My first reflow oven

[pldaniels]

Since I'm doing a lot of surface mount electronics, I was starting to get rather tired of soldering everything by hand, especially for general production work, so the choice was forced and it was time to look at making a reflow oven.

I was lucky and an online store had some 12" toaster ovens for sale at only $20 -new-, $9 delivery, ordered it last Friday and it arrived on Monday.

Let me first say that it's no surprise that this thing was $20... believe me the build quality was scary including the 240V wiring, a few of the crimps were just waiting to fall out :eek:

Anyhow, this oven is a 4 quartz element device (two top, two bottom), I ripped out the timer, thermostat (bimetalic) and proceeded to wire in my own 2-board controller system.

Board #1 is just a fairly generic opto-TRIAC (with zero crossing) driving a BT137 TO220 TRIAC, a snubber cap/resistor was also added though it probably isn't absolutely required with quartz heater elements (might be with wire wound ones). Using this the power can be turned on/off using DC logic levels and the built in zero-crossing detector (MOC6036) means I don't induce ugly noise on the power lines.

Board #2 is the brains, it is driven by an ATMega48 and the sensor is a PT100 type resistor (-50~550'C range). It's also got a 2400bps serial link to the PC for monitoring/logging purposes.

I'm using a 3 phase reflow process, rather than the more typical 4~5 phase. I just go from room temperature to 170'C at about 1K/sec, after which I pause for about 20 seconds and then proceed to push up to 220'C for no more than a total phase duration of 60 seconds, after that power is removed and the unit cools down.

One thing I found is that the solder reacts very quickly to infrared heat, however my temperature sensor is working off air temperature so to keep things better balanced I reflow on a metal tray and also put another metal plate between the tray and the top quartz elements. Using this diffuser the reflows were more controlled as the temperature of the components and boards were now more equalized with the solder.



Anyhow, that's about it... just thought I'd share. :)

[thanhTran]

Looks great! Great work as well!

I've thought about trying this, but the thought of buying & making stencial, buying solder paste, etc. really puts me back. What do you think?

-Thanh

[pldaniels]

I suppose it comes down to how much stuff you're doing.

Solder paste can be acquired rather cheaply. I know you can get it from places like DealExtreme for about $4 per 50g.

As for stencils, I use thin brass sheet which I etch as I would circuit boards.

It's not something I think I'd bother with for one off projects, though as something to help you churn out a dozen or more repeats of the same project at once, then yes, it's worth it.

If anyone's interested, I'm happy to post the schematics / code.

Paul.

[SteveM732]

When I helped build one of these (ATMega128 based) the oven did not cool down fast enough on its own and I would have to crack the door open slightly to provide cooling and allow the heating elements to kick in as needed to slow the cooling. Did you have a problem with this? We also did not need any clay insulation, the oven shell was replaced and it heats up just fine without too much radiated heat.

Even if you don't have a stencil, you can do parts like a QFN 56 or some of those parts with ground pads on the underside for cooling. Even with a professional quality stencil, fine pitch parts like a PQ-208 will smear a little between pins so the DIY should not feel like they need perfect application as the solder will tend to flow towards the pins rather than forming a bridge, unless you used too much paste.

A temperature controlled oven with a proper reflow temperature curve is the best to have, but in a pinch I have used a stock oven to do the soldering. Just put a small piece of standard solder on the PCB, when it melts your paste should be melted too (assuming similar melt points). But really, only do this in a pinch.

Circuit Cellar did an article (http://www.circuitcellar.com/library/print/0704/Lacoste_168/) on how to make these that could also be a good reference for those ready to make their own. But no need for a $150 oven like they used, the $20 ones work fine too.

[pldaniels]

I have also found the cooling to be the only part that needs to be manually done. I'll set it up so that a beeper goes off, or better yet, perhaps I should connect a solenoid up to the door and have it push it open.

I've ended up insulating the top and sides with 5mm thick balsa attached using RTV silicone. Since the oven isn't going past 210~220'C it should be perfectly good. I placed my hand on top of the balsa after a test run and it's just 'warm', so the job is being done :D

Paul.

[pldaniels]

Just took a look at that article you linked - you're right, a bit pricey ($150) considering this one was $20 and has the top and bottom elements. :)

My total reflow cycle is about 300 seconds, of which up to 240 are preheat, the last 60 are the actual reflow event.

Using the mega48 I've only consumed about 3K of code, which could probably be whittled down with some ingenious thinking but for now I'm more than content with it.

Rather than the LCD 16x2 matrix I'll just use the PC serial link... for now (not that it's even required since you can visually see that things are happening.

Paul.

[CrashingDutchman]

Nice Paul!

Regarding the clay, I remember when reading about building a Pizza oven that it can't be heathen (spelling?) up to much the first couple of times. It seems that the clay has to settle first.

This was also with a garden stove we bought a year ago.

So, you might want to try it again, but apply less heat.

CD

[pldaniels]

Crashing Dutchman,

I ended up going with 5mm thick balsa planks bonded to the metal using RTV silicone. Contact temperature of the metal is in the order of 200'C absolute max which is well below the charring/ignition point. I'll monitor it and see how it goes but it should work nicely as an insulator.

Next trick really for me now is to pull the 240V switching electronics out of the unit and place everything into a blackbox that just turns on/off the power via a normal powerpoint (or even just slice an extention lead in half) with only the temperature sensor coming from the oven. Will also pick up a solenoid to see if I can look at popping open that door at the end of the process.

Paul.

[pldaniels]

Has anyone tried doing two reflows, that is so you can do a double-sided board?

A lot of my production boards have components on both sides, namely because in order to get the compact dimensions R/C demands a lot of the time, it's my only choice (also cuts down on weight since you don't use excess FC4 board).

I suppose really, I should just try it and see :D

Paul.

[jeffs555]

I haven't tried it at home, but we do have boards at work built by a local assembly house. At least on our boards they count entirely on surface tension to hold the bottom parts on. You should probably just try it. You need to identify the side of the board where the components are most likely to stay on by surface tension and do that side first. Resistors and ceramic caps hold very well. Some of the chips that I thought would be too heavy stay on fine too because of the large pad area. I know there are glues you can use if the bottom parts won't hold by surface tension but don't have any experience with them.

PS You would have to be very careful when opening the door for the cooldown cycle so you didn't dislodge any components. A solenoid might be too abrupt. Perhaps a large servo would work.

[pldaniels]

Jeff,

Servo for the door... brilliance - now I've got a use for those huge "standard" servos that I never use ;)

Looking at most of my boards I can see that usually the IC/microcontroller is the highest profile component on the board, so it makes things a bit easier in that respect.

Got to love surface tension.

Paul.

[Gussy]

Nice work Paul, definitely interested in code/schematics, I have a bunch of ATMega48/88's here (among many others :D). I have heard the Black and Decker Infrawave ovens are good, but $20 is sure a bargain.
I have been planning on setting up one of these for a while, just don't have the time. Was planning on doing pretty much the same as you, but with a solid state relay, PID sort of thing.

For the double sided boards, when you do it for the second time, the first lot of solder will melt, but the surface tension should hold the bottom parts in place.
Try it and see :D

[pldaniels]

Well, I tried to do the 2nd reflow --- but then my oven died on me, we rather the temperature sensor started reading everything WRONG, must have broken one of the very fine wires on the PT100 device :(

Anyhow, I was smart enough to buy two of them - so I've just finished installing the new one (couldn't remove the old one as it was JB'd in - this one I just use RTV silicone).

Gussy, you can get those ovens from DealsDirect.com.au ---- they're horribly constructed but for $20 it gets you want you need :D

Originally I had planned on PWM'ing the 50Hz but now that I've done a few runs I've found that simply switching on/off periodically every second or so as required is more than ample, so a relay with a 100,000 contact rating would work fine, that said using the TRIAC method works nicely too and it has the added advantage of the built in zero-crossing detector.

[Brandano]

Just a thought, if surface tension isn't enough to keep the bottom components in place perhaps it's possible to cast them in a jig using automotive high temperature silicone caulking after the first side is made. Naturally after applying plenty of detaching agent. Another option would be to just pot the lot in epoxy, but that would ultimately add weight to the circuit and make it almost impossible to fix any faults. Never mind me, though, I am no expert :)

[pldaniels]

Replaced the temperature sensor tonight... all fixed in nicer and I've increased the sampling to 64 samples averaged (since they're about 700 max per sample, I can safely fit 64 of them into a single unsigned 16 bit int :) )

Anyhow, here's some pics of the 'controller box' I've made. All the parts on the controller brain are surface mount so that's why it appears rather blank.

I've got to attach the RS232 D-connector to the panel as well as a socket for the temperature sensor too. Later on if I'm grotesquely bored I'll put a 16x2 LCD matrix on the front.

Oh yes, also have to install a small 9V transformer to power the microcontroller, or I can just plug a jack in the back and power from an external brick/wart.

[pldaniels]

Next plans... add a servo for opening the door... and maybe I can incorporate a fan somehow?

One idea was/is to put a 80mm fan on the underside of the oven with a hole cut out and hinge in two half butterfly valves, so that they'll be closed under their own normal weight - but when the fan blows the pressure should open them up.... perhaps that's a bit crazy in the head?

Paul.

[ohmite]

Hi Paul,
Glad you are making progress on your oven. I'm on my second generation oven now (uses a convection oven).

Yes I have done double sided. Reflow the bottom first, as long as the components are mostly passives and small items like SOT-23, they will stay attached to the board from the surface tension of the solder. Larger items like SOIC and other ICs probably won't stay attached (I don't even bother putting them on, so I don't know the limit).

If you want a controller and can program a PIC18, I can send you a bare board to try. Here are couple of pictures of my latest generation. One is the controller inuse, the other is the oven. The last is a copy of one of the profiles from a reflow. Note the foil on the window, that helped quite abit to even the temperature across larger boards. Also improved the ability to warm the oven quickly.

On the cool down, I have an alert on my LCD that tells me to open the door. The requirement is that the cooldown is no faster that -6 degC/sec. Even at the worse case of a cold garage, I have never seen anything higher than -3 degC/sec opening the door. I log the temp profile on an on board EEPROM, so I can gauge the performance of my system.

Regards,
-Eric
www.eflightworks.net

Has anyone tried doing two reflows, that is so you can do a double-sided board?

A lot of my production boards have components on both sides, namely because in order to get the compact dimensions R/C demands a lot of the time, it's my only choice (also cuts down on weight since you don't use excess FC4 board).

I suppose really, I should just try it and see :D

Paul.

[pldaniels]

Ohmite,

It's, okay, I'm about done with my controller (will take photos just now). The PC link is good for logging the data (I don't display it graphically at the moment but it does still come over the serial link nicely).

Interesting to note that other than the (what looks like) fan on the side of yours, mine is fairly much the same with a 4-bar heater arrangement.

Your on/off switching as given in your graph is rather similar to what I'm seeing with mine as well, nice to see the correlation, puts my mind at ease.

I might well supplant the external RS232 link to the PC later for a 128x64 LCD matrix display ;) ... if I've got nothing better to do *cough*.
Paul.

[pldaniels]

Okay, obviously some minor tweaking with timer settings etc yet to be done - but everything's now packed in a box and sealed up... except I forgot to put the AVR programming port on the back *sigh*.

Here's a bunch of pics.

[SteveM732]

Thanks for sharing your progress and experiences. Some people are against surface mount as the parts can be difficult to solder, but a project like this demonstrates how easily a mini assembly line can be built.

I'm also glad to see you're using Linux, I can even see Gnome peeking out at the bottom of the screen shot.

[ohmite]

Paul,
You may have to redo the pads that have those three caps on them that went askew. Unfortunately you are dealing with the surface tension and how the big parts / little parts will move around as they go to where the most tension is. The best way to make the parts self-align while the solder is fluid (molten) is to have a pad for each component. That way surface tension will draw the part to the center of the pad and if there are two pads it will center the part between them. It is also recommended that you have a smaller trace between a component pad and a large ground plane or heat sink as the large copper area will pull heat away from the pad that you are trying to solder, usually causing those joints to be low on solder after you are done.

The other pads (around the other resistors, SOT-23 and SOIC) look great though!

That is some of what I've seen,
-Eric

Okay, obviously some minor tweaking with timer settings etc yet to be done - but everything's now packed in a box and sealed up... except I forgot to put the AVR programming port on the back *sigh*.

Here's a bunch of pics.

[pldaniels]

Steve,

Yes, an entirely linux (and some Solaris) shop here. Been using it since the early 90's (my 'other' job is an OpenSource and commercial software developer). I actually don't normally run GNOME/Metacity (preferring Fluxbox myself for speed) but on this machine I just installed the latest Ubuntu and left it at that, easy and does the job very nicely on the 1GHz celeron.

I agree about the move to surface mount. Most people seem to be rather afraid of it, however other than needing a good pair of tweezers and an acceptable soldering station (or now, a cheapish reflow oven ;) ), I don't see the problem. Personally I find surface mount a lot easier to deal with.

Eric,

Yes, I'll have activated the 'thermals' adjustment on the pads on the more recent versions of this particular circuit board, the only downside of course of having a pad-per-part is that if you ever want to drop in an additional part or such, you can't (eg, I'm cheating here by cramming in the 4 sets of decoupling caps ;)

Paul.

[jeffs555]

other than needing a good pair of tweezers

I have stopped using tweezers and gone to a homemade vacuum pickup tool for the smaller parts. It makes assembly much easier and faster. Total cost was well under $10 US.

I bought a cheap aquarium air pump and silicone tubing at a local pet store. I disassembled the pump and converted it to a vacuum pump. I soldered together 3 pieces of telescoping brass tubing to get a final nozzle OD of about 0.06 inches. I initially put a hole in the wider part of the brass so I could control the vacuum with the same hand. I kept uncovering the hole and dropping the part when trying to orient it to the board. Ended up covering that with tape and putting a hole in the silicone tubing next to the pump so I pickup with one hand and control the vacuum with the other. Works out much better that way. Someday might do a solenoid valve so I can control it with a foot switch.

The flat end of the 0.06 tube will pick up most components up to 8 pin SOIC's. Larger components would require a rubber tip, but I normally just put those on by hand or with tweezers.

[pldaniels]

Jeff,

Other than just the aquarium pump (which I have a lot of, plenty of tanks in this house too :D ), how did you modify it to become a vacuum pump?

Paul

[thanhTran]

Wow, great stuff!
Thanks for sharing Paul!
I also like to work with SMD too. It's faster and mostly don't require drilling. As long as components are not smaller than 0603, SOT23 or SOIC then I'm fine. Smaller than that I cant make PCB and can't see clearly :p

Thanh

[jeffs555]

I couldn't remember exactly, so I took it apart and took some pictures. I am sure the exact details would be different for different air pumps. This is labeled Airtech 2K3 and claims to be for up to 30 gal aquarium.

I considered several possibilities for converting to vacuum. First was to totally seal the case, and fashion some fitting for over the filtered air intake in the case. I have seen people post about that, but it seemed like it would be difficult. (1st pic)

Disassembling the case showed that there was a single intake hole on the orange pump body. Another option would be to put some sort of fitting into that. That would probably be the easiest if the pump construction was different than this one. (2nd pic)

Disassembling the orange pump body found that I could assemble it rotated 90 degrees. This moved the one-way valve and reversed the flow so that the normal air outlet became a vacuum inlet. Had to cut off a plastic locating pin, but that was all that was necessary. Could actually convert it back just by reassembling it the correct way. (3rd & 4th pic)

When I reassembled it after taking the pictures it would barely hold a 0805 resistor. It turns out I had bent the arm holding the magnet in front of the coil and it was too far away from the coil. I bent it back and it started working fine.

The last picture shows it holding an 8 pin SOIC.

[pldaniels]

Jeff,

Ah yes! of course - I should have known (having had to repair a few of those over the years).

Just got back from the workshop where I was calibrating the oven a bit more (within 2'C now). Should be starting full production tonight or tomorrow *celebrate*.

I'll upload the new schem/code/pcb's later.

Paul.

[pldaniels]

Jeff, going to go try mod one of my pumps with your idea... should be fun :D

Paul.

[pldaniels]

Okay, my modification works on the aquarium pump... nice thing is that I can continue to use it as an output pump as well for when I'm etching :)

I'm having trouble making a nice wand at the moment but I'll keep working at it.

Paul.

[pldaniels]

Going for gold

I think I'm about home now :D

I made my vacuum pump, to make it easier to control I used a prefabricated T-piece in the line. The line is a bit stiff as it's a good 5 years old, I'll go get some fresh stuff tomorrow if I can.

Also, here's a shot of the board after the reflow - the board was assembled with the vacuum tool, I can see it's going to take some practice but there does seem to be the "improved productivity" aspect to it.

Oh yes, notice that I moved the caps around on the board and now the surface tension doesn't yank things out of place :)

Paul.

[pldaniels]

Well, running into my first set of problems... seems I'm destroying my attiny's in the reflow but the other parts seem fine. It's been suggested that there's moisture in them (quite likely).

Anyone have a reasonably safe prebake profile to remove damaging moisture? I've heard everything from 24hrs @125'C to days!

Paul.

[SteveM732]

Well, running into my first set of problems... seems I'm destroying my attiny's in the reflow but the other parts seem fine. It's been suggested that there's moisture in them (quite likely).

Anyone have a reasonably safe prebake profile to remove damaging moisture? I've heard everything from 24hrs @125'C to days!

Paul.
A first step might be to modify your temperature curve to warm up to about 125C then hold that for a few minutes before continuing with the heat up. The longer your parts sit around unsealed (especially in a humid area) the longer you would need to bake the parts to drive out the mositure. Maybe just try 1 hour at 125C and see if that does the trick, if so do a whole lot of them at once.

[pldaniels]

Steve,

Further to my previous posts, it would appear that I am perhaps not reflowing hot/long enough as it would seem that a couple of the faults have been from shorted out components on the underside.

Sure is a lot of tweaking involved :eek:

I added a 20 minutes @ 80'C bake before... maybe then I can do another 20 minutes @ 125'C. Thank goodness at least the oven is now able to be controlled like this :D

Paul

[pldaniels]

Things are now going 'well'

Seems like the problems were mostly related to the reflow peak being too low and too short. I put the boards that were not functioning back in for the following reflow profile and they came out working 100%;

* 300 seconds @ 80'C
* Ramp to 100'C
* 100 seconds at 150'C
* 90 seconds to 220'C (the 220'C bit only lasts for about 30 seconds)
* cool-down.

An important change I made in the software also was to compensate for heater element overshoot (that is, the oven keeps getting hotter even after you've turned off the element, so your 220'C could go as high as 235~240'C)

Fortunately the adjustment was a very simple one; for every second that the element is on, you add 1'C to the actual oven temperature, up to 16 seconds (eg 16'C overshoot), then for each subsequent second the element is -off-, you deduct 1'C from the predicted oven temperature. Obviously those values have to be tweaked to each oven but it works pretty good on my oven now and the max I saw was about 222'C.

Just redrawing the controller and 240V board... then I'll post them up (in actuality, I botched the Eagle CAD file the other night so I had to start again :mad: ).

Paul.

[pldaniels]

First... a disclaimer.

WARNING

These schematics and circuit-boards are provided in good faith and come with absolutely no warranty and/or implied liability on my behalf.

If you are not experienced with mains power equipment do not attempt this project. A mistake in construction or handling could be lethal

Now, find attached the various schems and circuit boards, though I suspect people will make up their own circuit boards based on what gear they have available (eg pin-through etc).

If you want the source code, send me a PM, I won't post it here.

Paul

[pldaniels]

I'm going to have to make a new stencil for my SO8 type chips. The one I have at the moment leaves too much paste that gets caught underneath the chip and causes shorts. Oh well, all a very good learning process :D

Paul.

[billg]

I found a company that will produce a Kapton stencil on a 12" x 12" sheet for $25 and give a fast turn around. http://www.ohararp.com/Stencils.html
I have not purchased any stencils but have purchased GPS modules from them. Delivery was fast.
Bill Groft

[pldaniels]

Bill,

I'll probably be getting LaserArts/MountainModels to do mine - since I already send a lot of CAD work to them and import a lot of their gear for Australia.

Nice to see the Oharap people offering Kapton sheets, I wonder what its durability is like. The price is good - though I should email them too about what postage to Australia would be.

Paul.

[pldaniels]

Picked up some 18 AWG / 1.2mm diameter needles today from the local vet, ground off the tip and it makes for a perfect vacuum pickup even down to 0603's, though it takes a little bit to pick up SO8's (the widebody ones), most likely my vacuum's a bit leaky.

Etched some new stencils today with a much reduced solder paste area for the SO8 and immediately things were better with the units coming out of the oven.

Will post up pictures shortly.

Paul.

[CrashingDutchman]

I can highly recommend http://www.ohararp.com, Ryan is very helpful

[pldaniels]

Here's some more pictures.

Starting to ramp up the number of boards I do in each batch as my confidence in the setup grows.

[Brandano]

Nice. Now you just need to create a "pick and place" robotic arm, assemble the boards in a batch before the cutting and you have an actual construction chain :)

[pldaniels]

That'd be awfully nice wouldn't it... for now I'll just have to use the human robot :(

[Brandano]

Time to hack a couple of old printers? The biggest problem I can see is how to consistently pick the part with the proper orientation, or how to adjust for errors in the pickup. Perhaps a part could be picked up in more phases: pick up, place on rotating vacuum pad, pick up again once rotated to right position, place. I think I have some links on how to modify a webcam to make it into a low quality digital microscope. Here's one way using a lens from an old cdrom: http://www.engadget.com/2006/08/19/webcam-cd-rom-lens-instant-microscope/

[jeffs555]

Time to hack a couple of old printers? The biggest problem I can see is how to consistently pick the part with the proper orientation, or how to adjust for errors in the pickup. Perhaps a part could be picked up in more phases: pick up, place on rotating vacuum pad, pick up again once rotated to right position, place. I think I have some links on how to modify a webcam to make it into a low quality digital microscope. Here's one way using a lens from an old cdrom: http://www.engadget.com/2006/08/19/webcam-cd-rom-lens-instant-microscope/

Rotating is easy, you just have a servo to rotate the vacuum pickup needle. For highest speed you need a camera to measure the orientation and offset of the part then adjust the rotation and placement. If you only need medium speeds, you can use a squaring chamber for the parts. It is basically just a square box and after picking up the part you run the part against each side of the box to center and square it up on the pickup needle.
Here is a video of someone using a CNC mill to do SMT pick and place. Because of the slow speed of the mill this was just a demonstraton and too slow to be practical, but the principles could be applied to a purpose built medium speed machine.
http://www.youtube.com/watch?v=__dEMKzkLYc

[pldaniels]

Put through another batch just now, different board this time - reflow profile worked fine on it.

[pldaniels]

Jeff,

That homebrew P&P machine was quite interesting to watch... painfully slow, yes... but interesting.

paul.

[Brandano]

paul, just think about the same setup, but moving at the speed of an inkjet printer printing head. Needs some smart logic to handle the occasional dropped component, but perhaps the same machine could be used to dispense the solder paste on the boards (using a different needle) and doing away with stencils altogether

[pldaniels]

Brando,

Could be a possible use for my rather un-used vinyl cutting machine. It has XYZ control and I can control it via HPGL on the serial port... mmm... must not hack $500 vinyl cutter... (reason why I've _not_ used it to cut stencils is that it isn't much good below 5x5mm due to the nature of the cutting tool).

Paul.

[Brandano]

Btw, here is a nice, fast and accurate design that however requires some serious computing for pick and place operations: http://www.youtube.com/watch?v=xHuDvVa7mkw

[pldaniels]

Brandano,

Those are some serious machines... frightening in some ways :eek:

Paul.

[Brandano]

Oh, youtube is a great place to look up robotic applications. The layout is defined a "delta parallel", and is precise enough to be used in surgical applications, while it uses no sliding parts. Here's the patent it's based on:
http://www.google.com/patents?id=D0kjAAAAEBAJ&printsec=abstract
And here's another nice demo of a delta parallel robot:
http://www.youtube.com/watch?v=Du2f-EUDqio

[pldaniels]

I think that I'm going to have to add a "Press to start" button on my reflow-controller. This morning I powered up the workshop and I had left the oven on last night (not a problem since the cycle never repeats), however this morning I didn't know and I was sniffing around the workshop thinking "What the heck is burning!?". Obviously what had happened was that when everything powered back up the reflow cycle started again *whooops*.

Paul.

[pldaniels]

Good news everyone!

Well, for me it's more like Eureka!

I managed to get my vinyl cutter to cut out stencils for me on ~0.003" mylar and the results appear to be GREAT (I wrote some software in linux to do it).

Some photos attached.

... haven't been this excited in ages.

[Gussy]

That's looking really nice now, and the ATTiny13's are surviving??

[pldaniels]

Gussy,

Yes. My prior fears that it was due to moisture seem to have evaporated (oh I pun, I pun ;) ). The cause in all cases thus far was from excess paste getting under the body and forming a conductive bridge.

The solving of the stencil issue makes thins a lot better for me now - as I can experiment and develop at a faster rate, no need to wait days or weeks now. Next thing I have to do however is to get more of this stencil material.

Paul.

[pldaniels]

Using the new stencil making machine, I cut out the stencils for a double reflow setup (two sided board).

I did the DPAK side first,then the AVR side second.

Biggest issue i had was that it was a pain to apply the paste/parts with the components on the back.

The actual reflow was a non-event.

I've noticed with the new stencils there's a higher risk of disconnects due to the smaller amounts of paste being applied. I can probably safely adjust the pad sizes up a little bit more.

Paul.

[Brandano]

I imagine that you rewrote the code so that the machine would do all the cuts in a direction first, then all the cuts at 90 degrees to those, then the ones at 45 degrees etc. How does it handle cutting at 90 degrees when you have several cuts already on the vinyl blank? Does the new code allow you to make bigger stencils so that you may etch, stencil and populate a bigger board to cut out the individual boards as the last step?

[pldaniels]

Brando,

Actually, my mods were a bit simpler than that :(

With the trailing-edge free spin blades (0.25mm offset on mine, might try get some 0.1 or 0.05), it's best not to lift the blade until you've finished the cuts, else you're quite likely not going to get the corners to line up. All my code does is slows down the whole process so that it gives the steppers a bit more time to stop moving from their inertia before moving to the next cut, helps stop some misalignments, the code also turfs out HPGL codes that don't have any relevance to most lower end cutters like mine (see hpgl-distiller at http://pldaniels.com/hpgl-distiller ). The slowdown code is just a case of pausing a period of time before subsequent HPGL codes.

Next steps in the code are to allow me to reposition the output arbitarily across the page (useful when you want to put multiple boards onto the same stencil). I'm also wanting to see if I can do a pad size adjustment (specific for stencil cutting for PCBs), though this is going to be a bit more complex as I have to then pre-parse the HPGL file and determine all the closed loops etc. . . . nothing like a challenge ;)

(the next step after that is a visual display for the HPGL with the ability to edit/remove pads).

Paul.

[pldaniels]

Been improving the stencil cutting already this morning, new process goes like this;

Get Eagle to generate the HPGL via the CAM processor and use a negative tool diameter (-0.1mm) to increase the pad size slightly, you have to be watchful of it though as you can easily get the situation where your pads run into each other or you get a very fine separator that breaks during cutting.

Filter the HPGL via hpgl-distiller and dumping it direct to the cutter with a 250mS delay between each movement. I do the cutting process 3~4 times starting at 120g pressure, increasing to a maximum of 160g. Typically by 150g (3rd pass) the pads are falling out.

Paul.

[duhh]

still a bit slow compared to what castle creations machine does :)

http://www.rcgroups.com/forums/showatt.php?attachmentid=1976155

[pldaniels]

Got a video of my stencil cutter that I'll upload soon :D Then we can compare speeds ;)

Paul.

[Gussy]

You should start an Aussie stencil business, Ohararp style :D

[pldaniels]

Let me think about that....mmmm.... no. ;)

It's all good and fine doing it for oneself, where you can accept less than perfect quality or a few duds... and you've got the time to burn for it but turning it into a viable business, I think I'd end up buying a laser cutter and even then I'd be pondering the viability.

What I did try before was to do a vinyl cutting service, it was too awkward, too many format variances, too many tweaks, far too much time needed to get the right results :(

[pldaniels]

I'm thinking I should setup a dedicated WWW page on my site describing the whole setup now.

Also, it was my FIL's birthday today and he was given a smaller 2-bar (top and bottom) toaster oven... couldn't help but wonder how good it'd be as a reflow unit ;) . Just can't look at kitchen goods the same anymore.

Paul.

[darkith]

Paul,
You outsource your PCBs I assume? Sounds like that would be the slowest part of doing fast turnaround with your spiffy new setup!

David

[pldaniels]

David,

Correct, that is the slowest part now. Of course assuming that one has already decided to move a product into production the stencil creation (and subsequent remaking periodically to deal with the wear/damage) would have been a snagging point. Additionally, adjusting the stencil can let me build different configurations of a more generic board (different cap layouts etc). Of course, one could just do a generic stencil as well I suppose.

Anyhow, it's good to have this facility now sorted out :)

Paul.

[pldaniels]

Here's the link to my -short- video showing the vinyl cutter producing a stencil for me.

http://www.pldaniels.com/flying/videos/flash/playflash.html?video=stencilcut-001-360x288

[pldaniels]

Okay, so I got tired of keeping the PC on and I wanted to test a new creation anyhow ;)

Added:

*Arming button

*Serial LCD panel (uses a cheap/easy Tiny2313 @ 8MHz, powered directly from the 4 wire serial lead (+, -, RX, TX). I could just make it a 3 wire by dropping the TX line (relative to the device).

Anyhow, chatter aside, here's the photos. :)

[pldaniels]

Added a couple of tweaks this morning to the system, mostly just character arrangement and finding out why occasionally during the preheat phase it'd flash up showing [194] or some other temperature that'd be way above the wanted 165'C... turns out the timer interrupt was occurring just as I computed the 'desired temperature' but before I had a chance to check to see if it was excessive. Fix was easy, simply waited till after the interrupt had finished before recomputing the next desired-temperature value.

[pldaniels]

Just a bit of a status update...

Reflow oven now is going in 'production' and things are behaving very nicely overall. There's still some fine tuning to be done to minimize the out-balling of the solder but other than that I'm very happy. Double-sided (two reflows) is now par for the course. The only trick with doing double sided is that I have to put the boards onto a flat-plate rather than the open-mesh wire that I can use for the first side in order to keep the boards resting horizontally.

I've also moved my oven up to the top shelf right next to the 250mm exhaust fan, really keeps the fumes down :) ( one minor artifact I've found though is that the fan can induce a small vibration onto the oven if I'm not careful - haven't seen any troubles yet from it but it's something I'll watch ).

Also, today I received my order of 100 and 75 micron film (.004 and .003 respectively). For larger SOIC I'm happy to keep using 100 micron as it's quite a bit stronger than the 75. However I suspect as I move to MLFs I'll have to switch to 75 to prevent excess paste being put down.

[darkith]

You're a bad influence Paul. I spotted a toaster oven for $2.50 at a yard sale and couldn't resist. Probably won't get quite as fancy as you (no stencil cutter), but I want to try doing SOIC parts by hand with a syringe of paste...soldering them by hand isn't much fun.

D.

[pldaniels]

Darkith,

For $2.50, hard to pass up. What sort of oven is it? Got pics?

As for the stencils, you could always etch out your own, or make up a variety chip stencil (you know, with everything from SOT23-5 up to TQFP-100 etc) by one of those $25 services.

Paul.





Paul.

[darkith]

I think it's a "Toastmaster" brand. Nothing too special.

Hrm, etching stencils, eh? Something like toner on thin brass?
Getting a bunch of generic stencils done also sounds like a good idea.
So many projects...

David

[pldaniels]

To make it easier/faster to lay down individual boards and paste, I created a set of polymer clay casts.

What I do is rig up the board so that the -first- side is soldered up and then make the cast. This cast can then be used for -both- sides due to the fact that a blank board won't sink into the cast any further than a populated board.

So I insert the blank into the jig, first side facing up, stencil on the paste, reflow, then insert the partially assembled board back into the jig, component side down, apply the second side stencil and parts then reflowing the second time.

The polymer clay cost me about $28 for 800g and should last quite a while. Later I'll make multiple board jigs and film stencils to speed things up even more.

What's amusing is that I use the reflow oven to cure the polymer clay, fancy that, a multi-use oven. :)

[CrashingDutchman]

Very good idea! Man, I wish I was as productive as you are...

[Brandano]

Not automotive silicon caulking, but close enough. Probably a bit more expensive, but having a rigid molding probably is a bonus for this task? What are you using as a detaching agent?

[Ron W3FJW]

Pardon me jumping in here fellows, but I find the thread totally interesting even though I'm totally ignorant about this subject. I'm curious about the "paste" you're talking about. Can't see anything about adding solder to the process within the thread so, is it a solder bearing paste, and if so, where can it be purchased??
Thanks for any comments.

[SteveM732]

Pardon me jumping in here fellows, but I find the thread totally interesting even though I'm totally ignorant about this subject. I'm curious about the "paste" you're talking about. Can't see anything about adding solder to the process within the thread so, is it a solder bearing paste, and if so, where can it be purchased??
Thanks for any comments.
Solder Paste (http://en.wikipedia.org/wiki/Solder_paste), available just about anywhere you would buy soldering supplies.

[pldaniels]

Not automotive silicon caulking, but close enough. Probably a bit more expensive, but having a rigid molding probably is a bonus for this task? What are you using as a detaching agent?

Not using any agent to detach at this point, the board is removed prior to baking (which is a bit of a trick as it really has to come out straight else it'll deform one end).

I have thought about leaving the board in the polymer until it's cooked but then I haven't come up with a release agent yet ;)

The rigidity of the board is indeed useful, though I would have preferred it to be a substance like nylon-plastic for the wear characteristics, alas when I try to use nylon it's too difficult to get a good impression with sharp edges and not have the board get stuck (the nylon rapidly wicks between the component legs and locks it in).

Paul.

[pldaniels]

Very good idea! Man, I wish I was as productive as you are...

It's all lies and deception ;) A few selective photographs and a couple of expressive posts :p

Alas the truth is that it's vigilance towards the goal of shaving off as much time as possible from the build process with the goods at hand.

Ultimately a pick-and-place bot with a solder paste syringe would be the goal. Then I can just worry about replacing the parts reels and loading / unloading the reflow oven :)


One big issue that still defeats me though is how to speed up the loom attachment process (eg, soldering the wires to the PCB).

Paul.

[Ron W3FJW]

Solder Paste (http://en.wikipedia.org/wiki/Solder_paste), available just about anywhere you would buy soldering supplies.

Thanks for the link. It explains a lot about the process being discussed on this thread that I wasn't aware of. Last flux I bought (and still have/use) was about 15 years ago. Wasn't aware there was anything like this paste. I'll pick some up. It'll make life easier for me..
Thanks again... :)

[jeffs555]

Thanks for the link. It explains a lot about the process being discussed on this thread that I wasn't aware of. Last flux I bought (and still have/use) was about 15 years ago. Wasn't aware there was anything like this paste. I'll pick some up. It'll make life easier for me..
Thanks again... :)
You usually have to order solder paste, and shipping can be high because the best places ship it in ice. Solder paste has a rather short shelf life(6-8 months). Heat drastically shortens the life and it is best to store it refrigerated. What happens when it gets old is that the flux degrades. Instead of flowing smoothly into the joints, old paste tends to form tiny solder balls that roll around all over the board.

Here is one place to order small quantities online.
http://www.zeph.com/zephpaste.htm

[Ron W3FJW]

Thanks Jeffs. Appreciate the link. Have it bookmarked.

[pldaniels]

Still working on the new official WWW page for all that I've created/learned for reflowing stuff ... just takes so long to make a half-decent WWW site (esp since I'm busy with a billion other sites, electronics and email software).

I've found that the preheat/soak phase is perhaps one of the most important parts of the whole reflow - if you go too quick, you'll make the balling even worse, go too slow and things don't flow right to make thing worse it seems that different parts change the time/temp needed in some ways (like a DPAK can take a bit longer to get going on the reflow because of its greater thermal mass).

Paul.

[pldaniels]

Been a bit quiet around here - thought I'd see what everyone is up to.

I'm still needing to get around to making a better board holder for the paste/parts application. I've got the design in my head, trouble is just doing it :D

Paul.