DDC114: What does "±0.025% Reading ± 1.0ppm FSR, max" mean?

Hi Kazuya-san,

INL is the deviation of the transfer curve (input charge to output code) from an ideal line (in this case a best fit line). 

The INL value is given as the addition of two factors. Two break it from the example you give: ±0.01% Reading ± 0.5ppm FSR, typ

1. The first (±0.01% Reading) indicates that the larger the measurement value (reading), the bigger the potential error. I.e., if the input is very small, then this error will be very small error, but if the input is large we tolerate a much bigger one.
2. A second factor that is constant across the whole range (max input). In this case ± 0.5ppm FSR (FSR=full-scale range).

Hope this helps,
Edu

No problem at all Kazuya-san :)

And excellent question. No, it does not include that. The error is given respect to a best fit line which very likely will not go through the "zero" and yet, if the real transfer curve followed exactly that line, it would give zero INL. Actually the offset and the gain (slope) errors are given on separate specifications specific to it (offset and range) plus others further describing these in detail (like offset drift with temperature...). 

Kind regards,

Edu

Hi Kazuya-san,

Sorry for my late reply! I was traveling last week...

If one refers the INL to the end-point line (one that goes by the result/output of zero input and the full-scale input), then the end points have zero INL (the values for the real curve and the ideal line are the same by construction). But if one uses a best fit line then this is not the case. The line can be chosen in any way such that the resulting error follows (is close to) the equation ±0.01% Reading ± 0.5ppm FSR. There are many lines that probably give a result that meets this. For production, all what the production test system does is to find one line that makes the error (the deviation of the real curve to that line) fit in the max INL spec. I.e., that the error at any point is within the value expressed by the max INL value ±0.025% Reading ± 1ppm FSR.

But overall, you are right on the concept that the allowed error/deviation near the zero input is much smaller than at full-scale. In fact, for zero input, only 0.5ppm of FSR deviation is allowed, while at full scale it is basically 100.5ppm of FSR (0.01%). 

Not sure I follow the 2nd calculation example. If 1048576 is read out, then typ INL= ±0.01%*1048576  ± 0.5ppm*1048576 and must be within 100.5ppm of FSR

Regards,

Edu