HDA in theory

HDA in theory

The name used by Intel, High Definition Audio, is based on the 24 bits / 192 KHz sound support imposed by the norm. It’s an important HDA characteristic that only indicates what tomorrow’s numerical sound will be like. Indeed, for the moment content encoded in 24 bits / 192 KHz is restricted to Audio DVD and is only of interest to a small number of users. Another innovation introduced by HDA is 7.1 sound compatibility. Already included in several AC97 chipsets, HDA standardizes 7.1. In our opinion this is only a small detail, because the interest of a 7.1 system for a computer isn’t obvious and is even less interesting for an integrated chipset.

Just to review a couple of technical terms, 16 or 24 bits corresponds to the resolution of an audio file. During numerization, a value is attributed to each signal. In 16 bits each value varies from 0 to 65536 and in 24 bits from 0 to 16 700 000. When a signal doesn’t have a value corresponding to a whole number, the converter rounds the number arbitrarily to the next closest whole number. The sampling rate (44.1, 48, 96 KHz) corresponds to the number of values that the converter will process per second. In 48 KHz this number is 48 000 per second and 96 000 at 96 KHz. So theoretically, the higher the resolution and sampling rate, the closer the numerised version will be to the original or the sound the listener will hear.

We have to admit, however, that a simple Audio CD with tracks, a resolution of 16 bits and a sampling rate of 44.1 KHz seems adequate for the human ear. Based on this fact, we can also say that to increase the listening pleasure from a computer source, it would be much simpler and efficient to reduce noise (fan and power supply) and improve the quality of the material which reproduces the sound (speakers and audio chipset) rather than increase the resolution or sampling rate. We will return to these technical considerations in a future article and see if changing these parameters could actually increase sound quality restitution in computers.

One more interesting point is bandwidth. It isn’t fixed unlike that of the AC97 (11 MB/s) and is variable with a HDA controller. It reaches up to 48 MB/s in output and 24 MB/s in input with an HDA codec and doesn’t use this bandwidth unless it’s necessary. This flexibility is strengthened by the use of DMA multi-use channels, while the AC97 codec DMA channels are predefined. One benefit for example, would be if all DMA channels may be used for a single task it could be processed much faster.



We also noticed that with HDA, the audio part is better integrated to the system as the central chipset determines the codec internal functioning frequency (generally 48 KHz). Installation is also easier thanks to compatibility with Universal Audio Architecture (UAA). We will probably find an implemented driver for each HDA chip with Longhorn, which will allow basic audio functions without installation to a specific driver. The UAA driver doesn’t feature specific functions such as a control panel, EAX support or even Dolby Digital encoding.

This last function is in the HDA program but only as an option. We find it on a few motherboards such as the ABIT Fatal1ity AA8XE or even on most Premium line ASUSTeK motherboards based on the i915 or l’i925. DICE (Dolby Interactive Content Encoder) technology was included in NVIDIA’s APU, and is a software solution enabling the card to produce AC-3 encoded sound via a numerical output to send it to an appropriate decoder. Intel and Dolby developed a terminology, which determines the capacity of a sound card equipped with a HDA chip:

- Dolby Sound Room : 5.1 sound compatibility with an audio configuration for a single listener, Dolby Digital 5.1 decoding, Dolby Headphone, Dolby Virtual Speaker, Dolby Prologic II.
- Dolby Home Theater : 5.1 sound compatibility with audio configuration for several listeners, Dolby Digital 5.1 decoding, Dolby Headphone, Dolby Virtual Speaker, Dolby ProLogic II and Dolby Digital Stereo Creator for DVD creation.
- Dolby Master Studio : 7.1 sound compatibility with audio configuration for several listeners, Dolby Digital 5.1 and EX, Dolby Digital Live (DICE), Dolby Headphone, Dolby Virtual Speaker, Dolby ProLogic II and Dolby Digital encoding for DVD creation.

This terminology is completely lacking for motherboard manufacturers, which didn’t necessarily follow these considerations. We have noticed that manufacturers rarely communicate the presence of DICE software modules and the Dolby logo remains more than discreet on their products.

HDA in theory (next)

HDA in theory (next)

High Definition Audio architecture introduces the notion of flow and channels in the organization of audio data. Flow is the logical connection established by a DMA channel between the system’s memory and codec. This flow includes one or several components, called data channels, which are sent to a converter to be processed. For example, a stereo flow includes two channels, a left and a right. Each flow sample is doubled to represent both stereo channels.

The samplings are gathered when located in the system´s memory or transferred and can be separated to be sent to two different converters.



The flow is determined by its direction, either in or out. The specificity of “out” flow is that it can be attributed to several numerical and analogical converters, for example, to be available both from the stereo and headphone output. An “in” flow, however, is only intended for one decoder. An HDA controller can simultaneously support up to 15 in- and out-flows and it’s possible to have up to 16 channels integrated to each flow. All channels included in one flow must have the same sampling rate and resolution. Two flows, however, can include and provide channels with different sampling rates and resolutions. This wasn’t possible with the AC97.

The HDA also introduced (and imposes) Jack Sensing and Jack Retasking technology. With Jack Sensing the controller identifies a mini-jack plug when inserted. The principle is simple. Each mini-jack input and output has a switch, which is open when empty and closed when a jack plug is inserted. The system tells the driver, which then asks the user which type of media has been connected to the audio chipset; speakers, headphone, recording source, microphone, etc.



Jack Retasking then comes into effect. This technology assigns various tasks to an input or output. So it’s possible to transform lines and micro inputs to additional outputs, for example, for a 7.1 system. Or you can plug in two microphones, one on the line input and one on the micro input, because all entries with HDA codecs can theoretically benefit from microphone pre-amplification. We will see later that this isn’t necessarily true and that some HDA systems restrict Jack Retasking capabilities to the front panel. These ports are generally meant for stereo output and line input, which can be configured according to your needs.

Do not mistake Jack Sensing and Universal Audio Jack, even if they are intimately related. The Universal Audio Jack is indeed a system, which allows the controller to automatically detect the type of media connected to all inputs and outputs. It requires that the media be compatible with the norm, which is seldom the case, except for some of the latest headphones. HDA specifications as communicated by Intel only indicate Jack Sensing and Jack Retasking and don’t mention automatic detections. That doesn’t mean that HDA chips are compatible with Universal Audio Jack like the Realtek ALC880. The Universal Audio Jack also involves each mini jack connection needing adequate characteristics to become an input or output and a necessary pre-amplification system to be changed into headphone output or microphone input.

HDA chips

HDA chips

Two HDA chips are currently available on motherboards, the C-Media CMI-9880 and Realtek ALC880. Here are their specifications:


As both chips share a specific (and limiting) standard, they aren´t very different on paper. We noticed the low number of DS3D voices supported by the CMI-9880, but we will have to see what this means in practice. Both chips feature the Sensaura 3DPA, which isn’t very original. This engine supports the DS3D and EAX 1.0 and 2.0. Only the DS3D channel management is hardware, whereas all effects are entirely managed by software. The two chips can provide 7.1 sound, and Dolby Digital or DTS format are decoded by DVD software included with the motherboard.

Most of the boards have optical or coaxial S/PDIF inputs, either located on the motherboard next to the ATX bloc or on an internal bracket. For connections there are four mini jack outputs for 7.1 sound, and a line and microphone input. HDA chips are also able to support two front connections as is often found on most of the latest tower cases. With the Realtek ALC880 chip, the motherboard’s internal connections are configurable via the Jack Restasking function. Connections in the back have a fixed function, but still benefit from Jack Sensing.

There are two versions of the ALC880 available to motherboard manufacturers, the ALC880 and ALC880D. The only difference between these two chips is the presence of Dolby Digital encoding via the DICE, which allows the encoding of any flow in AC-3 and sends it to an external decoder via S/PDIF output. It is surprising to see that Realtek made two chips with different names for this function, which theoretically is entirely supported by the hardware. The CMI-9880 is also able to support DICE, but there is no denominational difference between chips which feature this option.

Reading and recording capacity are identical with both chips. The DAC supports up to 24 bits / 192 KHz, while the ADC is restricted to 20 bits (16 bits in practice) and 96 KHz. The ALC880 has 8 DAC 24 bits / 192 Khz at its disposal that provides 96 KHz with 8 channel tracks corresponding to a 24 bit / 192 KHz stereo track. There are three ADC 20 bits / 96 Khz entries, which support two line-in types or microphones and a general mixer allowing the mix of several flows in input. The C-Media CMI-9880 features 8 DAC 24 bits / 192 KHz and 4 ADC 16 bits / 96 KHz. With both chips, the only possibility for recording in 24 bits / 96 KHz is the use of the S/PDIF entry.

It’s important to note that with the C-Media chip, DAC supports a sampling rate of 48, 96 and 192 KHz but not 44.1 KHz. When reading a file in 44.1 KHz, as is the case with most MP3s and other similar files and audio CDs, the chip processes a 48 KHz resampling in output. This doesn’t happen with the ALC880 chip. It´s a recurring feature of C-Media chips and is present in the Audigy 1, which didn’t have a 44.1 KHz pass-through in DAC and processes internal resampling in 48 KHz. For most users and with middle range speakers, this is of little consequence. It’s just unfortunate that this choice necessarily involves a slight reduction in quality. It is important to note that this is a pseudo-numeric audio standard supposed to prevent an exact replications of 44.1 KHz content, such as Audio CDs on a computer. This isn’t something that limits this chip, because, as the chip datasheet indicates, the CMI-9880 0.98 version supports 44.1 KHz direct playing without resampling.

CMI-9880 : drivers

CMI-9880 : drivers

The CMI-9880’s main window offers the possibility to configure the type of speaker connected and allows adjustments of each settings thanks to a simple and pleasant interface. It is also possible to adjust the S/PDIF output sampling rate. (In our picture this area is faded, because we hadn’t yet connected the card S/PDIF bracket yet). The 7.1 Virtual Speaker Shifter control is specific to C-Media chips and allows reproduction of any stereo source with all speaker elements connected to the system. If this function is included in motherboard drivers, you can activate Dolby Digital Live encoding located in the bottom right of the window.


The Smart Jack window gives us information on connected hardware and allows reconfiguration thanks to Jack Retasking. The C-Media possibilities in this area are superior to the ALC880’s, because with the CMI-9880 one may entirely reconfigure all inputs/outputs as needed. With the Realtek chip this is only possible with the front panel’s two mini-jacks.


The mix window allows the adjustment of input and output levels.


The effect window gives us the possibility to configure the Sensaura engine environment if the user wants to apply an effect. The environment size is also configurable and has an influence on the range of compression and stereo enlargement. The right part of the window is devoted to the chip’s 10 band equalizer.

ALC880 : drivers

ALC880: drivers

The first window gives you access to environment effects and the chip’s 10 band equalizer.


Accessing the second window includes the possibility to adjust speaker configuration and indicates the available inputs and outputs.


The third window has a single 3D demonstration, giving you the possibility to try the Sensaura 3DPA engine option in positioning with three different effects.


The S/PDIF section is used to configure S/PDIF output.


Finally, accessing the last window allows you to configure mini-jack connections of the motherboard ATX bloc and front panel. Only the front panel mini-jacks are actually configurable with the Jack retasking function.

Input and output quality

Quality of input and output

To test these HDA chips we used RMAA 5.4 software for several objective tests combined with a Terratec DMX 6Fire 24/96 sound card. For subjective evaluation we used several listening sources. With RMAA 5.4 we can precisely and objectively check bandwidth, the noise signal report and distortion. The bandwidth graph indicates if the sound card reproduces the entire spectrum of sound frequency. A second test evaluates the card’s background noise and indicates the card’s sensitivities to certain IT interferences such as power supplies, monitors, etc. The signal/sound relationship is a very good evaluation of components, but it’s important to remember that this test is made without the signal.

In complement, we also use what we call the dynamic range. This value corresponds to a similar test but this time the signal is included. It gives a more precise idea of a system’s capabilities.

Another test concerns the total harmonic distortion or THD. With this accurate study we can see the harmonics when a simple but high level sinus wave (-3dB) is sent to the card. The IMD or Intermodulation Dirstorsion measures distortion and interference due to a combination of frequency and harmonics at the sound card output. Finally, Stereo Crosstalk measures possible interference between stereo channels.

We obtained the following figures with a MSI 925X Neo Platinum for the CMI-9880, and an ABIT AA8 Duramax for the ALC880. We also measured the analogical inputs/output of another CMI-9880 motherboard, the MSI 915P Neo2, and a motherboard with the ALC880, the Gigabyte GA-8ANXPD. We didn’t find any noticeable differences. You may have noticed that in addition to its capabilities, the Dolby Digital terminology is also supposed to certify a certain level of quality. In practice, the integration of HD Audio by manufacturers is able to reach the « Master Studio » level in terms of functions, but not for quality. So unfortunately, this terminology isn’t really used in practice.

Results in 16 bits 44.1 KHz


The bandwidth graph from the C-Media is slightly more stable than the Realtek chip. This chip, however, has less irregularity and is less restricted in high frequencies. The cut at around 30 Hz for the Realtek is completely normal.


In 16 bits / 44.1 KHz the Realtek ALC880 surpasses the C-Media CMI-9880 in all areas. For the latter, the bad results in total harmonic distortion and intermodulation are certainly due to the lack of pass-through for 44.1 KHz files.


We compared the two HDA chips to a recent AC97 chipset, the Realtek ALC850 found on many motherboards, and to a good quality middle range sound card, the Creative Labs Audigy LS. First off, we noticed the good results of both HDA chipsets, which show improvements compared to the AC97 (especially with the ALC880, which is nevertheless built by the same manufacturer as the 850 version). The comparison with the Audigy LS is, of course, less flattering. This middle range sound card with its good price / quality ratio luckily still has several advantages.

It’s important to note that the Realtek ALC850 was chosen because it has HDA options like Jack Sensing and Retasking and the Universal Audio Jack for front input. It corresponds to the last and ultimate version of the AC97, the 2.3.


Results in 16 bits / 48 KHz


In 16 bits / 48 KHz the C-Media has better results, which is a good start. The advantage, however, still goes to the Realtek ALC880 chip.

Results in 24 bits / 96 KHz


In 24 bits / 96 KHz both chipsets are relatively close. The computer, however, doesn’t often function in this mode.

Line input 16 bits / 44.1 KHz


Overall both chips’ line inputs are good with just a slight blowing sound. The ALC880 has better results than the CMI-9880 in reading and recording. We noticed that the C-Media’s recording level is very low even if we raise the level at both the source and input line. This can be remedied with a basic audio adjustment but this is somewhat bothersome. Micro inputs have approximately the same results and are of acceptable quality.

Output voltage


We measured the maximum output voltage for the main mini-jack port. This is important, because it’s synonymous with quality when it’s close to 2V. An especially low voltage was measured with the CMI-9880 compared to the ALC880. This is in correlation with the above results and observations.

Listening results

Overall subjective results concur with the objective tests. The ALC880 has a more pleasant sound to the ear and is more dynamic than the CMI-9880, which has just enough voltage to simultaneously provide a satisfactory dynamic and adequate power to be used with any type of speaker. Except for this weakness in the CMI-9880, we really didn’t notice any specific sound quality differences between the two chipsets. Sound is quite neutral and doesn’t benefit from any specific treatment or coloration.

3D sound and benchmarks

3D sound and benchmarks

We ran these benchmarks with the latest drivers available. All platform bios (all based on a Pentium 4 2.8 GHz Socket 775 or 478) were configured to obtain the same results without sound to have comparable results between all platforms once sound was activated.


The result with RightMark 3DSound speaks for itself. They show a good optimisation of C-Media driver compared to the Realtek’s, which requires more resources in DS3D and in DS3D+EAX as long as we remain below 16 simultaneous voices. After this, it’s the opposite, because the CMI9880 chip is restricted to 16 voices DS3D supported in hardware (that means that specific functions to the DS3D are processed by the chip, but that effects are at any rate processed by software). With the 32 voice test, the Realtek takes a slight lead. The improvements compared to the AC97 2.3 ALC850 chip, however, aren’t that great. HDA doesn’t bring anything revolutionary. The Audigy LS’ results, included here for comparison, show that a good sound card always keeps its advantage in this area.


Results in practice echo theoretical results with a slight advance for the ALC880. The Audigy LS shows that dedicated sound cards and especially the Creative Labs are always one length ahead in terms of CPU use for 3D sound. We obtained 3 to 9% higher performances with this sound card than a HAD codec. For comparison, changing from a Pentium 4 3 GHz to a Pentium 4 3.2 GHz with Far Cry improved performances by 6%.

This is due to drivers with better finishing touches and also to the presence of a DSP, which takes charge of DS3D management and also most environment effects. The HDA or AC97 chips however use the Sensaura engine in software.

In terms of positioning, the Sensaura engine provides excellent results and a quality similar to the Audigy LS. We noticed, however, a slight weakness in height effects (vertical source position in relation to the listener) that the Audigy doesn’t have. For effects, the advantage clearly goes to the sound card, which, thanks to DSP, are much more nuanced and features new EAX 3.0 technology.

Conclusion

Conclusion

High definition is currently only used with motherboards based on Intel’s latest chipsets. The Realtek, Media or Analog Devices chips will be adapted to other motherboards equipped with other chipsets for Intel and AMD processors. Like the AC97, HDA should become the new standard for integrated audio.

So what is the interest compared to the ageing AC97? Compatibility with 24 bit / 192 KHz sound in reading is indeed an improvement. But, let’s face it, it’s more in terms of marketing than real use. In fact, high definition audio content is in the minority. Most computer audio use is CD music (16 bits/ 44.1 KHz), encoded in MP3 or other formats (16 bits / 44.1 or 48 KHz) or Video DVD (16 bits/ 48 KHz). Only the future will tell us how this situation evolves. Chipset quality has been improved, but most users won’t feel sound quality improvement in going from a 16 bit/ 44.1 KHz to 24 bit/ 192 KHz source.

Above all this is a marketing strategy in that numbers have been increased. It’s true that in the world of computer sound it’s much simpler and less expensive to integrate DAC 24 bits / 192 KHz on motherboards rather than develop a real standard, which increases the sound quality for games and music.

The possibility to independently manage several audio flows and with access to dependant memory is undeniable progress. It permits simultaneous use without any media input or output problems without saturating the system and loss of quality. The possibility to manage flows with different sampling rates and resolutions is also an asset compared to the AC97.

Jack Sensing and Jack Restasking aren’t completely new as we saw in the ALC850 with the AC97 2.3. It simplifies connections and aids non-expert users. It is, however, just a detail and changing a subwoofer input with integrated chipsets isn’t for us the most transcendent and essential thing. This innovation is first and foremost interesting for motherboard manufacturers trying to reduce costs due to beginner users making the wrong connections. We can read in this document by Sigmatel that the main interest of the Jack Sensing is diminish customers support in resolving branching issues.

It’s important to note that sound quality has improved. This is a good point compared to the AC97, which often lacked in this area and is the main interest of HDA at this point. This of course rests on the condition that results with these chipsets are confirmed with all chipsets and motherboards. We haven’t noticed any important differences with the boards tested and this is good.

Possibilities for 3D sound remain quite basic with the Sensaura 3DPA engine. It is good enough for users who don’t play games often or know the quality of a good sound card. Like 24 bits / 192 KHz compatibility, compatibility with 8 channel sound is for us more symbolic. Take a look at the number of 7.1 speaker sets available on the market to realize its importance.

In the end, the HDA standard combines sensational marketing with welcome improvements such as better sound restitution and architecture similar to that of sound cards. The better sound quality is probably the most beneficial if users have adequate speakers or headphones. We can then only welcome the arrival of this standard replacing the AC97, and whose additional cost on the motherboard doesn’t seem to be exorbitant. Those capable of spending a little more ($50 to $100) can also benefit from the quality of a middle or high end range sound cards, which have functions and quality a step above.