Rated Power Output: 250 watts per channel into 8 ohms, 375 watts per channel into 4 ohms.
Dimensions: 19 in. W x 5 1/2 in. H x 12 1/2 in. D (48.3 cm x 13.3 cm x 31.8 cm).
Weight: 50 lbs. (22.7 kg).
Price: $2,200.
Company Address: 613 South Rockford Dr., Tempe, Ariz. 85281; 602/967 3565.
The Trans-Nova circuit design goes back to about 1984, when it was introduced by Acoustat (later bought by Hafler) in two solid-state power amps (the Models TNT120 and TNT200). "Trans-Nova" is a contraction of Transconductance Nodal Voltage Amplifier (U.S. Patent No. 4,467,288). When I first examined the circuits for these amplifiers, I was most impressed with the topology of the output stage. Other aspects of the original circuit design were quite elegant also.
The 9505, a third-generation Trans-Nova design, is aimed primarily at the professional audio market. It is the larger of two otherwise similar pro models and is rated at 250 watts per channel into 8-ohm loads. (The smaller 9303, rated at 150 watts per channel into 8 ohms, is priced at $1,300.) A reasonably sized package for its power out put, the 9505 has a front panel graced by a single on/off rocker switch. An indicator in the switch glows when the amp is turned on. Both balanced and unbalanced input connectors are provided on the rear panel. The interesting XLR connectors will accommodate either the usual mating XLR connector or a 1/2-inch phone plug. Speaker connections are made via two pairs of five-way binding posts. Three recessed slide switches select balanced or unbalanced in put mode, stereo or mono (bridged) operation, and connection or disconnection of the chassis to the third-wire ground. The AC line connection is via an IEC socket and mating power cord.
Inside the 9505 is a main p.c. board in a "C" shape, oriented with its long side to the rear. A large, rectangular, UI-lamination power transformer is situated in the opening of the p.c. board. The main filter capacitors for each channel are mounted on the short sides of the board, adjacent to the heat sinks. All of the input connectors and slide switches on the rear panel are mounted to the rear portion of the p.c. board. This is the first power amplifier I've seen that uses surface-mount parts for most of its signal circuitry and the low-power parts of its power supply.
If one considers the power supply as part of an amplifier, the standard "half-bridge" output-stage topology is actually a full bridge, consisting of four elements. These elements are the two output devices (or the equivalent, where multiple devices are paralleled for more power-handling capacity) and the positive and negative power sup plies. In the usual arrangement, the center point of the power supplies is grounded, and the load is connected between this ground reference point and the midpoint between the two output devices. In most designs, these output devices are driven as followers, with their input driving voltage slightly higher than the output voltage.
In the Trans-Nova design, the load is still connected between the same two points in the bridge. What's radically different is that the Trans-Nova uses the midpoint be tween the output devices as its ground reference and lets the center tap of its power supply move with the signal. In this arrangement, the output devices (MOS FETs in the 9505) are operated as common-source amplifiers with voltage gain, and their input driving signal is referenced to ground. The input signal required is much smaller, so the front-end driving circuitry can be operated from a much lower supply voltage than that needed for the output stage. As the 9505's excellent owner's manual points out, the output stage's voltage gain gives this stage approximately 10 times the power gain of a conventional follower circuit using exactly the same MOS-FET devices.
Negative feedback is taken from the out put point back to the input gates of the MOS-FETs. This inverting feedback converts what would be a high-impedance input to a low-impedance input. In other words, the output stage has been made into a transimpedance stage, or current-to-voltage converter. This stage is fed by the driver circuitry, which is configured as the output stage's complement, a voltage-to-current (or transconductance) circuit.
The driver stage is a newly developed circuit, DIABLO (Dynamically Invariant A-B Linear Operation), that is designed to provide up to 14 dB greater headroom than the usual Class-A driver stage. This extra headroom is needed because the output stage has four pairs of MOS-FETs per channel, used in their common-source mode. The appreciable in put capacitance of this arrangement calls for a driver stage that has more output current at high frequencies than the usual Class-A stage, with its limited 2-to-1 ratio of peak to quiescent current, can provide. To get around this limitation, the DIABLO circuit uses a complementary common-base first stage, direct-coupled to a complementary cascode-connected second stage.
At the input of the 9505, the phases of the signal are each buffered by a discrete circuit that consists of an N-channel J-FET source follower with a bipolar cur rent source. This is coupled into a complementary bipolar emitter follower. Grounding one phase of this buffer input changes the input from balanced to unbalanced; the balanced/unbalanced switch merely un grounds (or grounds) the negative input phase for balanced (or unbalanced) input configuration. For bridged operation, the stereo/mono switch establishes an inverted polarity signal path from the left channel's input (which doubles as the mono input) to the right channel's. An op-amp servo circuit monitors the amplifier output's DC level and applies any error to the ground end of a signal-voltage divider that feeds the positive input of what I consider the power amplifier proper (i.e., everything that follows this buffer).
As is often the case, the amplifier proper is embedded in a four-resistor, differential-to-single-ended circuit that incorporates two voltage dividers, one for each signal phase. The input of the power amplifier proper is a differential amplifier using a matched pair of N-channel J-FETs. The J-FETs' sources are connected to a bipolar current source whose drain outputs are coupled to a bipolar current mirror. One of the differential amplifier's outputs is direct-coupled to the input of the driver stage. Overall negative feedback is taken from the output to the inverting input of this differential amplifier.
The power transformer is somewhat unusual, having separate primary and secondary windings for each channel; each of the long sides of the transformer's UI core carries one such primary-secondary pair. This reduces the capacitive coupling be tween the high-current secondary windings as they move with the signal in respect to ground and to each other.
Measurements
Fig. 1-Frequency response.
Fig. 2-Square-wave response for 10 kHz into 8 ohms (top), 10 kHz into 8 ohms paralleled by 2 NF (middle), and 40 Hz into 8 ohms (bottom).
Fig. 3-THD + N and SMPTE IM distortion vs. power output.
Fig. 4-THD + N vs. frequency.
The test results cited here are for the left channel with unbalanced input. Any significant departure, for the right channel or balanced input, is noted.
Frequency response for open-circuit, 8-ohm, and 4-ohm loading at a nominal level of 2.83 volts (1 watt into 8 ohms) is plotted in Fig. 1. Bandwidth is very wide; further, the curves are very close together over the audio range, indicating a very low output impedance and consequent high damping factor. Rise and fall times measured 1.1 microseconds for an output level of ±5 volts into 8 ohms, yielding an equivalent bandwidth of about 318 kHz. Square wave response is shown in Fig. 2. For 10 kHz (top trace), rise time is sharp and fast. The addition of a 2-microfarad capacitor across the 8-ohm load (middle trace) causes ringing, typical of most solid-state amplifiers. The absence of tilt in the low-frequency trace (bottom) is indicative of excellent, extended infrasonic response.
Figure 3 shows total harmonic distortion plus noise (at 1 kHz) and SMPTE intermodulation distortion versus power. With the balanced inputs (not shown), THD + N was about the same as seen in Fig. 3, but IM distortion was a third to a half as much from 10 to 400 watts. Figure 4 shows THD + N versus frequency for low, medium, and high power into 4 ohms. Spectrum analysis (not shown) revealed that the second harmonic was the dominant distortion component over much of the power output range. When the second harmonic is dominant, harmonic distortion level is relatively constant with change in output, as can be seen in Figs. 3 and 4.
With the unbalanced inputs, crosstalk was more than 100 dB down up to 2.5 kHz; there was remarkable similarity between the right-to-left and left-to right directions. Crosstalk then increased at 6 dB/octave, reaching-86 dB at 20 kHz. With the balanced inputs, the symmetry between directions was not as good; the amount of crosstalk was some 2 to 10 dB worse, depending on frequency and direction.
For the 9505's balanced inputs, common-mode rejection ratio (CMRR) rose by approximately 6 dB/octave over the audio range. It started at-106 and-110 dB at 20 Hz for the left and right channels, respectively; it ended up at -54 and -60 dB at 20 kHz.
Output noise levels for the right (worse) channel were 314 micro volts wideband, 252 microvolts from 22 Hz to 22 kHz, 131 micro volts from 400 Hz to 22 kHz, and 130 microvolts A-weighted. The results for the left channel were about 10% to 20% better. The unit's A-weighted signal-to-noise ratio was-88.2 dB for the left channel and -86.7 dB for the right, relative to a 1-watt output into 8 ohms. The noise was satisfactorily low, mainly hum components induced by power-transformer flux. (There was also some audible mechanical hum emanating from the transformer.).
Output impedance was very low in both channels. Damping factor, referred to 8 ohms, was 670 from 20 to 500 Hz, decreasing to 615 at 1 kHz and to 100 at 20 kHz. Voltage gain into 8-ohm loads was slightly greater than 28.7 dB.
In the test of dynamic power, the 9505 produced 390 watts into 8 ohms at the beginning of the tone-burst signal and 380 watts at its end; dynamic headroom was 1.9 dB. For 4-ohm loads, output was 666 watts at the start of the burst and 648 watts at its end, corresponding to a dynamic headroom of 2.5 dB. Maximum undistorted output into a 1-ohm load with one channel driven was 48 volts at the start of the burst and 44 volts at its end, equivalent to peak currents of 48 and 44 amperes, respectively. Power attainable at the visual onset of clipping was 345 watts into 8 ohms and 553 watts into 4 ohms. Clipping headroom was therefore 1.4 and 1.7 dB, respectively.
The 9505's AC line draw was about 2 amperes. The current remained quite constant from cold turn-on to the point where the amplifier became quite hot during the power tests; this indicates excellent output-stage thermal stability.
Use and Listening Tests
During the review period, the equipment in my system included an Oracle turntable fitted with a Well Tempered Arm and an Accuphase AC-2 moving-coil cartridge, used with a Vendetta Research SCP-2C preamp. A Counterpoint DA-11A CD transport drove a Museatex Bidat or a Sonic Frontiers SFD-2 MKII D/A converter. Additionally, a Genesis Digital Lens jitter-reducing device was placed between the CD transport and the D/A converter. Other program sources were Nakamichi's ST-7 FM tuner, a Nakamichi 250 cassette recorder, and a Technics 1500 open-reel recorder. I used a Forssell balanced tube line driver with the Sonic Frontiers D/A converter and a Quicksilver preamp with the other components. Power amplifiers on hand were a Crown Macro Reference, a pair of Quicksilver M135s, an Arnoux 7B digital switching design, and a JoLida SJ 302A integrated tube unit. Loudspeakers used in the tests were B&W 801 Matrix Series 3s, each of which was augmented from 20 to 50 Hz by a subwoofer.
The Hafler Trans-Nova 9505 impressed me right away with its smooth presentation. The more I used this amplifier, the more I liked it. I found its ability to deliver excellent resolution and detail, without producing much edginess or irritation, endearing. Space, dimension, and air were excellent, as were tonal balance, bass definition, and impact.
"Resurrection," track 6 of Bourbon of Rosewater (Waterlily Acoustics WLA-CS 47-CD), yielded a sound so sweet, clear, and realistic that it was hard to imagine it sounding better. Similarly, on Mendelssohn's "Die Tageszeiten," track 7 of The Times of Day (Reference Recordings RR 67CD, an HDCD-encoded disc), the sound of an orchestra playing and men singing in a chorus was very palpably present.
Both in the lab and in my listening room, he Hafler Trans-Nova 9505 behaved just about flawlessly. I liked it very much. And although I didn't audition the less powerful 9303, I expect its sonic character is very similar to the 9505's.
-BASCOM H. KING
(Adapted from Audio magazine, Apr. 1996)
Also see:
Hafler XL600 Power Amplifier (Equip. Profile, Feb. 1989)
Hafler DH-330 FM Tuner (Nov. 1986)
Hafler XL-280 Amplifier (Nov. 1987)
= = = =