Signal that is a sine wave plus an offset

[0xDEADBEEF]

I have a signal that is a sine wave plus an offset. I would like to measure the dc component and possibly also subtract it from the signal later. The signal is very slow (10th of Herz) and the offset changes on the order of minutes.
I thought it should be possible to get a time average with an integrator, but the layouts that I find are all using periodic resets.
If I use the standard integrator with an opamp that has a capacitor as negative feed back, and I add a leak resistor over the capacitor, does this give me a time average of the input voltage with an exponential dependence on the history of the input signal? Are there better ways to do this?

 

[berkeman]

Can you just digitize the signal and process the data? You could always use a DAC if you need an analog output waveform after processing...

 

[0xDEADBEEF]

Can you just digitize the signal and process the data? You could always use a DAC if you need an analog output waveform after processing...

The sine signal is tiny and I use a lock in. Somehow I feel better with analog stuff (no digital noise, no clipping, no discretization error, no aliasing) especially since the task looks so simple.

 

[Bob S]

You are correct. For a leaky integrator, make sure the RC time constant is at least several times longer than the period of the signal you are trying to measure.

 

[f95toli]

The sine signal is tiny and I use a lock in. Somehow I feel better with analog stuff (no digital noise, no clipping, no discretization error, no aliasing) especially since the task looks so simple.

Analog electronics certainly has many advantages, but there are definately cases where digital electronics works better and I suspect this is one example. If the frequencies are so low noise shouldn't be an issue because there is not need for a high speed ADC, i.e. no 10 MHz clock signals to worry about etc. A good benchtop multimeter (e.g. a 3440A) shouldn't cause any problem with noise as long as you buffer it with a good pre-amp.

 

[vk6kro]

You could use an A to D converter and take a series of readings over a period of 10 minutes or so. It should include several cycles of the input DC variations.

You could display a moving average of the DC value, by averaging every 10 readings or so and feeding this information to a computer.

If the AC component is very small then it should just average out over a large number of readings.

Then pause and calculate the long term average voltage.

If you save these values you could input them into a spreadsheet and display them graphically.

 

[0xDEADBEEF]

[...]
If you save these values you could input them into a spreadsheet and display them graphically.

I am thinking about subtracting the signal actively to run the ac stage at higher gain. The digitizing does lower the need for very large capacitors and such and dc offsets will probably be a pain, but digitizing a measured signal in the middle of an amplifier chain leaves an awkward feeling.

 

[berkeman]

I am thinking about subtracting the signal actively to run the ac stage at higher gain. The digitizing does lower the need for very large capacitors and such and dc offsets will probably be a pain, but digitizing a measured signal in the middle of an amplifier chain leaves an awkward feeling.

I was thinking of suggesting that as well, so I'm glad you thought of it. You can digitize the full signal at the same time that you AC couple, amplify and digitize the small AC component. That would seem to give you the best overall information on the signal.

 

[vk6kro]

There is probably no need for the capacitors to be large.

The usual "voltage follower" unity gain Op amp circuit has a high input impedance and the RC time constants can be made such that capacitors like 0.1 uF will work OK to remove DC components but pass AC.
If you do this, you can have following stages operating at high gain and amplifying only AC.