av JM » 2020-02-01 08:24
Verkar vara svårt att hitta Valks avhandling. Publiserar abstract mm.
Initialt användes en Peerless högtalare i försöken men pga uppbrytningar i konen och därmed massiv distorsion byttes till SEAS L26RO4Y med betydligt stabilare kon.
Äldre analoga MFB lösningar använda av Philips och andra gav betydande negativa maskeringar. DSP lösningar verkar inte brottas med maskerings problemen. DSPns signalfördröjning vekar vara hanterbar.
Mätdata i avhandlingen pekar solklart på att någon MFB lösningen borde sitta i alla små seriösa audiofil subbasar.
Möjligen kan maskeringarna i äldre analoga MFB lösningar bidragit till att ljudet upplevdes som sämre?
Finns det dubbelblind lyssningstester eller simplare lyssningstest på DSP MFB subasar?
Vad tror ni?
Control of Voicecoil transducers Design and implementation of a Motional Feedback Loudspeaker Woofer
Master of Science Thesis
For the degree of Master of Science in Mechanical Engineering at Delft University of Technology
R. Valk November 13, 2013
Abstract:
This master thesis describes the dynamical behaviour of a loudspeaker woofer and the mod- elling of it. By the use of the knowledge gained from the model, a feedback controller is implemented in order to enhance the woofer’s performance in terms of acoustic total har- monic distortion (THD) and low frequency bandwidth. The first objective of the research is to build a woofer set-up that suppresses the THD below 1% during operation. The second objective is to evaluate diaphragm break-up and compensate for this break-up by the use of feedback.
The woofer is equipped with an accelerometer sensor. By the use of feedback, the motion of the voice-coil of the loudspeaker is controlled. This enhanced motion results into an acoustic enhancement.
Since the motion of only a single point on the woofer diaphragm is measured and used for feedback, only the local distortion is reduced. Even when feedback is applied, the surround of the woofer is radiating acoustic distortion. Initially a woofer is used that has some very specific characteristics in terms of acoustic radiation. The contribution of distortion by the surround is large with the chosen woofer. Therefore, the increase in performance measured on the accelerometer is not identical to the increase in performance measured by the microphone. Where the accelerometer is mounted, the motion of the diaphragm is improved, but the acoustic distortion radiated by the surround is not reduced.
After analysing the first woofer, the knowledge gained from the experiments is used to de- termine a better candidate for the experiments. The experiments are repeated on a second set-up.
Two important observations have led to the final result. One observation is that the surround of the woofer is a large contributor of the acoustic radiation. Choosing a woofer that is very uniform in terms of the distortion profile across the diaphragm, is advised. In that situation, when the motion of the location where the sensor is mounted is improved, the same holds for the surround of the woofer. Secondly, it is observed that when using a piezoelectric sensor, the sensor output is not only determined by the acceleration. Stress leading to deformation of the sensor is measured too. When the contribution of measured deformation becomes dominant over the measured acceleration, the magnitude of the signal is that of the deformation instead of the acceleration. For low frequencies, this leads to a limitation in terms of potential loop gain.
To some extend, both problems have been solved by the use of a different type of sensor mount and by a careful selection of the woofer. The second woofer used, has an acoustic distortion profile that is nearly identical across the membrane. This observation indicates that for the operating bandwidth of the woofer, the motion of the diaphragm is a close match to that of a rigid piston. Increasing the performance in terms of the motion of the centre, therefore leads to an increase in performance throughout the entire diaphragm. In order to solve the problem of the measured deformation of the sensor, a different type of sensor mount is designed. This sensor mount reduces the deformation of the sensor. This leads to a steeper roll-off slope in the sub-resonant frequency band of the woofer. This steeper slope makes it possible to design a controller, that leads to higher distortion suppression.
The achieved reduction of harmonic distortion measured on the accelerometer sensor is up to 22 dB, a factor 12,5. The acoustic reduction of harmonic distortion measured with a microphone is up to 22 dB too. In practice, the frequency band of this high suppression is narrow. The frequency band in which the suppression is over 17 dB, a factor 7, is between 40 Hz and 150 Hz
When the woofer excursion becomes large, the THD without feedback can be over 12,5%. Suppressing the distortion by a factor 12,5 therefore does not lead to the target THD of under 1%. The diaphragm break-up is evaluated, but a controller that suppresses this motion is not implemented. A controller would only compensate for the break-up effect on the diaphragm location where the sensor is mounted. The actual break-up in the diaphragm would remain.
Preface:
This research is inspired by a product developed by Philips. In 1970, Philips developed the MFB series of loudspeakers in which a woofer is equipped with a piezoelectric accelerometer. This accelerometer is used as a sensor to measure the acceleration of the centre of a woofer. That sensor signal is used for feedback control that makes the loudspeaker perform better by means of Total Harmonics Distortion (THD) in the first octaves and extend the low frequency bandwidth.
After reading on that series of loudspeakers, interest grew large on how that control might be enhanced by the use of modern digital electronics.
The Philips MFB series implemented feedback that reduced the harmonic distortion up to 10 dB, a factor 3,3. The Philips research facility [33], later on improved the motional feedback control loop and reported significant improvements over the control loop that was imple- mented in the regular production units. This enhanced control loop however, was never implemented in a commercially available product.
The enhanced control loop is different from the original circuitry in terms of the method used in order to achieve a high loop-gain. It uses two signals that are added to the sensor output. These two signals, depending on the frequency content of the input, can be dominant or sub-dominant over the sensor output. This method thereby creates a low frequency and high frequency masking effect. The frequency band in which the sensor output is dominant over the two masking signals is the control bandwidth of the woofer. This different approach to loop-shaping results in a loop gain, that is higher than the loop gain that was achieved before. In appendix A-23, more information about this improved control loop is available.
With the implementation of modern Digital Signal Processors (DSP), the new method de- veloped by Philips is not required in order to obtain a high loop gain. Due to the flexibility a DSP introduces, regular loop-shaping can be applied, as is done in the initial MFB series. Since the limitations in terms of analogue circuit design are reduced by the DSP flexibility, the remainder of limitations are the dynamics of the loudspeaker woofer itself. Both achieving a high loop gain and extending the control bandwidth, are the objectives of this research.
The research led to satisfactory results and was performed with great pleasure. Performing research on a topic that is close to the heart, eased the perceived workload.
Reflection and recommendations
Looking back on the the research, I am pleased with the final results and with the knowledge gained on this topic. The final measurements, as described in Chapter 10, show that the concept of MFB is one with potential when analysed properly.
The objectives set are not reached. Depending on the excursion of the woofer, the THD is not reduced below 1% under testconditions. The improvement in terms of THD are up to a factor 11, but even with that suppression, the THD can reach 4% at a 20 Hz high level reference signal. Tuning the controller can make the controller suppress the harmonic distortion by a higher magnitude, but at the cost of low frequency stability. The second objective was to control the diaphragm break-up. The controller however, was limited in control bandwidth due to stability issues. The analysis of the diaphragm break-up remains a interesting subject, but with the accelerometer sensor used, it was not possible to measure and control both the centre and the surround of the diaphragm.
JM
Annihilerar antimateria. Beauty is in the Brain of the Listener. "Kill your Darlings" => Scientific Progress.
