Thanks Dana. It helps, but I'm still unsure because I don't get some results that make sense.
I'm using an ADC and a Thermocouple component only. Is it enough? Cypress' example projects are more complicated, but basically, I just need to get the differential voltage (in uV) thanks to an ADC and then use the Thermocouple component to convert that in degres, right?
I tried changing the configuration of the ADC but my output is still not correct. Currently, ADC is 16bits and Input range is +/- 0.064V.
You have to follow the ap note, there is an attached project you can look at,
because of the need for cold junction comp and the inherent non linearity of
So to make a type T work is "as complicated" as a type K? That picture makes me so confused!
And how do I know if my thermocouple is high-end, mid-end or low-end? I only know that I have a type T thermocouple...
The picture shows an interface circuit for Thermistors and an interface
circuit for Thermocouples.
Regarding Hi, Mid, Low end, from the application note -
And pages 11 - 13 in application note describe in additional detail Hi, Mid, Low end performance
Thanks Dana, I have read the pdf yesterday and this morning but I have a limit knowledge in electronics so that makes it quite difficult to understand. Things are surely obvious to you but not for me. My questions may sound very dumb but I take the risk! Thanks for your patience.
So do we choose if we want high, mid or low end? Standard tolerance and accuracy is the same? Type T has a std tolerance of 1 degre so that makes it a low end?
And I thought that small piece of wires would be easy to add in my project... gosh I was wrong! :)
Design goals would drive the selection of what resolution and
accuracy you want. You can always start with low end, get it working,
then do an end to end signal path error analysis, to see if you can
achieve and report greater performance. Or start with specific design
goal and use error analysis to confirm it can all be done by PSOC,
or only partly.
If you are new to the tool take a look at the video series here on how
to use Creator. Wiring is quite simple, but if you are new to the tool
I understand how that can be challenging, especially doing busses
the first time and peeling out part of the buss connections.
Thanks Dana! I used the example project provided by Cypress. After trying different ADC settings, I seem to get a temperature close to the actual one, but that value doesn't vary much when I try to warm up the thermocouple. So, I'm very probably doing something wrong.
My ADC settings are 16 bits and Vref/2 for input right now. I tried using the low-end settings described in the AN75511 document but my result wasn't correct.
How do I do an end to end signal path error analysis?
You examine every component that touches the signal path.
The best approach is convert all errors to LSBs and then sum them up,
superposition governs doing this. Noise the exception, it has to be handled
So start with A/D, 16 bits, Vref 1.024 V. So one lsb = 1.024 / 65536 = ~ 15.6 uV.
The ref has an accuracy, convert hat error to LSbs.
An OpAmp has these errors –
3) Bias thru source R effects
7) Output loading
8) Temperature effect on all errors
You do this for anything touching the signal path.
Then you do the noise analysis, and add up all errors.
Attached are some ap notes covering some of these topics.
All AtoD Error10.zip 12.1 MB
If you are trying to build a lab grade instrument contact Keithly, they have
a free handbook that covers in depth passive component issues and
general metrology types of issues.
There is an approach used on instruments to achive high accuracy. Image
a DUT (your design) connected to a heated/cooled plate, and connected to
a ultra high signal generator, say over I2C or UART. Inside your DUT you have
a cal routine, and you send command to sig gen asking it to produce 1 V
exactly. Your DUT measures, and saves that. More generator requests are done,
and sweeps of V and temp are done. Then all this data is plugged into a least
squares error fitting function, or a power curve fiting function, and that produces
an equation with coefficients to use on all infield measurments. Note your design
has to have a T sensor in it as well.
This will take out essentially all error, except long term component drift. Takes test time,
but allows use of some fairly non precise components.