Adventures in home reflow - toaster oven conversion

[bit.cyber]

For anyone interested I've documented my example of converting a toaster oven into a reflow oven on my blog: www.lagrangianpoint.net
An oven sourced here in France and it uses the ControLeo2 contoller from whizoo.com. (Thanks to Peter at whizoo for the prompt replies to a number of different questions.)
There are some test temperature profiles detailed. Unfortunately no boards through the oven just yet as the stencils I ordered haven't yet arrived - always a crucial part missing in a build!

Here are the direct links:

• A reflow oven - the plan: lagrangianpoint.net/a-reflow-oven-the-plan
• The controller has arrived: lagrangianpoint.net/reflow-oven-the-controller-has-arrived
• The oven has arrived: lagrangianpoint.net/reflow-oven-the-oven-has-arrived
• Build basically complete: lagrangianpoint.net/reflow-oven-build-basically-complete
• Initial test runs: lagrangianpoint.net/reflow-oven-initial-test-runs
• Further testing: lagrangianpoint.net/reflow-oven-further-testing

I'll update this posting with definitive results once the stencils arrive, otherwise everything else is ready to go. Test boards - check, components - check, solder paste - check, stencils - no go (just yet).

[Dave]

You really went to town with the insulation. Amazing work, mate!

[bit.cyber]

I was a bit pressed time last night when I made the original posting. So for viewing pleasure, here are all the photos I've taken so far of the oven...

[bit.cyber]

And the few remaining...

[BFX]

Nice 
I'm impressed how good is your reflow curve.

[bit.cyber]

@Dave: in regard to the welding blanket... Maybe I just bought a crappy blanket - although it is rated to 600 °C and I'm fairly proud of the installation - but the insulation capabilities are not that great. For a while I was seriously thinking I should have instead bought a roll or two of ceramic wool.

But then my gloom lifted, at least somewhat, when I realised that the back wall has no shell between internal and external, i.e., they're one and the same thing. So in short: any heat saved is still going to go straight out the back. Unless I build an external cover, which would be possible, but I'm very unlikely to do.

C'est la vie when it comes to modifying existing equipment. I hope it shouldn't matter in the long run - although of course the mental note of not touching the back panel!

Anyways, temperature profiles - sans boards - to come tomorrow.

[bit.cyber]

So as promised here are the reflow oven's temperature profiles. There are two of each:

• Convection fan disabled, thermocouple in mid air
• Convection fan enabled, thermocouple in mid air
• Thermocouple wire away (as much as possible) from other wiring, thermocouple taped to sacrificial board

In the first four graphs the temperature profile differential is quite noisy. Particularly when the door is opened - go figure! I think the problem was that the termocouple wire was running in a loom with the low voltage control signals, and potentially it was picking up noise from the switch mode power supply(?). Or alternatively from the solid state relays.

The last two graphs have the thermocouple wire away from other wiring as much as possible. There's also a third trace being a 0.01 Hz low pass filtered version of the profile differential.

(Apologies for changing the axis lower and upper limits in the first four - poor style - although I was trying to understand the signal range. The last two have consistent and reasonable ranges.)

Overall pretty repeatable results, so looking forward to a couple of stencils to arrive such that I can try it out for real!

[bit.cyber]

I forgot to mention:

The cooling portion of the profile is very dependent on how far the oven door is open. The oven itself has a potentially nice feature whereby there's a "notch" to hold it open about 25-50 mm (1-2 inches). Based on nothing more than looking at the graphs - and perhaps unfortunately, but with no real evidence - I think the door should be held open a bit more. Real world testing might provide the answer...

[bit.cyber]

So the stencils for the (first) test boards arrived - yay! The stencils were from OSH Stencils and of the polyimide film type. I had no need to invest in stainless steel versions, and if the polyimide film works for test purposes then all the better! Solder was ChipQuick SMD291AX50T3 Sn63/Pb37 No-Clean T3 solder paste - liquidus temperature @ 183 °C - what I have at hand.

Thus to the question: does the oven work? Yes! 

The board in question is an NMdS HV Nixie PSU that I had made up a little while back - see: http://desmith.net/NMdS/Electronics/NixiePSU.html
I had ordered a couple of boards and two board's worth of parts, and therefore already have an operational version that I soldered by hand.

For viewing interest I've attached some photos. I don't have either a macro lens, nor a photographic microscope set-up, which is a pity! Upon looking at the solder joints with a 20x loupe under a 7x magnifying lamp they look good indeed. So it's a    result for a first board through an oven. Indeed where is the 'happy dance' emoticon when one needs it!

The temperature profile graphs will come in a following post.

(This "spare" board is basically the result of a glass or two too many of wine from the effort for the first board. For the first board I soldered both the IC and the electrolytic cap around the wrong way. The cap blew, and no doubt also did the switching converter. Thus I replaced both parts on the first board from the parts for the "second" board, and everything was fine. But it left me short for parts for this second board without ordering more parts... And thus ripe for a test through the oven!)

[BrianHG]

LOL, that PCB looks almost identical to my +5v to programmable 5-12v boost DC fan controller board.  Right down to the size, number & type of wire terminals & cap/mcu/mosfet/inductor placement & size.  The only difference is that I using a 8 pin pic to drive the mosfet with a 6 pin sip for programming...

[bit.cyber]

Interesting! Maybe a case of convergent evolution...

I had the boards manufactured from a freely available design from: http://desmith.net/NMdS/Electronics/NixiePSU.html
(all credit given to Nick de Smith at nick@desmith.net)

[bit.cyber]

As promised, find attached the temperature profile for the reflow soldering detailed in the last post.

The figure is much the same as the other figures, but I've not included the raw profile temperature delta/differential trace. It was simply too noisy, so only the 0.01 Hz filtered trace. The only explanation I have for the "noise" in the signal is that the oven now has its hood on (read: external case). I can't say why it has this effect, but it certainly didn't seem to affect the final result of what appeared to be a well soldered board.

Of course any comments or questions welcome.

[BrianHG]

Can you show me your thermocouple & processor connection/schematic & pin location?  Maybe we can see a simple reason or means to help clean this up easily...

[bit.cyber]

Thank you for the offer of help BrianHG! So... I'm using the ControLeo2 reflow oven controller - www.whizoo.com - and they don't make the controller circuit schematics available. I could ask to see whether Peter Easton would be interested in sharing.

Saying this, per the various graphs I've included in the posts there were differences when:

• I moved the thermocouple wire from the 'loom' of wires controlling the switching of the Solid State Relays (SSRs).
• Now putting the cover back on the oven.

With the cover off and the thermocouple wire away from anything else the noise level was OK, not great, but still what I might expect given the general layout of the system.
It would seem that putting the cover back on really changed the situation - I have no idea why.

In private communication with Peter, he confirmed that there is a level of software based filtering (I haven't taken the time to check), and indeed the single result so far has been good. So I'm not very concerned. I do have the data from a second equally successful run, using a different board/design, which I'll post in due course. We shall see. If it comes to it I'll be able to rework the controller software to include better filtering of the temperature signal, but am hesitating because I'm not sure what benefit it will bring versus the time spent for me to do this.

[bit.cyber]

And so to a second board - this time with closer pitch parts...

This is an SRAM memory expansion board for the Arduino Mega, which I make available here: http://lagrangianpoint.net/product/arduino-mega-sram-expansion-shield-pcb/. Also see http://lagrangianpoint.net/arduino-sram-expansion/ for details. (The circuit design is from Andy Brown - http://andybrown.me.uk/2011/08/28/512kb-sram-expansion-for-the-arduino-mega-design/)

The parts in question are:

• 44 pin TSOP 2 memory IC
• 20 pin TSSOP 20 octal D-type latch IC
• 3.2 mm x 1.6 mm LEDs
• 0805 capacitors and resistors

And the result? Again it looks fantastic!   
Now I haven't tested it, so judgement will be reserved, but typically I hand solder these boards and have never had a problem once everything is soldered. Certainly the soldering has the look of something better than what I'm able to achieve by hand. (I'll be the first to admit I'm not a professional technician/assembler, but I can solder, and still looking at this board the difference is obvious.)

Unfortunately I screwed up in regard to capturing the temperature data, so no profile for this board. However the experience so far is that the runs are quite repeatable. So mental note to self: don't stuff this part up next time! 

[bit.cyber]

The face palm was in regard to fact that I screwed the data capture aspect - for this particular real test board, i.e., to show results. As simple as not pressing the start button on the terminal program...

Otherwise I've so far done approximately 30 runs with the oven (sans components, but with sacrificial board), collected the data, reviewed the profiles, did a comparison and then tweaked. The graphs shown in earlier posts are representative, and are also repeatable across all the runs I've done. Alas I don't have so many test boards with components(!) to do a really proper job of tuning of the oven. However I've thought of this, as detailed in a different post: https://www.eevblog.com/forum/crowd-funded-projects/smd-reflow-test-board-interest/msg1175939/#msg1175939

Thus, just frustrating that I didn't do the data capture right for this second "real" run through the oven. As noted, I haven't tested the board yet - I need to solder the headers - but the visual inspection looks good.

[BrianHG]

In your setup, where are the thermocouples connected to?
The MCU board ADC input, or, a separate small board with the MAX31855 with a digital SPI interface?

[kkessler]

You did some beautiful work here.  I've been using just an old toaster oven for my reflow stuff, and, while it sort of does the job, it can't heat up quickly enough to get anywhere near the reflow profiles listed in the solder datasheets.

I've spent a lot of time eyeballing the Whizoo site, and it's nice to see a real world example of some else's successful build.  My big decision is whether to embark on this pretty involved build (which isn't cheap), or to buy a pre-built oven like the Qinsi qs-5100.

[bit.cyber]

@BrianHG: Included is a photo of the backside of the ControLeo2 board (from the Whizoo website) and it is a MAX31855 IC - confirmed via consulting the source code.

My initial suspects for the noise are: i) from the SMPS providing the +5 V DC rail (likely); ii) routing of the wire (likely); or iii) insufficient filtering on the board (maybe/maybe not).

In the latest run I've just done - being a true run with board and components to solder - the noise is +/- 3 °C. Given that the error in the MAX31855 is around 2°C and in the thermocouple probe around 3°C maybe I'm chasing something that doesn't actually exist...

[bit.cyber]

@kkessler: I have no idea in regard to the Qinsi QS-5100. It shows on the US Amazon as approx. US$525 - I don't know whether this is representative. I also don't know whether it's available here in France.

But in regard to costs for my build in EUR:

• Moulinex OX444810 oven                                 84.99
• ControLeo2 controller                                       75.00
• Kapton tape                                                        6.99
• High temperature wire (10 m)                           13.76
• Non insulated spade connectors                       11.30
• 500 W catridge heaters x 2                                10.51
• Fotek SSR-40 DA Solid State Relays x 3           26.70
• OMRON 2A Solid State Relays x 1                       3.34
• Aluminium sheet 300 mm x 200 mm x 1.5 mm  6.40
• Welding blanket 2 m x 1 m @ 600 °C                41.90
• Reflect-a-Gold tape                                           61.50
• Boom Mat                                                         47.64

So all up around €390. One shouldn't of course do a direct EUR to USD comparison because I think some of the parts would be cheaper in the US.

All in all I'm happy with the end result - it took a while to build it, but then again I like building stuff. The results have been great so far and it solves the problem that I had.

[ballanux]

Hi!

Great build and very detailed measurements! I'm just finishing my reflow oven, I have done the controller myself and I have been tuning the profile the last two days. I was going to post a question about the cooling rate and I saw your post, so let me ask it here 

You are measuring the temperature with a thermocouple in mid-air... but is this the correct way? I attached it to an old PCB to measure the temperature closest to what the components are seeing (see attachment)

This way I'm getting very slow cooling rates even if I open the door completely... Can this cause any problems? I see the recommendation of a maximum cooling rate of 4ºC/s but I don't see any minimum cooling rate recommendation.

The cooling rate I have is around 0.6ºC/s and I'm thinking to put a fan after opening the door to increase this rate... is this really a good idea? or could lead to thermal shock? The fans would have to be manually positioned after opening the door, but their speed could be regulated by the controller.

BTW, If I do the measurement with the thermocouple in mid-air, I get cooling rates very similar to yours, but I thought this was not the correct way 

edit: forgot to mention that I'm trying to do a Ramp-Soak-Spike type of profile as the PCB I'm doing has very generous planes and I want the temperature to be as even as possible. That's why my profile is a bit different to yours

[bit.cyber]

Hi ballanux!
In regard to thermocouple position, I moved to it being taped to an old board with Kapton tape. Much like what you included in your photo. (This was in the posts from around 07 April onwards.)

As for cooling rate - ideal, close to good and problematic - unfortunately I'm not that knowledgeable. For the oven I've retrofitted, the maximum temperature decrease is a around -1.0 °C/s, but then it quite quickly moves closer to approximately -0.5 °C/s. An early test run, with the thermocouple floating and the door immediately opened completely resulted in the attached profile. It would seem that the oven has enough thermal mass that beyond short term changes, differences in operating actions don't result in large longer term profile differences.

Mark Goldberg has detailed using fan assisted cooling on his oven page at: https://sites.google.com/site/markscontroleo2build/. Maybe this will provide some more information for you.

[ballanux]

Thanks! it is very useful, it seems that I should include a fan to have a decrease rate of 1 to 2 ºC/s and also explains that it creates better grain structure on the solder, which I didn't know 

[bit.cyber]

While making a reflow oven was fun and resulted in a good addition to the toolset available, the real reason was to be able to complete an LED design that is in the works.
Originally - and naively - I though that it shouldn't be a problem to solder a metal core PCB. Heat the board up, maybe quite a bit, and then "attack" it with a soldering iron. That feeling of 'oh oh' when things don't go to plan can be a sinking one... Particularly when on then thinks: now what? 

So with the oven now built and performing (approximately?) as it should, what better way to test it than on something more serious. Time to put the game face on!

Attached are the various photos taken and a graph of the temperature profile. And to the question: did it work? Yes   
(Really! Where is that happy dance emoticon when one needs it!)

More seriously: I had enlarged the pads - beyond the size from the LED data sheet - presuming that I would hand solder the LEDs. So they're not so accurately located, but c'est la vie - the life of a prototype board. Beyond this the board lights up like Clark Griswald's house at Christmas. 

[matkar]

@BrianHG: Included is a photo of the backside of the ControLeo2 board (from the Whizoo website) and it is a MAX31855 IC - confirmed via consulting the source code.

My initial suspects for the noise are: i) from the SMPS providing the +5 V DC rail (likely); ii) routing of the wire (likely); or iii) insufficient filtering on the board (maybe/maybe not).

In the latest run I've just done - being a true run with board and components to solder - the noise is +/- 3 °C. Given that the error in the MAX31855 is around 2°C and in the thermocouple probe around 3°C maybe I'm chasing something that doesn't actually exist...

Interesting. If I remember correctly from looking at the code some time ago it makes 5 samplings per second and averages the temperature. I'm surprised it fluctuates that much. You probably pick up some noise from the surrounding wires. Did you install thermocouple wire in parallel to power wires?