Toaster Reflow Oven Folly

[Pat Pending]

There are many good articles on toaster oven reflow and it looked as if a controller was essential to
regulate the temperature. I set out to build a controller from a MAX6675/Arduino hack.

The hack proved very useful for data collection. It showed

how slow a toaster oven is to respond at switch on.
how sluggish the rate of temperature increase is.
how much the temperature increases past the point the oven is turned off.
how the readings are influenced by airflow around the MAX6675.
It showed noise in the MAX6675 data, (+/- 20uV or +/- 0.5C)

I tried a PID control algorithm, but the noisy readings caused the derrivate term to go nuts.
A simple PI controller worked better but oscillated since there is a sizeable reaction delay.

It's now very apparent that a complex PID controller is simply overkill.

In the end my method is to let the oven heat to 226C and turn the heat off.
The temperature continues to climb to 230C, peak and then fall back to 226C.
I then fully open the oven door, the temperature rapidly falls through 183C (solder liquid phase)
where upon it crystalizes very quickly and a nice bright joint is formed.

This recipe spends 80 seconds above the solder liquid phase, I'm wondering if this is too long?

[marshallh]

Main problem I see is too-fast cooldown.

Refer to several part vendors' reflow profile guidelines and come up with one that satisfies most of them. You should peak just before 250 C, or slightly less if there is a large package that has a lower max temp.

Here is the profile I've used to solder several pcbs with large bgas:

[mikeselectricstuff]

If you stick to leaded solder, reflow profiles aren't nearly as critical as the documentation makes it appear, at least for small volumes and homebrew stuff.
Bear in mind these will have been specced for absolute optimum performance for people using gazillions pf parts, for whom tiny fractions of % yield matter a lot.

I've done many hundreds of boards with nothing more than setting a fixed power level of "not quite glowing" , watching for reflow then turning off and opening the door.
On a few occasions I've accidentally left it for a couple of minutes over the normal 3-4 mins, and apart from some darkening of the PCB. no obvious ill-effects.

[David_AVD]

Exactly the same method here.  I use a more upmarket toaster oven (Breville BOV800) that has 5 elements and that seems to give a fairly even heat distribution and reasonable heat curve.

[ftransform]

There is a recent post on hackaday about using a non oxygen environment for re flowing PCB's:


http://hackaday.com/2013/05/29/reflow-soldering-improved-with-carbon-dioxide/

possibly over kill but it is easy enough to do

[marshallh]

My mistake I assumed your were using lead free paste.

Leaded is even easier. Drop the temps by about 30 degrees and you're set.

[mikeselectricstuff]

There is a recent post on hackaday about using a non oxygen environment for re flowing PCB's:


http://hackaday.com/2013/05/29/reflow-soldering-improved-with-carbon-dioxide/

possibly over kill but it is easy enough to do

inert gas reflow can be useful  for leadfree as oxidisation is more of an issue at higher temps - some commercial reflow ovens have the option for nitrogen.

[Pat Pending]

Thinking that preheat and flux activation times were critical for successful joints  this is what happened on my first trial.

Manually it proved too difficult to maintain a constant preheat temperature (the main reason for this project)
so in the end, I warmed the oven to 180C and powered off until the temperature fell to 150C. -> see 1st photo
One final push raised the temperature to 225-230C to let the solder melt before opening the oven door.

Visually the joints look sound but under magnification it was a complete fail.

The volatile components of the solder had evaporated and the paste reduced to a solid.
When it eventually melted, it formed many tiny beads that did not coalesce.

Its harder to see but there is also a resin like fluid entrapping the beads.
I'm at a loss to explain what that is.

The picture shows an 0603 part - too much paste was used on this attempt.

The paste is Kester EP-256 63/37 - Date Nov 2011. It's old, but works fine with a faster profile.

[c4757p]

Its harder to see but there is also a resin like fluid entrapping the beads.
I'm at a loss to explain what that is.

...flux?

I'm thinking you heated it too fast (or too high - are you sure the temperature readings are accurate?), and the flux boiled, spraying solder paste.

[Kremmen]

Naah, that's what old paste does. You say it is vintage 2011. Well, pastes are not wines, they definitely do not age that well. This solder ball formation is a typical sign of past due paste and will only get worse.
Ditch the lot and get some new stuff instead. And store it in the fridge, that will prolong the usable life somewhat. Contrary to old wives tales, it will not poison your food supply but a plastic bag may be a good idea just for cleanliness' sake.

[IanJ]

Hi,

I recently built one based on an Arduino and AD595..........I didn't use PID as I simply followed a basic profile which I check against ever 2secs (y=mx+c between stages).
I used a cheap oven but used two heating elements from a slightly more expensive oven which I mounted up top rather than the original top & bottom setup.

My 1st test resulted in a cooked board.....
My 2nd successfully reflowed an op-amp (after adjusting many things).

Still have to play with timing, but here's what I have at the mo:-

int Time[5] = { 0, 90, 200, 230, 260 };  // Time (secs) Pre-heat start, Soak start, Reflow start, Reflow cool, Reflow finish
int Temp[5] = { 20, 150, 180, 220, 180 }; // Heat (DegC) to match the above timing

I followed a standard profile to start with but ended up stretching out parts of it due to too slow heat up.

In the end, the key I found is the placement of the thermocouple............I have it on a flexy which I can position immediately above the board. You can see the flexy in the pic below.

Will it be reliable for all sizes/shapes/density of pcb's..........who knows!

Ian.

[David_AVD]

I originally tried some (leaded) AIM paste and it seemed finicky to use.  Switching to a Multicore brand (from Mektronics) gave a much better result.

[mikeselectricstuff]

Results probably vary  more between different paste suppliers than difference in reflow profile

[David_AVD]

This is the paste we got from Mektronics. (no affiliation, just happy with it)

The previous one didn't flow nearly as well and there were always a fair few solder balls left afterwards.

[Mike Warren]

If you stick to leaded solder, reflow profiles aren't nearly as critical as the documentation makes it appear, at least for small volumes and homebrew stuff.
Bear in mind these will have been specced for absolute optimum performance for people using gazillions pf parts, for whom tiny fractions of % yield matter a lot.

I've done many hundreds of boards with nothing more than setting a fixed power level of "not quite glowing" , watching for reflow then turning off and opening the door.
On a few occasions I've accidentally left it for a couple of minutes over the normal 3-4 mins, and apart from some darkening of the PCB. no obvious ill-effects.

I've had similar experience. More than a thousand boards over the last 8 years or so. I bought a Beville convection toaster oven, planning to make a controller for it, but I found it very easy to get 100% success without a controller.

I switch it on at it's lowest temperature for 5 minutes to activate the flux, and then turn the temperature up to full and watch for the solder to flow. The moment it flows, I drop the temperature to minimum (it keeps climbing), wait about 30 seconds and then open the door and switch it off.

[Kremmen]

I have had success with a modified toaster oven controlled by a sequencing PID kiln controller fed back by a K-thermocouple. And contrary to what some say, a _properly tuned_ PID is just the ticket to an accurate temp sequence.
I moved all the heating elements to the ceiling to enable sufficiently fast temp ramp-up and double sided loads without risk of dropping the underside components (they do not reach liquidus). The thermocouple is an all-metal construction with a longish flexible probe that is in contact with the PCB to gauge the actual surface temp as closely as practical. There is also a circulation fan to mitigate hot spots and facilitate heat distribution.
The programmed sequence follows very closely the profiles specified by e.g. Altera and others http://en.wikipedia.org/wiki/Reflow_soldering
So far not a single failed batch. Of course it can be done all manually by Mk 1 eyeball as well, but the advantage of the sequencer is reliable repeatability.
Since the photos were taken, i have replaced the original grille with a more stable stainless steel grid and further improved the circulation fan. Those changes didn't impact the results but made handling somewhat easier.
The one manual phase remaining is rampdown after the reflow is complete. While the temp does not actually overshoot, it does ramp down a bit too slowly if the oven is left alone, so i usually open the hatch to let the PCB cool down more closely along the recommended rampdown.

The PCB is an example of a typical reflow job.

[Pat Pending]

Its reassuring to know that when the process is dialed in correctly that the results can be spectacular.
Could I ask how old the paste was in those cases?

As regards PIDs my oven is just plainly too slow to react.
As you hinted Kremmen, more elements is probably the key to a successful PID design.

My K-type is 1 to 2 millimeters above the PCB surface with the convection fan continuously running.
In terms of accuracy, the MAX6675 is within 1% of a nearby second K-type thermocouple & Fluke.

I ran another trial tonight with the profile attached. There was some solder beads but much less then
previously mentioned. In fact it was much more likely to see a few random solder balls resulting from excess paste.

Should I program the Arduino to slow down and extend the soak period or extend the
time above 200C or both, maybe new paste?

[ftransform]

True skill is BGA replacement with a blow torch.

[Kremmen]

Its reassuring to know that when the process is dialed in correctly that the results can be spectacular.
Could I ask how old the paste was in those cases?

If that was directed at me then the answer is, no older than a couple of weeks at most.

As regards PIDs my oven is just plainly too slow to react.
As you hinted Kremmen, more elements is probably the key to a successful PID design.

That and aligning the controller and oven time constants. Just from looking at the heaters it is easy to see that the controller pulls the power back well in advance of reaching the end of the ascending ramp. That is the D term in action, where the magnitude is relatively large but the time constant is slow, thus anticipating the behavior of the error signal very well. I seem to recall that the controller provided advanced configurability to account for the ascending and descending ramps separately, but this application does not need such sophistication.

My K-type is 1 to 2 millimeters above the PCB surface with the convection fan continuously running.
In terms of accuracy, the MAX6675 is within 1% of a nearby second K-type thermocouple & Fluke.

I ran another trial tonight with the profile attached. There was some solder beads but much less then
previously mentioned. In fact it was much more likely to see a few random solder balls resulting from excess paste.

Should I program the Arduino to slow down and extend the soak period or extend the
time above 200C or both, maybe new paste?

Soak should be 1-2 minutes at ~150C. Then a ramp up to ~215-220C (lead solders) so that the liquidus phase is not longer than 20-30 s max. Much longer and you run the risk of the intermetallic crystalline interface propagating excessively and making the joints unreliable/brittle. See this technical article for more details: http://www.solderingtech.com/articles/Intermetallics.pdf
New paste is always a sound idea.

[mazurov]

One advantage of PID controller is that you don't need to solder by eye and can cover the front door glass with reflective material (I use aluminum foil), which improves oven heating rate. I'm using a $30 4-element convectional toaster oven with Auber SYL-2372P controller and with door covered in foil I'm getting 2C/second ramp whereas bare glass gives me only 1C/sec. The foil adds about 5C overshoot though so the controller settings need to be adjusted accordingly.

The CO2 soldering link posted earlier in this thread is my project too. I since converted the oven to N2, results are similar.

[cwz]

Visually the joints look sound but under magnification it was a complete fail.

The volatile components of the solder had evaporated and the paste reduced to a solid.
When it eventually melted, it formed many tiny beads that did not coalesce.

Its harder to see but there is also a resin like fluid entrapping the beads.
I'm at a loss to explain what that is.

What it looks like here is a few things:

(1)  Those little balls are what the paste looks like under magnifications with the flux gone.
(2)  It looks like the flux didn't fully activate and that the heat was not even
(3)  #2 is usually because the paste was:  old, not mixed enough, or you didn't get good airflow over the whole thing

If you do any amount of hot air rework, you'll find out quickly that just heating something up is not necessarily a good way to get reflow.  Something really needs to break the surface tension of the solder and then it will quickly collapse.  If you ever need to see this in action, put some flux down around a QFN and heat it with a smaller hot air nozzle such that it moves around the perimeter of the part in a "wave".  It'll reflow much faster and more smoothly than if you just heat the part from overhead.

[Dreso12]

Your problem can be definitely the heat distribution inside the oven.

One big problem in reflow ovens is the temperature uniformity, it is necessary to have of moving hot air. I have recently find a nice post about this in this webpage:
http://www.nelectronics.dyndns.org/?p=297

It seems that even using an array of evenly distributed IR lamps the temperature distribution is not so good unless the oven is equipped with a fan.

[mark03]

Hi everyone, just found this thread.

When I started down the toaster-oven reflow path, I worried that the heat transfer in an average toaster oven was too much radiation (IR) and not enough convection.  Instead I purchased a commercial convection oven (from a failed Ben and Jerry's ice cream franchise) with the heating element and fan in a separate compartment behind the cooking chamber.  This was both a big mistake and a big opportunity.  With a similar heat output (1.5 kW) but lacking the direct IR path of the toasters, the temperature would not slew any faster than ~ 0.25 C/sec!  So, I retrofitted a 5 kW 220V nichrome duct heater and 3200-RPM centrifugal fan.  Now I have plenty of air and can do > 1 C/sec ramp-up.  I'm currently working on the controller.  The University of Exeter has a nice write-up on oven models and PID control, which seems to model my oven quite well:

http://newton.ex.ac.uk/teaching/CDHW/Feedback/PHY3128-CW070901-1.pdf

The primary motivation for this project is doing 1-mm BGA at home.  (Fine-pitch QFN and QFP are tedious by hand and I think the oven will be a lot more fun, but I have managed ok thus far, even some ground pads.)  So, I'd be curious if those with BGA experience could share how to choose a profile.  My paste mfg recommends a linear ramp, but the semiconductor vendors typically recommend profiles with a plateau (soak?) phase.  Which one should I give more weight to?

I'd also like to ask again, a question which came up earlier in the thread:  What is the range of acceptable TAL (time above liquidus)?  I am using leaded paste for now, although I do worry about lead-free BGA compatibility.  With 1 C/sec ramp up, 2 C/sec ramp down, and a melting point of 183 C, seems to me you are already at 45 seconds TAL minimum, longer if you follow the research on mixed leaded/lead-free assemblies and go to 220+ degrees.  How long is too long?

[Pat Pending]

Your problem can be definitely the heat distribution inside the oven.

One big problem in reflow ovens is the temperature uniformity, it is necessary to have of moving hot air. I have recently find a nice post about this in this webpage:
http://www.nelectronics.dyndns.org/?p=297

It seems that even using an array of evenly distributed IR lamps the temperature distribution is not so good unless the oven is equipped with a fan.

That's a really interesting link, the message I think is correct, either slow oven or uniform oven.

Heat distribution yes, but moving air in a toaster oven (from cold)...depends.

My Black&Decker oven had been wired for the convection fan to operate continuously. The thermocouple for the controller sat at the level of the wire slide in tray (mid rack position).
A second monitoring thermocouple sat beside it (< 1cm away). Both readings were within a few degrees C of the other.

Little did I know, but the air temperature 5cm lower in the oven was a full 35 deg C cooler, with a copper flood on the PCB its no wonder the solder balling appeared.
Maybe the oven needs to be preheated before use since operating from cold could be the cause.
 
In my case convection fans was a mistake. I don't use the convection fan any more and the temperature delta is an acceptable 15 deg C.

[Corporate666]

If you stick to leaded solder, reflow profiles aren't nearly as critical as the documentation makes it appear, at least for small volumes and homebrew stuff.
Bear in mind these will have been specced for absolute optimum performance for people using gazillions pf parts, for whom tiny fractions of % yield matter a lot.

I've done many hundreds of boards with nothing more than setting a fixed power level of "not quite glowing" , watching for reflow then turning off and opening the door.
On a few occasions I've accidentally left it for a couple of minutes over the normal 3-4 mins, and apart from some darkening of the PCB. no obvious ill-effects.

I've made tens and tens of thousands of boards using this exact method. 

The only times I've had a problem are when I forgot about the board for a few minutes and they came out tan colored... and even then, it was some popcorning of Atmel chips which hadn't been pre-baked before use.

I find storing susceptible parts in a dry environment to be infinitely more important than the reflow profile.  Can't imagine spending money on a reflow controller.