The mobile phone market continues to evolve from primarily voice communications devices to sophisticated organizational and entertainment “appliances.” With the arrival of smartphones, users enjoy feature-rich portability, such as an integrated MP3 players, video playback, video and still picture cameras, blue-tooth, and GPS–all with a touch screen interface.
Additionally, operating systems with true multitasking have arrived with seemingly limitless applications, resulting in a powerful handheld tool. The smartphone is truly a unique example of engineering innovation that is changing people’s lives.
Along with this functionality, smartphone users expect high performance. This article addresses a key area of handset performance:, the mobile phone's audio playback, specifically the MP3-player audio output to headphones or earbuds.
Audio consists of desirable content and undesirable content. An example of desirable audio content is music or movies, while examples of undesirable content are power-supply noise, harmonic distortion, cross talk, and data compression, which corrupts the listening experience. Noise can also be a part of the music content itself, but the audio IC headphone amplifier and the system surrounding that amplifier can be a major noise contributor if not designed carefully.
With respect to amplifier noise, the key parametric is signal to noise ratio (SNR). Higher SNR results in higher audio quality, while low SNR results in a noisier output. Let us first define SNR. Every audio output has a ‘noise floor,’ the inherent noise of the system and audio ICs.
Best board-layout practices can help reduce the noise floor, as can optimized IC designs. The goal is to maintain as much difference between the amplitude of the noise floor (undesirable content) and the music (desired content). Noise is most obvious during quiet passages between songs, or when playing content with high dynamic range where the difference between louder and quieter passages is higher, as is the case with classical music.
SNR is a ratio of desired content to undesired content. From an analytical standpoint, the equation is a ratio expressed on a logarithmic scale due to the vastly different amplitudes of the noise floor and desired signal.
where P signal is the average power of the desired signal (music) and P noise is the average power of the noise floor. Expressed as voltage:
where A is the root mean square (RMS) voltage of the signals.
In audio-IC design, there is often a reference used for A signal, which is the maximum output level at which the amplifier can deliver a 1-kHz sine wave into a 32 ? load at 1% total harmonic distortion + noise (THD+N). Note: some manufactures specify A signal under a no-load condition or at a higher THD+N to artificially elevate the SNR spec.
All IC amplifiers have an associated SNR depending on their design and layout. These amplifiers power external headphones when listening to music or movies on a mobile handset. Given that data than makes up the media is already highly compressed (usually MP3, AVI, or MOV formats), why is there a requirement for over 100 dB SNR headphone amplifiers, which is even higher than a CD (which has a theoretical SNR of 96 dB)?
First, data compression and noise are distinctly two different components of audio degradation. Data compression is based on lossy algorithms that reduce file size by eliminating or masking portions of content that are not easily heard by the human ear. You cannot recover the lost content as a result of data compression.
Further, you cannot reduce the noise that is inherent in the original recording. However, you can reduce noise that is added by the handset design, including the headphone amplifier IC.
The specified SNR of an amplifier is measured in the lab with a fixed numerator (A signal) value. It is important to note it is not a typical listening level. For example, many IC headphone amplifiers are capable of outputting over 30 mW into 32O. If turned up that loud, you can hear the earbud from across a room!
In reality, due to the close coupling of the eardrum to the earbud which is inserted in the ear canal, a reasonable listening level is in the range of 0.1 mW to 0.5 mW into 32O, depending on the efficiency of the earbuds. This is only a fraction of the full output power. Because SNR is the ratio of the signal to noise, and since the noise floor does not change, using these realistic listening levels lowers the apparent SNR to the listener.
As an example, an amplifier is specified at 105 dB SNR using an A signal 1 kHz tone at 30 mWRMS into 32?, with 1% THD+N. First, convert 30 mWRMS to VRMS :
PRMS = (VRMS )2 /R
or VRMS = v(PRMS x R) = v(0.03 W RMS x 32?) = 0.979 VRMS
So 105 dB = 20 log10 (0.979VRMS /A noiseRMS )
Or: inverse-log10 (105 dB/20) = .979 VRMS /A noiseRMS
so A noise = 5.5 µVRMS
This can also be calculated as:
105 dB/20 = log10 (0.979 VRMS /A noise)
10105dB/20 = 0.979 VRMS /A noiseRMS
A noise = 0.979 VRMS / 10105dB/20 = 5.5 µVRMS
Now that we know the noise floor A noise, let’s determine the SNR using the same amplifier, under the same conditions, at a typical listening level of 0.1 mWRMS . Again, using the equation:
First convert 0.1 WRMS to VRMS:
PRMS = (VRMS )2 /R
or VRMS =v(PRMS x R) = v(0.001 W RMS x 32?) = 0.179 VRMS
Now calculate the new SNR:
SNRdB = 20 log10 (0.179 VRMS /5.5 µVRMS ) = 90.24 dB.
Note this is about 15 dB less at typical listening level.
SNR measurements are a key indicator of the quality of an audio amplifier. Given that users now expect the same audio quality as found on MP3 players, special attention must given to this important specification.
About the author
Greg Davis is the for the Signal Conditioning Product Line at Fairchild Semiconductor Corp. Prior to Fairchild, Greg was the Senior Director of the Portable Audio Product Line at Leadis Technologies, and held various technical and management positions during his 15 years at Texas Instruments (TI), including systems engineer, product marketing engineer, systems manager, business marketing manager and product marketing manager. These positions spanned across many of TI's semiconductor groups, with a concentration in mixed signal analog.
Greg held several technical and management positions in the mobile communications industry at Motorola, Ericsson Radio Systems and ElectroSpace Systems, Inc. He graduated from Ohio Northern University in 1986 with a bachelor’s degree in electrical engineering with a focus on telecommunications and RF systems. He holds multiple patents, has been published in over 50 industry marketing and technical journals, and has chaired industry working groups.