Manufacturer's Specifications
Analog Section:
Power Output: 100 watts minimum rms, both channels driven, from 20 Hz to 20 kHz, at no more than 0.02% THD into 8 ohms.
SMPTE IM Distortion: Less than 0.02% at rated power.
Frequency Response at 1 Watt: MM phono, 20 Hz to 20 kHz, ± 0.02 dB; CD, 0 Hz to 125 kHz, +0,-3 dB. Input Sensitivity: MM phono, 2.5 mV; MC phono, 160 uV; high level, 150 mV.
Input Impedance: MM phono, 47 kilohms; MC phono, 100 ohms; high level, 47 kilohms.
Line Output Level: 150 mV into 47 kilohms at 1 kHz.
S/N: MM phono, 86 dB; high level, 110 dB.
Digital Section:
Frequency Response: 4 Hz to 20 kHz, ±0.5 dB.
S/N: 100 dB. Dynamic Range: 98 dB.
Harmonic Distortion: 0.003% at 1 kHz.
Signal Level: 0.5 V peak to peak.
Input Impedance: 75 ohms.
Optical Wavelength: 660 nm.
General Specifications
Power Requirements: 120 V a.c., 60 Hz.
Power Consumption: 560 watts, 680 VA; 720 watts maximum.
Accessory A.C. Outlets: Two unswitched (250 watts total), one switched (100 watts).
Dimensions: 17 in. W x 6 3/8 in. H x 17 3/4 in. D (43 cm x 16.3 cm x 45 cm).
Weight: 40.1 lbs. (18.2 kg).
Price: $1,250, including remote.
Company Address: 1250 Valley Brook Ave., Lyndhurst, N.J. 07071.
As a member of Sansui's premium Vintage Line, the AU-X911 DG is a flagship model for the company. In basic description, it's similar to other companies' flagship integrated amps from earlier in the digital era: A preamp section with a moving-coil pre-preamp, a more than minimal power amp, a D/A converter with both coaxial and optical digital inputs, a switching scheme that allows reproduction of digital sources with minimal analog processing intervention, and every fondly crafted detail the company is proud of.
This amp is impressive, no doubt about it. I have to point out right up front, however, that though this Sansui may outpoint competing models in certain key areas, the whole genre should be considered as just one step toward the digital future. The format is logical for 1990, when analog signals are the rule and digital signals the exception.
Once digital equalizers and ambience processors become commonplace, the AU-X911 DG and its rivals will still be useful-the Sansui, perhaps, more so than most. But by the turn of the century, they'll be competing against products designed for an even more digitally oriented era.
The most important single aspect of this amplifier's design is digital, however: The Linear and Direct D/A Conversion System (LDCS), or so-called one-bit D/A converter.
Sansui appears to have been the first company to offer and promote one-bit technology in the U.S. By now, several months after this amplifier's introduction, the basics of this technology are shared (under many names, and with considerable variety of possible implementation) by other companies; it will probably become the standard in short order, because of its potential for both lower price and finer performance, compared with even the fanciest multi-bit conventional processors.
The difference lies in using precise digital arithmetic to replace the often sloppy analog modeling provided by resistors in conventional D/A converters. For perfect linearity, the resistors would have to achieve phenomenally precise values. The 16-bit data stream is passed through an oversampling filter and then converted into a pulse-width modulated (PWM) signal. The process is called one-bit because the converter has only a single output (which is either off or on), as opposed to the 16 simultaneous bits necessary to define one instant of amplitude in the original data stream.
This PWM signal is digital in the sense that it is either full off or full-on, but analog in that the width (duration) of each on or off half-cycle is an analog of the instantaneous signal as represented by the original data stream. In determining the width of each pulse, the converter counts cycles of its internal clock, which is routinely far more precise than any resistive ladder. The PWM signal is then smoothed by a lowpass filter, transmuting the pulse widths into the instantaneous amplitudes of a regular analog signal.
While much of this may be familiar to readers who have been following digital audio, Sansui claims proprietary advantages in some areas. While acknowledging the importance of the so-called Multi-Stage Noise Shaping (MASH) technology (developed elsewhere but licensed to Sansui), Sansui's multi-stage feedforward circuit is said to reduce noise in the audio band still further. And, to control jitter in the signals arriving from unknown digital sources, the clock circuit D/A converter uses a lithium-tantalate oscillator said to achieve both the frequency accuracy of a crystal and the controllability necessary to adjust frequency for the available sampling rates (32, 44.1, or 48 kHz).
There are (need I say it?) separate power-supply transformers to isolate the digital domain from the analog; also, the two channels of amplification have independent circuit boards toward opposite ends of the chassis, whose heavy heat-sinking isolates these boards from the central power supply. Surrounding the circuit boards is a ring of shielded subchassis for the controls at the front, the low-level circuitry (including the D/A converter) at the sides, and the input and output switching (driven electrically from the front) at the back. The digital connections are all on the left side of the chassis; those for analog signals are at the right, with two sets of heavy-duty speaker terminals between them.
Control Layout
The analog input layout comprises jack pairs for phono ("MM/MC" is selected via the front panel), tuner, CD, and line, and loop sets for three tape decks plus an outboard processor. Of the three coaxial digital inputs, only the third is marked for DAT; there is an optical input as well, plus a coaxial digital output. According to the manual, the latter will not function with signals containing a copy-inhibit code or with data streams using the 44.1-kHz sampling rate of CDs. A front-panel indicator shows the current "Sampling Frequency" when digital inputs are in use.
That the designations for the line inputs are to be accepted or ignored, as you will, is made plain, both by the owner's manual and by the presence of a video-in and a video-out jack among the analog connections. Since none of the audio jacks are specifically marked as being for use with video soundtracks-and since the video jacks don't seem to do anything but hand the signal off to some other component-you are encouraged to consider how the jacks work rather than how they're marked.
The basic selectors are arranged in two banks. The top bank chooses the four digital sources via the built-in D/A converter, and the bottom bank chooses the four analog sources. Signal routing is further refined by a three-position "Source Direct Operation" switch: "Off" (normal operation), "Source" (a simplified signal path), and "Digital Direct" (input from the D/A converter to the power amp). In the "Source" position, the balance control, infrasonic filters, and muting (20-dB nominal attenuation) are bypassed; the "Digital Direct" option bypasses all of these plus the tone and loudness controls and the processor loop.
The "Source Direct Operation" switch is, oddly, interposed on the front panel between the monitor buttons for the analog decks and the recording selector. The latter has positions for "Tuner" and "CD" analog inputs (whose signals can therefore be recorded even while you're listening to something else), "Source" (whichever of the eight main buttons you have selected for listening), "Off" (so the recorders can't load the listening feed), "Tape-1" (dubbing to tape 2 and tape 3), and "Tape-2" (dubbing to tape 1 and tape 3). When the "Digital-3" (coaxial) source is selected, there is no output to the "Tape-1" (analog) connections, and the manual suggests use of these two sets to prevent accidental feedback if you want both digital and analog connections to the same DAT deck. There is no provision for monitoring from the output of a digital deck while it is recording, even if the deck itself has that capability.
The remote is very simple. It offers the eight selector buttons and four monitor buttons, plus up and down controls for volume. That's it--but that's sufficient. If you're serious enough about your audio to want the other options, you're probably enough of an activist about it to prefer a hands-on approach anyway. I do.
Measurements
Table I--Output impedance vs. frequency.
Table II--Input and tape output characteristics.
Table III--Linearity error of D/A converter.
Fig. 1--Frequency response via the phono inputs.
Fig. 2--Bass and treble control range, including effects of intermediate
settings.
Diversified Science Laboratories (DSL), who performed all of the laboratory tests on the Sansui AU-X911 DG, measured clipping at 123 watts into an 8-ohm resistive load, a comfortable 0.9 dB above rated power, and at 132 watts (another 0.3 dB higher) into 4 ohms. Dynamic headroom measured +2.1 dB (162 watts, as opposed to the 100-watt continuous rating). Dynamic power into 4 ohms was 1 dB higher yet, at 205 watts. With 2-ohm loads, dynamic power dropped somewhat, to 170 watts, and showed what could be interpreted as signs of instability. Still, this represents, at worst, average behavior for an amplifier section rated at 100 watts a side.
The output section's damping factor was 400 at 50 Hz, into an 8-ohm load, representing an impedance of 20 milliohms at that frequency. Output impedances at other frequencies are given in Table I. In measuring frequency response, DSL followed the IHFA-202 Amplifier Measurement Standard (now published by the EIA, which absorbed the IHF some years ago). This specifies a "standard" volume setting designed to simulate a working setting when the unit under test actually is used for listening purposes. In the AU-X911 DG with analog signals, this resulted in a setting of "31"--that is, with the volume knob set at the point calibrated as 31 dB below maximum gain. (A second calibration scale is used with the D/A converter.) These measurements were made from the power terminals--again, as called for in A-202-to exercise the entire signal chain.
The results showed a distinct though mild roll-off in the treble. Response was down less than 0.25 dB to at least 10 Hz in all these tests, but it dropped by that much at 13.8 kHz, to -0.5 dB at 18.9 kHz, to -1 dB at 29.1 kHz, and to -3 dB at 60.3 kHz. These figures aren't what one might expect from Sansui's specs, which may have been measured with a reduced input signal level and with the volume pot full open, a technique many companies use in their published specifications. Accordingly, DSL tried increasing the test level in 10-dB steps. At "21" (21 dB below maximum), the output was down 0.25 dB at 19.2 kHz and 3 dB at 80.4 kHz; at "11," it was down 0.25 dB at 33.8 kHz and 3 dB at 143 kHz.
Fig. 3--Loudness compensation; see text.
Fig. 4--Frequency response of D/A converter for left channel (solid curve)
and right channel (dashed).
This satisfies Sansui's spec, but it also proves the wisdom of the Standard's insistence that frequency response be measured "as used." In DSL's report, Edward J. Foster noted, "As you can see, treble response improves as volume is raised. This suggests that the input capacitance of the amplifier following the [volume] control is forming a lowpass filter with the impedance of the control. (Note how well calibrated the control is, however.)" With no measurable distortion at any frequency, either at 1 watt or at rated output, only one other point need be made about the lab data on the amplifier section. An error of about 0.5 dB in channel balance showed up, regardless of whether the centered balance control was included in the signal path or excluded via the "Source Direct Operation" option.
Frequency response of the MM phono section was very flat, as Fig. 1 demonstrates. The curve shows only the subtlest of treble rises and an equally subtle roll-off in the lower range. The curve for MC phono has a similar treble rise, to about 8 kHz, but then response rolls off to -0.7 dB at 20 kHz. The gentle downward slope of the low treble continues into the midrange, gradually becoming steeper as it goes. Response is down 0.1 dB (relative to 1 kHz) at about 400 Hz, 0.7 dB at 95 Hz, and 3 dB at 35 Hz-an unacceptable loss of deep bass. The suppression of unwanted infrasonics by this characteristic may be partial compensation if an MC cartridge must cope with warped records. Response of the MC phono section was down 16.5 dB at 5 Hz, whereas that of the MM section was down only 2 dB at 5 Hz. The infrasonic filter helped on warped discs, with a fairly vigorous slope of 12 dB per octave, though its low turnover point of 15 Hz limited its effectiveness to some extent.
In perusing the lab data of Table II, keep in mind that these results all were obtained following the IHF/EIA Standard. In some cases, disparities between the results shown and Sansui's specs probably can be attributed to a difference in technique. All are unexceptionable, in any event.
The tone controls were quite well behaved and gentle in operation. As can be seen in Fig. 2, the bass control shelves below 100 Hz, reaching maxima of about +5 and -5 dB, and treble runs to about the same maxima at 10 kHz. The loudness curve (Fig. 3) adds about 3 dB of treble above 5 kHz and 5 dB of bass below 100 Hz; it is unaffected by the volume setting over the 20-dB test range. Whether this fixed equalization deserves to be called loudness is a moot point.
It may be preferable to the volume-dependent variety, unless the latter can be calibrated to actual listening levels--which can vary widely in practice with so powerful an amplifier.
The main point to be made about the D/A converter is that its linearity was exceptional. The linearity errors indicated as "±0.1 dB" in Table Ill were significantly smaller than this figure, which represents the resolution limits of DSL's instruments under the given test conditions; in most cases, non linearity was, in fact, unmeasurably low. This is superb performance, in what is probably the most relevant of all D/A converter bench tests, by comparison to even the most elaborate multi-bit converter sections.
The D/A converter's frequency response had the usual slight roll-off at the top end, by about 0.5 dB without de emphasis (Fig. 4) and about 1 dB with de-emphasis. Channel balance was within ±0.2 dB, and A-weighted S/N, measured with respect to a 10-watt output (again, an average-listening maximum consistent with the IHF Standard) was 105.75 dB. Channel separation measured 84.5 dB at midband and was still all of 61.5 dB at 16 kHz-far better than is required for good stereo imaging.
In a word, DSL's bench tests paint a picture of an outstanding D/A converter coupled to an analog preamp/ switching/power amplification design that is beyond complaint in almost all respects. Curiously, both of the exceptions are in the most fundamental of all considerations, frequency response of the power amp at listening levels and of the MC phono section.
Use and Listening Tests
The listening focused-as, indeed, does the design itself-on the D/A converter. Sansui lent me a CD-X711 CD player to drive the AU-X911DG's digital inputs. With this player feeding both "Digital-2" and the CD (analog) jacks, I was able to compare the sound via the Sansui's D/A converter with that from the player's own. That I could discover no significant difference is not surprising when you consider the Sansui player is billed as containing the same LDCS D/A converter as the amp.
More revealing was a comparison between a conventional ladder decoder in another CD player and the LDCS decoder in the AU-X911 DG. The Sansui did seem significantly smoother and freer from the harshness I've come to associate with certain discs. However, because the CD player I used for comparison had no digital output, I was unable to make direct A/B comparisons as I had when using the CD-X711, so my judgments involved sonic memory--never an altogether trustworthy investigative tool. So while I question my own methods, I am satisfied that this Sansui amp offers a subtle but significant advantage over conventional designs.
Both sources were played through the AU-X911DG's power amp, of course, so any subtle warming of the top end by its inherent roll-off at my listening levels (through fairly sensitive speakers) applied equally to both. The sound, in fact, struck me as excellent, even taking that slight roll-off into consideration.
Among the best aspects of the design may be the switching. At first acquaintance, it seemed rather complicated, but the longer I used it, the more I became convinced that of all the products of this sort that I've worked with, this is the model best calculated to prove satisfactory in a true digital system-when such a system is available. It still doesn't allow for digital processor loops-you'll have to apply whatever digital legerdemain you want before feeding the signal to this Sansui amp-but neither does any other digital/ analog integrated amp I've worked with. And no model in that category can match Sansui's superb D/A converter.
-Robert Long with Edward J. Foster
(Audio magazine, Apr. 1990)
Also see:
Sansui Model CA-F1 Preamplifier (Jun. 1980)
Sansui AU-9500 Integrated Amplifier (Jun. 1974)
Sansui AU-D11 Integrated Amplifier (Equip. Profile, Feb. 1982)
Sansui Model SR-838 Direct-Drive Turntable (Feb. 1979)
Sansui PC-X1 PCM Digital Processor (Jan. 1984)
Jeff Rowland Coherence One Series II Preamp and Model 7 Mono Amp (Apr. 1990)
= = = =