There is no auto-zero function.

Figure 2 looks as if the bars sums to 100 % and the offset drift is clamped to less than 2 µV/°C; I guess this selection is done during testing. I suspect that changes of offset drift over time are not tested for.

This device is optimized for low cost.

Hi Rannie,

Clemens is correct on all of his points. 

Our plots are used to give the customer a reference as for what type of change can be observed across a certain sample of devices. These plots in no way guarantee a maximum or minimum for the datasheet EC table. If I were to give guidance to the customer, I would sufficiently guard band this design to ensure the design can tolerate changes in Vos over temperature.  

Generally, offset related specifications often look somewhat gaussian in shape (though not always). This means you can roughly calculate the statistical probability of finding a device within a certain number of standard deviation. What we can derive from the EC table: ~68% of devices will be within +-.53µV/°C. For this specific case, the "buckets" are pretty wide, so it is challenging to determine exactly how many devices were close to the 2µV/°C side of the bucket.

My best recommendation is to consider a number greater than +-2µV/°C for your worst case analysis. If I were doing this calculation myself, I would consider the maximum offset at 25C and the maximum offset over temperature to determine the fundamental limit for offset voltage. It is technically possible for a device to have a low inherent input offset voltage, and an especially large offset drift value, but these over temperature specifications in our EC table often consider the distribution of devices and their behavior over temperature. 

This means my ESTIMATED maximum would be +-2mV - +-1.6mV / (-40C - 25C) = +-6.154µV/°C

Understand however that this is still an estimated maximum value. It is likely that the customer will never see a part that is this high in drift, but I have no way to guarantee that for a device like TLV9064. 

The best way to think of things, every maximum or minimum in the datasheet adds cost to the device as TI needs to sufficiently guarantee that we do not sell a device which exceeds these limits. For this reason, the cost optimized TLV9064 will not have maximum specifications for quite a few specifications. If the customer does require a hard limit which is guaranteed by TI, it will likely come at a cost premium.

What temperature range is the customer using? 

Please let me know if the customer has any questions.

Best,

Jacob

Hey Rannie, 

Change over time is a much more challenging parameter to specify, but this is something that is tested in device characterization. The idea is to simulate the expected lifespan of a device, and observe the shift in some parameters.

We have a great article on this written by Marek Lis: 

IC long-term stability: The only constant is change

Effectively, the aging of the device will widen the gaussian distribution for zero centered parameters like Vos.

Please let me know if you have any questions here.

Thanks,

Jacob