Please use this persistent identifier to cite or link to this item:
doi:10.24405/9924
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.advisor | Sachau, Delf | - |
| dc.contributor.author | Hanselka, Jonas | - |
| dc.date.accessioned | 2020-09-01T07:51:19Z | - |
| dc.date.available | 2020-09-01T07:51:19Z | - |
| dc.date.issued | 2020 | - |
| dc.identifier.uri | https://doi.org/10.24405/9924 | - |
| dc.description.abstract | Die Steigerung des Komforts für den Menschen in Büros und Schlaf- oder Wohnräumen durch Lärmreduktion wird schnell problematisch, wenn die Lösung nicht mehr nur lokal an einem Ort einem Menschen zur Verfügung gestellt werden soll, sondern sich global auf einen ganzen Raum auswirken soll. Aktive Lärmreduktion erzeugt lokale Zonen der Ruhe mit begrenzter Ausdehnung, weshalb zur Beruhigung ganzer Räume, diese Räume in Gänze mit Sensoren und Aktoren belegt sein müssten und so die Nutzbarkeit und den Komfort des Raumes für den Menschen erheblich beeinträchtigten. Ein Lösungsansatz dazu ist es, das aktive Gegenschallsystem in den Transmissionspfad zu verschieben, den der Störschall in den Raum nimmt, es dort vollständig zu integrieren und so den Störschall am Eintritt in den Raum zu hindern. Bisherige Systeme, die diesen Ansatz verfolgen, sind, bezogen auf ein gekipptes, oder geöffnetes Fenster, nicht in den Fensterrahmen integriert, sondern erfordern Aufbauten vor dem Fenster in Form von Mikrophonen und Schallkanälen, oder Aufbauten über die gesamte Fensterfläche. Das im Rahmen dieser Arbeit entwickelte System integriert alle Sensoren und Aktoren vollständig in den Fensterrahmen. Dazu wird der FB-FxLMS-Algorithmus genutzt, der, anhand des Fehlersignals, das Stör- bzw. Referenzsignal schätzt und so die Installation von Referenzmikrophonen außerhalb des Gebäudes vor dem Fenster überflüssig macht. Die größte Schwachstelle dieses Algorithmus ist die Signallaufzeit zwischen der Sensierung am Fehlermikrophon und der Ausgabe des entsprechenden Gegensignals am Gegenschalllautsprecher. Um diese zu reduzieren werden verschiedene Maßnahmen angewandt. Durch optimierte, kolokale Positionierung der Komponenten und die Weiterentwicklung des vollständigen MIMO-Ansatzes zum Ansatz der vernetzten und überlappenden SISOSysteme, können der längste Signalpfad von über 1,63 m auf 0,22 m und die Anzahl der erforderlichen Rechenoperationen auf 9 ‰ reduziert werden. Zusätzlich wird die digitale Abarbeitung des Algorithmus durch parallele Verarbeitung auf mehreren Kernen der MCU so weit beschleunigt, dass analoge Tiefpassfilter nicht mehr notwendig sind. Mit diesem System werden in einem Raum mit 25,2 m3 Volumen bei breitbandiger Anregung mit weißem Rauschen 16 dB Reduktion des Schalldruckpegels erreicht. Zur kostengünstigen und energiesparenden Implementierung des Algorithmus wird dieser als Hardware-Schaltung auf einem FPGA umgesetzt. Dabei wird darauf geachtet, dass die Schaltung zur Optimierung hinsichtlich geringer Latenz, oder hinsichtlich geringem Platzbedarf frei skalierbar ist. So kann die Schaltung abhängig von der gegebenen Problemstellung für jede Ziel-Hardware konfiguriert werden. Durch die Umsetzung auf einem FPGA können auf dem verwendeten Testsystem Filterlängen von bis zu 131072 Filterkoeffizienten und Abtastraten von bis zu 3,8 GHz erzielt werden. Um sowohl die rechenoperations-optimierte Umsetzung des FB-FxLMS für mehrdimensionale Anwendungen auf einem MCU, als auch die latenz- und platz-optimierte Umsetzung für FPGA schnell und universell anwendbar zu machen, werden abschließend noch Vorschriften zur Abschätzung der erforderlichen MAC/s und daraus abgeleitet des erforderlichen MCU, bzw. zur Abschätzung der, je nach Einstellung, erreichbaren Abtastrate und daraus abgeleitet der erforderlichen Anzahl an CLB-, RAM- und DSP-Bausteinen vorgestellt. | - |
| dc.description.abstract | Hamburg is the city with the highest population density in Germany and as the commercial and industrial metropolis that it is, it is one perfect example for the noise pollution by machines with its negative impact on people. Therefore Hamburg measures the noise pollution caused by road-, rail- and air-traffic and by the port and publishes maps displaying the distribution of the noise pollution every five years. Regarding the road traffic, the latest figures from the year 2017 show that 140000 people in Hamburg are exposed to a noise level of 55 dB to 60 dB in a 24 h-average. These are 7.7 % of Hamburgs 1.8 million inhabitants. During nighttime between 22pm and 6am this number only decreases to 121000, which underlines the negativ impact even on the nights sleep. Contrary to this the number of exposed people to heavier noise in the range from 70 dB to 75 dB decreases from 24900 within the 24 h-average to 800 during nighttime, which shows that the heavy noise is a minor problem for restorative sleep. Clinical studies (e.g. from the cardiological institute of Mainz) proof the negative or in long-term even deadly impact of the noise on human. A studie with exposition to traffic noise of 15000 participants resulted in an increase of cardiovascular disease of 15 % compared to the non-exposed. These severe consequences are one reason Hamburg demands protective measures against noise pollution whenever a new building is planned. These measures have to fullfill a certain sound insulation on the exterior walls including the openings like doors and windows. Often the windows are replaced by thick glass panels and air conditioners. Although it is possible to significantly improve the sound insulation of partially opened/ tilted windows by active measures which rely on the destructive interference of soundwaves. It is even possible to completely include the components of such an active system into the reveal of a window. Like this the appearance and the functioning of the window are not disturbed. By further using a feedback controller with internally estimated reference signal instead of a feedforward controller with externally measured reference signal even the need of installing components outside of the building is avoided. The Multi-Channel feedback active noise control system using the Normalized Leaky Filtered Reference x Least Mean Squares (FxLMS) algorithm as it is decribed in ([Kuo96] Kuo, Morgan: Active Noise Control Systems: Algorithms and DSP Implementations, John Wiley & Sons, New York, 1996, pp. 33, 36 and 202) serves as a basis for the controllers developed during the experiments presented in this thesis. As a resulting system a system with 20 error microphones as inputs and 20 secondary loudspeakers as outputs is developed which reduces the noise passing through a tilted window by 16 dB in the frequency range of 150 Hz to 250 Hz. The window used for these experiments measures 1325 mm in heigth and 950 mm in width and has a maximal gap of the tilted casement of 80 mm and is installed in a transmission test bed between an anechoic transmitter room and a reverberant receiver room. To run this system with its 40 input/output channels and a longest soundpath of 1,63 m at the diagonal it is necessary to conduct the following changes with respect to the setup described in [Kuo96]. 1. On the one hand the setup as a traditional Multiple Input / Multiple Output (MIMO) system which uses every transfer path between every single secondary loudspeaker and error microphone could not be run in a stable state as the sound transfer paths are too long. These long transfer paths cause high delays between the sensing of the error signal and the output of the anti-noise signal at the secondary loudspeaker. On the other hand it is not a solution to simply drive every secondary loudspeaker with its accompanying error microphone as a Single Input / Single Output system ([Kuo96], p.195) because they influence each other. Therefore the system is set up with 20 overlapping systems with three inputs and one output each, which use the output of the two nearest neighbouring systems as input and avoid their destabilizing influence. 2. Each output signal of a 40-channel MIMO system based on the FxLMS algorithm needs 820 discrete convolution operations to be calculated. Considering high sample rates and long filter lengths (in the final setup it is 500 coefficients for each filter at 25 kHz) this means (820x filter length) of multiplications and additions during each sample. Conventional digital signal processors which work on a micro processor (MCU) basis are easily at their performance limit with this amount of operations. By adapting the calculation of the output signals the necessary number of convolutions could be reduced to 20 per output. 3. A discrete convolution operation is a series of independent multiplications which can be conducted in parallel. To avoid the parallel use of a high number of MCU which are designed to execute a lot of tasks which are not necessary for a discrete convolution it is more effective to design hardware dedicated to this task. Field Programmable Gate Arrays (FPGA) are the suitable hardware to do this. For this reason a platform independent, modular description in Very High Speed Integrated Circuit Hardware Description Language has been developed which can operate at different bit widths and can be configured to optimize for high sample rates, long filter lengths or low hardware resource usage. The design is able to provide a constant processing latency of only three clock cycles that is independent from the filter length. With this design sythesized to a Xilinx Kintex 7 FPGA filter lengths of up to 131072 coefficients or sample rates up to 4.5 MHz could be achieved. The noise reduction system developed whith respect to these three major considerations has all its components fully integrated into the reveal of the window in such a way that they have no impact in the functioning and the handling of the window. It has been named Active Noise Blocker because it prevents the noise from entering a room. The single papers written during the researches are published in english. | - |
| dc.description.sponsorship | Mechatronik | - |
| dc.language.iso | ger | - |
| dc.publisher | Universitätsbibliothek der HSU / UniBwH | - |
| dc.subject | Konstruktion eines Demonstrators | - |
| dc.subject | Optimierung des Algorithmus | - |
| dc.subject | Feedback FxLMS | - |
| dc.subject | Mehrdimensionale Regelung | - |
| dc.subject | FPGA | - |
| dc.subject | VHDL | - |
| dc.subject | Adaptive breitbandige Lärmminderung | - |
| dc.subject.ddc | 000 Informatik, Wissen & Systeme | - |
| dc.subject.ddc | 600 Technik | - |
| dc.subject.ddc | 620 Ingenieurwissenschaften | - |
| dc.title | Adaptive Lärmminderung an einem teilgeöffneten Fenster mittels mehrdimensionaler Regelung - Konstruktive und algorithmische Optimierung und Umsetzung als Hardware-Schaltung | - |
| dc.title.alternative | Adaptive Noise Reduction at a partially opened Window with multidimensional Feedback Control - Optimization of the Setup and the Algorithm and Implementation as Hardware Circuit | - |
| dc.type | Thesis | - |
| dcterms.dateAccepted | 2020-01-24 | - |
| dc.contributor.referee | Klauer, Bernd | - |
| dc.identifier.urn | urn:nbn:de:101:1-2020090209104529304256 | - |
| dcterms.bibliographicCitation.originalpublisherplace | Hamburg | - |
| dc.contributor.grantor | HSU Hamburg | - |
| dc.type.thesis | Doctoral Thesis | - |
| local.submission.type | full-text | - |
| hsu.opac.import | opac-2020 | - |
| hsu.identifier.ppn | 1735000566 | - |
| hsu.dnb.deeplink | https://d-nb.info/1216954089/ | - |
| item.grantfulltext | open | - |
| item.languageiso639-1 | de | - |
| item.fulltext_s | With Fulltext | - |
| item.openairetype | Thesis | - |
| item.fulltext | With Fulltext | - |
| Appears in Collections: | 2 - Theses | |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| openHSU_9924.pdf | 132.48 MB | Adobe PDF | View/Open |
CORE Recommender
Google ScholarTM
Check
User Tools
Items in openHSU are protected by copyright, with all rights reserved, unless otherwise indicated.