Another reflow oven build. What quartz heating element locations are best.

[clytle374]

So I just threw out my Goodwill toaster oven as it wasn't suitable.  I got 2 $20 ones from walmart, quartz elements and 1050W.  I was going to install all 4 elements in one oven, ovens are cheaper than just buying the elements. 

I have 20A (I live in a 120V part of the world)outlets on my bench, so I can probably run all 4 of the elements at once(technically, not properly, but PWM can solve that).  At least for a short time.  Being that the elements are both centered from to back, one on the top and one on the bottom, I am going to have to move them to add more.  I cracked the very end off one of the quartz tubes and then realized I can cut them down with a diamond wheel.  So I decided to build a insert for the oven with thin sheet metal.  This will allow much better insulation, reduced volume, and raw metal without holes for the element mounting.

Being that I keep reading about direct infrared heat causing issues I'm wondering if having one element on the top and 3 on the bottom is better?  Maybe all 4 on the bottom and move the rack to the top as heat rises?  Also can't find any good references as to if I want the PCB closer to the top elements or the bottom ones?  Another reason I would like to put all the elements in the bottom is that aluminum will spread the heat faster with less thermal inertia than steel,  but I think the quarts elements might melt the aluminum so if I put all the elements on the bottom I can make the bottom out of steel and the rest out of aluminum. 

Any advice?  Thanks

[RoGeorge]

Most PCBs these days are very small.  Unless you are working with very big ones, maybe use something smaller:
https://hackaday.com/2017/07/13/a-bright-idea-for-reflow-soldering/

I've tried a kitchen oven once, mine was underpowered.  Had to wrap its inside with Al film, and stop its fan.  It was about 2kW, 2 elements up, 2 down, yet it was not able to do steep enough temperature slopes.  Being rather big (the size of a plate) it was also cooling too slow, so had to juggle with hand opening the door.

Overall, didn't work very well, and the plan was to use something smaller next time, which I never did since.

Few random example that popped in the search:
https://youtu.be/6bxpy1Ret6Y
https://youtu.be/C7blZigaaaA

After all, it's a hot plate.  Pretty much everything works, even a pan with sand heated on a stove. 
https://hackaday.com/2019/06/05/solder-smds-with-a-pan-o-sand/

[clytle374]

Most PCBs these days are very small.  Unless you are working with very big ones, maybe use something smaller:
https://hackaday.com/2017/07/13/a-bright-idea-for-reflow-soldering/

I've tried a kitchen oven once, mine was underpowered.  Had to wrap its inside with Al film, and stop its fan.  It was about 2kW, 2 elements up, 2 down, yet it was not able to do steep enough temperature slopes.  Being rather big (the size of a plate) it was also cooling too slow, so had to juggle with hand opening the door.

Overall, didn't work very well, and the plan was to use something smaller next time, which I never did since.

Few random example that popped in the search:
https://youtu.be/6bxpy1Ret6Y
https://youtu.be/C7blZigaaaA

After all, it's a hot plate.  Pretty much everything works, even a pan with sand heated on a stove. 
https://hackaday.com/2019/06/05/solder-smds-with-a-pan-o-sand/

Thanks for the advice.  I almost went with the hotplate method.  The reason I asked about the elements on top is due to reports of top heating damaging some components, and removing them would allow the total volume to be smaller. I'm wondering if the lamp method works because it doesn't put out as much infrared. It is possible that I have made a mistake in choosing this design, but I'm too far into it to scrap the idea without giving it a shot.  Heating up shouldn't be an issue with 2Kw of quartz heaters.  Cooling will be just opening the door with a servo. 

[RoGeorge]

Not saying your choice is wrong.  There are plenty of good working reflow ovens like the one you want to build.  Only saying 2kW was not enough for mine.

Try a test first (without taking the time to modify anything), turn the oven full on, with nothing but a thermocouple and an empty PCB inside.  Use a DMM set to measure the temperature with its own thermocouple.  See if the oven can heat fast enough.

Another thing, there's no need to PWM a 2kW AC load, because thermal inertia is very big.  ON/OFF control will be enough.

I've used a triac + optodiac (optotriac) with zero crossing detection.  Don't skip on the optical isolation.  I've used MOC3063-M and BT139-600E connected like in the typical application shown in the MOC3063 datasheet (Fig.10).

Attached are the temperature charts measured inside my oven, interior was about 20x30cm (~8x12inch or so).  These temperatures were measured with the oven's door closed, 2kW total, all 4 heaters full on, with and without Aluminium foil inside.  The green trace (after its first peak) shows cooling with the door closed, then with the door open (then, after some cooling, it was powered again a few minutes later - the last part is irrelevant).

[MarkT]

I use only two elements on top, as I don't want bottom-side components falling off (not sure if this is really an issue), but I only do small boards.  Two elements makes for a fairly flat spatial distribution of temperature as they are spaced about the same as the element-to-board spacing.

With a mix of large and small components I use a low temperature pre-soak first to bring the large one's up to similar temperature.

I've noticed flat-flex connector locking tabs are most sensitive to radiant heat issues.

[clytle374]

Not saying your choice is wrong.  There are plenty of good working reflow ovens like the one you want to build.  Only saying 2kW was not enough for mine.

Try a test first (without taking the time to modify anything), turn the oven full on, with nothing but a thermocouple and an empty PCB inside.  Use a DMM set to measure the temperature with its own thermocouple.  See if the oven can heat fast enough.

Another thing, there's no need to PWM a 2kW AC load, because thermal inertia is very big.  ON/OFF control will be enough.

I've used a triac + optodiac (optotriac) with zero crossing detection.  Don't skip on the optical isolation.  I've used MOC3063-M and BT139-600E connected like in the typical application shown in the MOC3063 datasheet (Fig.10).

Attached are the temperature charts measured inside my oven, interior was about 20x30cm (~8x12inch or so).  These temperatures were measured with the oven's door closed, 2kW total, all 4 heaters full on, with and without Aluminium foil inside.  The green trace (after its first peak) shows cooling with the door closed, then with the door open (then, after some cooling, it was powered again a few minutes later - the last part is irrelevant).

Thanks! No the oven would not heat up fast enough with only 1kw, that's why I bought 2 ovens for the extra elements.   After putting the insert in it it will be 8" wide, 5" high, and 6" deep with koawool around all sides.  I'm going to try putting all 4 elements on the bottom with a bit of shielding.  The the bottom and up to the elements will be stainless steel and the rest will be .008" aluminum.  Looks like pulling the door all the way open is the best bet to cool down in time.  Might just use a weak spring and hold it shut with an electromagnet.  Thanks again.

[clytle374]

I use only two elements on top, as I don't want bottom-side components falling off (not sure if this is really an issue), but I only do small boards.  Two elements makes for a fairly flat spatial distribution of temperature as they are spaced about the same as the element-to-board spacing.

With a mix of large and small components I use a low temperature pre-soak first to bring the large one's up to similar temperature.

I've noticed flat-flex connector locking tabs are most sensitive to radiant heat issues.

mmm, I just went the other route.  Thanks, won't be my first screw up if this doesn't work out. Thanks!

[RoGeorge]

Found some build pics with the underpowered oven I've tried:


Oven, thermocouple, MCU board, LCD display, power supply, isolated mains controller


2 more elements are at the top side, not visible in this pic, total 2kW heating power


2 channels thermocouple reader PCB with 16bits ADC (schematic is from the ADS1118 devboard)


Power supply + thermocouple ADC + isolated mains


Oven power control schematic with optical isolation (MOC3063 + BT139)

[clytle374]

Found some build pics with the underpowered oven I've tried:

Thanks, I believe that I see your problem already... I did study this subject quite a bit.  Looks like your oven has the metal cased resistive elements in it.  They are horribly slow to change temp,  the quartz tube types will change temps in a fraction of the time.  I am playing with the idea to put those from the first oven in this one as well.  My thinking is they heat up so slowly I can ramp up with them and then switch to the quarts heaters for reflow.  Since the first ones have so much thermal mass they will still be heating while the quartz heaters come up to temp. Really tempted to do that to 'hedge my bets' so to speak.  That would basicly allow me to pump 3 Kw into the chamber at once.  Nice build work by the way, I love the SMD part on stilts.  I'm also going to have a lot of insulation in the walls of the oven.   Decided all the elements are going in the bottom due to ease of mounting.  Of all the things I'm second guessing that is the one.     

[RoGeorge]

The suspended SMD was because ADS1118 ADC apparently was keep behaving erratically, but thank you.  I've put the small PCB on a socket after I've changed 2-3 of them, only later to realize that the temp errors I was seeing were in fact reflections in the data wires.

After I've added some 33ohms series resistors to the SPI wires, in order to mellow down the edges a little, turned out all the ADS1118 I've suspected to malfunction were working just fine.  And so the suspended chip, because of repeated desoldering while keep replacing and testing them. 

Here we have 220Vac, and max 2kW per socket outlet, with about 5kW max for all power outlets together.  Don't know the wiring regulations for 110Vac.

With 3kW should be no problem to heat it all fast enough, if the power outlet allows that much. 
Good luck with the build. 

[clytle374]

As with all simple projects, I'm back like 2 months later and asking for advice.  After all the horror stories about not enough power I put the 4 ~500W quarts elements in and the 1000W metal cased element.  The plan was to only use the metal one in the beginning as a boost as it would retain it's heat when I went to the quartz elements, I don't have 3KW at my outlet anyway.  The boost element was unneeded and made PID awful to tune.  So i removed it, good thing I spent so much time modifying the program for it.

I started with the code from git "Tiny Reflow Controller"  changed the thermocouple controller, added menus to change the all setting and store them to eeprom instead.  Had to switch to a ATmega664P as the 328 was out of memory for the oled display even with stock code, libraries have changed I think.  I am not a programmer, so this was a large portion of the problem. 

Just thought I'd share and ask for advice here.  I've been looking into adding "take back half" as an improvement, but I don't think that is going to fix the overshoot much, if any.  I might removed the insulation under the oven chamber to dissipate some heat build in the steel under the elements.  Or switch to 4 500W halogen bulbs. 

The bottom half of the chamber is steel and the top aluminum flashing.  Surrounded with ceramic insulation.  I had to shorten the quartz tubes with a diamond wheel.  Was trying to mount the chamber and used sheet metal screws as stay bolts,  if it's good enough for a steam engine boiler... lol.

Here is a graph of my results.  I added a preheat ramp 2C per sec (here it is set to 125C) as the original code just switched the elements on and I got a 9C per second rise.  Then code then ramps up to 200C which then triggers the reflow portion, that is still original code.  Then I changed again to a ramp 2C per s until the original code cuts off at the 250C set-point -5C. 


I think the tune is as close as I can get it. But a 25C overshoot is almost guaranteed, or cutoff early and coast up to the setpoint.  Maybe add cutout parameter to shut the heat down early and then catch it when it levels off?   So am I screwed and need to change elements, or thing this can be made close enought?  Thanks.     

I found this and needed to share it here

According to Linear Technology’s Jim Williams, “The unfortunate relationship between servo systems and oscillators is very apparent in thermal control systems.” (Linear Applications Handbook, 1990).

[RoGeorge]

Looks good.

I would give it a try as it is, and tinker more only if you don't like the result.  Most of the components are not that sensitive to the temperature soldering profile.

The most sensitive I know are the electret microphones.  The elctret inside lose its dielectric polarisation at high temperature, just like a permanent magnet lose its magnetic field when overheated. Solder any electret microphones manually, and at an as low temperature as possible (usually specified in the microphone's datasheet).  They do not melt, but they become less sensitive, sometimes almost death to any sound.

[clytle374]

Looks good.

I would give it a try as it is, and tinker more only if you don't like the result.  Most of the components are not that sensitive to the temperature soldering profile.

The most sensitive I know are the electret microphones.  The elctret inside lose its dielectric polarisation at high temperature, just like a permanent magnet lose its magnetic field when overheated. Solder any electret microphones manually, and at an as low temperature as possible (usually specified in the microphone's datasheet).  They do not melt, but they become less sensitive, sometimes almost death to any sound.

Thanks for the encouragement.  I've spend the last couple days modifying code and making it less of a mess and more configurable.  Including playing with a few tricks to help with the overshoot, I'm far from a programmer, but enjoying it and getting to something I'm proud of.  Like boosting the PID set point and waiting for the temperature increase to hit the right rate then switching into the ramp function on top of it.  The cutting out early and letting it coast up to the next temp, if the change in temp during that time gets to small the program starts adding little bumps into the set  point just over the current value.    Going to finish a few things that are driving me nuts then add the door servo.  Here is my latest results.   

[RoGeorge]

Looks even better now.  If the door servo takes too much tinkering, maybe just emit a beep when it's time to open the door, and open it manually.  2-3 minutes is not too long to wait near the oven, and you would probably want to look at it anyway, to see the moment when the solder paste melt, and the components floating towards each toward the middle of their own pads.  That melting moment is very eye catching. 

Another reason to supervise the reflow in person, is that you can observe if something is not OK, so to correct the process for the next board.

And yet another reason would be that it's not good to let high power heaters unsupervised.  They are always a fire hazard, no matter how well they are built.  Sometimes very unpredictable things happen, like a pet or some other critter getting scared and tipping objects, a component failure, etc.

[clytle374]

Thanks, I tool the ramp function out of the soak section that the original code had.  I plan to add it as an option.  My understanding is that it was there due to most ovens not being able to make the ramp into reflow from the soak temp.    I absolutely will not be leaving the thing unattended.  I just want the servo on the door to be able to get some level of consistency in the cooling speed.  Getting the servo to work was easy, mounting it and the linkages will be more difficult.  I put the code up on github, but just as a remote backup until it is less embarrassing.  I'm not a programmer.  If it works out well I'll put the whole build on github. 

I've been thinking a lot about the safety aspect, even while watching it.  I used an optoisolator between the processor and the SSR, and using a battery for the SSR side to get 100% isolation. Battery is temporary, thinking a charge pump might be a good option as it can only fire the SSR if the processor is running.   I googled for a while and looks like that might be overkill as the SSR has an optoisolator inside?  Also added a watchdog to the code to be sure the processor is still running, refer back to "I'm not a programmer".   

I don't trust the SSR output to not fail shorted either.  Other than an external safety relay I guess the program catching it and setting off the buzzer is the only reasonable safeguard.  Me just zoning out while it's running is also a hazard that must be considered.