Ed Grochowski's Website

Modeling the MG1.6

Ed Grochowski

Written 3-1-2024

Introduction

MG16

The MG1.6 is a planar electromagnetic loudspeaker manufactured by Magnepan from 1999 to 2010. Unlike conventional speakers, the MG1.6 is a dipole, radiating sound equally to the front and rear. This enables the MG1.6 to reproduce sounds with a high degree of realism.

The downside of a dipole is that cancellation occurs between the front and rear waves when wavelengths become greater than the speaker's width. A dipole speaker must compensate by boosting low frequencies.

This article examines how the MG1.6 achieves the necessary boost.


Model

To understand the MG1.6, I constructed a model of the crossover and panel. The crossover can be modeled exactly from the schematic.

mg16schematic

The crossover's high-pass section has a -3dB frequency of 1.6KHz and a slope of 6dB/octave. The low-pass section has a -3dB frequency of 230Hz and an ultimate slope of 12dB/octave. The MG1.6's crossover has an unusually wide gap between the high-pass and low-pass sections.

The crossover model is combined with an idealized panel model as shown below.

mg16model
HP Crossover high-pass filter
LP Crossover low-pass filter
HG High-pass gain
LG Low-pass gain
Sum RMS sum
P1-4 Panel section
G1-4 Panel gain

The HG and LG blocks adjust the crossover's output by 11dB to reflect the lower efficiency of the tweeter relative to the woofer.

The panel consists of four sections separated by tuning dots. The left and right speakers have slightly different tunings. I modeled the right speaker by substituting each panel section with a mechanical spring-mass-damper which is equivalent to an electrical second-order high-pass filter.

For each section, the model assumes that the resonance frequency is inversely proportional to the square root of the area (mass). The gain is proportional to the area. I chose a Q factor to produce the desired response. These parameters are summarized below.

Section Area (in^2) F (Hz) Q Gain
1 60 99 1.8 -13dB
2 130 67 2.2 -6dB
3 145 63 2.2 -5dB
4 265 47 4.5 0dB

Simulation

The overall response is as follows.

mg16graph1
Green: Crossover low-pass, Red: Crossover high-pass, Blue: Crossover and panel, Teal: Ideal

The graph has five distinct regions.

  • Below 47Hz, the output drops like a rock
  • From 47Hz to 200Hz, the combination of the four panel resonances provides boost
  • From 200Hz to 700Hz, the crossover's low-pass filter provides boost
  • The output reaches a local minimum at 1.4KHz
  • From 1.4KHz to 5KHz, the crossover's high-pass filter contributes a 2dB rise

The 2dB dip at 1.4KHz is a typical speaker voicing that gives the impression of an extended frequency response albeit with some loss of coherency. The dip may be reduced by adding a small resistor to attenuate the tweeter.

The corner frequency of 700Hz is where the sound's wavelength exactly matches the speaker's width (19 inches).

The simulated speaker does a remarkably good job of compensating for dipole cancellation. Moreover, it does so with no additional components in the crossover. This is clever engineering.


Measurement

At midrange and treble frequencies, the measured response exhibits similar characteristics as the simulation. At bass frequencies, the effects of the rear wave and room resonances cause the measured response to differ. In addition, some 60/120Hz hum is present in my test setup.

mg16graph2
Green: Measurement (1 meter), Blue: Simulation (RMS sum), Red: Simulation (complex sum)

At bass frequencies, the wavelengths are long and so phase matters. The phase changes rapidly as each panel section passes through resonance. The complex sum attempts to model the combined output at the point seen by the microphone.

The null at 140Hz is caused by the reflection of the rear wave from the wall behind the speaker.

The excess output around 1KHz and above 8KHz is due to beaming as wavelengths approach the widths of the woofer and tweeter, respectively.

I think the measurement and simulation agree closely enough that the model can claim to be correct.