Toaster Oven reflow

[DaveW]

I've been looking at soldering at BGA and some higher density QFN packages at home recently and I started looking at using a toaster oven for solder reflowing. I couldn't find the oven used in the sparkfun writeup in the UK
http://www.sparkfun.com/tutorials/60
so I found this instead
http://www.amazon.co.uk/Andrew-James-White-Litre-Grill/dp/B003SE88Z4/ref=sr_1_1?ie=UTF8&qid=1295006018&sr=1-1-spell
The internal space is nice and roomy for any PCB I do at home



I took the side off to get to the controls. Very simple and the temperature control could be off as it's outside the oven but all the easier to modify!
I simply changed the thermostat out for an SSR-didn't even need to cut any wires!



A 3mm hole to get a thermocouple though and the modding is almost done



It does heat up quickly enough, 15-100 deg. C in 110 seconds, 15-200 deg. C in 246 seconds. I'll put a higher temperature thermocouple in there and get it under PID control with a PIC. Looks like a quick simple way to get into reflow soldering, judging by what other people have done should be easily capable to doing BGA packages which I've had to leave before,

[apex]

Hey, cool!
apex

[tyblu]

Very cool -- will enjoy doing this myself in the future. Is it pretty even heat, or is there a convection fan inside? (Does it matter?)

[Time]

hell, I have just done it on an unmodded toaster oven, thermo couple/meter lots with no problems

[Chasm]

The problems with such a conversion are usually the temperature probe and the distribution of the heat within the heater - and of course that you are playing with the mains.

The temperature probe problem can be solved by using something designed for this task, usually a thermocouple. But those are not necessarily cheap, and have some pitfalls. (You have to keep the other end of it at a constant temperature or know the current temperature to compensate for it.) OTOH cheap part bin solutions like a diode are not too stable and introduce nonlinearity problems.

Distribution of heat should be even. you don't want to burn one part of a board while another is not even melting. It can be improved with a fan -which of course has it's own pitfalls- or by using a oven that has top and bottom heating elements and thus heats up more evenly.

And then there is this mains thing... Let's say it this way:
Mains can kill you. Mains can burn down the house. Mains can kill your significant other.
And usually most important: Insurance companies have the habit of not paying at all when tampered with gadgets are involved.

So better know what you are doing before tampering with it.

[DaveW]

Very cool -- will enjoy doing this myself in the future. Is it pretty even heat, or is there a convection fan inside? (Does it matter?)

It does have a fan inside, you can see the back of it here
http://www.flickr.com/photos/21076036@N07/5354418734/#
I've tried the thermocouple in a few places and there seems to be some variation of temperature with height, but not horizontally so it should be able to keep a consistent temperature across a large board.

Chasm-this has top and bottom heating elements, four in total. Agreed on caution with mains-although this is one of the easiest mains modification I've ever done, the SSR just needs a 5V drive signal and provides full opto-isolation.

Good point with the thermocouple, the solution I'll be using is a thermocouple, with a to-92 temperature sensor to give me the cold junction temperature at the PCB

[JohnS_AZ]

Very cool (hot?)  

It's not my intention to slam your project, but offer the following as FYI ...

A lot of parameters that can be safely ignored when reflowing chip caps, resistors, and SOICs should be taken into account when trying to reflow BGAs and QFN/QFPs. Most importantly is the temperature profile. In a commercial reflow oven the temperature follows an exact and specific profile. In general terms, something like ....

PRE-HEAT ZONE - The board is SLOWLY brought up to an idle temperature, just below the activation temperature of the flux.  Most component guys recommend no more than 2 degrees C per second. The board will be brought up to about 170C in this zone.

SOAK ZONE (flux activation) - The board is then heated to the point at which the dehydrated flux chemically activates and does it's thing, and time is allowed for the entire board and it's components to reach the same temperature. The board will go from about 170C to 220C in this zone.

REFLOW ZONE - The board temp is then peaked into the reflow range of the specific solder paste being used. It was typically held there for another 30 seconds to a minute. The temperature depends entirely on your solder paste chemistry.

OUT RAMP (cooling) - Finally the board is SLOWLY cooled down to a temperature of about 90C over the course of another 3-4 minutes.

The importance of the gradual preheat is to avoid thermal stresses and fractures in the PCB and within the components.
The importance of the soak is to avoid solder balls and voids.
The importance of the controlled cool down is that you want it slow enough to avoid thermal stress, and yet fast enough that the solder can form the optimal granular structure to ensure a solid mechanical connection.

Now, this CAN be accomplished with converted toaster ovens, but I've never seen it done. The oven conversion projects I've seen published that pretty much drive the temperature directly to reflow, and then kill the power, make me grimace.  

There are lots of references on the web, but here's a pretty decent one...
http://www.rayprasad.com/home/rp1/page_70/smt_-_lead-free_reflow_profile_development_part_1.html

Good luck, and keep us informed on how it all works out!

[Chasm]

It does have a fan inside, you can see the back of it here
http://www.flickr.com/photos/21076036@N07/5354418734/#
I've tried the thermocouple in a few places and there seems to be some variation of temperature with height, but not horizontally so it should be able to keep a consistent temperature across a large board.

Nice, with 4 elements and the fain there should be a pretty even heat distribution.
Measure also with a (large) board inside to see if it changes the heat distribution.

Fan... Maybe you have to mod the fan to lower the airflow. But you'll know that if it blows small parts from the boards during reflow.

Chasm-this has top and bottom heating elements, four in total. Agreed on caution with mains-although this is one of the easiest mains modification I've ever done, the SSR just needs a 5V drive signal and provides full opto-isolation.

Yes, it is easy to do this in a safe way. After all this is a recurring problem. Still it is something to keep in mind.

I had some of the toaster oven projects form Germany in mind. Several of them use thyristors to archive a more flexible heating profile conforming to the requirements JohnS_AZ laid out. (Using gulse packet modulation or whatever the exact English term is, letting a couple full sinus waves trough, and then pausing for a while.).
Getting this right is a bit more complex than a relay.

[DaveW]

John-this is a good point. I try and stay away from the PCB side at work (don't want to ever get dragged away from the design and get stuck doing PCB layout), but I'll have a chat with the guy that does that part. I can easily put ramps in and from the tests I've done this oven should be more than capable of keeping up with those rates. With PID running I will be able to control the cooling as well. I have to confess I have no experience with BGAs at all, but hopefully the ramping rates and profiles should be similar to qfn chips.

Chasm-I did the measuring with a board (approx 300x120mm) inside on a wire rack and the temperature was even across it, although slightly different top and bottom. If this is a problem I should be able to fix that by adjusting the height inside the oven.
If the fan proves too powerful (looks unlikely) then I shall probably add some baffles to diffuse the airflow rather than mod the fan itself as it seems to help keep the temperature very even.
The solid state relay I'm using is quite capable of burst control (I think this is what you meant)-this isn't a mechanical relay, it uses mosfets and has a switching time in the microsecond range. I normally use triacs for this kind of thing, but I happened to have the SSR in the drawer and the faston terminals meant I didn't even have to cut the wires in the oven. I have used a similar system for the controller for an environmental chamber (415V, 15kW) before with excellent results.

[Chasm]

Burst control was what I was looking for, thank you.

The problem with many of the projects is that they try to get as cheap as possible.
Solid state relays like the one you used cost money, so do thermocouples and the AD595. Even space on PCB's costs money too. That can lead to unsafe designs.

OTOH there is eBay with both it's surplus and cheap Chinese reflow systems. So you actually have to take a look at the market and keep the cost at least somewhat under control.

Or you can make a business of it, take a look at auction: "reflow Oven TWS 800 SMT compact reflow and convection"  no link because it stinks so much of a firetrap....

[DaveW]

You aren't kidding about the AD595, nice chip though. I'll probably do a more discrete circuit as I've got some instrumentation amplifiers around, but useful chip to know. Looked up that reflow oven-looks terrifyingly similar to this homemade one! Thanks for everyone's input-some great links about the soldering profiles in particular
Hope it's been of interest,
Dave

[mikeselectricstuff]

You aren't kidding about the AD595, nice chip though. I'll probably do a more discrete circuit as I've got some instrumentation amplifiers around, but useful chip to know. Looked up that reflow oven-looks terrifyingly similar to this homemade one! Thanks for everyone's input-some great links about the soldering profiles in particular
Hope it's been of interest,
Dave

For a high-temp app like this, you can get away without cold-junction compensation - assuming your amplifier is at ambient temp, this isn't going to vary by anywhere near the other errors like evenness of heating. A simple instrumentation amp to boost the voltage, software lookup for linearisation   and offset by a nominal CJ temp would be fine.
BTW check that your SSR doesn't need heatsinking for the load you're running - most require a heatsink to run their full rated load.

[Jon Chandler]

I've been developing a thermocouple circuit using the TI TMP513.  It's designed for monitoring power supplies but it's perfect for this:

• I2C interface
• Built-in temperature measurement (for reference junction measurement) and 3 optional PN-junction measurements
• 12-bit ADC with a range of 40 mV for measuring shunt voltage (which is the perfect range for thermocouples)
• It's less than US$4 and available

I read the thermocouple voltage, do a table lookup to the nearest degree and interpolate to get an accurate reading.  The reference temperature is then added to the calculated thermocouple temperature to get the true temperature.  It's working well and calibrates nicely.

[Chasm]

Nice find!

[Jon Chandler]

I should also note about the TI TMP513....the shunt voltage ADC can handle positive or negative voltages so it's no problem if the reference junction is warmer than the thermocouple.

I've limited the measurement range in my circuit to -269 degrees C to + 760 degrees or so.  This keeps the table values in the range integer variables can handle.  That's plenty HOT for me!

[DaveW]

It would be good to put the electronics in the side of the oven, but that gets up to 65 degrees or so, so I will still add some cold side compensation-as you say though you could make this a lot simpler by keeping the electronics outside of the enclosure. This SSR is vastly overrated (about 4x) do I can get away without a heatsink-will probably bolt it to the side though.
That's a nice chip, like you say should be perfect for this; I think 760 degrees should be sufficient!

[Mechatrommer]

how about usb controllable oven from PC? with display of current temperature reading/ramp graph? think i want to build this, when i settle down.

[CafeLogic]

I just built a controller for griddle/toaster reflowing with graphing. It uses the TI ADS1147 which I highly recommend. I guess since there is a thread about it, I will finally get off my ass and shoot a video.

[Jon Chandler]

Here's a PDF describing the TI TMP513 thermocouple circuit and the registers needed to read the temperature.

I'll be posting an article soon on Digital-DIY.com with Swordfish Basic code for PIC18F-series parts.

[CafeLogic]

OK, finally prepared something. I will post schematic details and such later. I don't know how Dave finds the time to do this every week.










Videos:

[Chasm]

Nice!

One of the best conversions I've seen so far.

[Mechatrommer]

OK, finally prepared something. I will post schematic details and such later. I don't know how Dave finds the time to do this every week.

lol. do you use OS for the display? i can see the lcd interface is not the simple hibby jibby graph. is that menu-start-stop "soft buttons" functional?
cant argue with chasm more. this is the best i've seen so far, esp the UI.

[Zero999]

What about safety?

Have you connected the metal case to earth? I couldn't see any bonding on the photos and there's no way the case is double insulated from the mains; it's used as a heat sink and the thermal pad only provides basic insulation at best.

[CafeLogic]

What about safety?

Have you connected the metal case to earth? I couldn't see any bonding on the photos and there's no way the case is double insulated from the mains; it's used as a heat sink and the thermal pad only provides basic insulation at best.

Of course it is connected to earth, along with toaster shell. There is no thermal pad, The TRIAC is the insulated tab type. I'm not all that concerned about the shock hazard. I am concerned about the fire hazard, I wouldn't leave it unattended because there is no back-up if the TRIAC fails short. I thought about using a relay.

[CafeLogic]

lol. do you use OS for the display? i can see the lcd interface is not the simple hibby jibby graph. is that menu-start-stop "soft buttons" functional?
cant argue with chasm more. this is the best i've seen so far, esp the UI.

Thanks. NO OS. MPU is only 64K flash. I used the Stellaris graphics library for the buttons and some other stuff but the graph is a custom component. All buttons are functional. I will put up the source.