This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TPS54521: Factors influence Vout accuracy ?

Genius 3900 points
Part Number: TPS54521
Other Parts Discussed in Thread: TPS74401

Hi there

We need to use TPS54521 to step down 12Vin to get 3Vout with narrow DC accuracy of  ±1% tolerance, which is 5V±30mV.  Please note we are refering to the DC accuracy instead of Vout ripples. Vout ripple should be another topic in DC/DC design.

For our analysis, Vout = Vref * (1 + Rf_top/Rf_bot), as the accuracy of Reference Voltage is (0.8V ± 3%), and two Rf are 1% tolerance resistors, we know that all these three factors could deviate from its nominal value randomly. 

Mathematically, Vout would be positively maximum deviation from 3V nominal value when these three condition occur simultaneously: Vref is max (+3%), Rf_top is max (+1%), and Rf_bot is min (-1%).

Similarly, Vout would be negatively maximum deviation from 3V nominal value when these three condition occur simultaneously: Vref is min (-3%), Rf_top is min (-1%), and Rf_bot is max (+1%).

However, I guess this probability of these three condition happen at same time should be rather small.  Maybe there is more practical method to calculate the Vout DC range, but I do not know. 

Does my analysis correct ? Is there any other factors which will influence the DC Vout accuracy?

We may also use DC/DC + LDO scheme to achieve more accurate DC Vout. However, same question exist here for LDO (TPS74401 of TI), because  for user, the feedback and Vout settings are similar no matter for DC/D and LDO , both take the form of Vout = Vref * (1 + Rf_top/Rf_bot).

Any comment from TI experts will be appreciated. Thanks.

PS: 

After I investigated the error analysis of amplifier and ADC, to calculate TOTAL errors in the entire signal chain, for error of each componet, we usually use the square root of the sum of square of each error, rather than just add them up algebraically.

Maybe this idea would be applied to this question?