Consumer Audio Devices from TVs to Docking Stations to Personal Computers Benefit from the Use of FIR Filters in Providing High Quality Audio at a Reasonable Cost

By QuickFilter Technologies, Inc.

The U.S. market for consumer electronics goods is approximately $161B and growing at a rate of 8% per year (Consumer Electronics Association Digital America 2008). Television remained the biggest driver fueled by the growth of LCD TVs at 32% CAGR and Portable audio devices with a CAGR of 22%. To date, the key product differentiator has been unit price. As the market matures, differentiation will proceed along two vectors; one being unit price and the other being audio/video quality. In the past, these have been thought to have been mutually exclusive. Great strides have been made in addressing video quality, but audio quality has continued to be subpar in Digital TV and portable audio devices.

click to enlarge

Figure 1. Audio precision measurement of speaker transfer function.
Audio quality has often been sacrificed in order to achieve the cost structure necessary to be competitive, but as the market matures, audio quality will become an important product differentiator. The compromise in audio quality is often reflected by the manufacturers' choice of speakers. The speakers chosen by these manufacturers are constrained by their cost, the physical size of the equipment, and the manufacturing process in mounting the speakers. In making the design and cost tradeoffs, little attention is given to the speaker response over the audible frequency range (speaker transfer function) because it is thought that the equalizers in the audio processors will compensate for flaws in these lower cost speakers. This rarely happens. Listen to the soundtrack of a DVD played through a $3,000 HDTV and the same soundtrack played through a good quality sound system. The difference is very noticeable. With some minor changes, however, manufacturers can approach the audio quality of a high end sound system with very little to no incremental cost.

Many manufacturers believe that good audio quality is synonymous with maximum bass; however, audio quality is really the ability to reproduce the sound track exactly as it was recorded. To achieve this, the speaker response needs to be flat across the audible spectrum. Manufacturers of high-end stereo equipment have recognized ? and achieved ? this requirement. In many cases, these manufacturers ship their equipment with no bass or treble adjustments. Manufacturers of consumer electronics equipment can achieve this same performance by using Finite Impulse Response (FIR) filters that are low in cost but high in performance. Typical audio devices which will benefit from the use of FIR filters include:

  • Televisions, especially high definition TVs
  • Docking stations
  • Mid to low end stereo systems
  • Boomboxes
  • Headsets both wired and wireless
  • Personal computer audio equipment
An ideal speaker would be capable of reproducing all the audible frequencies at the same volume at which they were recorded. Rarely in life are things perfect, so the generally accepted standard is a variation of ±3 dB around a centerline. The 3 dB point is considered to be the point at which volume changes are barely perceptible. Even if a speaker transducer were to possess this characteristic, the process of mounting that speaker into a cabinet will change its response characteristics, so compensation will still need to be applied after the speaker is mounted in its final enclosure. The use of an FIR filter will simplify this compensation.

The graph shown in Figure 1 was taken with an Audio Precision 2700. It measures the frequency response of a name brand 40 inch LCD HDTV. It is clear that the response is not flat. The variation around the centerline (of-18) varies from 6 to 8 dB. The lower frequencies are overemphasized and tend to wash out the mid- to high-end ranges. Also, notice the significant drop in mid-range frequencies. The mid-range drop will affect the intelligibility of the dialog (particularly female speech). The goal is to bring response back so that it is within ±3 dB of centerline to ensure the frequencies are heard at the same volume at which they were recorded.

There are two methods that can be employed to equalize this speaker. The more traditional way is to use a five-band parametric equalizer comprised of five Infinite Impulse Response (IIR) digital filters. These are typically found in today's audio processors. There are two disadvantages to these filters. The first disadvantage is that they are relatively narrow band and must be placed strategically to cover the frequencies of interest. In addition, there are discontinuities at the overlaps which can cause distortion. The second disadvantage is that IIRs introduce phase distortion (different frequencies are delayed through the filter by different amounts of time). This again causes distortion to be heard.

click to enlarge

Figure 2. Docking station block diagram using the QF1DaF12.
An alternate method to achieving a flat response is to create an inverse of the transfer function and program it into a digital FIR filter. An FIR filter is broadband in nature and has absolutely no phase distortion. In theory, if you inverted every point of the graph in Figure 1 and reran the original test with the filter in series with the speaker, the resulting response would be a flat line across the measurement range. In reality, absolute perfection would require a filter with an extremely large number of taps which translates into very large die size (and thus high cost) as well as add significant delays into the audio stream. However, a 512-tap digital FIR filter will have a delay of <6 ms which is generally considered acceptable. Also, the resolution of this 512 tap filter is approximately 24Hz which means that it can perform reasonably down to about 100 Hz. This is a very cost effective filter and will yield a result that, while not perfectly flat, will meet the criteria of being within ±3 dB of a center line reference.

An FIR filter by its nature is designed to attenuate signals. In an audio application, this attenuation has the effect of reducing the dynamic range of the audio signal. The loss of dynamic range compresses the volume heard throughout the passage making the differences between soft passages and crescendos less dramatic with the overall effect being a less pleasing listening experience. In the past, to overcome this, the system designer would have to add a digital gain stage to make up for loss of dynamic range, thus adding cost and complexity to the design. With the introduction of the QF1Da512 (Audio SavFIRe) from Quickfilter Technologies, this is no longer the case. The Audio SavFIRe is the industry's first dedicated FIR filter to contain a gain and compressor stage that directly addresses the issue of dynamic range loss. The device supports up to 512 taps with 32 bit coefficients and user defined data word width from 12 to 24 bits, providing the user with superior audio quality. The Audio SavFIRe processes the equivalent of 50 million MAC instructions per second and supports sample rates from 1 Hz to 500 KHz, placing it in the class of mid-range Digital Signal Processors and exceeding the performance of most audio processors.

The QF1Da512 (Audio SavFIRe) costs less than $1.00 in volume, comes packaged in a 3 × 3 mm QFN and consumes only 2.8 mW of power providing the user with a clear price-performance advantage. The Digital Gain and Compression (DGC) block in the QF1Da512 (Audio SavFIRe) combines both gain and compression in one operational block. The compression is performed sample by sample with an instant attack and release. The slope of the compression curve is adjusted as the gain increases. The gain block is configured to provide up to 24 dB of positive digital gain. This combines to yield the maximum in dynamic range without distortion.

Using the Audio SavFIRe, a high-quality mp3 docking station can be designed with just a stereo codec, a microcontroller, and a simple power amplifier (see Figure 2.)

For an HDTV sound system, the system designer can use the Audio SavFIRe in co-processor mode, by placing it between the digital output and the digital input of the audio processor. Alternately (Figure 4), the HDTV manufacturer could exchange the Audio Processor for a high-quality (lower-cost) ADC. Coupling that with the Audio SavFIRe and a digital power amplifier would yield a digital audio subsystem of exceptional quality at lower cost than their current audio subsystems.

With this impressive new integrated circuit, manufacturers of consumer audio equipment can now take a significant step on the price performance curve.

H1 Eliminating Background Noise – The Next Best Audio Quality Improvement

By Lloyd Watts, PhD, Chairman, Founder & Chief Technology Officer, Audience

In recent years, advancements in noise suppression technology, and the willingness to design a second microphone into a mobile handset have enabled the last bastion of voice quality challenges heretofore unsolved — background noise suppression.

With some of the basic connectivity and network coverage issues solved, engineers are turning to new areas for audio quality improvement. With new solutions, like Audience's, offering noise suppression of up to 30 dB, there is a great opportunity in background noise suppression. With the addition of a second microphone into the handset to capture audio signals, and a powerful processor that emulates the human hearing system, mobile phone users are able to hear and be heard in places previously unimagined.

When choosing an advanced noise suppression solution, there are several audio quality enhancements that are now available. Transmit noise suppression in handheld mode and speakerphone modes, acoustic echo cancellation, and receive noise suppression with voice equalization that suppresses background noise from the incoming caller, and dynamically boosts their voice over noises in the local environment. Both objective and subjective measures are critical when evaluating solutions. Objective measures are defined in the ITU-T G.160 specification, and subjective measures in the P.835 standard. The G.160 standard is being amended to accurately capture the high levels of suppression that are being achieved in these new solutions. Improvements in Total Noise Level Reduction (TNLR), Signal to Noise Ratio Improvement (SNRI), and Delta Signal to Noise (DSN) are expected to be ratified by mid 2009.

The P.835 standard uses Mean Opinion Scores (MOS) ratings to measure audio quality improvements of advanced noise suppressors. The standard has been newly updated in Amendment I Appendix III calling for additional methods for testing of non-stationary noise sources.
Improving Audio Performance in Handsets

By Sanjay Voleti, Audio Marketing Director, National Semiconductor Corp.

Mobile handsets are packing in numerous new features such as video, MP3 playback, FM radio, GPS, and mobile TV driven by 3G networks. In addition, the rapid growth of handset sales over the last decade means that these devices are now used in a diverse range of environments (home, office, crowded airports, train stations, etc.) with different ambient noise levels.

This all points to the need for today's handsets to support higher quality audio while enabling phone conversations in all sorts of environments. Of course, all this must be done with minimal impact, if any, on battery life. There are several developments underway to address these trends.

For example, the latest audio subsystems now integrate ADCs, DACs as well as headphone and speaker amplifiers. This level of integration does not sacrifice performance — the ADCs and DACs support 24-bit audio and provide SNR performance of 100 dB while the speaker amps are capable of providing output power approaching 1.5W. Additionally, these audio subsystems provide very high PSRR at the GSM pulse frequency to eliminate the "bumble bee" noise that can interfere with surrounding audio equipment. Most of us have experienced this noise when using older cell phones in a car or near other speaker systems.

Handset manufacturers also want users to enjoy extended battery life when listening to MP3 music, FM radio or watching a video. New headphone amps based on Class G or H architectures allow the headphone to use the most optimized voltage supply, thereby helping extend battery life.

The most exciting developments are in the area of noise reduction. Everyone has either seen or experienced a phone conversation affected by a very noisy environment. Audio uplink and downlink noise reduction are directly focused on addressing this issue. The key here is to make sure that natural voice quality is preserved and power consumption is minimized.

Another issue handset makers wrestle with relates to the speakers. In order to support usage in noisy environments, most manufacturers like audio subsystems that provide very high output power. This is important for creating high SPL across various types of handsets with different form factors — sliders, bar-type, flip phone, etc. However, the speakers are very small and high output power can damage them. Newer audio subsystems solve this problem with built-in Automatic Level Control, enabling handsets with the highest audio SPL, and no speaker damage or distortion.

The ability to provide high quality audio for both entertainment and voice will become a key differentiator as handsets continue to support an increasing number of applications.