Heatermeter weird issue


 

Dennis R

New member
Hi Guys,
Im soldering the heatermeter 4.3 incl LCD and I have a weir issue.
All seems to be working but the TC still is a PITA....
after some re-soldering and testing, it is working except for one thing….
as soon as I physically tough the thermocouple wire, temp goes thru the roof.
I checked for bad connections but cant find any…
Anyone got an idea what can cause this?
 
that would be the easy guess but since I have made 4 heatermeters and this is the only one that does this... I'm talking about the outside shielding of the tc probe that I touch.
See pictures:
https://ibb.co/x2NYd0Z
https://ibb.co/4PCfMBs
Weird thing is that there is no connection to the outer wire... I tried with other TCs that I have and they all have this. My other heatermeters don't do this so there is something wrong with the hm..
 
I'd check that 1K SMD resistor to GND in the thermocouple circuit and make sure it is well soldered. It helps to eat the noise you're adding when you touch the braid. If that's solid then make sure the 1n and 10n capacitors (differential and common mode noise filters) are soldered well. Finally a therocouple amp with a loose leg can cause this as well.

Rejecting probe noise is a real problem with these mains-powered temperature sensors and there's always a little bit of magic. You can also upload the hm-noise firmware from the AVR Update page and that will add a new graph to the home screen that shows the raw ADC readings after a few seconds. That would tell you if there's always noise and touching it makes it worse, or if it is 100% fine until you touch it.
 
Thanks Bryan! I'll fire up the solder unit tomorrow to check the connections again. The amp is a real menace to get right, especially afterwards when all surrounding stuff is done.
Will report back asap!
 
Ok, did some soldering last days (haven't gotten around to it earlier) and I have 10 units soldered.
All 4.3 based on our zero w and the YC version.
6 work fine as far as I can tell with just some small testing.
1 doesnt indicate the tc but I probably have a loose connection somewhere I guess.
Need to check that...
The other 3 all have the same issue... When the tc is connected, the temp shows 25 deg (just above the temp of approx 20 it actually is). When I touch the tc wire it doubles to approx 55.
This behaviour is identical to all 3...
I really don't know where to look... Especially since the issue is identical on the 3 units.
 
I had this same issue when I was in Australia, operating on 220V wall power supply. Never really figured out the issue. try running it on 12V DC battery power, or try another power supply and see what happens?
 
I had this same issue when I was in Australia, operating on 220V wall power supply. Never really figured out the issue. try running it on 12V DC battery power, or try another power supply and see what happens?

I tried an 8pack AA batteries and you are right!
The issue is gone then but it still adds about 5 degC to the temp at room temperature.
Any idea which component does this or is influenced by the 220v adapter and isn't with a 12 supply?
 
The issue you see where the temperature increases when you touch the wires is due to some of the electricity of the circuit flowing through your body. The way the thermocouple works is by passing a small current through it and measuring the voltage drop. If some of that current flows through you and doesn't come back through the TC- wire, it counts as a drop and influences the temperature. You want to make sure that your probes are electrically isolated-- none of the circuit is exposed or uses the metal shield as a conductor as these will allow a path for the current to flow where it shouldn't. You can test this by checking for continuity from the TC+ or TC- wire to either the probe's braid or the tip of the thermocouple element. If it beeps, it's bad (for this circuit).

It may also be helpful to install the noise firmware (AVR Firmware -> Online Repository -> hm-noise) then go to the home screen and wait for the second graph to show up. It should be flat or bouncing between two values. If it is erratic or shows a sine wave then you've got a noise issue which comes from the power supply, power source, or coming in on the probe wires. You can also see a noise reading on the probe by going into the diagnostics display on the device. Navigate the menus to "Manual Mode" then hold the left button to enter the diagnostics display.

You might get a clue about the 5C offset there too. Note that the heat generated on the board influences the measured temperature as well, so if you're testing right after soldering the values will read high until the board returns to room temperature. I usually see a reading 3-4C high when I am testing a board when done soldering. If it stays 5C high even if left sitting overnight, eeeeee not sure where to go after that. Check the 3.3V at the "AVC" pin on the atmega. HeaterMeter expects that to be 3.300V because it is tying to calculate the absolute voltage coming from the amplifier. This can be accounted for if it is too far off by adjusting the mV/C setting int he configuration.
 
I will try to do it later today to see what I can figure out.
But reading the above still makes me wonder why there is no effect when using battery power and also why, when I made 10 units and all the testing items are identical, the results differ.
I don't think it's the tc for that reason since I use the same one...
 
since you have a connector on the thermocouple sensor, just disassemble and make sure the connections are good and where they strip the outer jacket back, they did not damage the insulation on the thermocouple leads where the shield stops and the couple leads emerge. I have had to carefully remove some outer jacket and make new connection when I have found this issue. Also make sure the thermocouple wires that are clamped in the connector are using all the wire. Clamping only a small portion of the wire can cause problems since the couple is generating a voltage. This can increase the issue with noise.
 
Good points Gary....
If the outer shield is touching the bare TC wire I could see that causing this issue, and loose tc wires in the connector could do the same. Only thing is, I think he said he tried a couple TC's.
I have to say it again for others who may read this thread, I always verify the TC's that I purchase are the ISOLATED type, I find them to be much less susceptible to these sort of issues and less noisy in general. The Thermoworks specs will tell you if they are isolated type, and when I bought generics on Amazon or Ebay I just emailed and asked if isolated before I purchase if it isn't noted in the listing.
 
Ok, did some checking.
First the download of the noise graph software on a good and a wrong one:
Good one without touching wire:

Good one with touching wire:

Wrong one without touching wire:

Wrong one with touching wire:

I have checked the temparature probe as well and there is a connection on the shield and the inner wires... so its not insulated:(

Having said this... the correct units do show effects when touching the wire (see above) but still function as normal.... why don't they all?
 
Having said this... the correct units do show effects when touching the wire (see above) but still function as normal.... why don't they all?
Huh, that is definitely interesting. The first image is how HeaterMeter expects them to look. The second is what it normally would look like if the power supply / power source wasn't great and you can see the curve is a ~50Hz sine wave matching the input power frequency. You can see that it is still centered around the right value but slightly off since the sample period includes slightly more than one full period. That could be addressed by using the "A/C input line noise filter", but obviously would be better to find a root cause. The challenge being that the full range of that graph is just 13mV (0.013V) peak-to-peak so it isn't some giant error that would be easily detected.

The others seem to always be showing a similar wave but not exactly matching the input AC voltage wave which is just boggling. You start by testing to see if it is a problem with all analog inputs or just the thermocouple by using a 1k resistor in place of a probe on one of the thermistor inputs. If you just tack a 1k resistor to the side pin on the mono jack and the pin on the board edge (with no probe plugged in), it should read 337.7F/169.8C. You could also put it in the J2 header between the pin you're testing (1-3) and GND. On the noise graph, the numbers on the right indicate which input it is sampling. 5 = Pit, 4 = FOOD1, 3 = FOOD2, 2 = FOOD3, 0 = BUTTONS. It also takes up to 10 seconds for the graph to cycle to the next update, so expect nothing to immediately happen when you click the number.


The scale of the graph will likely be 1/3rd the size, but given how giant the wave is on the bad ones I think it will still be obvious if it is a problem with all inputs or just the thermocouple and we can decide which direction to look once we know.
 
I will get a new 1K resistor.. I seem all out.
I tried to tag a SMD 1K to 2 wires but that really didn't work:(
A 220 ohm resistance gave 396 deg on all probes:

When I put a ET-73 probe in, the temp is fine (20degC) on the probe jacks... just really not on the TC..
 
The value doesn't matter really, we just need something that is low resistance, 1k or less, but not so low that it gets rejected by the sanity filters in HeaterMeter. Looks like you picked a winner right on the low end and the good news is that the issue doesn't seem related to the microcontroller's ability to measure the analog values properly, since you've got a perfect looking noise graph there.

But now the problem is I'm not sure where to look from here. Have you already tried reflowing the solder on the thermocouple amp legs? The only thing beyond that I can think of is check the 100uF capacitors on the board to make sure they have a good connection as well since it seems that somehow the power has that mains voltage wave on it still, even if it is just a new millivolts.
 
I resoldered all items on the TC part, including the capacitors and the legs of the amp.
The thing that is most worrying to me is that, of the 3 units with an issue, the graphs are exactly the same as a baseline as well as when I touch the wire.
This seems to me that the issue is identical as well...
Is there any way to measure the components themselves to exclude the possibility of having a bad batch of resistors somewhere?
 
In most cases you wont be able to measure components value while in circuit due to other devices connected to the circuit, the best you could do is compare readings on the bad unit with readings on a good unit (with power OFF), noted differences might help point you in the right direction.
Have you already tried cleaning the board (TC Amp) with alcohol and a soft brush? Excess flux on the board can make the probes wonky at times....
 
I was just doing exactly that while I received the message of the topic response:)
This is difficult to explain but I hope I can make sense:
left bottom (10k) is measurable and gives 10k as reference
left top (10k) as well
top center (1k) is 1k and below (10n) this is measured at 1100.
The top right (10n) and middle right (1n) give very high numbers (16M)
the one underneath (0,1u) the amp is unreadable and even gives negative readings sometimes.

Having said its pretty difficult to get proper readings...
The biggest difference that I could find in the readings is the upper left (10k) which measures 1k...
I really don't think I soldered the wrong one but would I measure a correct value on this resistor when connected?
Or is it another item nearby that influences it directly and can that one be the issue.
 

 

Back
Top