Jeff Rowland Design Group Coherence preamplifier & Cadence phono stage Transformers and the Audio Interface

Sidebar 2: Transformers and the Audio Interface

The majority of Jeff Rowland's new products are equipped with line-level input transformers. In the early days of stereo, the use of transformer-coupling components in the recording studio was the norm. Nearly all of the most prized recordings of the late 1950s and '60s were captured through a whole chain of these passive devices. Because at that time construction of good line-level transformers was more of an art than a science, quality was understandably variable. In addition, limitations in material quality, knowledge, and precision equipment led to products that were good at isolating one component from another and preventing unwanted hum, but often suffered from limited bandwidth, core saturation, audible hysteresis, and excess phase-distortion anomalies. Combined with high cost, these flaws led to a rapid reduction in the use of line-level transformers after the appearance of cheap silicon differential amplifiers, claimed to be "electronic transformers."

Unfortunately, the majority of these active-differential input topologies fell short of meeting the essential transformer requirement of rejecting unwanted in-band (and some out-of-band) spurious signals. Meanwhile, tremendous progress has been made in the past 25 years in transformer design, quality of core materials, and precise winding technologies. To be sure, many poor-quality audio transformers are still sold, but the few commercial reference-grade devices—such as those made by Jensen Transformers—are extremely consistent, and have largely addressed the performance drawbacks of the old days while enhancing the beneficial qualities inherent to these devices. But they're still very expensive.

The Jensen JT-10-KBD, used in the Coherence, sports a reported bandwidth of 0.3Hz-160kHz, and a deviation from linear phase (DLP) of less than 1 degrees across the audible band. (DLP is a true measure of phase distortion.) This is accomplished by taking what used to be a disadvantage of transformers—a degree of intrinsic leakage inductance and capacitive coupling between the primary and secondary windings—and shaping those reactances through specific winding techniques to form a second-order Bessel low-pass filter when the DC resistance of the device is taken into consideration. As a result, ultrasonic transformer resonances that can intermodulate with subsequent, poorly filtered amplifying stages are eliminated. Another key benefit of this inherent low-pass filter is good RFI/EMI rejection up to a claimed 30MHz.

Jensen states that their large, high-quality nickel-alloy cores make saturation a nonissue, even at inputs as high as +22dBu—far above any home-playback SPL. As a result, transformer-generated distortion is effectively quelled at any realistic signal level.

If all of this wasn't enough benefit from a simple passive device that requires no trimming and lasts a lifetime, the primary benefit of an input transformer is its outstanding common-mode rejection ratio (CMRR).

Transformers vs diff-amps: A fundamental limitation of most actively balanced input circuits compared to a good input transformer is revealed by examining a basic misconception of what constitutes a balanced interface. Semantics is part of the problem—balanced operation implies a circuit capable of robust common-mode rejection (CMR), regardless of its signal symmetry. (Signal symmetry is achieved at any point along the signal path when the positive and negative legs of a normal-mode signal are of equal amplitude but opposite polarity with respect to ground reference.) It's also important to distinguish between balanced topology within a component, where symmetry can play a key role, and the limited value of symmetry through the analog interface. If perfect-differential signal symmetry is maintained throughout a component—no small feat—benefits such as improved power-supply rejection and reduction of even-order distortion can occur.

Unfortunately, many active or discrete interface designs were built based on the erroneous assumption that signal symmetry alone ensured good balanced operation, and thus a high CMRR. In reality, common-mode rejection is not an inherent advantage of signal symmetry. Hence, a purely symmetrical interface is not inherently balanced. On the other hand, fully differential balanced systems done right, such as Rowland's new gear, maintain both good symmetry throughout the signal path and robust CMR at the interface.

Transformers can have an edge over active circuitry because CMR within the interface is due to the relative ratio of the common-mode impedances between the output and input circuits of two components—cables included. Only when the source impedances from the positive and negative leg of each channel to ground are tightly matched can unwanted in-band hum and noise be rejected. In addition, any balanced interface scheme is least sensitive to degradation when the ratio of the pair of source impedances and pair of load impedances are widely different from one another. A typical diff-amp has an input impedance range of from around 5k ohms to 50k ohms, resulting in a relatively narrow ratio with typical source impedance values in real systems. As a consequence, even small imbalances of just a few ohms between the two legs of a signal, as are often encountered in the field, can markedly degrade CMR.

The disparity between real-world performance and lab measurements illustrates the problem. When most components with diff-amp inputs are tested with perfectly balanced source impedances, such as from the output of an Audio Precision set or with the two inputs tied together, the high CMRR touted by the diff-amp vendors can indeed be achieved. This often leads to surprise when some of these products are inserted into difficult, "noisy" systems but still exhibit hum and buzz problems.

By contrast, a good input transformer has extremely high common-mode input impedance of 50 megohms or more, ensuring very high actual CMRR regardless of the degree of imbalance of the source. Herein lies the transformer's fundamental advantage; indeed, even when driven by a single-ended component with a source impedance imbalance of 1k ohms or more, the ratio of this figure relative to 50 megohms is still so high that 90dB of real-world CMR is still possible. Thus, a modern transformer-coupled input qualifies as a truly universal interface capable of delivering consistently high performance, whether the source be single-ended or balanced.

Of course, there's much more to this story, and issues like proper "Pin 1" termination are equally important considerations in interface performance (footnote 1). Also, THAT Corp. recently introduced a new, patented active input circuit that's said to offer most of the benefits (excluding galvanic isolation) of a true transformer, and for a fraction of the cost.—Shannon Dickson

Footnote 1: See the "Technical Discussions" link at Jeff Rowland Design Group's website.—Shannon Dickson
Share | |

Enter your username.
Enter the password that accompanies your username.