In today's world of increasing electrical noise, higher-speed logic, and more-sensitive circuits, the need for high-frequency noise filtering is especially critical. Noise can enter a circuit through external connections, such as the input power source or other transmission lines within the circuit. In addition, undesired noise may be created by switchmode power conversion or other power-system circuits. The generation of electromagnetic emissions from the system being designed is of equal concern. These applications require high-frequency noise filtering to comply with the myriad industry or governmental standards and regulations (FCC, IEC, MIL-STD, CE, etc.). Fig. 1 illustrates how the power filter minimizes the effects of noise sources.
This article will explore basic high-current filtering techniques and cost-effective methods of achieving desired noise-reduction levels. We will examine currents from 50 A to over 400 A because these levels are typically above common chassis wiring. The conductors used for this level of current are either bus bars or wire larger than 8 AWG (8 mm2). We will consider only high-performance filters (those achieving 40-dB insertion loss or more) because this level of filtering cannot be easily obtained by simple series impedance or inductance, and achieving 40-dB reduction is the minimum requirement for most applications. Fig. 2 shows some typical high-current feedthrough capacitors; their physical size increases with current rating.
High-current feedthrough filters have characteristics different from lower-current feedthroughs. They are jam-nut-mounted to a panel and have a massive cross-sectional area through electrode compared with lower-current feedthroughs. Lug attachment is best accomplished with threaded connections. It is important that the electrode nuts be capable of high torque to prevent failure and to provide minimal resistance for the wire connection. Solder and quick-disconnect tabs are best relegated to lower-current feedthroughs. Figs. 3a and 3b show, respectively, a feedthrough capacitor with bus-bar and lug attachment, and a feedthrough capacitor with lug attachment and wire-crimp terminals.
The key component of all high-performance low-pass filters is shunt capacitance. Suitable levels of filtering can generally be achieved using a C-Type filter in which the filter can remain compact and lightweight, with scaling remaining relatively independent of the higher current levels. Alternatively, a π filter that uses inductors may be necessary if there are large differences in impedance between the line and the load or higher insertion loss is required. However, as current levels rise so does the size and weight of the inductors. For example, an inductor rated for 40 A can weigh over 25 lbs., and when increased to 200 A the weight is more than doubled. Due to the large size and weight, inductance is not a preferred solution for most systems, especially aerospace or mobile applications.
A simplified circuit schematic of a capacitor is shown in Fig. 4. The intended capacitance is indicated by C. We should take note of three parasitic constituents: ESL (Equivalent Series Inductance), which reduces or eliminates high-frequency filtering action; ESR (Equivalent Series Resistance), which limits high-frequency filtering action; and EPR (Equivalent Parallel Resistance), which causes leakage current flow and is not typically an important factor in filtering effectiveness, though leakage must be accounted for and remain at acceptably low levels.
A capacitor has two critical filtering performance parameters: its cut-off frequency and self-resonant frequency. The cut-off frequency is usually based on 3-dB of insertion loss, and is closely approximated by:
where fco is the desired cut-off frequency in megahertz for a filter of C capacitance (in picofarads).
For a C-Type filter, the insertion loss increases by 20 dB for each factor of 10 (decade) increase of frequency above fco. The insertion loss of a capacitor would be 30 dB at about 31.6 times the cut-off frequency. So, for a 220,000-pF (220 nF) filter capacitance, the fco would be roughly 28.8 kHz. The insertion loss would be 20 dB at 288 kHz, and 30 dB at 912 kHz. These frequencies scale inversely with capacitance, so a 22-nF capacitor would have an fco of 288 kHz, and a 30-dB insertion loss at 9.12 MHz.
The insertion loss and frequencies mentioned above are accurate up to approximately 40 dB of filtering performance. However, the second critical performance parameter of self-resonant frequency limits is the increase of insertion loss with higher frequencies. The self-resonant frequency is the result of resonance of the capacitor, C, with the parasitic series inductance, ESL. This inductance is caused by internal inductance and any lead attachment lengths. The governing equation for self resonance is:
SRF = Self-resonant frequency in Hz
C = Capacitance in F
L = Equivalent total series inductance in H
The inductance for a leaded capacitor is difficult to reduce to less than 4 nH, especially in high-current applications. The SRF for a leaded 220-nF capacitor in this example is calculated to be approximately 5.4 MHz. One goal of a feedthrough capacitor is to substantially eliminate this lead inductance, and any self-resonant effects.
Fig. 5 shows the filtering performance of both leaded and feedthrough-type 220-nF capacitors. Filtering performance begins as essentially identical for both capacitors, from the cutoff frequency up to the SRF of the leaded capacitor at ~4 MHz. The feedthrough continues to increase and maintain insertion loss at the higher frequencies, whereas the leaded component tends to have a symmetrical curve with an inflection point at the SRF.
After a brief peak of filtering performance at the SRF, the filtering action of the leaded capacitor decreases approximately at the same rate as the rise of performance, making a V-shaped insertion-loss plot. Conversely, the feedthrough capacitor has effectively maintained insertion loss and shows little signs of the effects of ESL.
Fig. 5 highlights how the discrete leaded capacitor performs well only at low frequencies, while the feedthrough capacitor has high filtering performance to 1 GHz and beyond. These results demonstrate that leaded capacitors are suitable for filtering up to 1 MHz or possibly 10 MHz, while feedthrough capacitors extend the high insertion-loss-filtering range beyond 1 GHz.
The maximum filtering action for a capacitor is typically limited by the ESR of the capacitor. An ESR of 0.02 Ω will tend to limit insertion loss to about 62 dB. Most high-quality filter capacitors will perform to this level. Fig. 6 shows the ESR of the NexTek 220-nF filter tested in Fig. 5. Notice how the ESR limits the maximum insertion loss, as is also reflected in Fig. 5 by the plateau at about 60 dB at frequencies greater than 100 MHz.
The insertion-loss curve in Fig. 5 shows a first-order shunting-capacitor filter. The characteristics are a cutoff frequency based solely on the capacitance value per equation 1, and an insertion-loss slope of ~20 dB/decade. If more insertion loss is required there are three choices:
Increase the capacitance, thereby reducing the cutoff frequency.
Include a series inductor to increase the slope of the insertion loss curve. The value of the inductor must be optimized to work with the value of the capacitors, and should be designed according to Butterworth or Chebyshev criteria.
Use small-value inductors to assist with increasing the maximum insertion loss. These inductors can be approximately 10% of the optimized inductors.
For high-power applications with higher insertion-loss requirements, the most efficient circuit is a π filter, as these use only one series inductor and two relatively compact capacitors.
Increasing capacitance is the preferred method for gaining higher insertion loss at lower frequencies due to the minimal impact in terms of size, volume, and economics. This is especially true for dc circuits of 200 V or less. Increasing the capacitance would shift the green performance curve in Fig. 7 and both curves in Fig. 5, to the left, increasing the insertion loss to lower frequencies.
To increase the slope of insertion loss a substantial series inductance is required. For example, to enhance the performance of the previously discussed feedthrough capacitor a Butterworth π filter could be constructed with two 220-nF capacitors and a 1.2-mH series inductor. The performance of this filter would increase dramatically and the fco would be reduced to about 14 kHz. The result would be 30 dB insertion loss at 46 kHz, or approximately three times the fco frequency. However the size and weight of the inductor will be significant, especially for higher current applications. The optimized π filter insertion loss is shown by the blue line in Fig. 7.
Inductors are available that have ~10% change in inductance from no current to full current. The volume of such an inductor is approximately:
Vin3 = 0.01 × current × µH (3)
- or -
INDUCTOR EDGE DIMENSION FOR BUTTERWORTH π FILTERS
Vcm3 = 0.16 × current × µH
where Vin3 is the approximate volume in cubic inches and Vcm3 is the volume in cubic centimeters, current is the current rating of the inductor in Amperes, and µH is the desired inductance.
To increase the filtering at high current and low frequencies, larger inductors are required. For low-current filters, a 1-mH inductor is very easy to manufacture and integrate. When the current level exceeds 50 A the inductor presents physical and economic challenges. As current-level requirements enter the 200- to >400-A range the inductor becomes a significant problem.
The Table shows the inductance edge dimension in cm for optimal inductors at various frequencies and current ratings required for a Butterworth π filter (the cube root of the volume calculated in Eq. 3 highlights the substantial size and weight required with the use of inductors). In comparison, the volume of a feedthrough capacitor filter does not increase dramatically with an increase of either current or capacitance.
An alternative to the optimized inductor is to use an undersized inductor of approximately 10% the Butterworth value to increase the slope and also maximum insertion loss. This is the red curve in Fig. 7. The insertion-loss plot shows a pronounced dip before the steep insertion-loss slope begins.
NexTek has improved compact high-current filters for many applications. The feedthrough filter offers high filtering action and fits large-lug and wire and bus-bar attachments.
WATCH THE WIRES!
Since large wires and bus bars can apply significant force to the filter, the NexTek feedthrough capacitor uses mechanically decoupled capacitors that withstand rugged environmental conditions such as shock and vibration, temperature shock, and atmospheric pressure changes.
RULE OF THUMB
A properly configured filter should provide the required insertion loss versus frequency and handle the through-current and applied voltage with adequate margin. A low-pass-filtering circuit should be centered on high-current feedthrough capacitors if significant insertion loss is required above 10 MHz, and in cases where the conductor is carrying 50 A or more. The simple feedthrough capacitor will provide the most compact and highest-frequency filtering solution.
In order to filter out the lowest frequency of concern, the maximum capacitance value should be used. If more insertion loss is required, then series inductors can be used.
Each wire leading to and from the filter can act as an antenna and create a coupling path which will bypass the filter and defeat higher-performance filtering.
Use shielding to realize the full benefit of the filter by separating the electrical noise from one side of the filter to the other.
Use discrete capacitors and inductors to filter frequencies below 1 MHz.
Identify the lowest frequency of concern and begin with the value of capacitance to provide at least 20 dB at that frequency.
Increase the capacitance as needed to increase filtering, consistent with the compatibility of the circuit.
Use the largest value of capacitance that will work in the circuit to cover the lowest frequency of concern.
Consider a π or higher-order filter if increased capacitance is not compatible with the circuit, or if attaining even greater insertion loss is required over what can be achieved with a feedthrough capacitor.
Try to eliminate the restriction of increased capacitance, by either filtering at a dc bus to eliminate ac leakage current, or reducing the noise at the source.