This board is no longer available on tindie- it seems. Do you have plans to fix the issue in revA and re-inlist it for sale? I also think a smaller version with two or four channles using MAX31850K similar to Unified Engineering board would be a nice supplement to your 10 channel board that would install on T-962 and/or T-962A. These ovens need a much better thermocouple reading to help regulate the heat and I would think there would be many who would purchase them given that the price of these ovens have dropped down quite a bit recently. Thank you for your amazingly clear and to the point videos and documentation.
I always somehow end up making silly mistakes like that too, both in hobby and professional projects. At least in my job we go through a long schematic review process, which helps to catch majority of these issues. Great project by the way!
Excellent work!!! When you run short of I/O and especially in this case where you are sharing a single input pin and you could have bus contention between the interface chip outputs, it's always best to use a dedicated logic decoder like the 74HC138 to generate the selects. Using two 74HC138s to generate the 10 chip selects will guarantee that only a single interface chip is active at any time. The 138 has 3 address inputs so to get the 4th address bit without any extra logic, just use a positive chip select of one 138 wired to a negative chip select of the other 138. Tie the remaining chip selects active and now you have 16 safely operating selects from 4 ESP outputs. If you're not using a dedicated decoder and have any potential for multiple outputs to speak at the same time, it's standard practice to add a small series resistor in line with each device to limit the current. A small value like 470Ω will limit current to safe levels if you have bus contention and it's small enough to not affect rise time (and thus speed) by very much. 330Ω was another common value back in the days of parallel buses.
@@voltlog The problem is that a design which is able to damage itself thanks to a random crash is less than professional. Hardware should be engineered to work properly under all circumstances. That's what separates professional designers from amateurs. Just because you can, doesn't mean it's a good idea or desirable to do. It's possible to do all kinds of undesirable things. If you want to represent yourself as a professional, then this is a golden opportunity for you. 😉
If you don't have enough pins for CS, just use a 4-to-16 decoder IC (e.g. 74HC154), or two 3-to-8 decoder ICs (e.g. 74HC138). If you don't want to pay $60 for 10 MAX31855 ICs, you can multiplex the thermocouple inputs with analog MUXes (e.g. 4051), and use fewer MAX31855 ICs, being mindful of the conversion timing requirements.
I'm not too experienced with high precision TC measurements but there are a few concerns that will introduce errors when muxing them like: leakage currents from the mux, introducing tiny thermal junctions in the signal path, reducing the accuracy of the cold junction compensation. These might be fixable through careful layout of the board to avoid thermal gradients and by using a more expensive mux chip but I don't want to go there. A good mux is expensive just as well and I would be spending more time trying to solve the mentioned issues.
I also tried to see if you published a video showing this card in action investigating the heat profile inside the oven- I couldn't find it. Did you pursue this any further?
wow outstanding very informative 👍🎉 When I mistake any design in circuit amd order for print them, and when I test them I just do add insulated jumper wire
Just thinking out loud. I was wondering if you could multiplex the thermocouples and then only use one or maybe two of the thermocouple interface chips and save some money. I assume the thermocouple are slow to respond so reading them sequential or by two banks would be in msecs. Again I probably don't know what I am talking about... but that has never stopped me before. I look forward to your follow-up video/s. I have the same reflow oven.
I'm not too experienced with high precision TC measurements but there are a few concerns that will introduce errors when muxing them like: leakage currents from the mux, introducing tiny thermal junctions in the signal path, reducing the accuracy of the cold junction compensation. These might be fixable through careful layout of the board to avoid thermal gradients and by using a more expensive mux chip but I don't want to go there. A good mux is expensive just as well and I would be spending more time trying to solve the mentioned issues.
Thanks for posting, nice job and very few changes required on the board - like others have said there is always something, its the law! Not sure what accuracy you were looking to achieve, but I can tell you that the cold junction compensation will not be at all accurate. If you only need +- a few degrees then OK, but you'll struggle to achieve anything consistently better than that. I've been designing TC measuring boards for years the CJC bit is the hardest to nail! Great job though
Accuracy... I don't know, even 5 degrees C of error wouldn't be that bad as I'm mostly interested in temp gradients that span more than 10 deg C. But I'm hoping the max31855 with the built in CJC will perform better.
"Little bodge never killed nobody", a hit song by Fergie from 2013. Input only -pins labeled just like any IO pin, grouped with them... That's just asking for trouble. To be fair, electrical type is "Input", instead of "Bidirectional", but as those aren't visible by default, does anyone bother to check them? I wouldn't have checked, as I didn't have any reason to think those last pins weren't usual GPIOs, until you mentioned it. I haven't used ESP32 all that much, but would probably use those pins at some point and blame my own code for things not working. I'll go add that same note on my library, to avoid that mistake when I play with my ESP32 modules in near future. Thanks for sharing.
could have used something like a shift register or bin to hex chip for saving GPIO for accessing CS pin.(as far as i know, only one of em go high at a time). Also the reason why I have 100 pcs of 74595 in my inventory.
I also made my own esp32 symbol, due to these input only, and the adc 1 and 2 banks, as you cannot read from adc2 while wifi is active. There is also conflict pins for the hall sensor and specials like the pair differential pins as well as the 2 dacs.
Sometimes the bugs are very difficult to find. It took me a while to notice that not all ESP32 ADCs work when wifi enabled. It took a while to figure out why everything was ok when on computer and why ADC seamed to read garbage after enabling wifi for remote interface. I think more than often the first prototype has some kind of error (at least my prototypes).
I did a similar mistake while making a smart wall clock and connected my bin to hex converter to input only pins of esp8266🤦♂️. they should rename those from gpio to gpi (general purpose inputs :p).
Would it work to use one thermocouple interface ic and switch between thermocouples? I'm not seeing any reason why not but maybe I'm missing something.
one cannot learn without bodges - I just received 30 pcb from jlcpcb and I realised I messed up with labels on them - thx that only labels - not the real connections ;-)
