Source of dispensing solderpaste?

Started by SteveW, May 23, 2014, 02:54:13 PM

Previous topic - Next topic

SteveW

My Dima dotmasters are up and running - but can't dispense my standard stencilling paste. Any recommendations for a make & supplier of specifically dispensing paste? (smaller solder balls, more flux in the mix)?

Cheers

phonoplug

Ok first of all disclaimers... I am not familiar with your specific bit of kit nor how much replacement dispensing heads might cost, however...

In my ignorance I would be very cautious about using any old paste in something like that. I would only use ones specifically recommended or profiled for the machine/head.

I know that for some other machines, different solder pastes have different profiles which the machine needs to be set to for it work reliably, and more importantly not to damage or clog up the head.

SteveW

noted :)
The heads come apart, and aren't particularly complex, there's an Archimedes screw in a tube. It's not positive displacement, so if the pump stalls for whatever reason, you just spin a column of solder paste around in a tube - not ideal, but I seem to have got away with it - you can just remove the needle (which opens up a huge hole, then stick cleaning grease stuff in a cartridge instead of solder paste, and pump it through 'til it runs clear. An 'after' test did a board happily with cleaning grease, so with any luck...

Was mostly wondering if there was a favourite paste out there for this sort of thing. Solderconnection offered these (no connection, they just answered my web query)

111323 CL78 T5 10CC 62/36/2 25GM SYRINGE                      £ 13.20 EACH
Web link to the above product:
http://www.solderconnection.com/P15/ALPHA_RMA_CL78_DISPENSE_SOLDER_PASTE.html

152822 OMNIX 338T T4.5 SAC405  40GM SYRINGE             £ 14.98 EACH
Web link to the above product:
http://www.solderconnection.com/P1169/ALPHA_OM338_LEAD_FREE_DISPENSING_SOLDER_PASTE.html

which seemed like bearable prices for leaded and lead-free respectively. I expect to use rather less of the stuff, compared to short-run stencilling, which is really quite wasteful.

(Replacement pumps seem to be about £1700. I bought the more recent Dotmaster for £50! They're not the current version, but only the software seems to have improved - and it'll do the job OK as is. I thought 'yay, a jewel case, maybe it's a CD and supports a modern OS'. Nope - two floppies, and nothing newer than Win98. I feel right at home...)

Gopher

I'm a little jealous, where/how did you score these badboys?

SteveW

Been trawling ebay - First one was £400, second was almost unused, for £50 - he didn't list it too well, but there's obviously not a lot of competition for them. Seems strange, they're quite spiffy (as long as you get one with an auger pump, not just a syringe).
They're built like a battleship, mind - 70Kg. They're not complicated at all - a massive chassis, some rods & linear bearings, stepper motors and drives, and belt drive for the axes. Then the pump!
The movement isn't super-fast, it doesn't really need to be, since it's just going from pad to pad. You set the PCB height, but there's freedom so it doesn't mash the needle through the board, and there's a finger next to the needle, that does land, so the needle will be at the right height regardless of board warp. Add a camera for fiducial spotting (manual in this version) and a G-code interpreting 8-bit micro, and that's it. I need to write a bit of Python to convert pick&place files into fully formed dot files, since teaching it 90s-style is unacceptably dull).
Build quality seems exceptional - stiffness comes from massive amounts of steel, rather than super-cunning design and modern materials, though.

If you do score one, and have it delivered, try to get them to tether the head in one corner with cable ties or something. With power off, the head moved freely, and mine had slammed into an endstop and broken an optoswitch such that it wouldn't home, and grumbled).

It'll never be as fast as stencilling on a production run - but the lack of setup, cleanup and stencil costs all has an appeal. If I can convince it to do local re-pasting for BGA rework (on both the BGA and the PCB) I'll be delighted.

Mike

Quote from: SteveW on May 23, 2014, 07:26:06 PM
I need to write a bit of Python to convert pick&place files into fully formed dot files, since teaching it 90s-style is unacceptably dull).
Surely from the paste gerber, not the pick/place data?

SteveW


Surely from the paste gerber, not the pick/place data?
[/quote]

My thinking is not...
Parsing the gerbers isn't trivial, especially if I've been creative with pad and paste shapes. Add to that the need to stagger dots on things like adjacent QFP pads, to stop the dots coalescing because they overhang a little, and I'm more tempted to just build a library of dot patterns, per placed component. That'll also (maybe?) make it easier to tune things as I learn what works.
Trying to flood-fill rendered pads to a given density and obeying not-too-close adjacency rules just sounds like hard work... I could probably be conviced otherwise, though!

Gopher

I had a play with FAB3000, a piece of software for parsing gerbers to create pick and place files, stencil layers, BOM's work instructions etc. I don't think it currently does anything specifically for dispensing as yet, but I would not be surprised if such a feature turned up in a later version, or if such a thing is possible using its scripting support. It is currently lacking in a few areas (90 degree rotations being one of them) but seemed quite nice to use unlike Graphicodes monstrosity.

One of the many reasons we chose our new machine was because it can be fitted with dispensing heads at a later date if the need arises. As I understand it, pasting data is derived from pick and place location data and the rules you define for each package. I think the same is also true for the very expensive MY500 printer from MyData, so it would seem to be the most popular approach. Technically printing paste should be less wasteful and much more flexible, especially on a board with a mixture of device types and sizes where you might otherwise use stepped stencils and other such black magic. However when things get that complex , what I have understood of the dotmasters abilities suggest perhaps such situations are out of its league.

SteveW

A quick update - solder printing is proving superb, for low volume stuff.
You do have to manage the room temperature, though - the paste behaves profoundly differently at 5 degrees and 20 degrees, and the machine takes most of a day to come up to temperature from when I turn the heaters on (it's coooold in the unit).
0402s hold no fear, but I'm still working on getting QFNs to not be overpasted - moving down another gauge of needle will probably do the job.
I bartered an assembly job against some software effort, and can now generate files to feed the Dotmaster straight from Altium with a little python scripting.

Upsides:
Far lower usage of paste compared to stencilling,
far less cleaning of stencils,
alignment just involves fiducials and a camera (not automated (yet))
no stencil cost or storage.
Pad area and solder volume are decoupled. If I want to overpaste a pad because a pinmount socket goes in it, then it's easy.
Machine can cost much less than even a crappy stencil printer (but that may not be repeatable)

Downsides:
Slower than stencilling, for big runs (but faster than an RV1s, so who cares, after the first panel. Probably faster than an RV4, too).
Takes up space
Requires stable temperature
Takes compressed air (in miniscule quantity).
I haven't yet got the smallest pads to work - but that's just because I haven't lathed up an adapter widget for the fine needles I bought...

SteveW

Getting better (until now, 0.4mm DFNs and QFNs have been a pain). Dots are a bit wiggly because this is a board from a guy who does his own PCB layout, so I have to hand teach the dot placement. Imported jobs form Altium or Eagle are spot-on. This board used 27ga needle - 0.23mm ID. I've got 28 & 30ga needles, but they need to be run warmer.
The run-together dots generally sort themselves out during reflow, as long as there's not wild overpasting. My python script offsets adjacent pads, to reduce the problem - but I CBA when hand-teaching.

Things I've learned so far:
Paste matters: Super fine particles, and dispense-specific. I'm currently using
Item No: SP-021-018 40G 1 14.98 14.98 20.0%
152822 OMNIX 338T T4.5 SAC405 40GM SY
FINE PARTICLE LEAD FREE DISPENSE SOLDER PASTE
from www.solderconnection.co.uk

Temperature: That board was done at room temperature, but if you're pushing fine needles, temperature does make a substantial difference. Dispensing at 5oC is right out - the paste neither flows nor sticks to the pads

Dwell time: For these small dots, I have to hold the needle over the pad for 0.2s after dispensing, or the paste doesn't always stick.

Anyway, just dumping this on the net, in case anyone else is playing this game.

The pic has 0402, 0.4mm DFN and a microphone that would be slightly tricky on a stencil.

Gopher

Great stuff Steve, I like it. I wonder has anyone ever adapted a 3d printer or one of those cheapo "3040" cnc machines into a dispenser?

SteveW

The mechanics are easy enough, you just need 0.1mm accuracy over your working area.
The solder pump really shouldn't be that hard, although making / sourcing the archimedes screw won't be trivial. The software is definitely not hard, unless you want to roll in vision, for fiducial spotting and / or checking that a pad looks different after you've visited it. Measuring the volume of solder paste deposited would be great, but I suspect really quite hard.

SteveW

The other cool thing - I tried reballing a BGA with it - worked a treat.
One chip I did with small dots and a preform ball tweezered onto each pad. The other, I just did with bigger dots, and cooked it to make bumps. Both reflowed fine, settled nicely and worked. They were only 0.8mm pitch NAND flashes, but I can't see why smaller won't work.
I can also paste straight onto assembled boards for rework, which has been handy once, and looks very cool.

Mike


I wondered whether you could use something like a standard woodscrew as an auger - ideally with something like a torx or hex head for easy coupling.
As regards temperature, a simple temperature-controlled heater on the needle shouldn't be too hard to do - Am I right in assuming you're using stainless needles ? If so, then clamping an ali. block round the needle shouldn't be too hard.


SteveW

You might need to grind the woodscrew into a more perfect cylinder, but yeah, worth a try. I don't know what the internals of my pumps look like, and I'm reluctant to srtip them for fun...

Given how slowly the printer responds to changes in temperature (many hours), I think that it's more the temperature of the paste in the pump and the channel to the needle than the needle itself - or both are involved. There's also limited room to clamp thermal stuff to the needle, as it would need to avoid the foot that sits behind the needle to hold it off the board by a couple of thou. (filing the shim washers down to compensate for the needle length variations is _not_ something that I think any homebrew solution should duplicate!)
Now I keep ambient temperature at around 22 degrees, things just work. If I need the paste to be really runny, half a day at 30 degrees does the trick, but it's foul to work like that.