This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TMP117: Averaging vs Not-Averaging

Part Number: TMP117

The datasheet discusses repeatability for averaging vs not-averaging the results.

The datasheet lists LSB as 0.0078℃, or 7.8m℃

Then 7.3.2 discusses this difference of averaging, stating that:

As shown in the noise histograms of Figure 6 and Figure 7, the temperature result output has a repeatability of approximately ±3 LSBs when there is no averaging and ±1 LSB when the device is configured to perform eight averages.

But looking at these graphs, the distribution is over ±70m℃...which is ±10LSB given that 1LSB is 7.8m℃

I think this is a problem with the datasheet. Please confirm.

Regards,

Darren

  • Hi Darren,

    Figures 5 and 6 represent a single device read a hundred times. Figure 7 is variation in 100 devices. 

    thanks,

    ren

  • Please help me understand the reference in DS section 7.3.2. where it says “from figure 6 and seven...+-1LSB...”

    I don’t see how to read the graph for this +-1LSB repeatability? Figure 6 was stated as +-3LSB, and it is easy to see this from the graph.

    Figure 7 is a distribution over 100 devices with averaging on...and error can range from -70 to 60 mdegC? For different devices? Then what is the range of error for a single device when using averaging?

    Meanwhile figure 6 is 100 measurements of a single device...where the error can range from +-20mdegC for any single measurement? Then what mind of range would this be if 100 devices were measured?

    The real question is, when performing averaging, what is the realistic performance “noise-free resolution” you can expect vs non-averaged samples?

  • Hi Darren,

    There isn't a graph of the +/-1LSB conditions. It wouldn't be a very interesting graph. Graphs 5 and 6 are for conditions worse than the recommended condition. This is to show clearly how bad it can be when our recommendation is not followed. It should also be noted that the Repeatability spec of +/-1LSB is a Typical only.

    Regarding figure 7, absolute error can vary +/-70mC across devices and still meet our 0.1C spec. This graph is further detail of how good the accuracy really is. Figures 5 and 6 are relative error; the device variation seen in Figure 7 is removed and not relevant to Figures 5 and 6.

    thanks,

    ren