Line-level transformers in High-End Audio
Mark J Medrud, Design Engineer
Jeff Rowland Design Group, Inc.
17 March 1995
To some, transformers epitomize the early obstacles to high fidelity: Abruptly limited bandwidth short changed both the ends of the frequency spectrum and smeared transients. Limited power bandwidth in particular hampered output transformers, resulting in compressed dynamics, masses of harmonic distortions from hysteresis(1), low-frequency saturation(2), phase distortions and more.
This brings us to an important distinction. When most people think of the sonic characteristics of transformers, they are actually remembering the gross distortions of power output transformers of yesteryear – not the sound of studio grade line-level input transformers. Indeed, they are nothing alike.
Witness the continued re-releases of the superb and cherished classical recordings from the mid-sixties – it’s not just the performances, but the marvelous recording quality that keeps them alive. It’s also a safe bet that they were made with not one, but several transformers in the signal path.
Excellent line-level transformers have long been available to professionals in the recording studio world but they have always been vastly too expensive for mainstream high-fi gear. They remain too expensive for any but the finest high-end audio components to this day.
Even among modern line-level transformers, the Jensen models(3) chosen for Jeff Rowland Design Group’s new products(4) are stand-outs. Here’s are some of the reasons they are so close to perfect:
1) Very expensive high-percentage-Nickel alloy cores. Virtually all commercial grade transformers use low-carbon steel or silicon steel. Though affordable, or even cheap by comparison, they have a lot of hysteresis as demonstrated by their fat BH loops(5). Nickel, on the other hand has a desirably narrow BH loop.
2) Huge core size compared to the power handling needed. The microwatt to milliwatt power levels seen at line-level would normally require only a tiny core. However, very low frequencies at high levels demand big cores to avoid the response roll-off and distortion of saturation. In order to handle deep sub-bass at outrageously high levels without complaint, these transformers are massively bigger than would otherwise be indicated.
3) Super-wide passband. Like most any high-end audio amplifying stages, these transformers have a few extra octaves of frequency response at each end. That’s 0.5Hz to 180KHz!
4) Nearly perfect group delay and hence excellent phase response in the audio band. Before his death in 1989, Deane Jensen pioneered the use of a Bessel(7) lowpass function to define the shape and characteristics of the transformer’s upper frequency limit.
Through seeming magic, he created computer modeling techniques which exploit the inherent winding resistance, inter-winding capacitance and stray inductance in the transformer to create a second order Bessel filter rolloff. The result is a transformer that will pass a squarewave that looks perfect on an oscilloscope: No droop, overshoot, ringing or other wigglies – ultra-rapid rise and fall times with flat tops and bottoms.
Audio transformer design is not the simple, strict engineering exercise it ought to be. More an art than a discipline, creating the finest examples contain subtleties of technique which resist reverse-engineering and mass production. Like most good things in high-end audio, a it takes a lot of care and skill to make good ones.
OK. Very good line level transformers aren’t damaging to the sound. How do they manage the miracle of subjectively improving the sound so dramatically? That is harder to explain definitively – and possibly not fully understood. Let’s start with why recording studios love them so much:
1) Transformers break ground loops. Since the transformer’s primary and secondary aren’t electrically connected, no ground loop currents can flow. Hair- pulling, aggravating ground loop related hum and noise problems disappear.
2) Transformers have ultra-high common mode input impedances (inputs-to-ground), unrivaled by active differential circuits. This is also a result of the galvanic isolation provided by the insulated, shielded primary windings. Direct connection to a differential (balanced) amplifier’s inputs doesn’t guarantee perfectly balanced operation – transformers do.
It should be noted that many studios have been seduced by the low cost and great looking specifications of the solid-state "diff-amps" sold as transformer replacements over the years. The harsh reality is that there is a huge gap between laboratory measurements and real-world performance.
Most studios can’t afford the extravagance of transformers over diff-amps. Nevertheless, there is a rapidly growing cadre of perfectionists in professional audio who are returning to transformers as the ultimate signal interface.
These usability issues should be important to we audiophiles as well, but our list pretty much starts and ends with sonics. Ease-of-use is important to a manufacturer supporting a product, but some of the more radical "tweaks" consider dealing with difficult hook-ups a red badge of courage. They wouldn’t dream of compromising the sound for convenience. Fortunately, line-level input transformers improve the sound as well.
In these modern times, we live in a veritable soup of RFI/EMI(8). Sources from microwave ovens to TV stations contribute, but CD players are often the most egregious polluters. Some argue that RF interference should have no affect on an audio signal. That would probably be true if active components had infinite bandwidths and could accurately track any signal regardless of frequency. Of course they don’t, and out-of-band interference "sees" them as a jumble of the resistances, capacitances, inductances and diode-junctions that make up electronic circuits of all types. "The hidden schematic" is the new jargon to describe this phenomenon.
We all know that non-linearities cause harmonic and Inter-Modulation (IM) distortions. Diodes are extremely nonlinear around their turn-on voltage knee – so much so, that this characteristic is exploited in the frequency mixers that make superheterodyne radios possible. IM is just like superhet: two frequencies subjected to nonlinearity result in two new frequencies – the sum and difference of the two originals.
We don’t care much about the sums, but consider the zillions of possible differences in our RFI ridden environment. For every pair that is separated by an audio frequency there is a difference signal folded down into the audible range. For example: 1.001MHz and 1.000MHz RFI components sneaking into an amplifier mix to a difference of 1KHz. You can’t hear 1MHz or 1.001MHz, but you sure can hear 1KHz. Speaking of superhet, the popular AM radio test(9) used to sniff out RFI emitting components demonstrates the same principle.
The harmonic distortion caused by the tiny amount of hysteresis in a nickel- core transformer is mostly second order. This would normally be suspect as just another euphonic distortion except that it is diminishingly small. Throughout most of the audio range, the total distortion is much lower than .001%. Likelier, we audiophiles appreciate the ways that these transformers block RFI so well.
1) Bandwidth limiting: The 180KHz bandwidth of the transformers passes only frequencies that the subsequent active circuits can follow accurately. In other words, the outside world’s RFI doesn’t get a chance to act on the hidden schematic and can’t mix down into the audible range.
2) Bessel function roll-off at the upper bandwidth limit: Not only are Bessel filters "phase linear", but they are ideally damped to eliminate ringing (signals amplified by resonance) of the out-of-band garbage. The bandwidth is gracefully limited in a way that blocks the interference without intruding on the music.
3) Faraday shielding. The transformer’s primary and secondary windings are separated by an electrostatic shield that shunts the remaining capacitively coupled interference currents to ground.
4) Magnetic shielding. The entire transformer is encased in a mu-metal(10) shield that keeps out magnetic interference.
Of course, nothing absolutely eliminates RFI. However, the right transformer can be a simple and elegant way to purify the desired signal. They do a terrific job of isolating the sound from the hostile world outside so that the active circuitry can work its best.
It is important to look beyond the lack of distortion and the great specifications of these transformers. They are also so subjectively transparent that the underlying characteristics of the electronics behind them are more readily apparent in critical listening. The quiet around the elements of the soundstage (both in time and space – dimensionality) is so dramatic that it reveals by exposing, rather than by exacerbating. The tiniest contributions of other components in the signal chain are unmasked.
(1) Hysteresis, in this case, refers to the tendency of the magnetic flux in a transformer’s core to not respond precisely to the magnetizing force of the current in primary coil. For instance, the predilection of the core to stay slightly magnetized even when the current in the primary coil has ceased.
(2) The upper limit of a transformer’s core to contain magnetic flux is called saturation. Above saturation, additional magnetic force causes little increase in flux density. The flux density in a transformer’s core is inversely proportional to frequency of the signa1 applied to the primary coil: Halve the frequency and the flux density doubles.
(3) See the following data sheet for specifications of the Jensen Transformer model JT-10KB-D.
(4) Jeff Rowland Design Group’s Coherence Preamplifier II, Model 2 and Model 6 power amplifiers employ Jensen line-level input transformers. The Coherence also uses Jensen line-level output transformers.
(5) B is the notation for magnetic flux density and H is magnetic force. Plotted together, they graphically illustrate hysteresis. A large separation between the magnetizing and demagnetizing curves (a "fat" loop) shows much hysteresis and a small separation (a "narrow" loop) approaches the ideal of no hysteresis.
(7) Of the many functions that describe electrical filter characteristics, Bessel filters are among the most highly damped. They also employ the minimum damping necessary for ideal time domain response.
(8) Radio Frequency Interference is a subset of all Electro-Magnetic Interferences. RFI is specifically mentioned since the discussion specifically describes the consequences of EMI in the range of radio frequencies.
(9) RFI in the AM broadcast band can be detected with an inexpensive portable AM radio. Tune it to an unused frequency and turn up the volume until substantial noise is heard. Various "squawking", "chirping" or "buzzing" sounds will be heard in the presence of interference from fluorescent lamps, desktop computers, most CD players, and a variety of other appliances.
(10) Mu-metals (mu, "mew", from the Greek letter µ) are a family of magnetic metals specifically optimized for magnetic shielding.