, it was returning to within 0.050 degree from the original position. All it took was a $50 driver, a cheap hollow-shaft NEMA14 bipolar 0.9-degree motor I received from Madell (don't ask) and 24VDC PSU. Coil currents around 1.0A were working just fine.

Of course, all this positioning capability is only good when you have vision feedback for parts alignment. I haven't read the whole forum yet and am not sure where you guys are at with respect to bottom vision and pattern matching (reza?). In my experiments I was playing with commercial vision libraries. For the large QFP-206 (0.5mm pitch package) I found that 0.050 degree accuracy was bare minimum. The A-motor with DM422 was able to position down to 0.005 degree as measured via vision feedback and I can try to dig up a video of that process. (see below)

All of the above applies only to the A-axis with direct drive and no gearing whatsoever (read no resistance). I haven't even thought of trying such high order microstepping anywhere else on the machine (X/Y/Z) because I don't believe the motors w/o feedback can position reliably. Also we should keep in mind that at such high orders of microstepping (1/100+), we will be pushing speed limits of the motion controller if we want the shafts to move fast. By the way, what's the maximum pulse rate can your board deliver?

Cheers,
Kirill

EDIT: Found one of the old videos where I was testing the proof-of-concept vision alignment method using rather expensive commercial pattern matching algorithm. In this program I was manually jogging the head away from perfectly aligned position and then clicked "Align" button repeatedly. In each "align" attempt the program was trying to compensate misalignment by measuring deviation in X/Y/A and commanding the machine to move the head and rotate the A-axis respectively. If I recall correctly, the stepper driver on the A axis was set to 1/200 micro and it was driving 0.9 degree NEMA14 motor you can see in other videos of that vintage (5-6 years ago). The pattern matching algorithm was reliably detecting angles of around 0.005-0.010 degrees using sub-pixel resolution of course. https://www.youtube.com/watch?v=9bODRom1jrs

[l0wside]

After hours of reading, I am still confused.
I am planning to build a small batch (100 parts) of a sensor in the next couple of months. I could either outsource the job or spend the same money on a Liteplacer. My tendency is towards the latter.

My design includes two finepitch (0.5mm) components, one in QFN, one TQFP. By now, is there a reliable way (be it in the main software or some fork) to place these? Or should I rather save my money and go for one of the commercial machines like Neoden or TVM802? My ambitions to manually place 200 components is close to zero.

Max

P.S.: mrandt, we are both located in the same city. Is there a chance to see your machine some day?

[wwwenrico]

Hi,

any update on this issue?

Can the machine reliably pick and place integrated circuits from a tray?

By integrated circuits I mean packages of 0.5 mm picth: QFP-XX, SO-XX, QFN-XX

Enrico Migliore

[Pixopax]

Since it does not have vision functions for detecting part rotration in thge tray it will not work.

The discussion about rotation precision of the A-Axis if useless as long as the software cannot us the upcam for nmeasuring the part rotation.
I only use the machine for placing my 0603s and a few other small parts, no ICs at all, as it is not precise enough without upcam support.