Need help choosing right capacitor

Discussion in 'General Electronics Chat' started by Tom88, Oct 14, 2011.

  1. Tom88

    Thread Starter New Member

    Oct 14, 2011
    4
    0
    Hi, I'm currently building a 10Hz low pass filter after an integrator stage to reduce the DC noise before it enters an ADC. It is currently around 20mVpp and I would like to reduce it to around +-1mvp. The filter is a 4th order butterworth. I know which values of capacitors I need to use (from microchips filter lab program). However, I need help which type of caps to use. I've been researching but there is no real concrete answer I could find. I'm leaning toward polypropylene but if anyone has any input or a guide they follow, any reply would be greatly appreciated. Thanks.
     
  2. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Polypropylene would be a good choice.

    It would be a good idea to simulate the circuit before actually building it, to get an idea of how well it will actually work.
     
  3. colinb

    Active Member

    Jun 15, 2011
    351
    35
    Can you elaborate on the reasoning for the education of us less knowledgeable?

    Are there models for the different types of capacitor (polypropylene, ceramic, tantalum, Al electrolytic, ...) such that the simulation would help choose capacitor type?

    I am very new to SPICE but I guess you could at least model the capacitor with a series and parallel resistor for the ESR and capacitor leakage. I think you could define a capacitor model with this subcircuit rather than actually putting these resistors on the schematic?
     
  4. #12

    Expert

    Nov 30, 2010
    16,322
    6,818
    I don't know anything about simulating, but I know that the dissipation factor is the reason. One of the inescapable defects in capacitors is that the dielectric absorbs some of the energy. The poly caps are the best at not losing energy in their dielectric.
     
  5. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Have a look at one of Steve Bench's pages:
    http://greygum.net/sbench/sbench102/caps.html
    He tested a number of caps constructed from different materials; the results are rather interesting. An ideal cap's response would be a single diagonal line that was perfectly straight; but as you can see in Steve's results that several of the caps tested exhibit non-linear results. Note the polyester cap shows two traces in the middle. The ceramic LV has a very non-linear response.
    More caps on this page:
    http://greygum.net/sbench/sbench102/caps2.html

    LTSpice includes a number of models for electrolytic and tantalum capacitors, but it's far from complete. Capacitor manufacturers might include some data on parasitics in their datasheets, but for broadband response you really need to do your own testing. Generic PSPICE capacitors are ideal; no parasitics. Linear Technology's LTSpice has provisions for adding parasitics (if you know them) for inductors & capacitors.

    Parasitics on real-world components can really trip you up, particularly at higher frequencies - even with very small components. A number of years ago, we did some testing on 0805-size SMD multilayer ceramic caps in the pF range (these things are pretty tiny; about like this => [] on a 17" screen). A cap that tested around 56pF at low frequencies (<20MHz) measured around 120pF or so by the time we reached 500MHz due to the parasitics; or nearly tripled in value. We were using a Hewlett-Packard material analyzer that could test components up to around 20GHz; you could buy a very nice vehicle for what it cost.

    As I mentioned previously, Linear Technology's LTSpice has provisions for modeling a number of parasitics within a passive device; you can get to these by right-clicking on it, as in this example:

    [​IMG]

    As you can see, I didn't enter all of the possible parasitics for C1, but it's certainly a better approximation than leaving them at the defaults (ideal).

    The more elements you add to a simulation, the better it will represent a real-world circuit. However, you can reach a point of diminishing return, where the amount of time required to build and run the simulation far exceeds what it would take with the real components on a test bench.

    SPICE simulations are great for testing an idea for a starting place, to be relatively certain that you're not going to burn stuff up.

    One big thing people frequently forget to model is the inductance of wiring. PSPICE wiring is ideal; real wiring isn't. Even a 1" straight piece of AWG-24 jumper wire has about 24nH inductance. That doesn't seem like much, but it adds up in a really big hurry. Even the leads on a TO-220 package for a transistor or MOSFET have 4.5nH-8nH inductance; people forget that when they try to make a high frequency switch, and then wonder why they have high-frequency "ringing" going on.
     
    Last edited: Oct 14, 2011
    colinb likes this.
  6. #12

    Expert

    Nov 30, 2010
    16,322
    6,818
    Why are you elaborating on MHz effects for a 10Hz filter?
     
  7. gootee

    Senior Member

    Apr 24, 2007
    447
    50
    As stated, polypropylene are usually good (low "dielectric absorption"). So are polystyrene and teflon, usually.

    It depends on the values needed, though, since those types are only cheap or small for relatively-small values.

    If you need larger-value polypropylene caps, look for them at places like madisound.com or parts-express.com. Otherwise, the little red WIMA box caps come in small dimensions for polypropylene, from places like mouser.com.

    If you can, also make sure that your polypropylene caps are of the "non-inductively wound" type.

    Modeling capacitors in SPICE: The ESR can be a gotcha, especially with electrolytics. It can change dramatically with both frequency and temperature. The Cornell Dubilier website had a great applet for showing that data, for some of their capacitor lines, and also generated a spice model taking those effects into account!

    For simple sims of electrolytics, at least calculate the ESR for the frequency of interest, using the fact that the DF spec usually stays somewhat constant over frequency but is easily related to ESR. Tan(Delta) @ f = DF = 2 x Pi x f x C x ESR. ESR = DF / (2 x Pi x f x C).
     
Loading...