# What is the highest practical value for DC voltage divider

I am using a high side load MOSFET switching circuit as a low current voltage divider. I am trying to minimize the current from the battery V1.

• How do I determine what is the highest practical value of R2, R3, and R4 can be? Can I use Megahoms values?
• This is for DC measurement and the output Vdiv goes to microcontroller ADC. Also, would it help to have a buffer going to the ADC?

Thanks

• It depends most on the SNR ratio spec but -60dB is easy -120 might be noisy. Jan 26 '21 at 4:36
• Data sheet for the ADC says " The maximum recommended impedance for analog sources is 10K". So, if i add opamp buffer, how large can R3 and R4 be? Jan 26 '21 at 4:43
• R2 can be 1M. Maybe more. I believe 1M is the largest I have used in production designs for this kind of thing. In order to stay below 10k as requested by your ADC datasheet, you need R3 in parallel with R4 to be less than 10k. If you add a buffer, then you can use Meghohms. But you need to pay attention to buffer specs such as input offset voltage and input bias current. These will cause errors. Another thing to consider is that A you should put a capacitor on Vdiv to make a low-pass filter, and B if you calibrate, you may get away with R3 and R4 as is... Jan 26 '21 at 5:36
• (without any buffer) and in any event the act of trying to calibrate will tell you whether the system is reasonable and consistent. Jan 26 '21 at 5:37
• R1 is larger than seems necessary or reasonable. The 2N7000 is not the best choice for 3.3V logic operation. The BSS138 will switch more reliably at 3.3V. The 2N7000 is OK for 5V logic. Jan 26 '21 at 5:40

How do I determine what is the highest practical value of R2, R3, and R4 can be? Can I use Megahoms values?

It really depends on how much noise you can tolerate in your design.

This chart can help you calculate the noise, first, you need to know the low pass cutoff for your ADC (every ADC needs a low pass filter or it will alias) and the Nyquist cutoff is lower than half the sampling frequency.

Let's say the Nyquist is 10kHz, we would use the green line and look up the resistance:

1MΩ is 10uVrms 100kΩ is 5uVrms

So that is about as much noise as you'd see on the ADC for a 12bit ADC and 3.3V range;

3.3V/2^12= 805uV

So 10uV isn't going to matter too much but for a 16 bit AC that has a 50uV per bit, you might start to notice it.

This is for DC measurement and the output Vdiv goes to the microcontroller ADC. Also, would it help to have a buffer going to the ADC?

The ADC also pulls current which changes the measurement, so the higher value you make the bridge the easier it is for the ADC to change the DC value (think of the ADC as a resistor pulling current, the current can be found in the datasheet usually listed as bias current or input current). If DC values are important, then buffering with an op-amp can limit this problem.

• thanks all. i'll try out your suggestions. Jan 27 '21 at 4:20
• I tested with R2 as 1M, keeping R3 and R4 as is. PIC16FLF1826 10 bit ADC reading error was sufficient for my application. I am reading 3V and 9V battery voltages to see if they fall below some noncritical threshold. The high R values made my acquisition time about 1msec (using data sheet eq-n). So, i guess, for non-critical appl-n high R should be ok. Thank you all. Jan 30 '21 at 20:58