Voltage Integrator

[dknguyen]

Hi. Is there any way (discrete circuit or IC) that can time-integrate a voltage input? I seem to be only be able to find current integrators that output a voltage.

Preferably, the voltage integrator would also output a voltage so I can chain a couple of them together to do double or triple integration.

I can do it in digital, but due to quantization and noise I am looking for an analog solution. I also think analog integration would get rid of much of the noise without having to use filtering compared to digital integration right?

 

[Nigel Goodwin]

I suggest you try asking EXACTLY what you are trying to do, rather than a fairly vague question?.

 

[dknguyen]

Integrator

I have a accelerometer that outputs an analog voltage that outputs with a scale factor of say, 500mV/g. I want to integrate the output (or double integrate the output) so that I can get the velocity (and displacement).

I can do it through digitally sampling the output voltage, but was wondering if there was an analog circuit or device capable of doing it (since then method could inherently filter out noise without the use of filters, remove a lot of quantization noise, and would offload a lot time-sensitive of work from the processor).

The basic process I was going to use whether it was analog or digital was to time-integrate the voltage (once or twice) and then digitizing the result and digitally multiply the result by the scale factor (once or twice) to give me real-world phenomena (velocity or displacement) being measured.

I also have a rate gyroscope that outputs a velocity as an analog voltage which I would also like to time-integrate so that can get the total angular movement. Its the same idea as above.

The only integrators I have been able to find work by integrating a current and outputting the result as a voltage. I need a circuit that integrates a voltage and outputs the result as a voltage so that I can cascade them together to achieve multiple-integrations.

 

[Styx]

I have a accelerometer that outputs an analog voltage that outputs with a scale factor of say, 500mV/g. I want to integrate the output (or double integrate the output) so that I can get the velocity (and displacement).

You are going to suffer alot with this, double integration will cause you alot of problems any slight offset will be integrated to a slope after the 1st integrator and then into a quadratic by the next so if you are not careful you can get it running away very easily

I can do it through digitally sampling the output voltage, but was wondering if there was an analog circuit or device capable of doing it (since then method could inherently filter out noise without the use of filters, remove a lot of quantization noise, and would offload a lot time-sensitive of work from the processor).

Yes there is an analogue cct it is called an "integrator"
**broken link removed**
Why do you say "inherently filter" an integrator is effectively a low-pass filter and even if it was implemented in digital it would filter as well.
You wont really be removing quantization noise since I guess some where down the line you will still be passing this signal through an ADC.
IF anything it will be more stable in the digital domain due to OPAMP offsets and noise outputs. IF it was a single integrator then sure analogue will do nicely (equally digital), but cascading integrator after integrator will cause problems since the output of the 1st integrator will contain the integrated signal+OPAMP noise+OPAMP offset (this offset will then get integrated by the 2nd)

Sure you can put a nice arrangement of variable resistors down to tune the offset but there is alot of time involved in calibrating it to remove the offsets - 2 off as oposed to only one calabration for the digital domain. Equally any variations in temp or other variables (the state of solder-points) will increase noise injected into the analogue implementation

The only real arguement for having it in the time-domain is limitations in the digital (space,polling etc)

The basic process I was going to use whether it was analog or digital was to time-integrate the voltage (once or twice) and then digitizing the result and digitally multiply the result by the scale factor (once or twice) to give me real-world phenomena (velocity or displacement) being measured.

I also have a rate gyroscope that outputs a velocity as an analog voltage which I would also like to time-integrate so that can get the total angular movement. Its the same idea as above.

The only integrators I have been able to find work by integrating a current and outputting the result as a voltage. I need a circuit that integrates a voltage and outputs the result as a voltage so that I can cascade them together to achieve multiple-integrations.

Why do you need a "current" ?

 

[eng1]

The output of an op amp integrator is the integral of the input voltage and not the current. Of course you won't use the basic circuit... you need some improvements...
A fast op amp that can be used for integrators is the LF411 and its offset can be adjusted with a pot.

 

[dknguyen]

Integrator

Hmm...yeah...the offset was something that just occured to me last night. When I was planning to completely digitize all the stages, it would have been fairly easily to periodically recalibrate for the offset...doing so in analog would require making the circuit more complex.

Hmm, yeah that circuit seems pretty obvious to me now. I have been working with a current-integrator circuit and trying different (it is the circuit you have without the resistor.) I had forgotten about the virtual ground and that it would maintain Vin across the input resistor while providing a corresponding current flow. That's what I meant by I needed a current...the op-amp part of the circuit integrates a current...but adding the input resistor converts Vin into an appropriate current to integrate. The virtual ground slipped my mind. KISS!

I figured an analog integrator would filter noise better than a digital integrator because it uses infintely smaller intervals than digital integration. I also thought I could minimize quantization noise by sampling less often- sampling/resetting the analog integrator output every ms rather than sampling the sensor output every usec (if I chose that as my time-interval for integrating). I also didn't want to load down my DSP so much with oversampling.

Maybe I should use an analog integrator for the first integration, and digitize any integration beyond that.

 

[Hero999]

The output of an op amp integrator is the integral of the input voltage and not the current.

Simple, add a transresistance amplifier to the input.

A fast op amp that can be used for integrators is the LF411 and its offset can be adjusted with a pot.

Why do you need a fast op-amp?

Unless you're using it for high frequencies (which we're not doing her) the capacitor in the feedback loop indroduces a huge low frequency roll-off far greater than the op-amp's.

 

[eng1]

Simple, add a transresistance amplifier to the input.

Why? His input is a voltage