PCT2075 temperature sensor accuracy
I read datasheet for PCT2075 temperature sensor and there are some contradicting data about accuracy for me.
Page 21:
accuracy ±1°C (Tamb = -25°C to +100°C)
accuracy ±2°C (Tamb = -55°C to +125°C)
Page 23:
figure 24 looks much better than those numbers. The accuracy looks like ±0.3°C over the whole range (nowhere near the ±2°C value).
So, how to understand this?
What is actually the accuracy?
What I care the most about is range between 0-50°C.
UPDATE:
I have contacted NXP support and received these answers:
The accuracy of the measurement is dependent upon the definition of the environment temperature, which is affected by different factors: the printed-circuit board on which the device is mounted; the air flow contacting the device body (if the ambient air temperature and the printed-circuit board temperature are much different, then the measurement may not be stable because of the different thermal paths between the die and the environment).
The accuracy for the PCT2075 in the range from -25C to +100C is stated to be MAXIMUM within the range from -1C to +1C, so, no matter how is the design of the board or environment where the sensor is mounted, all the time should never be beyond this range.
In typical applications, with a good board design and typical environments, the real accuracy should looks something like in the Figure 24, where there are no big deviations from the real temperature.
As you can see in section 8.2 of the device datasheet, the accuracy of the measurement is more dependent upon the definition of the environment temperature other than just the device itself.
You understanding is correct, the -1C to +1C inaccuracy could be caused mostly by factors outside of device, but it is expected, as you can see in Figure 24, that the actual inaccuracy is much more closed to the +/-0.3C range. What you can see in Figure 24 are the typical values, max values can go up to +/-1C range.
These answers are to me as unclear as the datasheet is.
They say that inaccuracy of ±1°C is caused mainly by external causes, not the device itself. That device is basically inaccurate as ±0.3°C.
And then say that ±0.3°C is typical value and ±1°C is maximum value.
There's something wrong about that, isn't it?
1 answer
The text values in the specifications section are always the real specs. Unless otherwise stated (and that's quite rare), graphs are suggested values, typical ranges, variations due to a single variable, and the like. They are NOT specifications.
The only accuracy you can count on is what it says in the Static Characteristics section of the datasheet that you quoted: ±1 °C from -25 °C to 100 °C, and ±2 °C from -55 °C to 125 °C. This is over the 2.7 V to 5.5 V supply range.
Note that there are several variables here effecting the offset of the reading from the true temperature. Three obvious one are part to part variation, temperature, and supply voltage.
The labeling for the graph you showed could be better, but it seems to be trying to show the typical variation as a function of temperature alone. Remember, THIS IS NOT A SPEC. It's more a rough guidance to give you some feeling for the tradeoffs.
It is interesting that they show the supply voltage range as 2.8 V to 5.5 V, which eliminates the lowest 100 mV of the recommended supply range. Perhaps the 2.7 to 2.8 V supply range is really pushing the limits, and eliminating that reduces errors as a function of temperature.
I also interpret that graph as implying that while the supply voltage is somewhere within the stated range, it is held constant for the purpose of the graph.
Summary
The text gives you the real spec worst case error.
The one graph you quoted says that
if:
1 - You hold the supply voltage constant somewhere from 2.8 to 5.5 V.
And
2 - For a single part.
Then
Here is the typical variation of the error as a function of temperature alone.
These answers are to me as unclear as the datasheet is.
It can seem that way because they contain too much babble. Only the very last sentence matters and clearly states what is going on is:
"What you can see in Figure 24 are the typical values, max values can go up to +/-1C range."
In other words, ±1 °C is the error you have to design to. Everything else might give you some intuition, but is ultimately irrelevant.
1 comment thread