Confusion about the meaning a parameter

Below is a part from this accelerometer:

For +/-2g the Scale factor tolerance is given as 10mg. There is also 0.25%. This is in parentheses as 1.25(0.25%) V/g. Is 10mg derived from 1.25(0.25%) V/g?

I'm completely lost about what scale factor tolerance is.

Imagine scale factor is x V/g and the scale factor tolerance is %y what does that mean in terms of accuracy? Is this tolerance an absolute value of changes with the g?

I would be glad if someone can explain meaning this parameter in a simple manner associating with accuracy. What can be said about the minimum g can be measured regarding this parameter and why?

In an analog system there is a sensor, and then some way to convert it to volts. Observe the measurement range in g's

±2g which is the input and 0 to 5V which is the output. Somehow they convert the input to the output.

So what they are really doing is taking ±2g and mulitplying it by 1.25(V/g) which would give you ±2.5V (if you look at the units they cancel out and leave you with volts.

They then add 2.5V to the ±2.5V range, which gives -2.5+2.5= 0V and +2.5+2.5=5V to get the 0V to 5V range.

The analog system can be represented by a gain, and then an offset, which is a common format in analog systems. No gain is perfect, however (due to resistance mismatches and other sources of constant manufacturing error.) So the 1.25 can be ±1% or 0.99 or 1.01 of 1.25. So one unit could have a 1.25*0.99= 1.2375(v/g) gain and another unit could be 1.25*1.01= 1.2625(v/g)

This means your knowledge of the original signal will vary by 1% without calibration. If the scale factor is calibrated, then the scale factor tolerance will not matter and it can be multiplied out.

• The scale factor tolerance is given for the full span of +/-2g. 0.25% of 4g is 10mg. The bias of the accelerometer on the datasheet is as much as 1mg so 20ug (DC) would require calibration. – Spehro Pefhany Jul 13 '18 at 1:22
• Because someone told me: "if you wanna mesure 20μg you need another sensor since the accuracy is 1%". How are these two things related? I hope I could articulate where I'm stuck at. – user1999 Jul 13 '18 at 1:33
• The scale factor error must be added to the bias error (including temperature drift). Then there is noise, nonlinearity, hysteresis etc. You need to be much more specific about you are trying to measure. – Spehro Pefhany Jul 13 '18 at 1:41
• @newage2000 The 1% accuracy one had higher SNR so lower resolution was possible, but better calibration is needed if you want that accuracy which is unlikely possible due to hysteresis on repeatability, but at least you have the linear resolution for small changes. – Sunnyskyguy EE75 Jul 17 '18 at 15:30

Gain error is 0.25% FS or 10mg whichever is greater + 100ppm/‘C.

Initial Offset (bias) error is 10mg but stability is +/2 mg + 0.1 mg/‘C.

Resolution is 100 ug

noise is 18*20uV in 400Hz BW

20 ug is below threshold and well below noise level.

The CRI models have much greater SNR but worse gain error.

• It seems -2 people cant understand so explanation is possible – Sunnyskyguy EE75 Jul 14 '18 at 4:15