Parasitics & Capacitor Selection, Part 1B: Leakage

First, we need to consider how far within the direct current leakage (DCL) limit the typical part will be. Unlike the range of capacitance values in a given lot, which have a normal distribution, most capacitor technologies have conduction mechanisms that contribute to leakage with a DCL capability far below the limit. However, the distribution of DCL within a lot will have a noticeable tail on the high side of the distribution, so the mode of the population can easily be 10x better than the specification limit.

Next, we need to look at the effects of voltage and temperature. Let's refer back to our leaky capacitor bucket from part 1A in this series, in which the fluid represents the amount of charge the capacitor is carrying and the properties of the various dielectric materials comprising it. If the bucket were only half full, the speed of the flow would be reduced and would taper off to a trickle at even lower applied voltages.

This effect is often nonlinear and can be quite dramatic for some dielectrics. For example, thin dielectrics operating at high field strength (e.g., tantalum dielectrics) have a logarithmic response to voltage derating, which means that parts being operated with 50% or more derating can have up to two decades lower DCL in a given application. However, for many dielectrics, the effect is much less dramatic.

For instance, a Class 1 ceramic dielectric (e.g., NP0) will have an Ohmic response to applied voltage, so halving the voltage would double the IR. That isn't exactly two decades, but it also isn't too shabby. Similarly, Class 2 ceramics (e.g., X7R) will start off with an Ohmic response around the rated voltage, but they will exhibit diminishing returns on IR increases at very low applied voltages.

As for temperature, the DCL for all technologies tends to increase by a factor of 10 between room temperature and maximum operating temperature. Conversely, lower temperatures typically result in lower leakage current. Exceptions include film capacitors and polymer counter-electrode tantalum and aluminum; you'll want to check the specifications.

One last thing to acknowledge is that DC measurements are considered steady state and time independent, but in every instance, the metaphorical bucket must be filled up before we can measure its leakage. As such, the capacitor will have high initial leakage when first switched into the circuit, but it will decay exponentially with a time constant that's dependent upon the capacitance value and load resistance. For example, for a given load (1 kΩ is a typical test reference), this can amount to a delay of 1–5 minutes for capacitors between 10 µF and 100 µF to settle into their steady state leakage current.

In sum, the primary takeaways for design engineers are:

  • Always check the component specification limits for leakage in any battery-limited application, and make sure to budget for leakage at your maximum operating temperature. But don't be surprised if your typical achieved leakage is two decades or more lower than expected.
  • For low-leakage designs, seek out the special series specifying lower DCL or higher IR. Incorporate additional voltage derating into the design, and keep an eye on thermal management.

Primary takeaways for process engineers include:

  • Your capacitors will be supplied as buckets with holes, so take care not to make the holes bigger. This can occur as a result of thermomechanical damage or electrical overstress. To avoid the former, check thermal profiles on reflow or wave solder, take overheating precautions during rework, and select mechanically robust components, such as ceramics with flexible terminations or lead frame assemblies that minimize shock or TCE mismatch. To avoid the latter, check for correct polarity, and make sure ESD and EMI events are budgeted for and closely monitored.
  • When troubleshooting, DCL can often be a good diagnostic tool. In reverse polarity, tantalum dielectric has near-Zener characteristics, so small changes in voltages can elicit large changes in DCL. Additionally, voltage step stress can help determine if there are Ohmic and non-Ohmic contributions to DCL or IR, which can help pinpoint the fault current mechanism.
  • The hole does not always have to get bigger. Some dielectric/electrode combinations (e.g., tantalum and plastic film technologies) have self-healing properties that will ensure a long lifetime, as long as the parts are operated within their specifications.

9 comments on “Parasitics & Capacitor Selection, Part 1B: Leakage

  1. Scott Elder
    February 25, 2014

    Hi Chris,

    I recently happened upon the AVX capacitor modeling software SPICAP-3 and SPICalci8.  These are great tools.  I wonder why they don't model the leakage issue you discuss?  They certainly include the “free” series R and series L, voltage and temperature coefficient.

    Also, companies like Ambiq Micro and NXP make products where capacitor leakage is everything!  Ambiq's products run on nanoamps.  I suspect the consequence of capacitor leakage will be growing in importance with new low Iq products that operate with solar or micro-watthour batteries.

  2. amrutah
    February 25, 2014

    @Scott: Yes I hve also seen this kind of issue with capacitors from murata.  The leakage is not built into the model.

  3. samicksha
    February 26, 2014

    I guess DC leakege, the capacitance may be intentional for instance EMI filter capacitors.

  4. chirshadblog
    February 27, 2014

    @samicksha: Yes it's the DC leakage. Any solution for this ?   Trusted and a proven one.          

  5. chirshadblog
    February 27, 2014

    @amrutah: What is the main cause for this ? Any specific reason you have figured out ?           

  6. amrutah
    February 27, 2014

    @chirshadblog: I don't know why it is not modeled.

      The only thing that I can think off is, device characteriszed leakage nos. might be very inconsistent w.r.t. to voltage (voltage coefficient) and temperture (temperature coefficient). 

    May be @Chris Reynolds can throw more light on this.

  7. Davidled
    February 28, 2014

    I image that the higher nonlinear mathematics model presents the leakage current for R and L function unless  someone has experiment  data of leakage corresponding to R and L..

  8. RedDerek
    March 4, 2014

    Nice set of articles on capacitors. Clearly the non-ideal part in the designer's tool chest.

  9. Chris-R
    May 28, 2014

    This is a good question – in many models, the parallel resistance representing can be included, but in normal operation this is is a very high resistance. For most technologies, it will reduce as temperature increases but will still be high – but if a capacitor is damaged, so that the IR drops (or DCL increases) beyond specification, then this can have the effect of increasing te dissipation factor (DF) which could addect the model. But this would be a special cause, and not a factor in normal operation.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.