TPS92520-Q1: output current accuracy calculation need some detail data

Hello Zhiyuan,

1) The tolerances on each resistor in the divider can be fairly high, but they are matched resistors so the ratio is tightly controlled, which is what is important.

2) This is a forced offset that is only done at the start of turn on to ensure it starts up.  There is an offset voltage on the V-I converter amplifier and that can be positive or negative, therefore we force the offset to ensure that the signal in the control loop is positive at the start of the turn on sequence.  It gets bypassed once the current is high enough to make the signal positive. 

3)  input bias current are insignificant compared to the resistor divider and doesn't contribute significantly to accuracy.

Each part is tested to ensure that it meets our +/-4% accuracy at full scale. 

-fhoude

Hello Zhiyuan,

Yes, this it is correct that control loop is tested at ATE and that the only thing you need to adjust for is the current sense resistor tolerance.  Keep in mind that is just one test conditions of 6V on the output and at 24Vin, but the  limits are over process and temp.  You will see that at lower DACx setting the % error increases.  This is expected, because as the voltage on the V-C converter input pin gets smaller, the off set voltage (Vos) of the amplifier become a larger ratio of the signal controlling the loop.  I have included a diagram that shows basic calculations for the loop.  

In the datasheet we can see that on figure 6-2, where at lower CHxIADJ Counts (Which sets VIADJ in about diagram by programming of DAC) the % error starts to drop quickly.   There are other sources of error, like gain, which might not be linear.  Generally speaking the gain error is more dominant at higher CHxIADJ counts.  You may want to center  your CHxIADJ based on typical value of ~600 because the % error is lowest there and makes the % error distribution more centered.  See figure 6-2 below.  

You can address Vos in the effects by the following generalized equations.  

Please send me the actual excel spreadsheet so I can look at it to verify.  You can send it via email if you worried about proprietary information.

-Francis Houde

Hello Zhiyuan,

The ATE test data shows that the best accuracy is at full scale, but you have to keep in mind that the limits are 6 sigma.  This data is a summary of the distribution of many tested parts with extra wide guard band that considers process and temp. 

Figure 6-2 is the assessment of one part showing the general curvature of the device over the CHxIADJ range.  I would use the ~600 CHxIADJ of the ATE typical test data as the location to center my accuracy for each part.  

-Francis Houde

Use the ATE data for accuracy but know that is super conservative and 6 sigma boundry limits.  The figure is to help with centering for calibration to minimize errors over all set points.