28
Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist [email protected] Hakon Strande Program Manager WMDG hakons @ microsoft.com Microsoft Corporation

Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist [email protected] Hakon Strande Program Manager WMDG

Embed Size (px)

Citation preview

Page 1: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Media Advances:High-Fidelity Audio from Integrated Audio Components

David RoachAudio [email protected]

Hakon StrandeProgram ManagerWMDGhakons @ microsoft.comMicrosoft Corporation

Page 2: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Session OutlineSession Outline

In Search of Better AudioWhy

Terminology

Attributes of Good Audio

Audio is Part of the SystemCommon Complaints

Considerations

Testing

Resources

Page 3: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Why Focus on Fidelity From Integrated Audio?Why Focus on Fidelity From Integrated Audio?

Much can be gained with minimal effortDesign board with audio in mind

Our motivation for providing these guidelines95% of PC audio solutions are integrated

Microsoft would like the Windows PC to achieve parity with Consumer Electronics fidelity

Microsoft Universal Audio Architecture initiativeStandardizing external and integrated audio around established technology specifications

Integrated audio UAA compliant solution is Intel High Definition Audio

Has great potential if device and board implementation is done with fidelity in mind

Page 4: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

TerminologyTerminology

In order to understand and communicate about audio, consistent terminology must be used

Volt

Decibel (dB)

Hertz

Octave

Frequency Response

Signal To Noise Ratio (SNR)

Total Harmonic Distortion plus Noise (THD+N)

Page 5: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

What Is A Volt?What Is A Volt?

The expected line input and output level of a PC is 1 Volt RMS for PCs using 5 Volts DC for the analog audio power supply (AVdd)

PCs using 3.3V power supplies for AVdd are expected to put out at least 0.707 Volts RMS

RMS stands for “Root Mean Square”, and is a method for measuring voltages that corresponds more closely to the perception of the human ear

1 Volt RMS: Is the same as 1.414 Volts Peak

Which is the same as 2.828 Volts Peak-to-Peak

Page 6: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

What is a Decibel or dB?What is a Decibel or dB?

A Bel was named after Alexander Graham Bell, the inventor of the telephone; It is a unit which defines a doubling of loudnessA decibel is 1/10th of a BelA decibel is commonly expressed as a ratio of two values0 dBV is referenced to 1 Volt RMS; Each doubling or halving of voltage is approximately 6 dB dB FS (Full Scale) is a measurement of digital signals relative to the maximum possible value

0 dB FS = maximum value possible, all other levels are expressed as minus

The average human ear has a dynamic range of over 10 Bels, or 100 dBVoltage is linear scale, dB is log scaleA range of 60 dB is a ratio of a thousand to oneA range of 120 dB is a ratio of a million to one3 dB is the smallest volume step that a typical untrained listener can easily discern; Trained listeners can identify smaller steps

Page 7: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

What Are Hertz?What Are Hertz?

Hertz is an measure of frequency expressed in cycles per second

A pure sine wave has a tone at only one frequency

A complex waveform will contain combinations of many different frequencies at any one time

The typical range of human hearing is 20 Hertz (Bass) to 20 kiloHertz (Treble)

Page 8: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

What Is An Octave?What Is An Octave?

An octave is a doubling of frequencyThe average human ear has a frequency response of over 10 octavesHertz is linear scale, octave is log scale10-octave ISO (International Standards Organization) center frequencies

31.25 Hz62.5 Hz125 Hz250 Hz500 Hz1000 Hz2000 Hz4000 Hz8000 Hz16000 Hz

Note that the ear has log response for both frequency (octaves) and volume (db), but our test equipment normally uses the linear scales such as Volts and Hertz

Page 9: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Attributes of Good AudioAttributes of Good Audio

High Dynamic Range (DR)Can accommodate both soft and loud signals

Adequate HeadroomAbility to reproduce peaks without distortion

High Signal to Noise Ratio, or SNRFreedom from noise

Low Distortion, or THD+NFreedom from distortion

Full-range frequency responseMatch the range of human hearing

Low channel-to-channel crosstalkMinimize leakage between audio channels

Page 10: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

High Dynamic Range, or DRHigh Dynamic Range, or DR

The dynamic range, or DR, is the ratio between the loudest signal that can be reproduced accurately and the softest signal that can be reproduced accurately.

A dynamic range ratio of 100 dB means that the loudest signal is at 0 dB (typically 0 dBV or 1 volt RMS) and that all measured noise signals are no greater than -100 dB (typically – 100 dBV or 10 microvolts)

A dynamic range of 100 dB is a minimal goal for media-centric designs; This requires a high-quality codec with greater than 16-bit resolution along with a good layout

For 100 dB dynamic range performance, all stray noises together must equal less than 10 microvolts RMS, or 28 microvolts peak to peak

Page 11: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Adequate HeadroomAdequate Headroom

Digital signal overload, or clipping, sounds very bad

To avoid clipping, a normal signal should be roughly 20 dB below the maximum output level, so that clipping does not occur on peak material

Running a signal at a level higher than this will likely result in clipping, and should be avoided

Page 12: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

High Signal to Noise Ratio or SNRHigh Signal to Noise Ratio or SNR

The signal to noise ratio, or SNR, is the ratio between the typical signal level and the softest signal that can be reproduced accurately.

DR = SNR + Headroom

For example, 80 dB SNR + 20 dB headroom = 100 dB of dynamic range

SNR and DR are often confused, and the term SNR is often used when the term DR should be used

Page 13: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Low Total Harmonic Distortion or THD+NLow Total Harmonic Distortion or THD+N

A pure sine wave has a tone or energy content at only one frequency; a complex waveform will contain combinations of many different frequencies at any one timeDistortion is a change or deviation between input and output signalHarmonic Distortion is deviation from a pure sine wave, which includes the energy at frequencies which are multiples (harmonics) of the original signal. For instance, a sine wave of 100 Hz has harmonics at 200 Hz, 300 Hz, 400 Hz, and so on Harmonic distortion is generally undesirable, but is harder to hearthan noiseTHD+N measures both harmonic distortion and noise. If a system is very noisy, then the SNR and the THD+N will probably be the same. For a well designed system, SNR will usually be better than THD+NTHD is mainly influenced by the codec and amplifier designsNoise is mainly influenced by both the components and the circuit layoutA THD+N of -90 dB is a minimal goal for media-centric designs

Page 14: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Frequency ResponseFrequency Response

The frequency response is the bandwidth of the audio passband, and should match the range of human hearing, that is 20 Hertz to 20 Kilohertz; The edges of the frequency response are the frequencies at the upper and lower end where the level drops below -3 dB compared to normal; A flat response curve is desired (less than +/- 0.5 dB or less)

-3dB down points define bandwidth

Hertz Octaves

Page 15: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Audio is Part of the SystemAudio is Part of the System

Everything affects audio!Audio sub-section Layout

Audio Power supply

CD/DVD analog outputs and power supply

CPU Power supply

External power supply (laptop)

Battery charge circuit (laptop)

Case grounding

Jack grounding

EMI suppression techniques

Page 16: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Common ComplaintsCommon Complaints

Noise sourcesNoise from mouse movement

Noise from stepper motors on disk drives

Pops and clicks due to uni-polar power supply

Inadequate power supply traces

Poor grounding.

Inadequate power supply

Microphonics

Ground loops

Page 17: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Different Types of NoisesDifferent Types of Noises

There are many different types of noisesHissing noise or white noise is typically steady state, and usually derives from the thermal noise of the components in the subsystem

Occasional or intermittent noise usually comes from other components, such as mouse movement, hard disk noise, CD stepper motors, RF circuits, memory strobing, video activity, and switching power supplies

Be sure to check noise while other components are in use

Reduce this noise by using components designed for low noise operation

The choice of audio codec is critical for low noise

Page 18: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Passive ComponentsPassive Components

The way that a resistor or capacitor is constructed or the materials that are used can add noise or distortion to a circuit

Not all passive components are suitable for audio

If a polarized capacitor is being used as a coupling capacitor, make sure that it is biased properly

For microphones, be sure to use a low-leakage capacitor, as any leakage may cause a DC offset which will be amplified by the microphone preamp

Avoid using resistor, capacitor, ferrite, or inductor packs which have more than one device in a single package. These are a possible source of crosstalk

Avoid using capacitors with varying tolerances or that have microphonic characteristics

Page 19: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Considerations for Laptop and Considerations for Laptop and All-in-One ModelsAll-in-One Models

Models with built-in speaker systems haveadditional considerations

Typically the larger the screen, the louder the expected listening volume

Loud listening volumes require a well-filtered power supply capable of providing peak currents in excess of the maximum current consumption of the power amplifiers

If the power supply is not well-filtered and regulated, or if the power supply traces are not wide enough, then the output amplifier can become unstable at high volumes

Page 20: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Considerations for MicrophonesConsiderations for Microphones

Microphones typically require extra preamp or gain stages, which add between 20 dB and 62.5 dB of additional gain

For external microphones, make sure that the microphone bias supply is super-clean (i.e., noise-free)

Keep microphone connections to a minimum length, preferably within an inch or two of the codec

For longer microphone runs, use a microphone with a built-in preamp of 20 dB or more; Generally avoid preamps which are not inside the microphone or near the microphone

When wiring for stereo microphones, maintain maximum isolation between bias supplies to ensure low crosstalk; this is especially important for beam-forming or phased-array microphones, as excessive crosstalk will cancel out the beamforming effect

Page 21: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Power Supply ConsiderationsPower Supply Considerations

Use a codec with a good PSRR (Power Supply Rejection Ratio) to minimize noise being induced from the power supply. This is especially important for designs which don’t have a dedicated voltage regulator.

Be especially careful with power supplies which provide mic bias voltage. Any noises on this supply will be amplified by the microphone preamp. For instance, a small 10 microvolt noise on the supply, amplified by a typical 40 dB preamp, will result in a noise floor

of -60 dB, which cannot pass WHQL requirements.

Whenever possible, leave a stuffing option for a voltage regulator even if a voltage regulator is not planned as part of the final design. This can prevent an expensive board re-spin late in the design cycle.

Try to always bring up digital power supply before analog power supply, and return analog power supply to zero before removing digital power supply. Fast turn-on or turn-off of analog power supply will cause pop noises.

Page 22: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

ISOLATE!ISOLATE!

HD Audio link allow a 5-wire serial connection between the codec and the controller

Take advantage of this to locate the codec as close as possible to the jack, and as far away as possible from other circuits in the computer

Use an isolated voltage regulator designated exclusively for analog audio power supply, as close to 5V as possible

Try to eliminate any coupling (magnetic, capacitive, or electronic) between the audio circuitry and any other circuits in the computer

If possible, build a “guard ring” around the audio circuitry, and don’t allow any other circuits within this area

Avoid routing audio traces near other circuitry, especially high-speed digital circuits

For external analog connections use twisted-pair shielded wires with the shield connected only at the input end; avoid using ribbon cable or unshielded cables for analog connections

Locate audio circuitry away from wireless LANs such as 802.11 and Bluetooth

Page 23: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Star GroundingStar Grounding

Star ground is accomplished by running a separate trace for each analog ground (return) shown on the schematic back to a central grounding point underneath the codec. Ground planes should only be used for shielding, and should be attached at a single point. Ground planes should never carry any current.

Connect digital ground and analog ground together directly underneath the codec

Ensure that there are no other “ground loops” or additional return paths, as these can cause noises to be induced into the audio

Take special care with shielding around the jacks, as well as EMI and ESD circuitry. Shields must be grounded in only one place, and connected directly to the main ground underneath the codec.

Ground return traces must be capable of handling as much current as each associated supply trace. An overly small ground return trace can cause noise due to the voltage drop over the resistance of the trace.

Star grounding is counter-intuitive for board designers who are unfamiliar with audio layout best practices

Use Optical SPDIF output or use transformer for coaxial SPDIF Out

Page 24: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

TestingTesting

Be sure to test under dynamic conditions

Listen to and measure SNR on both record and playback paths

While moving the mouse

While running Disk Defragmenter on each hard disk

While copying large files to and from CD/R

While copying large files over Wi-Fi and/or LAN

While exercising memory

While exercising video

Page 25: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Call to ActionCall to Action

Start at the beginningStart with the understanding that audio must be designed in from the beginning, not added in at the end

Consider your performance targetConsumer equipment for the living room has Signal to Noise Ratios of 115 dB or higher, but PCs are typically about 85 dB and can be as low as 65 or 70 dB

Utilize new Microsoft Designed for Windows required standardsHigh Definition Audio, part of the Microsoft Universal Audio Architecture initiative removes the digital format barriers which previously limited computer audio to 48kHz and 20 bits

Use good design and layout implementationsEven when using codecs with high-quality analog circuitry, audio performance depends on good design and layout to realize good quality audio

Take care at every stageTo realize living room quality, care has to be taken at every step of development, including schematic design and especially layout

Page 26: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Additional ResourcesAdditional Resources

EmailSigmaTel: hdaudio @ sigmatel.com

Microsoft: uaa @ microsoft.com

For assistance in your audio designContact SigmaTel or your codec vendor for guidance and assistance

Use reference designs where possible

Attempt “Golden Layouts”Once you have the perfect audio design, re-use it

Do not re-use designs that you know have audio issues You will continue to have the same issues.

Page 27: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG

Community ResourcesCommunity Resources

Community Siteshttp://www.microsoft.com/communities/default.mspx

List of Newsgroupshttp://communities2.microsoft.com/communities/newsgroups/en-us/default.aspx

Attend a free chat or webcasthttp://www.microsoft.com/communities/chats/default.mspx

http://www.microsoft.com/seminar/events/webcasts/default.mspx

Locate a local user group(s)http://www.microsoft.com/communities/usergroups/default.mspx

Non-Microsoft Community Siteshttp://www.microsoft.com/communities/related/default.mspx

Page 28: Media Advances: High-Fidelity Audio from Integrated Audio Components David Roach Audio Evangelist droach@sigmatel.com Hakon Strande Program Manager WMDG