High resolution sound : why and how?

High resolution sound : why and how?

Since the release of the first Audigy and its supposedly 24 bits, the general public audio field for computers has constantly increased its level of sound quality. For processors, the most important marketing point was the GHz, the bandwidth in GB/s and a number of pixel pipelines for 3D graphic cards, so why not also offer us several bits and a couple of dozen additional KHz for new audio solutions? All users know that a 34 bit wide bus memory is better than a 32 bits wide bus and 32 bits graphics are better than 16 bits, etc. Therefore the audio field manufacturers adopted the same principle. But does this increase in bits and KHz hold any interest whatsoever for the average computer user who is going to listen to some music, play, watch a video DVD or even, in the worst scenario case, attempt a little bit of audio creation.

In order to give an answer to this question, it is imperative to go back to the origins and explain what these numbers mean.

To begin with, when the signal is numerized it appears shaped like a wave, like this to the converter:


To be numerized, this wave will be split in several periods: it is going to be sampled. This sampling is more or less intense according to the chosen rate: 22 000 sampling per second at 22 KHz or 96 000 at 96 KHz for example. The sound wave looks like this graphic that has been deliberately roughly decomposed:


The numerization will have to split the graph in several precise points, which are going to determine the rest of operations :


Once the signal per signal split is made, one value is attributed to each signal. This is where the resolution intervenes: in 16 bits this value varies between 0 to 65535 (in reality in audio between -32768 and +32767) and in 24 bits between 0 and 16 777 215 (in reality between – 8 388 608 and + 8 388 607). When a signal doesn’t correspond to a whole number, the converter chooses the closest one. Each accommodation therefore induces an error or a distortion compared to the original signal. These errors are called quantization.


Once the values are attributed, the graph is rebuilt via a reconstruction filter. During this reconstruction, the converter applies a low pass filter that eliminates the highest frequencies (these are the ones generating the most mistakes). The higher the number of values previously rounded up is, the greater the number of high frequencies that will have to be removed.


The number of rounded values, and so of " errors " is directly connected to the chosen sampling rate. It is therefore clear that theoretically, the higher numerization intervals we have, the more accurate the reconstituted graph will be, compared to the original. In the same way, the broad panel provided by the 24 bits to assign this value increases the chances of accuracy. Now the point is that to benefit from this profusion of nuances you need to have not only an excellent ear, as well as a performing and silent restitution configuration, this is rarely the case with computers. Furthermore, even if these sound cards contain such resolution and sampling rates, the quality of the content doesn’t often provide us with the possibility of using it. This means, in effect, that some of the audio DVD have DTS or MLP in 24 / 96 KHz tracks or even Stereo HD in 24 / 192 KHz tracks but this type of use is still more than confidential for computers.

To counteract this argument, some manufacturers emphasize the fact that the 24/192 support also improves the 16/44.1 sound quality restitution. Indeed in its theorem, Shannon determined that the highest frequency is equal to half the sampling rate. For example with the 96 KHz sampling rate, the highest frequency reproducible is 48 KHz. This is what we call the Nyquist frequency. It is important to note that the difference between the sampling rate (96 KHz) and the maximum " numerisable " sound frequency (48 KHz) are both expressed with the same unit but don’t represent the same thing.
If we follow the same logic, without questioning the Shannon/Nyquist duet of course, with a maximum sampling rate of 48 KHz as used so far, we were restricted to a 24 KHz frequency. Today, thanks to the generalized 96 KHz including integrated chipsets, it is possible to reach high frequencies of 48 KHz. Knowing that the normal human adult ear uses a bandwidth of 20 Hz to approximately 17 KHz (if you are curious you can listen to this file including a 15 KHz wave) and that our speakers have a bandwidth of 40 Hz to 20 KHz for the best of them, our question is: what is the interest of such a frequency at this stage of reproduction?

We can, nonetheless, appreciate the theoretical dynamic range increase provided by high definition sound. The 16 bits/ 44.1 KHz is indeed restricted to 96 dB whereas the 24 bits is able to reach a dynamic range of up to 144 dB. Representing the difference between the highest and the lowest possible level, this value has a direct influence on sound quality.

Finally, this does not mean denying all interest for the 24-bits/ 192 KHz sound and for all the possible uses. In fact, where the resolution and sampling rate increases are interesting, is during sound recording and processing, and not in its reproduction. Indeed, resampling operations are delicate when one is dealing with sinusoids as close as possible to the maximum frequency (22.05 KHz in 44.1 KHz for example). The point of working in high resolution and in 96 or 192 KHz is then obvious because it enables the avoidance of the distortion of high frequencies and leads to a more subtle application of effects. But nothing stops you from resampling the final result in 44.1 KHz once the numerization and process are over. A few people are, in reality, capable of telling the difference between two sounds which have a common source in 24 bits / 192 KHz and one played with this resolution and the other one re-sampled in 16 bits /44.1 KHz.

So we feel that for most computer users, the 24 bits/ 192 KHz sound is currently a marketing device. At the present time, this possibility may interest those who want to start audio creation but these users would probably rather choose a product made by EMU, Cream or Lynx, for instance.

Audigy 4 Pro: drivers

Audigy 4 Pro: drivers

Nothing new as regards drivers. Are included the "good ole" system of separated modules that can be launched from the systray. Apparently it just is not possible for Creative Labs to provide us with a unified interface like the one provided by NVIDIA, Hercules or Terratec for example. It is true that when developing a graphic equalizer using no less than 10 MB of RAM memory, it is clear that they might understand the consequences of a unified solution. So if we simultaneously launch the EAX control panel, graphic equalizer, speaker configuration panel and the surround mixer the grand total of RAM memory used is 60 MB. To compare, the DMX 6Fire 24/96 unified interface uses a little less than 5 MB of memory and the Realtek ALC880 16 MB.


The new Creative Entertainment Center quality and ergonomics are very good. One can even use a remote control to play music or video files or even to watch photos.

Inputs and outputs quality

Inputs and outputs quality

To test the Audigy 4 Pro we used the RMAA 5.4 software for objective tests combined with a Terratec MIC8 sound card as well as several sound sources for subjective tests. With RMAA 5.4 we can precisely and objectively check bandwidth, the noise signal report and distortion.

The bandwidth graph indicates whether the sound card reproduces the entire spectrum of sound frequency. The second test evaluates the card’s background noise and indicates the card’s sensitivities to certain IT interference such as power supplies, monitors, etc. The signal/sound relationship is a very good indicator of components but it’s important to remember that this test is made without the signal.

We also used what we call the dynamic range. This value corresponds to a similar test but with the signal. It gives a more precise idea of a system’s capabilities. The following 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 with the sound card output. Finally, Stereo Crosstalk measures possible interference between stereo channels.

We used the following test configuration:

- Abit AA8 Duramax Intel i925 motherboard
- Intel P4 3.6GHz@2.8GHz processor
- 2 GB DDR2
- HDD Maxtor Diamond Max Plus 9 120 GB
- NVIDIA GeForce 6800 GT graphic card

Results in 16 bits / 44.1 KHz


The bandwidth graph shows slight distortions between 5 and 10 KHz, but nothing really important.



Results in 16 bits / 44.1 KHz are excellent. The noise signal level isn’t extraordinary but this sampling rate has never been the Audigy strong point.



The Audigy 4 Pro provides better performances than cheaper or previous generation sound cards and therefore it meets our expectations. We have noticed however that the progression compared to the Audigy 2 ZS isn’t extraordinary.

Results in 16 bits / 48 KHz


Results are sensibly better in 16 bits / 48 KHz because unlike with the 44.1 KHz the Audigy doesn’t process any resampling operation for this sampling rate.

Results in 24 bits / 96 KHz


It is in 24 bits / 96 KHz that the Audigy 4 Pro really shows what it is capable of. It is the first general public sound card to reach over -100 dB for the noise signal report in our test bench. The performance increases with the already excellent Audigy 2 ZS isn’t extraordinary.

Output quality: subjective tests

Output quality: subjective tests

For games, since the Audigy 4 Pro is almost identical to the Audigy 2 our expectations were quite high. And it came up to our expectations, because the excellent restitution quality is combined with efficient and subtle positioning and effects management. With a hardware reverberation engine and an effortless multiple effect application the Audigy 4 Pro is like the previous Audigy a perfect sound card for games seeking a good sound quality. This is partly due to the DSP’s intrinsic qualities but also to the lack of competition in that area. Most of the competitors’ sound cards feature the Sensaura or the QSound (recently bought by Creative Labs) 3D and effect engine. It is technically reliable but it alters the positioned audio flows too much.

For Video DVD, the main criteria are the decoding quality, the type of decoding provided, the sound quality and tautness and the volume attainable without losing in quality. From that point of view, the Audigy 4 Pro has the same base as the Audigy 2 ZS. Dolby Digital EX decoding is still very efficient and perfectly uses the power of the sound card sound. DTS ES tracks are also included and this is where the sound card restitution quality comes fully into its own. This format’s quality is sensibly better than the Dolby Digital. The Audigy 4 Pro decoding parameters also provide the opportunity to determine the sound card’s attitude whether a DTS source or a Dolby Digital 5.1 source is read with a 6.1 or 7.1 speaker set: it is also possible to point surround channel sound only to the surround speakers, only to the rear speakers or even to the four speakers simultaneously.

For music sound quality, the Audigy 4 Pro provides an excellent restitution of basses and mediums and ostensibly softens trebles (it is worth noting the presence of an AES-17 filter). It allows the sound to remain within the lines defined by the THX certification, but feedback (in French Larsen) and high strident treble enthusiasts will be dissatisfied. Sound is generally quite taut and is clearly enriched to satisfy above all the DTS and Dolby Digital decoding.

Line input 16 bits / 44.1 KHz


The Audigy 4 Pro shows excellent qualities for analogical and numerical recording. Sound isn’t overly distorted and background noise is reduced. However, compared to the Audigy 2 ZS there is no significant increase.

3D sound and benchmarks

3D sound and benchmarks

The Audigy, LS or 2 ZS or 4 Pro sound card’s interest is that it doesn’t include a simple chip with a software engine, but a real DSP which process all the work in hardware. This considerably reduces the central processor work and also provides very pleasing effects. The economical aspect of the Audigy 2 and 4 Pro is verified in practice by this benchmark. It shows a clear advance in this field.



Results observed in the Benchmarks are verified in practice with an obvious increase in performance for the Audigy 4 Pro that doesn’t however stand out compared to the Audigy 2 ZS. The positioning quality is still irreproachable.

It is important to remind you that all Audigy sound cards support the EAX for all versions, 3.0 and 4.0 included. As a comparison, here are the improvement of the last EAX version included in Thief and Hitman 2:

- Multi Environment: With the EAX 1.0 and 2.0, the environment effect is restricted to a determined geometrical area. From now on, each sound possesses its own effect. For instance, you can be in a room and hear the sound of a gun firing as affected by the size of the room, the walls and floor materials, whereas the sound of the gun firing coming from the room next door will be affected by the characteristics of that room (a corridor for example). The Audigy is therefore able to manage and mix 4 different environments.
- Environment Reflection: the Environment Reflection brings the same improvements as the WaveTracing and A3D 2.0. It means that the sounds don’t only come from the source but are also redirected by the environment. It is the primary reflection and reverberation control.
- Environment Morphing: it is one of the most significant improvements brought by the EAX 3.0, previously named Dynamic Morphing. With the EAX 2.0, each environment was assigned to a specific place in a game and there was no real difference between an open-air place and a cave for example. Thanks to the Dynamic morphing, environments can be combined in certain areas and one takes over the other when the character is getting deeper inside the cave.
- Environment Panning: it completes the Environment Morphing because it allows the possibility of adding an "apprehension" environment of the place coming up, like the tunnel you are about to enter.
- Environment Filtering: This name, with its graphic connotations, corresponds in fact to a technique allowing for the improvement of parameters to control tonal nuances for each environment.

Conclusion

Conclusion

The Audigy 2 ZS Platinum Pro was already quite complete in terms of functionalities and was clearly above the other sound cards in terms of sound quality. The interest of this new version named "Audigy 4 Pro " seems to be quite limited.

It is nonetheless a very complete high end product. Its functionalities are superabundant and the sound card can be used equally for musical creations, games or home cinema. The number of inputs and outputs and their variety, the presence of ASIO 2.0 drivers, the 32 bits engine effect, the full support of 24 bits / 96 KHz sound, the advanced MIDI management thanks to SoundFont are all very good points for listening to or creating music. The environment and positioning quality effects added to the EAX 3.0 and 4.0 make of this sound card a perfect product for games. Finally is also possible to use this sound card for an HTPC (home theatre PC) thanks to the remote control, the DTS ES and DD EX decoding. Sound quality is also excellent in the objective and subjective tests.

Considering the lack of content and of sound process quality guarantee in our computers, the 24 bits / 192 KHz sound remains for the moment an extra marketing device. This situation may possibly change in the future but it certainly does not provide one reason to buy this sound card rather than another. Unless you have a weakness for Audio DVD.

The last point is the biggest downside of this sound card is its price: around US $250. The Audigy 4 Pro is a very high end product but this price doesn’t seem to be fully justified according to the possibilities and technical qualities observed. So the last question is: who really needs all of this?

From a more global point of view, we have noticed a stagnation of general public sound card functionalities whereas integrated solution are getting better without of course being able to provide the same level of quality (see our article on the HDA).