Best configuration for 4 thermocouples

[Sothis-A]

First, thanks to the community I was able to successfully mod a T962C version 2 (the one with the exhaust cuffs for vent hoses)! I made a custom PCB with 4 MAX31850K thermocouples and a USB interface onboard, and was able to flash the new firmware without much fuss.

Tonight I did my first practice run with the 'custom 2' profile and SAC305 paste. The displayed temps were WAY lower than the target curve and didn't reach 200 deg C before cooldown. However, just before the fans kicked on I heard some popping sounds inside which subsided shortly after. Upon opening the drawer I was shocked to see that solder flowed and were brown with slight charring on the pad edges. The silkscreen was yellowed, too. Clearly the oven temps were actually way too hot!! The resistors I tested with were all soldered nicely.

After letting things cool a bit I went into the bake mode menu and stumbled on what was going on. My L & R thermocouples (which are ambient like stock) were registering 50C, but my X1 and X2 thermos, which were in vias on the boards, were still measuring 100C. My assumption is that the PCB X thermos got much hotter than ambient and so the oven controller followed these temps (the code apparently follows the hottest one if its significantly higher than the others), but was displaying the L+R average (ambient) on the screen, which were lower.

This makes me wonder, what is the best configuration for accurate temps? Should I swap and make the PCB thermocouples L and R and the ambient ones X1 and X2? Or is the way I have it OK, I just need to calibrate L and R to account for them not having the PCB's thermal mass? Or should I get rid of the ambient ones and make them all able to be attached to the PCB? Or something else even? BTW, all of the thermocouples are within 1 deg of one another measuring ambient and agree when I bring the tips together or measure ice water.

Thanks so much everyone!

[GitLang]

Maybe the best way is to follow the adage of "the best part is no part". Why are you trying to use 4 thermocouples?. The 2 originals work absolutely fine, as witnessed by the many boards produced by a lot of us here and elsewhere. The Maxim chip interface seems to be a necessity though as the original thermocouple interface is poor. Can you tell us what "The resistors I tested" are?

[Sothis-A]

Good point. I'm trying to use 4 simply because of the "Better Thermocouple Interface" mods from the wiki; I assumed there's a good reason for the extra 2.

The resistors were just some 0805s I placed around the panel in various places to see if there were obvious temperature variations.

[GitLang]

The reason there are 4 is because the Maxim interface board has 4 inputs. Doesn't mean you have to use them!. I think 4 may help get a better average of the whole cabinet, but I don't think the PCB cares about individual thermocouple temperatures, just the temperature at each to-be-soldered pad. Any pad spending enough time in the right air temperature will reach that temperature, no matter if it is a feather or a block of copper. Maybe start with 2 thermocouples. Ensure it all works OK and if so add the other two in one by one. From memory, I believe the software can use the average for temperature control.

It sounds to me like your temperature feedback was reading lower than reality. That would account for the temperature display showing lower than the setup, and the actual conditions being too hot. Can you get debug information over the serial port or USB?. It would be good to log measurement of all thermocouples individually at different sets points in bake mode. If possible, an independent thermocouple and meter as a check at the same time would be useful. If that all looks OK, then from cold get set up as above and put the set point at 150C or similar and see what they read as they heat up and if they undershoot or overshoot the endpoint. If the PID is set really badly you may even see oscillation.

On a small, semi DIY prototyping oven like this I wouldn't, personally, even think of trying to measure board temperature. I have always believed this system is an air temperature control system and the board will follow on, lagging the air temperature a bit. Don't set your degC/sec slopes too high, and think air temperature. Like most temperature control of air, you have a cooling input and a heating input. You can't control without the cooling element as the natural cooling is way slower that the control loop needs to be. But it is very sensitive to PID settings. It took me a while to get it right after fitting the Maxim based board, but once right it performs great. Great, for this oven, is lagging no more than about 15C behind setpoint. It's not a problem as we're not interested in the set point curve, only the actual temperature curve. If it all happens a bit after the setpoint, what matters?. Just keep it slow, within the component soldering spec and leave a good plateau about 20 degC below liquidus. I've always got away with 60 to 90 seconds at 165 degC.

I've wandered a bit, but I'm just trying to make the point that a reliable measurement of cabinet air temperature (for this oven) is vital, and a temperature curve which reflects what is needed to do the job, is way more important than it lagging 20 degC. In an industrial soldering system you will see radiation dominant control rather than air temperature and it then become valid to monitor the actual PCB temperature. On our little oven we have to control the air and know that we will lag behind a bit, but other than a small amount of direct radiation heating, nothing on our board is going to exceed that air temperature. A very different situation to the IR tunnels which are almost 100% radiation dominated. I'll shut up now :)

[Sothis-A]

Thank you so much! This is great information to have and it all makes sense. As for getting the debug information via USB, in theory yes. That's the reason I built a USB interface on the same board as the TC interface...I had visions of computer control, temp logging and more profiles. But...I don't have that setup quite yet. Is there software for this you can recommend? I think I've seen one on GitHub but I haven't messed with it yet.

I just did some playing around in 'bake mode' like you suggested. First I ran it at 100 deg C with the two TCs I used for the PCB not attached to anything. All 4 sensors were consistent.

I repeated this test again but with 2 TCs in the same holes I used last night. Things stayed close, with the PCB temps lagging behind ambient temps until around 50 deg C, at which point the PCB TC's over took them...and kept going. When we got to 100 deg C, the PCB sensors were both reading around 150 deg C! I used the TC on my Fluke meter which agreed with the PCB TC's readings. Then I realized that I'm using black PCBs which are supposed to absorb more infrared energy.

So, I went and got some green PCBs and repeated the test. This time the PCB TCs read 120 deg C when ambient was 100 deg C. Again, the Fluke agreed.

To your point, that the PCB temp can't get hotter than the air temperature, I wonder if what I'm seeing is unique to the C model (the big one)? Is that the one you use?

That got me thinking that my original test boards were both small (6x8") relative to the size of the chamber. So I grabbed a 17"x 10" black PCB and performed the above test again. This time, the readings all stayed similar except for a bit of thermal inertia in the PCB where they got around 110 deg C. It's very interesting and kind of suggests at least in this oven, radiation does play a bigger role?

[mikeanton]

@Sothis-A I can tell you from having a lot of experience in profiling ovens that the PCB temperature can certainly get much higher than the air temperature. This is because it can absorb IR energy sometimes much better than the air can. On my oven, the temperature of the board will reach the highest temperature just after reflow, when the oven starts to cool down. Probably because it is removing heat from the elements, and the PCB hasn't caught up yet.

The only reliable way to get the profile you want on the PCB is to measure the temperature there with another meter, and adjust the air profile until the temperature reads correctly on the PCB. This usually take a number of reflow cycles to achieve through trial and error. Different colors, and patterning on the board changes the profile you need to use, so I tend to profile the oven for each design I am building.

[GitLang]

@Sothis-A Yes, you're right, I'm sure. There's no way it can be all air temperature, there has to be a radiation component. Anywhere there is a heat source, some cooler things with mass and some gas you are going to get radiation, convection and conduction, and even the best CFD program has a job to tell you exactly what the resultant temperature of the masses will be. The industrial ovens I have known¹ have large areas of IR radiating heaters and not much air temperature. Our DIY ovens have not had a lot of thermal analysis go into the design. Two or Four linear heaters seem to do a better job of heating the air than direct radiation heating the board. For the industrial oven, yes you would take feedback from the board temperature and do test runs until you get it right then put the 200000 others through :)

So far, the biggest SM component I have used is a 30A inductor about 22mm square with two pads around 12mmx6mm and the smallest is 0.5mm pitch chips and no problems or rework yet. The biggest batch I have done is 8 boards, done 4 at a time (100mmx80mm). I always go just by common sense and the procedure I outlined above and in more detail in a post several months ago. I haven't once taken the temperature of the board, not seen any charring or burning or dark flux whatsoever. I could have been very lucky. I maybe only just creeping over the liquidus by a few degrees and leaving a metallurgical nightmare behind, I don't know.

¹ my experience of industrial PCBA isn't up to date though, I've been retired a number of years now. That could explain the Abe Simpsonism creeping into my posts. Sigh.

[Sothis-A]

Thank you @mikeanton & @GitLang!! It's been very tricky so far!

Just to see, I decided to swap my X1/2 (PCB) and L/R (air) TCs so the machine followed the PCB temp and the machine sort of follows the profile better but this oven has A LOT of thermal lag (my bet is the PCB's thermal lag compounds this). Despite lessening the ramp slope and upping the temps to spend more time in reflow, I now have the opposite problem in that I'm getting some good joints but other (large) areas aren't soldered. I keep upping the profile temps to spend more time in reflow. Part of this could be that I'm using the profile from my paste manufacturer (MG 4900P) but it has the profile peak briefly at 224 deg C, but apparently SAC305's liquidus temp is 220 deg C. That seems way too close, even with oven overshoot, to have enough time to reliably wet.

Have either of you tried the code modification to eliminate the PWM of the heaters? Apparently the T-962C's SSR chokes the heater current pretty bad with PWM. I found a few references for this and will likely try it.

The fans, pointing out, are pretty useless for cooling on this unit too. I'm wondering if turning one around so that it draws air in while the other exhausts isn't worth a try. Or stronger fans.

Thanks again guys, and @GitLang I appreciate the humor along with the great advice.

[mikeanton]

@Sothis-A I use GC10 SAC305 solder paste, which I highly recommend if you are looking for an alternative. It claims a liquidus temperature of 217C. A reflow temp of 224C is closer to what you would use for leaded solder (I usually used 215C for eutectic leaded which reflowed at 183C). Usually I shoot for a peak reflow temp of 250C for SAC305. Generally you don't want to go over this, as many components don't like much higher than this, and I just shoot for hitting it, and not really staying there for any length of time. Usually I'm above 217C for around a minute, which works pretty well usually, unless there are some very large components. For large component boards, I've found it helps to have an initial preheat of around 70C for at least a minute, to get the temperature of the oven and board to not start from cold. Then try to get the soak temp to knee over at 150C, which I often have to overdrive the air temp by about 20C or so to hit, and ramp during the soak period up to 200C before starting the reflow. Usually I target soak from 150C to 200C to be about 90 to 120 seconds in length, longer helps to get the board to a more even temperature before you start reflow.

I don't have the control thermocouples on the PCB as that would be very difficult to manage for production, and the temperature can vary greatly across the PCB anyhow. I usually profile the board by picking a pad on the board that is well tied to top copper, or inside planes, like a tab on a large transistor, or a power pad under a package. Then keep adjusting the air temp profile, until the PCB temperatures match the desired profile. You will find that different parts of the oven measure differently, so I tend to put the PCB in the same location and orientation as when I profiled the board.

I also added small diameter wires running from the front of my oven to the rear that are tensioned with springs, and these are what the board sits on. They are raised off the bottom of the drawer by about 6mm, or 1/4". This helps to decouple the PCB from the drawer, so that the PCB can heat up faster. Even just putting the PCB on standoffs that are screwed into the drawer through the slots would work too. The wires just make it easier to accommodate different PCBs without having to move the standoffs. You can find a picture of this and further discussion about this here on another thread: #228 (comment).

[GitLang]

@Sothis-A Mikes's advice on temperature's is spot on. The trouble with such high temperatures on a cheap DIY oven is that because the control is bad, you may set a peak of 245 degC or similar and hit 15 degC above that which may toast some components. Whatever you choose, trying to peak at 4 degC above the liquidus on this oven is unlikely to solder many joints.

Do you absolutely have to use lead free?. Advantages - it has no lead. Disadvantage - high liquidus, up to 40 degC higher than leaded. Also has poor wetability. That can be improved with alloying elements but the cost can be increasing the liquidus. Alloys containing bismuth have lower liquidus but have their own problems.

Normal 63/37E melts at 183 degC hence my advice to allow a plateau around or below that temperature before peaking for the final flow. That allows the flux to fully flow and the metal to be ready to melt during that final kick, which in my case is 210 degC for 20 seconds at peak. I didn't realise (or I'd missed the fact) that you were using lead free. Again, same experienced as Mike with the set point having to be set above what you need. I also keep the board off the tray, I use an M3x6 and nut (St/St) in each mounting hole. I also preheat the oven at 50 degC for 10 minutes, put the boards in and continue for another 10 minutes.

I'm using ChipQuik TS391AX250

[Sothis-A]

@mikeanton That was you (the wire standoffs)!! That was a great idea. I almost stole it after seeing it but opted for some adjustable brackets instead:

As luck would have it, for bigger boards, the thumb nuts that attach the bracket to the drawer are the perfect height to support the inner PCB with that little bit of screw protruding. Too early to tell if this will work in the long run but its promising so far.

Thank you for the paste suggestion, I'll look into it! Your info also confirms I'm way too low, temp wise. Surprising that MG's recommendation is so low for the same alloy.

[Sothis-A]

@GitLang OK, that likely explains the current issue of being too low temp! And ya, I do have to use LF to adhere to EU regulations. Funny, I switched to lead-free solder right at the start of my electronics journey. I still have about 1/2 a roll of Kester leaded, which was my first and only I ever bought! Despite not having any problems with the LF solder, I would love to switch back.

You both mentioned the preheat and I've noticed that, the preheat temp makes a big difference in how well the oven follows the profile. I've been trying 50-60 deg C, but I accidentally started the ramp test and the oven got to 90 deg C before I was able to stop things and load the correct profile. That nearly matched, even on the ramp up to reflow which 50 - 60 deg C really struggles to do.

[sevstels]

I saw it and I'm sitting in shock. Don't understand, why go through all this pain? Everything has been done long ago and works perfectly well: https://en.gradient-sg.com/t962/ Yes, it's hard to read a lot of letters, it's easier to do bullshit.

[HattonLe]

Just finished my upgrade project for an original T962 oven. Completely rewrote the code in C++, designed and installed a I2C dual MAX6675 thermocouple board, reworked the PCB to make the JTAG interface functional again, implemented the cold junction sensor but for use as an ambient temperature sensor. Also added a USB serial interface with external reset\program push buttons, replaced the Optos with zero crossing ones, and replaced the system fan with a blower one. Oh and I upped the EEPROM storage to 16kb as well.

So why did I go through all this pain.... because I have the best T962 oven on the planet.

If anyone is interested in my I2C Dual MAX6675 Thermocouple board I may be able to produce a few for sell on Tindie together with some instructions on modding the T962 PCB to swap over from the analog thermocouples to digital I2C.

[sevstels]

dual MAX6755 thermocouple board

Are you sure? MAX6755 cannot measure temperature and does not work with thermocouples