I've written about this issue several times in this forum, but got the answer that TC-08USB was a 'value for money' converter, and there were no plans to re-design it.
The CJ compensating layout has a very poor thermal design.
The CJC thermistor is placed in air right in the middle between the 8 input terminals with no direct thermal connection to any of the inputs.
Therefore it is not possible to achive stability better than the temperature fluctuations inside the converter case.
If the temperature gradients inside the TC-08 is e.g. 1°C between say input termial 4 and 8, then the readings will also differ 1°C between those two inputs no matter how uniform the inputs are and how high class the thermocouples have.
As the end user of the converter, one can do a couple of things to improve the accuracy.
1) Place the TC-08 in an environment, as stable as possible. Cover the converter with a towel so fast temperature fluctations around the converter is damped.
2) After powering up the converter (start of the logger program), wait app. 20 - 30 minutes before taking measurements. If you have the opportunity to place all sensors in the same, stable temperature (e.g. an ice-point or a thermostated bath), observe via the graph option when the readings are stable enough for your application.
3) If you have access to a calibrated reference thermometer and stable baths or a dry block calibrator, make correction tables via the scaling function. Do it in steps of e.g. 50 °C
4) An ultimate but tedious solution is to drill some ventillation holes in the converter case and then mount a small 5V fan just above the CJC thermistor. This will force the temperature inside the box to be distributed more uniformly and the 'warm up' time to be reduced. The fan is powered with 5 V directly from the USB socket.
Regarding no. 4), I have uploaded some pictures onto the web at:http://imageshack.us/g/638/100506xxtc08modifik23.jpg/
orhttp://img638.imageshack.us/slideshow/w ... ifik23.jpg
Perhaps these can be an inspiration.
For information, the small yellow wires was attached to micro NTC sensors which I have used to measure temperature differences between the eight inputs.
Some of the pictures shows a highly insulated box with TC-08.
This solution I would not currently recommend, as measurements have shown far too long time constants for stability.