I'm working on a PCB for a temperature controller. The device reads temperature from a thermocouple to regulate the temp in a heater system.
It is using the MAX31856 thermocouple converter IC for its versatility, and it is talking to an Arduino. The sensor works, but I constantly get Over/Under Voltage Faults. I can ignore these in the code, but I want to improve the circuit. I followed the example schematic for the MAX chip, but still have issues. Any ideas are awesome. Gerber_REV 2.zip (139.9 KB)
If it is picking up interference then it is often to do with the physical wiring. Have to tried checking this, using shielding, filtering
Your schematic is very busy. It is often easier to understand if broken down into labeled sections on sperate sheets with the first sheet giving the links
Thanks for the schematic idea. Not sure how to do that in my EDA software, but I'll look into it. (Next revision will be on a bigger area so I can separate the different functions.
Yeah, in practice the thermocouple have 20-30ft leads right next to ac for the heaters, but I'll see if I can isolate them more.
Doing some more research, I found that the chip is of poor design, and will probably look at others made by other companies.
The general idea I got from the post, is that the IC suffers from grounding issues, because it manipulates the ground side of the thermocouple, so if the thermocouple is bonded to the shielding, then it doesn't work.
Thermocouples come in many flavors and designs. Just for an example a popular Type K may come in a sheath type design. The sheath may be stainless steel or Inconel or other alloys. Assuming a sheath type the actual TC junction may or may not be grounded to the sheath. This illustrates what I am talking about. Next the thermocouple extension wire is also important, there is shielded and unshielded TC extension wire. I always made sure to run all of my TC leads in a separate conduit away from my 480 VAC heater power lines.
The lower part of the image shows the three wire Type K shielded TC connectors. Above the TC connectors are the TC transmitters converting a range of temperatures to a 4 ~ 20 mA loop for data acquisition. Anyway here nor there noise depends on the use of grounded or ungrounded TC types. Thermocouples do not have a ground side, they have a + and - output.
My bad with the terminology. Yeah it manipulates the - side of the TC. (Common mode?). The thermocouples are connected to the shielding, grounded, and won't allow me to connect the shielding to the IC ground because it creates some sort of ground loop.
Further progress:
I added a huge 220uf capacitor to the power line of the MAX ic, and 1uf capacitors where the original 0.1uf "filter" caps were. This has improved the stability of the chip greatly. I'll see if there a limit to too much capacitance, but in the hope of a worst case noisy scenario, I'm going to see how much capacitance I can add.
The transition from T/C material (Chromel - Alumel) to copper will create a local T/C junction that by design of most T/C IC's must be near the physical IC.
You should take a look at the many T/C tutorials on the internet to get a better understanding of how they function and their circuit requirements.
added a huge 220uf capacitor to the power line of the MAX ic, and 1uf capacitors where the original 0.1uf "filter" caps were.
I doubt the 220µF is doing much. The 1 and 0.1 are likely causing your difference. This is because at high frequencies, the 0.1 will actually outperform the 220.
Actually at 1,000 degrees C a type K thermocouple produces 41.276 mV not uV. Anyway, low level voltages are produced and supceprable to noise. Where the actual TC wires connect to copper is where CJC (Cold Junction Compensation figures into things.
It's still a pretty small voltage to deal with especially with PCB runs and relays in the path.
As for the reference junction, I don't see how the OP will handle that. Assuming the T/C type is K then the T/C material cannot be soldered to the board, so the reference junction will be in a yet unknown location. We also don't know what range the OP is interested in and what accuracy is required.
Its a plastic injection molding hot runner controller.
400+-10 degrees f is acceptable.
The cables the connect the controller to the mainframe to the mold create at least 3 cold junctions multiple yards away from the IC, so accuracy isn't the biggest concern. The reading dropping out all together is the issue.
OK an injection molding machine. Most injection molding machines I have been around use a sheath type thermocouple. Many include an armored cable in the industrial environment. They look like this.
The example just happens to be a Type K dual thermocouple. Here nor there, it's a sheath type TC.
The Maxim 31856 is designed to work with an ungrounded thermocouple meaning in the case of a sheath type thermocouple the actual TC junction, the hot junction, is not grounded or connected to the sheath.
Note the resistance is below an Ohm telling us the actual TC junction is grounded to the sheath. Not a good choice for a MAX31856. Typically an ungrounded TC will read 10 Meg Ohm or just open circuit as seen below.
How can you have three cold junctions? You come off the thermocouple using a correct plug or socket. You use TC extension wire all connections and TC connectors all the way to your Max31856 TC connection point and that should be the only cold junction. Thermocouple to the 31856 should all be done using TC extension wire. Three cold junctions makes no sense to me?
Ok so many of us have been around injection molding machines and understand what a hot runner.
I have a question and a comment:
What is the goal here? Are you trying to add a controller to a machine without a controller? Repair? Academic project?
We are not trying to be mean here but if you worked for me as showed my that schematic I would immediately tell you to go redraw it. It is not easily understandable in its current form. The sections should be drawn so experienced folks can look at it and immediately understand the function and details. This goal may not be attainable for all designs but it certainly is for this one.
The goal is to replace the QTC-15 controllers used in the shop. They seem to fail for no reason, and ppe has less than extensive information on the product. It often shows error codes that we can't find anywhere on the internet. http://www.ppe.com/17pdf/0870.pdf
My work uses external mainframes for hot runner controllers, so it would be possible to build our own. http://www.ppe.com/17pdf/0871.pdf
I'm a high school student with no formal training, so this is mostly a hobby for me (work is paying for the time and materials for this project though). I'm not entirely sure what the plan in for the device, but it will probably just used in house and not sold as a product.
I'm working on a much more organized schematic now.
The thermocouple system is kind of a mess. I guess there aren't multiple cold junctions. However there is the grounded, and hot junction of the thermocouple, then the cold junction, then connections via han connectors into copper wire into the mainframe via another han connector, into a back plane then into the controller card. None of it is thermocouple metal, so not ideal. I guess they aren't cold junctions in the way I thought though.
I would suggest looking at Watlow PID controller and a SSD relay from some well known MFG.
The thermocouple system is kind of a mess. I guess there aren't multiple cold junctions. However there is the grounded, and hot junction of the thermocouple, then the cold junction, then connections via han connectors
Not sure what a han connector is. However you can purchase T/C connectors and panel jacks to keep the T/C cold junction right at the IC.
So here's what i suggest.
Forget the PCB for now
Purchase a Max 6675 from Amazon and maybe an AD8495 board.
Get them both working with the Arduino and test it at or near your machine.
I would eliminate the relay and get two measurement circuits.
Make a list of device "minimum requirements". You can make another list of "would likes".
Now the next step will be to "harden" the inputs so ESD etc doesn't blow them out. Well let you catch up before we start that.
If you haven't guessed by now, the real world of electronics is kind of ugly and a reliable design must be able to survive the expected and unexpected electrical signals on your inputs and outputs.
BTW Don't feel bad, you made a rookie mistake, never go right to a PCB, always test first. I will say for a high school student you've done amazingly well
I just saw a photo of the QTC board. Can you tell us what U2 is? and are the input pins going directly to the card edge pads?
How would that work? Can you hook two thermocouple reading chips up to the same thermocouple.
Hopefully the heater circuit wasn't too confusing. Its a single heater per controller, but it is powered by two phases of 3 phase 240V so a relay is required on both legs to isolate the heater in case of a short to ground.
I'll look into more protection for the thermocouple system, but I need a chip that for sure works with at least type k and j thermocouples, so I may see if I can get the 31856 chip to work. The heater side is all optically isolated so that side should be okay.
That's not recommended. Can you post a picture of the thermocouple? Earlier I posted images of sheath thermocouples is that what you have because you can always get dual thermocouples.
Your MAX31856 works with a wide range of thermocouple types including J and K types. You never mentioned if your thermocouples were grounded or ungrounded types?
I can't get into most of the molds right now because they are in use. However this is these are the kinds of heaters in use. The sprews and manifolds just have raw thermocouple in them like this.
Most of the thermocouple are grounded, but I haven't checked to see if all of them are. I can make the 31856 chip work with a grounded TC as long as the DC side of the card is floating compared to ground. Am I correct that the lack of ground reference makes the shielding ineffective though?
The variety of thermocouple types the 31856 can use is the biggest reason for me to try to use it. Some of the error check functionalities such as open TC are very nice too.
