
- •Contents
- •Acknowledgments
- •Preface
- •What a Crossover Does
- •Why a Crossover Is Necessary
- •Beaming and Lobing
- •Passive Crossovers
- •Active Crossover Applications
- •Bi-Amping and Bi-Wiring
- •Loudspeaker Cables
- •The Advantages and Disadvantages of Active Crossovers
- •The Advantages of Active Crossovers
- •Some Illusory Advantages of Active Crossovers
- •The Disadvantages of Active Crossovers
- •The Next Step in Hi-Fi
- •Active Crossover Systems
- •Matching Crossovers and Loudspeakers
- •A Modest Proposal: Popularising Active Crossovers
- •Multi-Way Connectors
- •Subjectivism
- •Sealed-Box Loudspeakers
- •Reflex (Ported) Loudspeakers
- •Auxiliary Bass Radiator (ABR) Loudspeakers
- •Transmission Line Loudspeakers
- •Horn Loudspeakers
- •Electrostatic Loudspeakers
- •Ribbon Loudspeakers
- •Electromagnetic Planar Loudspeakers
- •Air-Motion Transformers
- •Plasma Arc Loudspeakers
- •The Rotary Woofer
- •MTM Tweeter-Mid Configurations (d’Appolito)
- •Vertical Line Arrays
- •Line Array Amplitude Tapering
- •Line Array Frequency Tapering
- •CBT Line Arrays
- •Diffraction
- •Sound Absorption in Air
- •Modulation Distortion
- •Drive Unit Distortion
- •Doppler Distortion
- •Further Reading on Loudspeaker Design
- •General Crossover Requirements
- •1 Adequate Flatness of Summed Amplitude/Frequency Response On-Axis
- •2 Sufficiently Steep Roll-Off Slopes Between the Filter Outputs
- •3 Acceptable Polar Response
- •4 Acceptable Phase Response
- •5 Acceptable Group Delay Behaviour
- •Further Requirements for Active Crossovers
- •1 Negligible Extra Noise
- •2 Negligible Impairment of System Headroom
- •3 Negligible Extra Distortion
- •4 Negligible Impairment of Frequency Response
- •5 Negligible Impairment of Reliability
- •Linear Phase
- •Minimum Phase
- •Absolute Phase
- •Phase Perception
- •Target Functions
- •All-Pole and Non-All-Pole Crossovers
- •Symmetric and Asymmetric Crossovers
- •Allpass and Constant-Power Crossovers
- •Constant-Voltage Crossovers
- •First-Order Crossovers
- •First-Order Solen Split Crossover
- •First-Order Crossovers: 3-Way
- •Second-Order Crossovers
- •Second-Order Butterworth Crossover
- •Second-Order Linkwitz-Riley Crossover
- •Second-Order Bessel Crossover
- •Second-Order 1.0 dB-Chebyshev Crossover
- •Third-Order Crossovers
- •Third-Order Butterworth Crossover
- •Third-Order Linkwitz-Riley Crossover
- •Third-Order Bessel Crossover
- •Third-Order 1.0 dB-Chebyshev Crossover
- •Fourth-Order Crossovers
- •Fourth-Order Butterworth Crossover
- •Fourth-Order Linkwitz-Riley Crossover
- •Fourth-Order Bessel Crossover
- •Fourth-Order 1.0 dB-Chebyshev Crossover
- •Fourth-Order Linear-Phase Crossover
- •Fourth-Order Gaussian Crossover
- •Fourth-Order Legendre Crossover
- •Higher-Order Crossovers
- •Determining Frequency Offsets
- •Filler-Driver Crossovers
- •The Duelund Crossover
- •Crossover Topology
- •Crossover Conclusions
- •Elliptical Filter Crossovers
- •Neville Thiele MethodTM (NTM) Crossovers
- •Subtractive Crossovers
- •First-Order Subtractive Crossovers
- •Second-Order Butterworth Subtractive Crossovers
- •Third-Order Butterworth Subtractive Crossovers
- •Fourth-Order Butterworth Subtractive Crossovers
- •Subtractive Crossovers With Time Delays
- •Performing the Subtraction
- •Active Filters
- •Lowpass Filters
- •Highpass Filters
- •Bandpass Filters
- •Notch Filters
- •Allpass Filters
- •All-Stop Filters
- •Brickwall Filters
- •The Order of a Filter
- •Filter Cutoff Frequencies and Characteristic Frequencies
- •First-Order Filters
- •Second-Order and Higher-Order Filters
- •Filter Characteristics
- •Amplitude Peaking and Q
- •Butterworth Filters
- •Linkwitz-Riley Filters
- •Bessel Filters
- •Chebyshev Filters
- •1 dB-Chebyshev Lowpass Filter
- •3 dB-Chebyshev Lowpass Filter
- •Higher-Order Filters
- •Butterworth Filters up to 8th-Order
- •Linkwitz-Riley Filters up to 8th-Order
- •Bessel Filters up to 8th-Order
- •Chebyshev Filters up to 8th-Order
- •More Complex Filters—Adding Zeros
- •Inverse Chebyshev Filters (Chebyshev Type II)
- •Elliptical Filters (Cauer Filters)
- •Some Lesser-Known Filter Characteristics
- •Transitional Filters
- •Linear-Phase Filters
- •Gaussian Filters
- •Legendre-Papoulis Filters
- •Laguerre Filters
- •Synchronous Filters
- •Other Filter Characteristics
- •Designing Real Filters
- •Component Sensitivity
- •First-Order Lowpass Filters
- •Second-Order Filters
- •Sallen & Key 2nd-Order Lowpass Filters
- •Sallen & Key Lowpass Filter Components
- •Sallen & Key 2nd-Order Lowpass: Unity Gain
- •Sallen & Key 2nd-Order Lowpass Unity Gain: Component Sensitivity
- •Filter Frequency Scaling
- •Sallen & Key 2nd-Order Lowpass: Equal Capacitor
- •Sallen & Key 2nd-Order Lowpass Equal-C: Component Sensitivity
- •Sallen & Key 2nd-Order Butterworth Lowpass: Defined Gains
- •Sallen & Key 2nd-Order Lowpass: Non-Equal Resistors
- •Sallen & Key 2nd-Order Lowpass: Optimisation
- •Sallen & Key 3rd-Order Lowpass: Two Stages
- •Sallen & Key 3rd-Order Lowpass: Single Stage
- •Sallen & Key 4th-Order Lowpass: Two Stages
- •Sallen & Key 4th-Order Lowpass: Single-Stage Butterworth
- •Sallen & Key 4th-Order Lowpass: Single-Stage Linkwitz-Riley
- •Sallen & Key 5th-Order Lowpass: Three Stages
- •Sallen & Key 5th-Order Lowpass: Two Stages
- •Sallen & Key 5th-Order Lowpass: Single Stage
- •Sallen & Key 6th-Order Lowpass: Three Stages
- •Sallen & Key 6th-Order Lowpass: Single Stage
- •Sallen & Key Lowpass: Input Impedance
- •Linkwitz-Riley Lowpass With Sallen & Key Filters: Loading Effects
- •Lowpass Filters With Attenuation
- •Bandwidth Definition Filters
- •Bandwidth Definition: Butterworth Versus Bessel
- •Variable-Frequency Lowpass Filters: Sallen & Key
- •First-Order Highpass Filters
- •Sallen & Key 2nd-Order Filters
- •Sallen & Key 2nd-Order Highpass Filters
- •Sallen & Key Highpass Filter Components
- •Sallen & Key 2nd-Order Highpass: Unity Gain
- •Sallen & Key 2nd-Order Highpass: Equal Resistors
- •Sallen & Key 2nd-Order Butterworth Highpass: Defined Gains
- •Sallen & Key 2nd-Order Highpass: Non-Equal Capacitors
- •Sallen & Key 3rd-Order Highpass: Two Stages
- •Sallen & Key 3rd-Order Highpass in a Single Stage
- •Sallen & Key 4th-Order Highpass: Two Stages
- •Sallen & Key 4th-Order Highpass: Butterworth in a Single Stage
- •Sallen & Key 4th-Order Highpass: Linkwitz-Riley in a Single Stage
- •Sallen & Key 4th-Order Highpass: Single-Stage With Other Filter Characteristics
- •Sallen & Key 5th-Order Highpass: Three Stages
- •Sallen & Key 5th-Order Butterworth Filter: Two Stages
- •Sallen & Key 5th-Order Highpass: Single Stage
- •Sallen & Key 6th-Order Highpass: Three Stages
- •Sallen & Key 6th-Order Highpass: Single Stage
- •Sallen & Key Highpass: Input Impedance
- •Bandwidth Definition Filters
- •Bandwidth Definition: Subsonic Filters
- •Bandwidth Definition: Combined Ultrasonic and Subsonic Filters
- •Variable-Frequency Highpass Filters: Sallen & Key
- •Designing Filters
- •Multiple-Feedback Filters
- •Multiple-Feedback 2nd-Order Lowpass Filters
- •Multiple-Feedback 2nd-Order Highpass Filters
- •Multiple-Feedback 3rd-Order Filters
- •Multiple-Feedback 3rd-Order Lowpass Filters
- •Multiple-Feedback 3rd-Order Highpass Filters
- •Biquad Filters
- •Akerberg-Mossberg Lowpass Filter
- •Akerberg-Mossberg Highpass Filters
- •Tow-Thomas Biquad Lowpass and Bandpass Filter
- •Tow-Thomas Biquad Notch and Allpass Responses
- •Tow-Thomas Biquad Highpass Filter
- •State-Variable Filters
- •Variable-Frequency Filters: State-Variable 2nd Order
- •Variable-Frequency Filters: State-Variable 4th-Order
- •Variable-Frequency Filters: Other Orders of State-Variable
- •Other Filters
- •Aspects of Filter Performance: Noise and Distortion
- •Distortion in Active Filters
- •Distortion in Sallen & Key Filters: Looking for DAF
- •Distortion in Sallen & Key Filters: 2nd-Order Lowpass
- •Distortion in Sallen & Key Filters: 2nd-Order Highpass
- •Mixed Capacitors in Low-Distortion 2nd-Order Sallen & Key Filters
- •Distortion in Sallen & Key Filters: 3rd-Order Lowpass Single Stage
- •Distortion in Sallen & Key Filters: 3rd-Order Highpass Single Stage
- •Distortion in Sallen & Key Filters: 4th-Order Lowpass Single Stage
- •Distortion in Sallen & Key Filters: 4th-Order Highpass Single Stage
- •Distortion in Sallen & Key Filters: Simulations
- •Distortion in Sallen & Key Filters: Capacitor Conclusions
- •Distortion in Multiple-Feedback Filters: 2nd-Order Lowpass
- •Distortion in Multiple-Feedback Filters: 2nd-Order Highpass
- •Distortion in Tow-Thomas Filters: 2nd-Order Lowpass
- •Distortion in Tow-Thomas Filters: 2nd-Order Highpass
- •Noise in Active Filters
- •Noise and Bandwidth
- •Noise in Sallen & Key Filters: 2nd-Order Lowpass
- •Noise in Sallen & Key Filters: 2nd-Order Highpass
- •Noise in Sallen & Key Filters: 3rd-Order Lowpass Single Stage
- •Noise in Sallen & Key Filters: 3rd-Order Highpass Single Stage
- •Noise in Sallen & Key Filters: 4th-Order Lowpass Single Stage
- •Noise in Sallen & Key Filters: 4th-Order Highpass Single Stage
- •Noise in Multiple-Feedback Filters: 2nd-Order Lowpass
- •Noise in Multiple-Feedback Filters: 2nd-Order Highpass
- •Noise in Tow-Thomas Filters
- •Multiple-Feedback Bandpass Filters
- •High-Q Bandpass Filters
- •Notch Filters
- •The Twin-T Notch Filter
- •The 1-Bandpass Notch Filter
- •The Bainter Notch Filter
- •Bainter Notch Filter Design
- •Bainter Notch Filter Example
- •An Elliptical Filter Using a Bainter Highpass Notch
- •The Bridged-Differentiator Notch Filter
- •Boctor Notch Filters
- •Other Notch Filters
- •Simulating Notch Filters
- •The Requirement for Delay Compensation
- •Calculating the Required Delays
- •Signal Summation
- •Physical Methods of Delay Compensation
- •Delay Filter Technology
- •Sample Crossover and Delay Filter Specification
- •Allpass Filters in General
- •First-Order Allpass Filters
- •Distortion and Noise in 1st-Order Allpass Filters
- •Cascaded 1st-Order Allpass Filters
- •Second-Order Allpass Filters
- •Distortion and Noise in 2nd-Order Allpass Filters
- •Third-Order Allpass Filters
- •Distortion and Noise in 3rd-Order Allpass Filters
- •Higher-Order Allpass Filters
- •Delay Lines for Subtractive Crossovers
- •Variable Allpass Time Delays
- •Lowpass Filters for Time Delays
- •The Need for Equalisation
- •What Equalisation Can and Can’t Do
- •Loudspeaker Equalisation
- •1 Drive Unit Equalisation
- •3 Bass Response Extension
- •4 Diffraction Compensation Equalisation
- •5 Room Interaction Correction
- •Equalisation Circuits
- •HF-Cut and LF-Boost Equaliser
- •Combined HF-Boost and HF-Cut Equaliser
- •Adjustable Peak/Dip Equalisers: Fixed Frequency and Low Q
- •Adjustable Peak/Dip Equalisers With High Q
- •Parametric Equalisers
- •The Bridged-T Equaliser
- •The Biquad Equaliser
- •Capacitance Multiplication for the Biquad Equaliser
- •Equalisers With Non-Standard Slopes
- •Equalisers With −3 dB/Octave Slopes
- •Equalisers With −3 dB/Octave Slopes Over Limited Range
- •Equalisers With −4.5 dB/Octave Slopes
- •Equalisers With Other Slopes
- •Equalisation by Filter Frequency Offset
- •Equalisation by Adjusting All Filter Parameters
- •Component Values
- •Resistors
- •Through-Hole Resistors
- •Surface-Mount Resistors
- •Resistors: Values and Tolerances
- •Resistor Value Distributions
- •Obtaining Arbitrary Resistance Values
- •Other Resistor Combinations
- •Resistor Noise: Johnson and Excess Noise
- •Resistor Non-Linearity
- •Capacitors: Values and Tolerances
- •Obtaining Arbitrary Capacitance Values
- •Capacitor Shortcomings
- •Non-Electrolytic Capacitor Non-Linearity
- •Electrolytic Capacitor Non-Linearity
- •Active Devices for Active Crossovers
- •Opamp Types
- •Opamp Properties: Noise
- •Opamp Properties: Slew Rate
- •Opamp Properties: Common-Mode Range
- •Opamp Properties: Input Offset Voltage
- •Opamp Properties: Bias Current
- •Opamp Properties: Cost
- •Opamp Properties: Internal Distortion
- •Opamp Properties: Slew Rate Limiting Distortion
- •Opamp Properties: Distortion Due to Loading
- •Opamp Properties: Common-Mode Distortion
- •Opamps Surveyed
- •The TL072 Opamp
- •The NE5532 and 5534 Opamps
- •The 5532 With Shunt Feedback
- •5532 Output Loading in Shunt-Feedback Mode
- •The 5532 With Series Feedback
- •Common-Mode Distortion in the 5532
- •Reducing 5532 Distortion by Output Stage Biasing
- •Which 5532?
- •The 5534 Opamp
- •The LM4562 Opamp
- •Common-Mode Distortion in the LM4562
- •The LME49990 Opamp
- •Common-Mode Distortion in the LME49990
- •The AD797 Opamp
- •Common-Mode Distortion in the AD797
- •The OP27 Opamp
- •Opamp Selection
- •Crossover Features
- •Input Level Controls
- •Subsonic Filters
- •Ultrasonic Filters
- •Output Level Trims
- •Output Mute Switches, Output Phase-Reverse Switches
- •Control Protection
- •Features Usually Absent
- •Metering
- •Relay Output Muting
- •Switchable Crossover Modes
- •Noise, Headroom, and Internal Levels
- •Circuit Noise and Low-Impedance Design
- •Using Raised Internal Levels
- •Placing the Output Attenuator
- •Gain Structures
- •Noise Gain
- •Active Gain Controls
- •Filter Order in the Signal Path
- •Output Level Controls
- •Mute Switches
- •Phase-Invert Switches
- •Distributed Peak Detection
- •Power Amplifier Considerations
- •Subwoofer Applications
- •Subwoofer Technologies
- •Sealed-Box (Infinite Baffle) Subwoofers
- •Reflex (Ported) Subwoofers
- •Auxiliary Bass Radiator (ABR) Subwoofers
- •Transmission Line Subwoofers
- •Bandpass Subwoofers
- •Isobaric Subwoofers
- •Dipole Subwoofers
- •Horn-Loaded Subwoofers
- •Subwoofer Drive Units
- •Hi-Fi Subwoofers
- •Home Entertainment Subwoofers
- •Low-Level Inputs (Unbalanced)
- •Low-Level Inputs (Balanced)
- •High-Level Inputs
- •High-Level Outputs
- •Mono Summing
- •LFE Input
- •Level Control
- •Crossover In/Out Switch
- •Crossover Frequency Control (Lowpass Filter)
- •Highpass Subsonic Filter
- •Phase Switch (Normal/Inverted)
- •Variable Phase Control
- •Signal Activation Out of Standby
- •Home Entertainment Crossovers
- •Fixed Frequency
- •Variable Frequency
- •Multiple Variable
- •Power Amplifiers for Home Entertainment Subwoofers
- •Subwoofer Integration
- •Sound-Reinforcement Subwoofers
- •Line or Area Arrays
- •Cardioid Subwoofer Arrays
- •Aux-Fed Subwoofers
- •Automotive Audio Subwoofers
- •Motional Feedback Loudspeakers
- •History
- •Feedback of Position
- •Feedback of Velocity
- •Feedback of Acceleration
- •Other MFB Speakers
- •Published Projects
- •Conclusions
- •External Signal Levels
- •Internal Signal Levels
- •Input Amplifier Functions
- •Unbalanced Inputs
- •Balanced Interconnections
- •The Advantages of Balanced Interconnections
- •The Disadvantages of Balanced Interconnections
- •Balanced Cables and Interference
- •Balanced Connectors
- •Balanced Signal Levels
- •Electronic vs Transformer Balanced Inputs
- •Common-Mode Rejection Ratio (CMRR)
- •The Basic Electronic Balanced Input
- •Common-Mode Rejection Ratio: Opamp Gain
- •Common-Mode Rejection Ratio: Opamp Frequency Response
- •Common-Mode Rejection Ratio: Opamp CMRR
- •Common-Mode Rejection Ratio: Amplifier Component Mismatches
- •A Practical Balanced Input
- •Variations on the Balanced Input Stage
- •Combined Unbalanced and Balanced Inputs
- •The Superbal Input
- •Switched-Gain Balanced Inputs
- •Variable-Gain Balanced Inputs
- •The Self Variable-Gain Balanced Input
- •High Input Impedance Balanced Inputs
- •The Instrumentation Amplifier
- •Instrumentation Amplifier Applications
- •The Instrumentation Amplifier With 4x Gain
- •The Instrumentation Amplifier at Unity Gain
- •Transformer Balanced Inputs
- •Input Overvoltage Protection
- •Noise and Balanced Inputs
- •Low-Noise Balanced Inputs
- •Low-Noise Balanced Inputs in Real Life
- •Ultra-Low-Noise Balanced Inputs
- •Unbalanced Outputs
- •Zero-Impedance Outputs
- •Ground-Cancelling Outputs
- •Balanced Outputs
- •Transformer Balanced Outputs
- •Output Transformer Frequency Response
- •Transformer Distortion
- •Reducing Transformer Distortion
- •Opamp Supply Rail Voltages
- •Designing a ±15 V Supply
- •Designing a ±17 V Supply
- •Using Variable-Voltage Regulators
- •Improving Ripple Performance
- •Dual Supplies From a Single Winding
- •Mutual Shutdown Circuitry
- •Power Supplies for Discrete Circuitry
- •Design Principles
- •Example Crossover Specification
- •The Gain Structure
- •Resistor Selection
- •Capacitor Selection
- •The Balanced Line Input Stage
- •The Bandwidth Definition Filter
- •The HF Path: 3 kHz Linkwitz-Riley Highpass Filter
- •The HF Path: Time-Delay Compensation
- •The MID Path: Topology
- •The MID Path: 400 Hz Linkwitz-Riley Highpass Filter
- •The MID Path: 3 kHz Linkwitz-Riley Lowpass Filter
- •The MID Path: Time-Delay Compensation
- •The LF Path: 400 Hz Linkwitz-Riley Lowpass Filter
- •The LF Path: No Time-Delay Compensation
- •Output Attenuators and Level Trim Controls
- •Balanced Outputs
- •Crossover Programming
- •Noise Analysis: Input Circuitry
- •Noise Analysis: HF Path
- •Noise Analysis: MID Path
- •Noise Analysis: LF Path
- •Improving the Noise Performance: The MID Path
- •Improving the Noise Performance: The Input Circuitry
- •The Noise Performance: Comparisons With Power Amplifier Noise
- •Conclusion
- •Index

70 Crossover Types
Figure 4.9: Frequency response of Solen split 1st-order 1 kHz crossover; both filter outputs plus the sum, with one output reversed (phase inverted).
In passive crossover design, phase-inverting one of the outputs is extremely simple; just swap over the two wires to the drive unit in question; everything is done in the box, and no-one is any the wiser. Active crossovers, however, will need to use some kind of phase-inverting stage.
Figure 4.9 demonstrates that with one of the phases reversed we still get a dip, but it is now only 1.2 dB deep; information on this crossover scheme is scanty, but that is presumably how it is supposed to work. It might be possible to reduce the deviation from perfect flatness by partial cancellation of the dip with a response irregularity in one of the drivers. It is however still difficult to get enthusiastic about a crossover with 6 dB/octave slopes.
First-Order Crossovers: 3-Way
It is difficult to make a 1st-order 3-way crossover because the slow 6 dB/octave slopes do not provide adequate separation into three bands across the audio spectrum. There is in any case little point because drive units capable of handling the wide frequency ranges inherent in such a crossover would probably be equally suitable for a 1st-order 2-way crossover setup.
Second-Order Crossovers
The use of a 2nd-order crossover promises relief from the lobing, tilting, and time-alignment criticality of 1st-order crossovers, because the filter slopes are now twice as steep at 12 dB/octave.Avery large

Crossover Types 71
number of passive crossovers are 2nd-order, because they are still relatively simple, and this simplicity is very welcome, as it reduces power losses and cuts the total cost of the large crossover components required. Neither of these factors applies to active crossovers; the extra power consumption and the extra cost of making a 4th-order crossover rather than a 2nd-order crossover are very small.
All 2nd-order filters have a 180° phase-shift between the two outputs, which causes a deep cancellation notch in the response at the crossover frequency when the HF and LF outputs are summed. Such a response is of no use whatever, and the standard cure is to invert the polarity of one of the outputs.
With a 2nd-order crossover using Butterworth filters, this gives not a flat response but a +3 dB hump at the crossover frequency. As we shall see, the size of the hump can be much reduced by using a frequency offset; in other words the highpass and lowpass filters are given different cutoff frequencies. An offset factor of 1.30 turns the +3 dB hump into symmetrical amplitude ripples of ±0.45 dB, which is the flattest response that can be achieved for the Butterworth crossover by this method.
Second-order crossovers have much less sensitivity to driver time-misalignments because of their 12 dB/octave slopes. Vance Dickason [2] has shown that for a 2nd-order Butterworth crossover, a 1 inch time-misalignment gives errors of only fractions of a dB, while a 2 inch misalignment gives maximal errors of 2 dB. The corresponding figures for a 1st-order crossover are 4 dB and 10 dB respectively.
The frequency-offset technique can also be used to reduce the effect of time-alignment errors on the amplitude/frequency response.
A 2nd-order crossover gives better, though by no means stunning, protection of the drive units against inappropriate frequencies, less excitement of unwanted behaviour outside their intended frequency range, and less modulation distortion. Since one output has to be inverted to get a usable amplitude response, the outputs are in phase instead of 180° phase-shifted, and so there should be no lobing error, i.e. tilt in the vertical coverage pattern.
Second-Order Butterworth Crossover
The 2nd-order Butterworth crossover is perhaps the best-known type, despite the fact that it is far from satisfactory. A classic bit of crossover misdesign that has been published in circuit ideas
columns and the like a thousand times is shown in Figure 4.10. You take two 2nd-order Butterworth filters with the same cutoff frequency, one highpass and one lowpass, and there you have your two outputs. As before, the summing device represents how the two outputs add linearly in the air in front of the loudspeaker.
As Figure 4.11 shows all too clearly, this does not work well; in fact “catastrophic” would be a more accurate description. Each filter gives a 90° phase-shift at the crossover frequency, one leading and one lagging. The signals being summed are therefore a total of 180° out of phase and cancel out completely. This causes the deep notch at the crossover frequency seen in Figure 4.11.
Since a 180° phase-shift is the root of the problem, that at least can be eliminated by the simple expedient of reversing the connections to one of the drivers, normally the high-frequency one of the pair. Figure 4.12 shows the result—the yawning gulf is transformed into a much less frightening +3 dB hump centred at the crossover frequency.
In a passive crossover this reversed connection can be hidden inside the speaker enclosure along with the crossover components, but in an active crossover the issue is more exposed. You will need to either

Figure 4.10: A doomed attempt to make a crossover using two 2nd-order Butterworth filters.
Crossover 2nd-order Butterworth 1kHz-Small Signal AC-13-Graph
norm
5.0
0.0
−5.0
−10.0
−15.0
−20.0
−25.0
−30.0
−35.0
−40.0
100.0 |
freq |
1.0k |
10.0k |
|
|
|
Figure 4.11: The frequency response resulting from the in-phase summation of two 2nd-order Butterworth filters: a disconcerting crevasse in the
combined response. The dashed line is at −3 dB.

Crossover Types 73
Figure 4.12: Second-order Butterworth crossover, with the phase of one output reversed; the crevasse has become a more usable +3 dB hump. The dashed line is at −3 dB.
build a phase inversion into the active crossover, which again effectively hides the phase reversal from the user, or specify that one of the power amplifier-speaker cables be reversed.Alot of users are going to feel that there is something not right about such an instruction, and building the inversion into the crossover is strongly recommended.
Clearly our +3 dB hump is much better than an audio grand canyon, but accepting that much deviation from a flat response is clearly not a good foundation for a crossover design. That hump is going to be very audible. It might be cancelled out by an equalisation circuit with a corresponding dip in its response, but there is a simpler approach. Looking at Figure 4.12, it may well occur to you, as it has to many others, that something might be done by pulling apart the two filter cutoff frequencies so the response sags a bit in the middle, as it were. There is no rule in crossover design, be it active or passive, that requires the two halves of the filtering to have the same cutoff frequency.
Figure 4.13 shows the result of offsetting each of the filter cutoff frequencies by a factor of 1.30 times. This means that the highpass filter cutoff frequency is changed from 1.00 kHz to 1.30 kHz, while the lowpass cutoff becomes 1.00/1.30 = 0.769 kHz. Crossover now occurs at −6 dB. With the phase inversion, the offset factor of 1.30 turns the hump into symmetrical amplitude ripples of ±0.45 dB above and below the 0 dB line; this represents the minimum possible response deviation obtainable in this way. Now the amplitude response is looking a good deal more respectable, if not exactly mathematically perfect, and it is very questionable whether response ripples of this size could ever be audible. You may wonder if the frequency-offset process has rescued the response with outputs in-phase; the answer is that it is not much better; there is no longer a notch with theoretically

74 Crossover Types
Figure 4.13: Two 2nd-order Butterworth filters with 1.30 times frequency offset, in normal and reversed connection. With the phase of one output reversed, the +3 dB hump is smoothed out to a mere ± 0.45 dB ripple. The normal-phase connection still has a serious dip 9 dB deep. The dashed line is at −3 dB; the two filter cutoff (−3 dB) frequencies are now different.
infinite depth, but the there is a great big dip 9 dB deep at the bottom, and such a response is still of no use at all.
The frequency offsets required for maximal flatness with various types of crossover are summarised in
Table 4.1 at the end of this chapter.
The power response for the 2nd-order Butterworth crossover with no frequency offset is shown in
Figure 4.14; it is a perfect straight line. This qualifies it as a constant-power crossover (CPC); it is a pity that the power response is of much less importance than the pressure response. It is important to realise that the power response is the same whether or not one of the outputs is phase reversed. When the two outputs are squared and added to get the total power, the negative sign of the reversed output disappears in the squaring process.
When looking at this power response plot, it is important to appreciate that each crossover output is still shown at −3 dB at the crossover frequency, meaning the power output from it is halved. The two lots of half-power sum to unity, in other words 0 dB.
Earlier we saw that a frequency offset of 1.30 times was required to get near-flat amplitude response. How is that going to affect the perfectly flat power response of Figure 4.14? The answer, predictably, is that any change is going to be for the worse, and Figure 4.15 shows that there is now an 8 dB dip in the power response.

Figure 4.14: Power response of 2nd-order Butterworth crossover with no frequency offset—the sum is perfectly flat at 0 dB. Each crossover output is still at −3 dB at the crossover frequency, meaning the power is halved. The two half-powers sum to 0 dB.
Figure 4.15: Power amplitude response of 2nd-order Butterworth crossover with 1.30x frequency offset—rather less than perfect, showing an−8 dB dip at the crossover frequency.

76 Crossover Types
As mentioned earlier, 2nd order filters have a 180 degree phase-shift between the two outputs. This is shown in Figure 4.16, where the phase of the sum lies exactly on top of the trace for the lowpass output. The phase of the sum is that of a 1st-order allpass filter, the same phase characteristic as that of the 1st-order crossover. This is inaudible with normal music signals. For reasons of space, phase and group delay plots are from here on only given for the more interesting crossover types.
The summed group delay for the inverted connection is shown in Figure 4.17. It has a level section at 226 usec and a gentle peak of 275 usec just below the crossover frequency. Note that the level section shows less group delay than the 1st-order crossover.
When we were looking at 1st-order crossovers, you will recall that it was said that no other crossover could solve the waveform reconstruction problem, i.e. to get out the waveform that we put in after summing the outputs. How does a 2nd-order Butterworth cope with the square wave reconstruction problem?
The answer from Figure 4.18 is that it fails completely, and inverting one of the filter outputs makes things even worse. The phase-shifts introduced by the 2nd-order filters make it impossible for reconstruction to occur. While it is not very obvious from Figure 4.18, the first and second cycles of the simulation are not quite identical; since we have a circuit with energy-storage elements (capacitors) and we are starting from scratch, it takes a little time for things to settle down so you obtain the
result for continuous operation. In some cases it is not uncommon for 20 cycles to be required to reach equilibrium. Not every writer on the subject of audio has appreciated this fact, and major embarrassment has resulted.
Figure 4.16: Phase response of 2nd-order Butterworth crossover with one output inverted; the outputs are always 180° out of phase. The phase of the sum lies exactly on top of that of the lowpass output. Highpass output shown before inversion.

Figure 4.17: The group delay response of a 2nd-order Butterworth crossover, one output inverted.
Figure 4.18: Attempted reconstruction of a square wave by a 2nd-order Butterworth crossover without offset. Total failure!