Im interested on getting long analog integration times, the continuous-time integrator drifts due to DC offset errors such as voltage offset, bias current and offset current, limiting its useful integration time. I turned my attention towards switched capacitor integrators which employ only capacitors and MOS switches, my hypothesis is that due to the lack of resistors, current errors can be minimized, leaving only voltage errors, which means lower drift.

There is a lot of info on the web about switched capacitor integrators and how great they are to manufacture at the IC level, however, I have not found a source that mentions the practical integration times that can be achieved using switched capacitor integrators VS continuous time integrators. Im aware that the opamp used for the integrator is critical, but opamp aside, do switched capacitor integrators have lower drift than continuous time integrators? If so, is it worth building a discrete switched capacitor integrator using capacitors, MOS switches (Like the 4066) and opamps?

  • \$\begingroup\$ Have you performed an error analysis, using a Voffset of 1 millivolt? \$\endgroup\$ – analogsystemsrf Sep 13 '18 at 4:39
  • \$\begingroup\$ Do you mean 1mV offset on the opamp? \$\endgroup\$ – S.s. Sep 13 '18 at 4:42
  • \$\begingroup\$ What kind of components do you plan on using? If you get into the thermal drift range for your length of integration time, thermal differences will need to be included in your error budget \$\endgroup\$ – Voltage Spike Sep 13 '18 at 4:44
  • \$\begingroup\$ Thats a good point. On the opamp side I was thinking on something like the LTC1151, for the switches something like a 4066 or similar, and for the caps 5% tolerance or lower. \$\endgroup\$ – S.s. Sep 13 '18 at 4:46
  • \$\begingroup\$ What signal are you wanting to integrate? \$\endgroup\$ – Andy aka Sep 13 '18 at 7:21

What's great about switched capacitor (SC) is that, when properly configured, they allow the effect of a variable resistance when driven with a variable clock rate. Which is good for things you want to tune. If you're an IC manufacturer, it's a powerful technique to produce interesting on-chip functions.

What's bad about SC is that no switch is perfect, charge injection from the clock into the output node can introduce an uncertainty into the design not present when using simpler components.

To complete an integrator, you need an amplifier, which will have offset voltage, bias and offset currents, and variations of those with temperature and time, regardless of whether currents are being controlled by resistors or SC.

In summary, there's no reason a SC design, especially one built from discrete components, should be any better than an equivalent resistor one, and in fact reasons it should be worse.

An SC one could be made better if you can take advantage of the new topologies the concept of a flying capacitor could allow. A flying capacitor can move charge from one voltage to another with minimal (not zero) influence of the voltage. While this is similar to what a cascode stage would allow you in a conventional design, it's much more flexible, and can allow you to sum currents directly and eliminate some amplifiers.

| improve this answer | |
  • \$\begingroup\$ Why would a discrete switched capacitor integrator have worse performance than a resistor continuous-time integrator? Am I wrong to think that opamp input bias and offset current errors would be minimized by using capacitors instead of resistors? \$\endgroup\$ – S.s. Sep 13 '18 at 17:08
  • \$\begingroup\$ @S.s. Yes, I think you would be wrong to think that. Switches have charge injection, a source of offset current errors, resistors don't. If you keep the same topology, where SCs replace resistors 1:1, then I don't see how the extra uncertainties can result in a better performance. If however you can simplify away one or more opamps by using flying capacitors, something resistors can't do, then you may get an improvement. \$\endgroup\$ – Neil_UK Sep 13 '18 at 19:43
  • \$\begingroup\$ Ohhh, I see, I didnt think about charge injection. Thanks Neil \$\endgroup\$ – S.s. Sep 13 '18 at 20:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.