Common mistakes in electronic design
By Craig Hillman, PhD, DFR Solutions -- EDN, 12/14/2007
Electrolytic capacitors
[/quote]Although designers love electrolytic capacitors because of their high capacitance, they hate them because they fail over time. This love/hate relationship has led to a range of methods for derating and predicting lifetime. What are the best approaches? It all depends on whether you are derating voltage, ripple current, or temperature. With voltage derating, remember that electrolytic capacitors work best when you apply voltage to them. With no voltage, they have no dielectric and no capacitance. Although electrolytic-capacitor manufacturers have over the last five years improved these capacitors' low-voltage performance, try to avoid voltages below 25% of the rated voltage. At the other end of the spectrum, designers create capacitors by applying voltages 150 to 200% greater than the rated voltage. In addition, applied voltage tends to have minimal influence on lifetime. Because of this fact, the derating guidelines specify a maximum applied voltage of 80 to 90% of rated voltage, although some manufacturers apply 90 to 100% of rated voltage.
Once you target a desired life cycle for your design, you can decide on the appropriate temperature derating. The industry-accepted equation is a doubling of life for every 10°C drop in temperature. Although some questions exist concerning the accuracy of this model, designers must be aware of three nuances. The first is that this life equation is relatively conservative—at least for reputable capacitor manufacturers. Vendors often define “lifetime” as 1 or 0.1% of failed parts, as opposed to the more standard MTTF (mean time to failure), which might yield a 63% failure rate. If your design lies between these extremes in desired lifetime, then it should be OK. Second, few applications experience constant temperatures. Users turn computers on and off, the sun rises and sets, and other similar temperature-affecting conditions occur. Make sure to incorporate variations in temperature into any lifetime calculation. Finally, all bets are off if there is elevated temperature due to an adjacent component, such as a resistor or a MOSFET. Some indications show that a highly localized temperature increase more quickly induces failure than the industry model predicts. Keep hot components away from electrolytics.
Ripple current on electrolytic capacitors is an odd electrical parameter. Designers tend to ignore or forget it in most bill-of-materials calculations. Remember that “equivalent” capacitors are not equivalent when it comes to ripple-current ratings. And manufacturers can “uprate” ripple current. Some companies allow applied ripple current to be 150 to 200% of rated ripple current. They achieve this flexibility because ripple current primarily increases capacitor temperature, and vendors often specify capacitor lifetime at rated temperature and rated ripple current. The lower the temperature at which the design can operate, the higher the uprating margin on the ripple current.[/quote]
By Craig Hillman, PhD, DFR Solutions -- EDN, 12/14/2007
Electrolytic capacitors
[/quote]Although designers love electrolytic capacitors because of their high capacitance, they hate them because they fail over time. This love/hate relationship has led to a range of methods for derating and predicting lifetime. What are the best approaches? It all depends on whether you are derating voltage, ripple current, or temperature. With voltage derating, remember that electrolytic capacitors work best when you apply voltage to them. With no voltage, they have no dielectric and no capacitance. Although electrolytic-capacitor manufacturers have over the last five years improved these capacitors' low-voltage performance, try to avoid voltages below 25% of the rated voltage. At the other end of the spectrum, designers create capacitors by applying voltages 150 to 200% greater than the rated voltage. In addition, applied voltage tends to have minimal influence on lifetime. Because of this fact, the derating guidelines specify a maximum applied voltage of 80 to 90% of rated voltage, although some manufacturers apply 90 to 100% of rated voltage.
Once you target a desired life cycle for your design, you can decide on the appropriate temperature derating. The industry-accepted equation is a doubling of life for every 10°C drop in temperature. Although some questions exist concerning the accuracy of this model, designers must be aware of three nuances. The first is that this life equation is relatively conservative—at least for reputable capacitor manufacturers. Vendors often define “lifetime” as 1 or 0.1% of failed parts, as opposed to the more standard MTTF (mean time to failure), which might yield a 63% failure rate. If your design lies between these extremes in desired lifetime, then it should be OK. Second, few applications experience constant temperatures. Users turn computers on and off, the sun rises and sets, and other similar temperature-affecting conditions occur. Make sure to incorporate variations in temperature into any lifetime calculation. Finally, all bets are off if there is elevated temperature due to an adjacent component, such as a resistor or a MOSFET. Some indications show that a highly localized temperature increase more quickly induces failure than the industry model predicts. Keep hot components away from electrolytics.
Ripple current on electrolytic capacitors is an odd electrical parameter. Designers tend to ignore or forget it in most bill-of-materials calculations. Remember that “equivalent” capacitors are not equivalent when it comes to ripple-current ratings. And manufacturers can “uprate” ripple current. Some companies allow applied ripple current to be 150 to 200% of rated ripple current. They achieve this flexibility because ripple current primarily increases capacitor temperature, and vendors often specify capacitor lifetime at rated temperature and rated ripple current. The lower the temperature at which the design can operate, the higher the uprating margin on the ripple current.[/quote]