Noise suppression in PWM-driven DC motor

Abstract

A system for reducing Radio Frequency Interference (RFI) produced by motors which are powered by Pulse Width Modulation (PWM). It is believed that the PWM induces radiation at the PWM frequency and harmonics. The invention shields power cables which carry the PWM pulsed current, and provides capacitive shunting of one of the cables to ground.

Description

BACKGROUND OF THE INVENTION

The invention concerns suppression of radio frequency interference produced by a DC motor in which speed is controlled by Pulse Width Modulation (PWM).

FIG. 1 illustrates a motor M used in the prior art, such as to power a blower in an automotive air conditioner. A vehicle battery BAT powers the motor. The motor is contained within a conductive housing H. The motor contains brushes and a commutator (not separately shown) which cause electrical arcing and sparking, and produce high-frequency Radio Frequency Interference (RFI).

The RFI can interfere with communication devices, such as radios, tape players, televisions, ham radios, and cellular telephones. Measures taken to reduce the RFI include connecting the housing H to the negative terminal, B-, of the motor, as indicated by wire W, and installation of a filter which includes inductors L1, L2, and capacitor C1.

The motor shown uses a simple type of speed control, illustrated by the adjustable resistor R. Adjusting the resistor modifies the voltage presented to motor terminals B+ and B-, by the voltage divider principle However, such speed controls, although simple, are inefficient, because power is dissipated in the resistor. Not only does this dissipation represent wasted power, but the wasted power is dissipated as heat, which poses a heat elimination problem.

To avoid these problems, the inventors have investigated Pulse Width Modulation (PWM) for controlling motor speed.

OBJECTS OF THE INVENTION

An object of the invention is to reduce radio frequency interference in pulse-width controlled motors.

SUMMARY OF THE INVENTION

In one form of the invention, a PWM controller is used to remotely control a motor. The system was found to produce RFI at the PWM's base frequency, and at harmonics, as well as at other frequencies. The invention includes a filter system for reducing the RFI.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 illustrates a prior art motor, filter and speed control;

FIG. 2 illustrates a test wherein a different speed control was used for the motor of FIG. 1;.

FIG. 3 is a plot showing results of a noise test of the apparatus of FIG. 2;

FIG. 3A is a plot showing results of a noise test of apparatus similar to that of FIG. 2, but in which the housing H is left floating, and in which chokes L1 and L2 have been removed;

FIG. 4 illustrates one form of the invention;

FIG. 5 is a plot showing results of the test of the apparatus of FIG. 4;

FIGS. 6 and 7 emphasize topographical aspects of FIG. 4;

FIG. 8 is a plot showing results of the test of the apparatus of FIG. 4, but with capacitor C2 removed;

FIG. 9 illustrates three current paths taken by currents when capacitor C2, housing H, and the braided shield BR are grounded at different points;

FIG. 10 illustrates an alternate embodiment of the invention;

FIG. 11A illustrates one form of the invention;

FIG. 11B illustrates a modification of FIG. 11A;

FIG. 12 illustrates noise performance of the apparatus of FIG. 11A. The bottom part of FIG. 12 is a continuation of the top part, beginning with the frequency of 10 MHz;

FIG. 13 illustrates noise performance of the apparatus of FIG. 11B. The bottom part of FIG. 13 is a continuation of the top part, beginning with the frequency of 10 MHz;

FIG. 14 illustrates a model of capacitor C2 in FIG. 11A;

FIG. 15 is a hypothetical data sheet of impedance-versus-frequency, of a capacitor;

FIG. 16 illustrates a plot of the imaginary part of impedance of a series LC circuit;

FIG. 17 illustrates a plot of the imaginary part of impedance of a two-element ladder network, with each "rung" containing a series LC network, of the type shown in FIG. 16;

FIGS. 18A, 18B and 18C illustrate impedances for three LC networks;

FIG. 19A illustrates the plots of FIG. 18, combined into a single coordinate system; and

FIG. 19B illustrates the plots of FIG. 19A, but with an envelope E added.

DETAILED DESCRIPTION OF THE INVENTION

The inventors modified the apparatus of FIG. 1 by adding a Pulse Width Modulation Controller, labeled PWM, as shown in FIG. 2. The PWM includes a heat sink HS, which is electrically grounded. The PWM generates a train T of square wave pulses. The width W of the pulses is modulated, thereby modulating the energy delivered to the motor M, thereby controlling motor speed. The frequency of the pulses was 20 KiloHertz (KHz).

The inventors tested the PWM system for noise production. FIG. 3 shows results. One significant feature of FIG. 3 is that significant noise begins near 20 KHz, at point A in the Figure. Since this noise begins at approximately 20 KHz, which is the PWM base frequency, this noise is perhaps attributable to the PWM modulation, and its harmonics.

One possible explanation lies in the connection of the housing H to the negative motor terminal B-, as shown in FIG. 2. When the switch SW closes, the housing is pulled to ground. When the switch SW opens, the housing is pulled to the battery potential. Thus, the housing alternately jumps between ground and battery potential at a frequency of 20 KHz, and can act as a radiator.

The inventors found that merely disconnecting the housing from the negative motor terminal B- did not remove the noise. A test of the system shown in FIG. 2, but with (a) the connection between B- and the housing H eliminated (that is, with the housing left floating), and (b) chokes L1 and L2 eliminated, produced the results shown in FIG. 3A.

The results in FIGS. 3 and 3A, at low frequencies (i.e., below about 10 MHz), are somewhat similar, indicating that grounding the housing H in FIG. 2 may not reduce noise. (The elimination of the chokes L1 and L2 for FIG. 3A is not a significant factor at the low frequencies. These chokes are designed to block high-frequency noise, in the range of 10 MHz to 100 MHz, and above. Consistent with this, as FIG. 3A indicates, the elimination of the chokes does increase noise at these high frequencies.)

ONE FORM OF THE INVENTION

FIG. 4 illustrates one form of the invention. A hollow, braided shield BR surrounds the conductors feeding B+ and B-. The shield is about 12 inches long. This shield need not completely cover the exposed conductors, but about two or three inches of conductor, at each end, are left exposed, as indicated by dimensions D.

Since the RFI wavelengths intended to be suppressed are very long, compared to the exposed conductors, the exposed conductors act as poor radiators. (In general, for good radiation, the radiator, or antenna, must be a significant fraction of a wavelength.) The braided shield BR is grounded, as indicated by wire W1, which connects to the ground of the heat sink HS. The braided shield is also connected to the housing H, as indicated by wire W2.

The positive power lead, B+, is coupled to the housing H, and thence to ground through the braided shield BR, through a capacitor C2. Preferred values of the components shown in FIG. 5 are given in the following TABLE 1.

TABLE 1______________________________________Component Preferred Value______________________________________L1 4.0 micro-HenryL2 4.0 micro-HenryC1 0.001 micro-FaradC2 500 pico-Farad______________________________________

The housing H is not connected to the negative motor terminal B-, contrary to the situation shown in FIG. 2.

Test Results

The system of FIG. 4 was tested for RFI noise production. The test was similar to that which produced the plot of FIG. 3, and results are shown in FIG. 5. In the frequency range spanning from about 15 KHz to about 6 MHz, indicated by range 6, significant reduction in measured noise power was found. Some examples are given in the following TABLE 2.

TABLE 2______________________________________ Noise (FIG. 3) Noise (FIG. 5) ReductionFrequency (dB Units) (dB Units) (dB Units)______________________________________20 KHZ 97 38 5930 KHZ 96 18 780.5 MHz 72 33 391.0 MHZ 62 29 336.2 MHZ 50 32 18______________________________________

The data are given in units of dB microvolts, referenced to one microvolt. For example, a measured voltage of 100 microvolts corresponds to 20 LOG (100 microvolts/1 microvolt), or 40 dB. Consequently, the reductions in TABLE 2, given in dB, correspond to the fractional reductions given in TABLE 3, below.

TABLE 3______________________________________Reduction in dB Fractional Equivalent______________________________________59 1/89178 1/794239 1/8933 1/4518 1/8______________________________________

Thus, for example, at 30 KHz, the reduction by 78 dB corresponds to a reduction by a factor of 7,942.

Characterization of Invention

1. It is believed that grounding of the housing H, through the braided shield BR in FIG. 4, reduces radiation of RFI by the housing.

It is believed that grounding of capacitor C2, as shown in FIG. 4, provides a shunt path to ground for RFI which would otherwise escape onto lead B+. This RFI is generated by the arcing commutator. Capacitor C2 is intended to shunt frequencies in the range of about 50 to 100 MHz, and above.

2. FIG. 6 shows the apparatus of FIG. 3, when the switch SW is open. A simplification is indicated by the arrow. Capacitor C2 is in parallel with the battery.

FIG. 7 shows the situation when switch SW closes. Capacitor C2 remains in parallel with the battery, but now bridges the power leads B+ and B- of the motor. The system alternates between the situations of FIGS. 6 and 7.

Significant Features

1. In addition to testing the configuration of FIG. 3, the inventors also tested the same configuration, but with capacitor C2 is absent. The test results are shown in FIG. 8.

Comparison of FIG. 8 with FIG. 5 indicates that the capacitor C2 does, in fact, reduce noise in the range below 10 MHz. This is perhaps surprising, in view of the fact that the impedance of C2 equals 1/jw (C2), wherein j is the imaginary operator, w is radian frequency, and C2 is the value of the capacitance. At 10 MHz, this impedance is 1/(2.times.PI.times.10.times.10.sup.6 .times.500.times.10.sup.-12), which equals about 32 ohms, which is not considered a short circuit.

Further, the impedance of the capacitor increases, as frequency decreases. For example, at 1 MHz, the impedance is about 320 ohms, which is ten times the value at 10 MHz (32 ohms).

This inverse dependence of impedance on frequency seems to be inconsistent with the reduction in low-frequency noise (ie, below 10 MHz in FIG. 8) which is caused by addition of C2. This inconsistency perhaps indicates that the lumped-element model shown in FIG. 4 does not actually describe the behavior of the system, and that waveguide and transmission-line modes of operation are at work.

2. The invention is expected to be used in motor vehicles. Commonly, motor vehicles are designed so that the entire metallic body and frame act as an electrical ground. Therefore, it may be thought that the configuration of FIG. 9 is feasible.

That is, it may be thought that (a) housing H, or (b) braided shield BR, or (c) capacitor C2, or any combination of these, can be grounded locally to the vehicle, at respective grounds GR1, GR2, and GR3. However, such is not the case.

One reason is that noise currents flow through the grounds. These noise currents eventually find their way to the PWM local ground, in order to complete their respective circuits. FIG. 9 illustrates three exemplary paths P1, P2, and P3 for the noise currents.

These paths act as loops, which act as antennas, which radiate noise. The radiating ability of a loop depends on its area. Thus, to keep radiation small, the loop area should be kept small. In general, grounding the three components as shown in FIG. 9 will produce relatively large loops.

In contrast, grounding as shown in FIG. 4 will produce relatively smaller loops. Therefore, it is preferable to ground in the configuration shown in FIG. 4, wherein the housing, shield, C2, and the PWM are grounded at a common point.

3. It is possible to ground the housing, capacitor C2, the braided shield, and the PWM at a point other than the PWM's local ground. Point P1 in FIG. 4 represents such a point. In this case, it is the commonality of the grounding point for four components which is significant, rather than the location of the common ground point at the PWM.

4. The housing H is not necessarily a structure which is added to a self-contained, independently operable motor, although it could be. Instead, housing H provides support for the motor, as by supporting the bearings which support the rotor, and so on.

5. The PWM frequency of 20 KHz was chosen in order to lie outside the range of average human hearing. The PWM controller generates a hum at the pulse frequency, which can be objectionable when the invention is used in an automobile. (A somewhat analogous hum is found in the noise produced by magnetostriction within power transformers, such as the 60 Hz buzz commonly heard in electrical equipment.)

It would be preferred to run the PWM at a lower frequency, such as 10 KHz, because the solid-state electronics within the PWM are much more efficient, perhaps by as much as 40 percent, compared with 20 KHz.

6. The preferred motor draws about 25 amps, from a battery of 12 volts.

7. The invention can be used with motors which produce RFI which interferes with nearby electronic equipment. Preferred value of capacitor C2 was given above. However, C2 can range from 400-600 pico-farads.

8. Definitions. The following definitions are relevant. Other definitions for these terms can be found.

"AM frequencies" are defined as the frequency band used by AM radio, and extends from about 500 KHz to about 1600 KHz.

"FM frequencies" are defined as the frequency band used by FM radio, and extends from about 85 MHz to about 118 MHz.

"Low side chopper" is a term sometimes applied to the PWM controller shown in FIG. 2. The term is derived from the concept that the current is "chopped" to form the square wave train T, and that the chopped current is supplied to the "low side," or ground, of the motor.

Extending the terminology of the previous paragraph, the capacitor C2 extends from the "high side" to the housing, and thence to ground. Capacitor C2 can be called a "high side filter" or a "high side shunt" Current is commonly delivered to the motor M in FIG. 2 by the leads shown. However, the current is the active agent in delivering power; the leads provide a path for the power to travel. It is the current which generates noise, and the current which travels through the braided shield BR.

The PWM controller is "remote" from the motor. The PWM controller could be contained within the same housing H as the motor, but it is not. The PWM is located such that power leads of length between 6 inches and 2 feet are required to reach the motor.

Alternate Embodiment

FIG. 10 illustrates an alternate embodiment. A single inductor is used, and a ground strap GR replaces the braided shield.

Additional Embodiment

FIG. 11A illustrates one form of the invention as described above, and FIG. 11B illustrates a modification. In FIG. 11B, capacitors C2A and C2B, connected in parallel, replace capacitor C2 of FIG. 11A. An additional capacitor C2C is shown in FIG. 11B, and represents optional capacitors which can be added, for purposes which will later be explained. FIG. 11B also illustrates an additional capacitor C3, which is not shown in FIG. 11A.

FIGS. 12 and 13 illustrate test results obtained from the apparatus of FIGS. 11A and 11B, respectively. FIG. 13 indicates that the noise performance of the apparatus of FIG. 11B is significantly better than that of FIG. 11A.

In the following discussion, the inventors will propose some explanations as to why the added capacitors in FIG. 11B provide the improved noise performance. However, it must be recognized that, from an electromagnetic emissions perspective, the overall system shown in FIG. 11B is extremely complex, and cannot be explained in a simple manner.

For example, within the motor housing H exist multiple radiating sources (not shown), at multiple locations, which radiate at multiple frequencies. A truly accurate analysis of the situation would require a solution of Maxwell's equations, subject to boundary conditions, which include the motor case, which is not, in actuality, a simple geometric shape. Such an analysis would be very time-consuming, and may, in fact, be impossible to attain. Instead of such detailed analysis, the inventors offer some general principles which suggested the configuration of FIG. 11B.

One simple explanation of the relatively high noise performance of the apparatus of FIG. 11A lies in the fact that capacitor C2 is not actually a pure capacitance at the frequencies of interest, but contains inductance as well. That is, capacitor C2 can be modeled as shown in FIG. 14, which shows an inductance L in series with a capacitor C.

For a given capacitor, the size of the inductance L can be estimated from data sheets supplied by the manufacturer. FIG. 15 illustrates a hypothetical data sheet. The data sheets plot impedance, Z, versus frequency. In general, the plot will display a minimum point, such as at frequency F. At the minimum point, the capacitance C in FIG. 14 is considered to be at resonance with the inductor L.

At resonance, the capacitive impedance, which equals 1/jwC, equals the inductive impedance, jwL, wherein j is the imaginary operator used in complex numbers, w is radian frequency, C is capacitance and L is inductance. By equating the two expressions (1/jwC-jwL), the value of L can be obtained, because the values of C and w are known.

The presence of inductance, L, is not desired. One reason is that the impedance of the inductor is directly proportional to frequency, as indicated by the expression jwL, given above. As frequency increases, the impedance of L increases. But this increase is opposite to the desired impedance between P1 in FIG. 11B and the motor housing H. It is desired that, instead, the impedance be minimal, in order to shunt signals from point P1 to the housing H, which is grounded. Therefore, it would be desirable to reduce, or eliminate, L.

From a more quantative perspective, the total impedance Z of the series L and C of FIG. 14 is jwL+1/jwC, as indicated in FIG. 14. If L and C are given nominal values of one Farad and one Henry, respectively, as indicated in FIG. 16, then the impedance Z can be plotted as shown in FIG. 16, which plots the imaginary component of Z.

The plot of FIG. 16 was obtained from the software package Mathematica, available from Wolfram Research, Champaign, Ill. The code at the top of the Figure was that given to Mathematica.

At frequencies below F1, the imaginary component is negative, indicating that the impedance is capacitive. At frequencies above F1, the imaginary component is positive, indicating that the impedance is inductive. Further, the inductive impedance increases, as frequency increases, which is not desirable, as explained above.

If two capacitors C2 (both containing a C in series with an L as in FIG. 14) are placed in parallel, as in FIG. 17, their total impedance, Zpar, is computed by the parallel impedance rule, given in Equation 1, below. ##EQU1## The impedance of the inductor L is given by Equation 2, and the impedance of the capacitor C2 is given by Equation 3.

Z.sub.L =j.omega.L (2) ##EQU2## Their total series impedance is given by Equation 4. ##EQU3## Replacing Z1 in Equation 1 by its equivalent from Equation 4, and replacing Z2 in Equation 1 by its equivalent, which is not shown, but is the same as that in Equation 4, but with subscripts changed from "1" to "2", produces Equation 5. ##EQU4## Expanding Equation 5 produces Equation 6. ##EQU5## Multiplying numerator and denominator by j produces Equation 7. ##EQU6## Grouping terms produces Equation 8. ##EQU7## Assuming that the two capacitors are identical, so that C1=C2 and L1=L2, results in Equation 9. ##EQU8##

FIG. 17 is a plot of the imaginary part or FIG. 9, as a function of frequency. (No resistance was included in the equations above, so that equation 9 contains only imaginary parts.) FIG. 17 should be compared with FIG. 16.

Plainly, in FIG. 17, at a given frequency, the impedance is one-half that of FIG. 16. For example, at about 4.2 radians/sec, the impedance in FIG. 16 is about 4.0 ohms, but at the same frequency in FIG. 17, the impedance is about 2.0 ohms.

Restated, two goals have been achieved. One is that the inductive impedance (where the imaginary part of Z is positive, and occurs for frequencies above F1 in FIGS. 16 and 17) has been reduced, at every frequency, compared with FIG. 16. In effect, the vertical axis in FIG. 17 has been shrunk by a factor of one-half, compared with FIG. 16. This reduction promotes signal transfer from point P1 in FIG. 11B to the housing H, by reducing impedance of the signal path.

The second goal is that the capacitive impedance (where the imaginary part of Z is negative, and occurs for frequencies below F1 in FIGS. 16 and 17) has now been reduced, at every frequency, compared with FIG. 16. This reduction also promotes signal transfer from point P1 in FIG. 11B to the housing H.

Both the capacitive impedance and the inductive impedance of the single capacitor C have been reduced in FIG. 11A.

Additional Perspective

Another perspective can explain the functioning of the invention. FIGS. 18 and 19 illustrate an approximate method of determining a filter function for a ladder network. In FIGS. 18A-18C, the value of L is the same at 1.0. The value of C changes. In FIG. 18A, C is 1.0; in FIG. 18B, C is 3.0; in FIG. 18C, C is 5.0. Each Figure shows a plot of impedance, and the Mathematica code used to obtain the plots.

The plots are combined in FIG. 19A. If one draws an "envelope" E around the plots, as in FIG. 19B, one obtains an approximation of the filter function which may be expected from the parallel combination of the three networks in FIGS. 18A-18C. Thus, a band-pass filter having a bandwidth represented as B in FIG. 19B is obtained. This bandwidth B is wider than that for any of the networks shown in FIG. 18.

Based on a combination of the techniques described above, capacitor C2 in FIG. 11A was replaced by two capacitors C2A and C2B in FIG. 1B. These capacitors were 500 and 1,000 pico-Farad in value. In addition, other capacitors, represented by C2C in FIG. 19B, can be added, to broaden the bandwidth B in FIG. 19B.

From another point of view, capacitor C2 in FIG. 11A, which was stated in Table 1 as having a value of 500 pico-Farads, is augmented in FIG. 11B by another capacitor of value 1,000 pico-Farads.

Capacitor C3

Capacitor C3 in FIG. 11B connects from the motor lead B-, which is connected to the negative terminal of the automobile battery to the grounded housing H. The reader is reminded that lead B- is not connected to the battery at all times, because of the cycling of switch SW shown in FIGS. 9 and 10. When switch SW opens, lead B- becomes disconnected from the battery.

Lead B- in FIG. 11A was suspected of contributing to the radiated noise, partly because of its periodic disconnection from the battery. Capacitor C3 in FIG. 11B connects this lead B- to the grounded housing H, and acts to shunt noise signals to ground. Capacitor C3 is preferably about 500 pico-Farads in value.

Significantly a single 500-pF C3 sufficed. No additional capacitor, such as C2B, was needed, as capacitor C2A required.

Numerous substitutions and modifications can be undertaken without departing from the true spirit and scope of the invention. What is desired to be secured by Letters Patent is the invention as defined in the following claims.

Claims

1. In a motor which is given a current supplied on leads from a remote PWM controller, and which radiates RFI, the improvement comprising:

a) a first capacitor (C2A) connecting between a lead (B+) and ground; and

b) a second capacitor (C2B) connecting between said lead (B+) and ground.

2. Improvement accordance to claim 1, in which the first capacitor has a value of about 500 pico-Farads, and the second capacitor has a value of about 1,000 pico-Farads.

3. Improvement according to claim 1, and further comprising:

c) a third capacitor, connecting between a second lead (B-) and ground.

4. In a motor having two power leads and a noise-reduction capacitor (C2) connecting between a first of the leads and a ground, the improvement comprising:

a) a second capacitor in parallel with said noise-reduction capacitor.

5. Improvement according to claim 4, and further comprising:

b) a third capacitor connecting between a second of the leads and a ground.