# How to properly voltage divide an AC/differential signals

What is the proper resistor configuration for voltage dividing an AC or differential source?

For simplicity sake assume i would want to reduce a 5v pk-pk perfect sine wave to half 2.5v pk-pk @ 5kHz. The voltage divider is to be read by a differential ADC. Note that resistor values to get the desired output in the image might be wrong since im not sure how to balance as its not the typical voltage divider anymore (B and C)

I have seen all this circuits and it would seem they all do the job (using simulations) but i would like to know which one is best for what,

what if the signal is not perfect for example like a lets say a microphone where the frequency can vary and voltages too. (assuming again the ADC has a sampling frequency way above the required)

• This is just a simulation, note that real scope has the shield connected to mains ground. So your example wouldn't work with real components, most probable the scope would smoke. Jan 10, 2021 at 16:07
• I don't think you need the ground, just 2 or usually 3 resistors. And yes careful with the scope ground Jan 10, 2021 at 16:34
• @MarkoBuršič So which would be the proper way of doing it? Jan 11, 2021 at 8:24
• @PeteW 3? so the second configuration but instead of R5 and R6 there would be a single 300K ohm resistor? Jan 11, 2021 at 8:25
• B is best but R3 needs to be twice that of R1 and R2 to get the desired attenuation (the mid point of R3 will be at "ground" so each leg only sees half of R3). If P and N are from low impedance sources (opamps) then resistors should be much smaller, in the 1-5K ohm range. Jan 17, 2021 at 21:15

A is good because it uses the smallest number of parts. A is bad because the upper output is higher impedance than the lower one. B fixes the bad problem with A, but now you need to match the series resistors precisely to maintain accuracy.

C treats the input as two single ended signals. You will need precision resistors again, and you need to match two pairs. If there is any imbalance in the input ( not perfect differential input), C will attenuate the common mode the same as the differential mode. In contrast. A and B do not attenuate the common mode at all. Thus, the common mode rejection is a factor of two worse in A and B.

There is no right answer. I would probably use B.

Note that 300k resistors are most likely way too big, adding noise and probably offsets to the error budget. You should probably use smaller values.

• I apologize about the values o the resistor they were just the default values when i placed them into the schematics. Speaking of values are the values in B correct to get half voltage? Jan 17, 2021 at 22:08
• 0.1% resistor is common and matching shouldn't be an issue. Jan 18, 2021 at 0:42
• Yes will be using 0.1% tolerance, but should R1,2,3 have correct resistance value of 300k to have 1/2 attenuation? Jan 18, 2021 at 14:40
• I would also use B Jan 21, 2021 at 1:42

Unfortunately your requirements are insufficient for commercial use.

There are specs such as CMRR, Lightning transient withstanding voltages and RF noise as well as tolerances to be defined.

The proper design must consider voltage breakdown for each resistor and the maximum transient voltage. Worst case is 8kV due to a lightning transient as the meter has rugged gap suppression for this rating but most people use 5kV for an upper limit and filtered transients rarely go above this. The most reliable solution is a ceramic or epoxy coated hybrid attenuator, which is large enough to prevent surface creepage due to dust and humidity. It might even have a diode bridge for rectifying to single supply ADC input.

These professional solutions tend to be expensive. If you chose 2000V rated resistors, they might look like this.

A DIY solution might look like this with the 1M made from a string of 1/4 W Resistors. Simulation here

For an integer ratio 800Vpp to 5Vpp or 160:1 divider with LPF at 2kHz also demands the use of a CMOS Op Amp for Rail to Rail

• Thank you for your answer sir, but i am more concerned on the configuration rather than the resistor specs, 300k is just a place holder value they could be 50 ohms for all i care. Of course using a buffer would always be best, but if we take out the option of using a buffer and all thats left is those 3 configuration which would be best. Would C be a catch all? or there are situations where i must use C than B, what are those scenarios. Jan 21, 2021 at 8:27
• There is no advantage to C other than CMRR of a lightning transient for which you can add a CM choke then an MOV Jan 21, 2021 at 9:02

It all depends very much on the application. It's important to remember, that 300k resistor is never actually 300k, they are all a bit different. So in B and C you immediately get poor CMRR- calculate it, and you will see that for 1% resistors you are suddenly in an unacceptable range. In fact, 0.1% would be a compromise as well.

Option A seemingly doesn't introduce common mode noise, but as a transmission line it's not too perfect. Although, if you put 300k as a termination, maybe you don't care. The common-mode voltage is different- again, you might not care, I don't know.

My personal preference for such signals is to always use amplifiers. It's more expensive, I know, but it saves a lot of time otherwise spent "searching for ghosts".

Any circuits that are meant to process differential signals must be symmetric in the physical implementation, not only on the schematic: otherwise, the parasitic impedances will be unbalanced and will act as common-mode-to-differential converters, degrading the signal integrity. Even at audio frequencies such unbalanced parasitics will cause hum and interference pick-up. Thus circuit A is a no-no just on those grounds, unless you don't care about signal integrity. Even in audio applications you'd want to maintain CMRR better than 50dB, and thus the unbalance at entire signal bandwith, parasitics included, must be on the order of 1:1000. You'd need to be using 0.1% resistors, or a film hybrid and laser-trim it.

Furthermore, with any such divider, if it is to work on transmission lines (i.e. signal transmission circuits with controlled impedance), then it must terminate the input line with a suitable impedance, and must provide and output impedance suitable for the load. Even for "low" frequencies, a transmission line setup may be beneficial, since it provides a well-behaved purely capacitive load (IIRC) to the driver.

If differential signal doesn’t have common ground with adc then C type is good idea because it provide references voltage to adc. If your signal have common ground with adc and has center voltage of vcc/2 I think B or C type is good idea because it reduce amplitude of both signal so the center of signal still the same. That make more sense when you probe to see both signal.