How to scale down signal to 2.5V without affecting measuring

[cyberblak]

high voltage measurement

Hi all,

I got a system in which I need to measure a highly noisy 48V signal (this is the output voltage of a switching power supply). Since it's a noisy environment, I thought of using a difference amplifier. My question if how do I need to scale down the signal to 2.5V without affecting the measure quality? If I use a voltage divider circuit and then enter this signal in the diff. amp, the linearity will be affected, isn't it? Is there a way to scale it down while still keeping a high CMRR and linearity?

Thanks,

CyBeRbLaK

 

[instruite]

Re: high voltage measurement

cyberblak:

got a system in which I need to measure a highly noisy 48V signal (this is the output voltage of a switching power supply). Since it's a noisy environment, I thought of using a difference amplifier.

In our design to eliminate noise we normally have a low pass RC filter (with time constant of around 1ms) after the voltage divider and directly feed it to the microcontroller
If you want to measure only the DC value of the supply than i guess this idea will work for you too

 

[IanP]

Re: high voltage measurement

Use differentilal amplifier in its basic configuration:
It allows you to divide high voltages, and in your case you want to reduce 48V to 2.5V, so the values of resistors will be 160kΩ and 8.2kΩ.

Connect 160kΩ resistor between positive pole of 48V supply and (+) of the opamp.
Connect 8.2kΩ resistor between (+) of opamp and its 0V.
Connect 160kΩ resistor between negative pole of 48V supply and (-) of the opamp, and last,
Connect 8.2kΩ resistor between (-) of opamp and its output (negative feedback).

The voltage reduction ratio will be 8.2:160 ≈ 0.052, so the voltage will be reduced from 48V to 2.5V.
To filter out noises you can connect a 100nF capacitor between (-) of the opamp and its output (in parallel with 8.2kΩ).

Regards,
IanP

 

[cyberblak]

Re: high voltage measurement

In our design to eliminate noise we normally have a low pass RC filter (with time constant of around 1ms) after the voltage divider and directly feed it to the microcontroller

The problem with the voltage divider is that the signal picks up a lot of noise along the way towards the microcontroller. It is very difficult to filter out the impulse components that gets into this signal. I can't use a very low corner frequency since this measure is used to feed a PI regulator, it needs to react quickly.

Connect 160kΩ resistor between positive pole of 48V supply and (+) of the opamp.
Connect 8.2kΩ resistor between (+) of opamp and its 0V.
Connect 160kΩ resistor between negative pole of 48V supply and (-) of the opamp, and last,
Connect 8.2kΩ resistor between (-) of opamp and its output (negative feedback).

This exactly the solution that I thought of in the first time, and it sounds good to me. However I read that many amplifiers doesn't behaves well when used with a gain below unity. Any idea about this?

 

[IanP]

Re: high voltage measurement

Frankly, I don't think there is a problem in working with gains<1 .. or I should say I haven't experienced anything like that ..
Regards,
IanP

 

[instruite]

Re: high voltage measurement

The problem with the voltage divider is that the signal picks up a lot of noise along the way towards the microcontroller. It is very difficult to filter out the impulse components that gets into this signal. I can't use a very low corner frequency since this measure is used to feed a PI regulator, it needs to react quickly.

Well I think with 1 ms RC time constant RC time constant there should not be any problem
The ADC will not be continuously monitoring the signal there will be some cycle time for that
Even in our critical designs we have minimum cycle of 1ms to measure any ADC input so i guess it should not be any problem
If you need more help post the exact details of your requirements then i guess other guys can also help you better

 

[cyberblak]

Re: high voltage measurement

You're right about the time constant. I need it to be 0.5ms which is really not far from what you suggested. The main problem with the RC filter is that it's first order and thus it gives only -20dB of attenuation at 20kHz, which is the main component of the noise.

I'll rather use a third order active filter, that'll give -60dB of attenuation, that should be sufficient.

Thanks anyway