Enter the New Digital Domain
ACCULINEAR 18-BIT TECHNOLOGY FROM ONKYO.
Everything has limitations. Including conventional CD players.
That's because 16-bit digital processing simply isn't accurate enough to retrieve all the data that's on a disc. So some of the music is lost.
Onkyo's linear 18-bit technology, on the other hand, assures you that all the musical information gets processed. So you don't lose anything. Even the subtle clues that tell you about the space the music was recorded in. And how well the engineer chose the microphones.
---Dual Acculinear D/A Converters with calibrated accuracy to the 4th Significant
Bit.
But getting all the data off the disc is only the first step. Getting it to your ears is at least as important.
That's why Onkyo developed the Acculinear D/A Converter. And individually calibrates each one to minimize crossover distortion. This unprecedented accuracy means you'll be able to listen to music, even at low levels, and still hear the delicate harmonic structures that distinguish a Gibson guitar from a Martin.
---Exclusive Opto-Coupling Modules use light to transmit digital data and
reduce harshness.
Onkyo's extensive use of optical transmission techniques instead of conventional wiring further increases musical enjoyment. Proprietary Opto-Coupling Modules at critical circuit junctions eliminate Digital Signal Interference (DSI) and its consequent metallic harshness. So you can enjoy the sound of the Philadelphia Symphony without wondering if the entire string section was playing aluminum violins.
The power supply combines low impedance/low loss transformers, regulators, and capacitors for high stability and isolation.
---- Dual transformers in the DX-7500 thoroughly isolate digital
and analog stages.
In addition, the critical D/A converters benefit from Opto-Drive, a new Onkyo technology that uses LED/phototransistor arrays for the ultimate in current stability and operating accuracy. Which means that any sonic variations you hear will be in the music, not in the disc player.
And the best part? We didn't reserve these technical innovations for one outrageously expensive flagship model. All the musical benefits are affordable.
Yes, this is the New Digital Domain.
Audibly significant technology.
Enhanced musical enjoyment.
The New Digital Domain.
Starting at less than $600. Enter it today at your Onkyo dealer.
ONKYO. 200 Williams Drive, Ramsey, NJ. 07446 201-825-7950
If you would like further information, write for the White Paper on digital technology.
Technical White Paper
Theory and Application of Linear 18-bit D/A Converters in Digital Audio Products
Onkyo was the first manufacturer to show a Compact Disc player employing linear 18-bit digital-to analog converters. This is indicative of Onkyo's contention that the use of such converters significantly improves both the measured and sonic quality of digital audio products. The use of 18-bit converters clearly makes obsolete earlier 14- and 16-bit architectures. Moreover, unlike designs in which 16-bit converters are switched to yield pseudo 18-bit resolution, linear 18-bit conversion does not introduce subtle conversion artifacts. Onkyo believes that linear 18-bit conversion will become the industry norm in high quality digital audio products, and we take pleasure in presenting this White Paper.
The Compact Disc Format--Some General Notes
The Compact Disc (CD) standard specifies a 16-bit word length and 44.1 kHz sampling rate. Together they yield audio fidelity with a flat frequency response exceeding 20 kHz and a Signal-to-Noise (S/N) ratio greater than 96 dB. Recently, manufacturers have introduced 18-bit CD players of various designs.
Although discs themselves will maintain a 16-bit format, 18-bit players will provide greater playback fidelity for the 16-bit recordings. The reason for this lies in flaws inherent in digital-to-analog (D/A) converters. Except in theory, 16-bit converters cannot fully decode a 16-bit signal without a degree of error.
When eighteen bits are derived from the disc data and converted through 18-bit conversion, errors can be reduced and reproduction specifications improved. In order to realize the full potential of audio fidelity for the end user, the signal digitization and processing steps must have a greater dynamic range than the final recording.' However, there are several ways of performing 18-bit conversion, some better (and more expensive) than others. Onkyo firmly supports true, linear 18-bit conversion in which 18-bit converter chips are employed.
The Compact Disc Sampling Rate
The Compact Disc specifies a 44.1 kHz sampling rate. Some critics have suggested that the CD's sampling rate be increased from 44.1 kHz to 100 kHz or so. This higher sampling rate would provide a higher frequency response. For example, the CD's flat frequency response would be extended from 20 kHz to 50 kHz. Unfortunately, the extended frequency response is beyond the range of human hearing.
The only advantage to a higher sampling rate is the decrease in demands on the anti-aliasing filter preceding a digital audio system, and the anti-imaging filter following it. The need to sharply limit audio energy at frequencies higher than half the sampling frequency dictates the use of brickwall filters to preserve a 20 kHz audio bandwidth with a 44.1 sampling rate. And analog brickwall filters introduce phase nonlinearities. However CD players can easily avoid the problem by using digital oversampling anti-imaging filters;
their phase nonlinearities are negligible. And even the problem of brickwall filters on the input of professional recorders are eased with the development of oversampling anti-aliasing filters. An increased sampling rate per se does not offer an improvement in CD echnology.
The Compact Disc Quantization Word Length
Quantization assigns a numerical value to an audio sample; it is that value which is stored on disc. The conversion of the waveform to a binary value, and back again, is the function of the analog-to-digital (A/D) and digital-to-analog (D/A) converters respectively. The number of bits in the binary word determines the accuracy of the assigned value. The number of bits thus ultimately determines the S/N ratio of the system.
The choice of 16-bit words for the CD standard was determined primarily on the availability of 16-bit digital-to-analog converters and the fact that longer word lengths would diminish the playing time of discs. However affordable 18-bit converters have been developed and they offer a chance to improve the playback of discs. While no technology can yield more than 16 bits of music from a 16-bit CD, the use of 18-bit conversion provides better utilization of the 16 bits from the disc. Specifically, 18-bit conversion technology overcomes problems in 16-bit converters that limit their decoding of the information coming from a disc. When correctly done, 18-bit conversion thus improves the amplitude resolution of the player by ensuring a fully linear conversion of the disc's 16-bit signal, yielding faithful reproduction of the signal encoded on the disc. If 16-bit conversion is a dirty window, then 18-bit conversion is a clean window.
Why 16-Bit Conversion Isn't Enough
As we have seen, a Compact Disc stores 16-bit audio data words. Logically, most CD players employ one or more 16-bit digital-to-analog converters. However, D/As present a weak link in the signal chain;
they are not always good at generating the analog voltages which represent the digital words read from the disc. In particular, low amplitude distortion is sometimes introduced because, simply, 16-bit converters are not ideal for reliably converting 16-bit data.
The difficulty is understandable. A 16-bit D/A must determine which of the 65,536 output analog voltages corresponds to the input digital word within the 20-microsecond sampling period. Several problems can interfere with the decision. For example the distance between steps may not be exactly equidistant; that would lead to nonlinearity in the conversion. One solution is a D/A converter with more bits' worth of conversion. An 18-bit D/A, for example, has 262,144 increments, exactly four times as many output levels as a 16-bit converter. Nonlinearities are correspondingly smaller, and the sound from the CD player theoretically better.
The intent of 18-bit D/A conversion technology is not to somehow improve the data from the disc, but rather to make better use of that data. In other words, 18-bit technology attempts to overcome problems in 16-bit converters which may limit their decoding of the information coming from the disc. An analogy may be made to oversampling: Increasing the sampling rate does not create new information, it simply allows better use of what existed already.
In other words, an 18-bit converter gives better 16-bit conversion. In fact, the two extra bits of a linear 18-bit converter would not even have to be connected to yield improved 16-bit performance. Justification of Increase in Word Length As noted, a 16-bit word can represent 65,535 amplitude increments; an 18-bit word represents 262,143 increments. Increasing the quantization word length by two bits at the conversion stage results, primarily, in an increase in S/N ratio. Simultaneously, any quantization artifacts are diminished. A properly dithered 16-bit system has an excellent S/N ratio, and quantization artifacts are handled such that resolution may be obtained at levels even less than that of the least significant bit. When the CD was standardized, 18-bit A/D and D/A converters were very costly; so, there seemed little sense in increasing the word length. Thus the Compact Disc standard was set at 16 bits.
Oversampling makes 18-bit conversion possible. It solves the obvious dilemma of coming up with 18 bits, when the output from the disc is only 16. When the 44.1 kHz, 16-bit signal from the CD is oversampled, both the sampling frequency and number of bits are increased-the former because of oversampling, and the latter because of the multiplication which must take place. For example, the output of an oversampling filter may be 176.4 kHz and 28 bits. Normally, only the 16 most significant bits are used, for conversion through a 16-bit D/A converter, and the rest are discarded (in some designs they are used for noise shaping).
A linear 18-bit system uses 18 of the bits from the output of an oversampling circuit, instead of just 16.
When proper oversampling techniques are used, those two extra bits do indeed convey useful amplitude information, albeit at levels below the first 16 bits.
A Pseudo 18-Bit Conversion Method
A linear 18-bit D/A converter chip is an expensive part. Some manufacturers are unwilling to incorporate that cost into their designs. An alternative is so-called pseudo 18-bit methods, in which an 18-bit D/A converter is not used at all. Rather, a 16-bit converter is "floated" to operate on 18 bits.
The 18-bits from the oversampling filter's output are wired through switches to the inputs of a 16-bit D/A converter. When all 16 bits are being used to convey a signal (as in the case of a high amplitude), the upper 16 bits are applied to the 16 bit converter, as usual. However, when the two upper bits from the oversampling filter are not being used to convey a signal (a lower amplitude signal), the 18 bits are shifted downward, so that the unused bits are ignored, and the 16 lower bits are utilized instead, as shown in Figure 1. Through bit shifting, a 16-bit converter may thus handle an 18-bit input.
Figure 1. Floating 18-bit conversion using 16-bit D/As, bit switching and
attenuation.
However, bit switching multiplication upsets the weighted values of the binary word. Consequently overall system gain must be constantly shifted as well. Specifically, whenever the two lower bits are shifted in, the word's value has been effectively multiplied by four, thus the gain of the signal must be reduced by one-fourth to compensate. An analog gain block downstream of the D/A performs this function. A one-fourth reduction is required as follows: In the decimal system, when a digit is shifted to the left one place, the result is ten times larger. In the binary system, a shift results in a doubling of value. A shift of two places (two bits) quadruples the value. Thus when the lower bits are shifted up, the amplitude is four times too big; therefore the output must be attenuated by 1/4. Similarly, a one-bit shift would require a gain reduction of 1/2.
This adaptive 18-bit conversion may be considered as a dynamic noise reduction scheme in that the signal is being expanded at the D/A converter. The benefits result from the fact that the residual noise of the converter, as well as its conversion nonlinearities, will be proportionally reduced. Looked at another way, a four-times higher analog output has been achieved without increasing the D/A's residual noise and conversion error. When the gain is reduced by 1/4, the noise and conversion errors are reduced by 1/4.
There is an increase in S/N of 12 dB and distortion is reduced by 1/4.5 However, this method often produces unwanted conversion artifacts which are discussed on page 6.
Another Pseudo 18-Bit Conversion Method
Some CD player conversion systems employ 4 D/A converters. This, too, is intended to provide a cleaner conversion of a CD's 16-bit output than any single 16-bit converter alone could provide.
An oversampling digital filter and 16-bit D/As are employed, as well as bit shifting. However in this design, two D/As are used per channel-one for the positive half and one for the negative half of the output bipolar analog waveform. Again, the reason is to improve low amplitude resolution. In this case, the design specifically addresses a problem known as crossover distortion.
Crossover distortion occurs at the zero cross point between the positive and negative voltage swing of the waveform. The converter must switch all of its digits (from 01111111111111111 to 1000000000000000) as polarity changes. Crossover distortion can be alleviated by a providing a D/A for each waveform polarity.
In that way, the total switching of digits never occurs. Of course, the digital signal must be split between the two D/As, but an upstream processor chip handles this switching digitally. Theoretically, this can be glitch-free, if properly implemented.
The circuit's signal flow is shown in the following block diagram.
Figure 2. Floating 18-bit conversion using bi-polar 16-bit D/As, bit switching
and attenuation.
An oversampling filter outputs 18-bit words to the demodulator. It directs the appropriate portion of the waveform to the correct D/A. The two waveform halves are joined at the output by a differential amplifier.
An attenuator compensates for the gain change caused by shifting.
When the signal is of high amplitude, greater low amplitude resolution is not required so there is no bit shifting, and straight 16-bit conversion takes place. Both polarity D/A converters reproduce both halves of the waveform, and they are combined in a differential push-pull mode.
However when the signal drops below -12 dB, the circuit reads the most significant bit (MSB) and observes that when the bit is high, the word has positive polarity, and thus must be directed to the positive polarity D/A converter. Likewise, when the MSB is low, the negative half D/A gets the word. Also, when the signal is below -12 dB, the second most significant bit ceases to change, and in fact it is always low when the MSB is high, and it is high when the MSB is low. The word sent to the 16-bit D/As may be shifted down two bits. The D/As thus receive bits 3 through 18. Since the gain of the signal increases when bits are shifted, the attenuators must be switched in, to proportionally reduce gain.
The 17th and 18th bits from the oversampling filter have been utilized in reproducing the waveform, instead of being truncated and discarded. Moreover, because dual D/As are used to convert the bipolar waveform, crossover distortion has been avoided. This can benefit reproduction of low level audio signals. However, again, artifacts may intrude.
Disadvantages of Pseudo 18-Bit Conversion Although pseudo 18-bit conversion methods such as those described above offer advantages, they may also promote problems such as switching noise and subtle gain matching errors, an inherent flaw in any arrangement when bits are switched. In fact, in analyzing the performance of one pseudo 18-bit CD player, two researchers concluded, "...16-bit precision gain matching is surely out of the question and so (possibly inaudible) conversion errors will occur whenever bit shifting is invoked."6 Thus "floating" conversion systems can introduce errors in the output signal because one must compensate for the resulting shifts in amplitude caused by shifting bits.
It isn't possible to get all the benefits of an 18-bit converter with a 16-bit converter, no matter how a 16-bit converter is manipulated. This is the problem: When the bits are shifted, it is difficult to immediately and simultaneously shift the gain of the analog static offset will become apparent when the switching takes place.
Until the first 14 bits are occupied, the output is four-times its nominal amplitude, so the 1/4 attenuator is used to compensate. When all 14 bits are occupied, the output voltage is maximum. When the 15th bit turns on, the bits are shifted, and the attenuator is removed. The output increases at normal gain until full 16-bit voltage is reached. The difficulty is that the attenuator could introduce a static error, owing to component tolerances. It would not be significant over the first 14 bits, affecting only attenuation ratio. But when switched off (or later, on), the difference between the attenuator's error and the absolute value of the output would create a glitch in the waveform where the attenuated and non-attenuated signals are joined.
The glitch would always be present at -12 dB regardless of signal frequency. The question of course, is whether the increase in S/N ratio outweighs the possibility of switching distortion. At any rate, pseudo 18-bit conversion will always run the risk of introducing error into the analog output signal.
Linear 18-Bit Conversion And Its Advantages
Until recently, affordable 18-bit converters were not available. However recent advances have resulted in a practical 18-bit D/A converter. This integrated circuit is the only true 18-bit D/A currently in mass production and is employed in several of the newest Onkyo designs. The chip provides a harmonic distortion specification of 0.0008% for a 1 kHz sinewave: This is an unprecedented accomplishment.
Some specifications of the new 18-bit D/A chips exceed the ability of test equipment to measure them. For example, the D/A contains a test circuit to monitor settling time of the current source. This minimizes glitches to the extent that resultant distortion is below measurement levels and can be analyzed only through computer modeling.
Although a conventional binary-weighted architecture is employed, several special design considerations were instituted to attain the desired accuracy. The most significant three bits are made up of seven individual current sources; this reduces thermal errors. Bits four through sixteen are made up of unit-valued current sources which feed the R-2R ladder network of the digital filter, as shown in Figure 3.
Currents for bits seventeen and eighteen are derived from the unit-valued source. The relative gain of the three upper bits can be adjusted against the total weight of the fifteen lower R-2R bits by trimming the scale-down network.
Figure 3. Complete Hybrid 18-bit D/A
The chip itself is a forty-pin hybrid with a special divided layout to provide full control over laser trimming of the upper and lower bits. Supporting circuitry is located outside the chip to help maintain the chip's thermal balance, a critical concern in D/A design. Accuracy to the least significant bit requires precision equivalent to four parts per million.
Onkyo CD players with 18-bit conversion make use of these new D/A converters. The architecture of this design, as shown in Figure 4, is markedly simpler than pseudo 18-bit conversion schemes.
Intuitively, one may correctly surmise that the simpler signal path promotes better performance.
Simply put, the less manipulation of the analog signal, the better the fidelity.
Figure 4. True Linear 18-bit Conversion (ONKYO)
Other Onkyo Advances
In addition, Onkyo CD players utilizing 18-bit conversion employ dual converter chips, one for each audio channel. This ensures perfect phase linearity between channels. In some other players in which one D/A converter must service both channels, interchannel phase distortion is inevitable as the data from one channel must wait while the other channel's data is being converted. In addition, the practice of alternating channel data through one converter may introduce switching noise. By using dual 18-bit D/A converters, each converter is responsible only for its channel's data, and interchannel phase distortion and noise are eliminated.
The use of 18-bit D/A chips results in player specifications clearly superior to 16-bit designs. As described above, an 18-bit converter has four times greater resolution than a 16-bit converter. In other words, 18-bit conversion of a 16-bit signal provides for a 12 dB increase in S/N ratio while processing the data. This improvement is clearly measurable in a CD player with 18-bit conversion. Of course, there is no risk of gain matching errors as with pseudo conversion methods.
In summary, these chips set a new standard for digital signal conversion. In fact, the primary sonic limitation in a CD player using linear 18-bit converters is the quality of the equipment used to make the master recording itself.
Oversampling Output Filtering
To maximize the performance potential of 18-bit converters, Onkyo employs oversampling output filtering. In an oversampling filter, audio samples from the disc are subjected to computation which implements digital filtering of the audio signal. Additional audio samples are generated between the original samples; hence, the output sampling rate is increased. This additional data creates a more linear waveform and shifts unwanted modulation noise to an extreme supersonic frequency, where it can be removed without audible effect.
Eighteen-bit Onkyo CD players employ either four-times or eight-times oversampling circuits. The latter circuit is particularly significant and generally found only in the most costly players. In an eight-times oversampling circuit, seven new audio samples are computed for each input sample. In other words, the output sampling rate is raised to 352.8 kHz. The accuracy of this filter is precise, yielding pass band ripple on the order of 0.00001 dB. In addition, the stop band suppression is greater than 120 dB. Few filters have ever achieved these results. In addition, modulation artifacts are shifted to a band centered at 352.8 kHz where they are easily removed with an analog low-pass filter. Finally, the multiplication algorithms in the oversampling filter provide amplitude resolution in excess of 16 bits, thus providing 18-bit words with fully meaningful data for the output 18-bit D/A converters.
Conclusion
Onkyo is firmly committed to linear 18-bit D/A conversion in its digital audio products wherever costs permit the inclusion of these devices. This design approach offers clear advantages to both existing 16-bit and pseudo 18-bit conversion methods. The introduction of 18-bt conversion, coupled with carefully implemented oversampling filtering, perhaps represents the single greatest advance in digital audio technology since the introduction of the Compact Disc in 1982. Onkyo is proud to have been the first to introduce this technology to the U.S. marketplace.
References
1. Joel H. Halbert, Mark A. Shill, "An 18-Bit D/A Converter for High Performance Digital Audio Applications," AES Preprint, October, New York, 1987.
2. Ken C. Pohlmann, Principles of Digital Audio, Howard W. Sams, 1985.
3. David Ranada, "The Two-Bit Difference," High Fidelity, January, 1988.
4. John Vanderkooy, Stanley P. Lipshitz, "Resolution Below the Least Significant Bit in Digital Audio Systems With Dither," JAES, March, 1984.
5. Ken C. Pohlmann, "Eighteen Bit Conversion," Audio, October, 1987.
6. Stanley P. Lipshitz, John Vanderkooy, "Are D/A Converters Getting Worse?," AES Preprint, Paris, March, 1988.
For a free reprint of this paper, write to: 18 Bit Technology c/o Onkyo USA Corporation.
ONKYO.
200 Williams Drive, Ramsey, NJ 07446 201-825-7950
(Audio magazine, Oct. 1988)
Also see:
Onkyo Acculinear 18-bit CD player (Nov. 1988)
Onkyo AccuPulse CD player (Jan. 1992)
NAD Monitor Series CD player (Feb. 1988)
Onkyo DX-300 Compact Disc Player (Sept. 1984)
Revox B226 CD player (Apr. 1987)
Onkyo--Artistry Sound (Apr. 1991)
= = = =