In a digital audio landscape, transformers are the new tubes. But there’s much, much more to transformer design than perceived ‘warmth’. Jensen Transformers’ president Bill Whitlock talks cores, phase distortion, and the enduring legacy of Deane Jensen.
Story: Mark Davie
There’s a fresh fracas to rival the Loudness Wars — we’re on the slippery downslope of the Battle for Harmonics, and it’s getting muddy. At first, we were happy with the three T’s: the harmonic triangle of Tubes, Tape, and Transformers — each far more than a prosaic electronic component or pitstop in your signal chain, but a potentially prized colour in your tonal palette.
Lately, we’ve seen some hysteria over hysteresis. Most of us wouldn’t have the foggiest clue what it is, but we want it, that’s for sure. I’ve been reading one-sheets for transformer-laden gear that never mention noise isolation, but can’t wait to slip in a Trademarked reference to hysteresis-induced harmonics. And just recently, news of a glitch plug-in came across my desk. Its name? Hysteresis [take a look at our news pages for more].
But have we overdriven our common sense so far as to dig up coal and declare it gold?
Bill Whitlock certainly thinks so. He was the president of Jensen Transformers until just recently when the company was bought out by its number one customer, Radial Engineering. He was also a lifelong friend of founder Deane Jensen, and has continued his legacy of great transformer design through an intimate knowledge of the component’s role.
As far as Jensen is concerned, high levels of hysteresis have never been a desirable trait. “Deane was convinced that the goal in music reproduction was to accurately reproduce that oscilloscope waveform,” said Whitlock. “The main reason people use a transformer in the first place is to get rid of ground noise: hum, buzz and all those other issues. So if you do try to get rid of noise, I think it’s extremely critical you put a transformer in the signal path that’s not going to ruin your signal quality. If I had to put the goal of Jensen in a single word, it’s ‘transparency’; to make transformers that sound like they’re not there.”
It’s a good point. Do we want components that do their job as cleanly and efficiently as possible, or components with inferior performance that imbue a subjective ‘sense’ of warmth? If the popularity of Radial’s JDI box is anything to go by, we value transparency even if we don’t realise it.
Bill Whitlock has always followed in Deane Jensen’s footsteps, but in the way a cleanup hitter treads the same bases as the leadoff batter; he gets everyone home.
The first time the pair met was during a job interview at Quad Eight where Jensen, the chief engineer at the time, was interviewing Whitlock as his potential replacement. Jensen was practically edging towards the door, itching to start up Jensen Transformers.
It was 1971, and Whitlock had just moved to California, the audio engineering hotbed where UREI had just dropped the 1176 peak limiter, and Quad Eight was building arguably the best consoles of the day.
Jensen had worked at Quad Eight since 1968, and before that, a stint at UREI. But it was his work as a consultant to Wally Heider Studios that led him to look deeply at transformers. The studio was complaining about an edgy sound from their consoles they didn’t like, and charged him with unearthing the issue. It eventually led him to the transformers, and after taking some measurements, he didn’t like what he saw.
Jensen didn’t have breakthrough expertise in winding coils — that had been happening for 50 years — or alloying metals (though he did develop expertise later). Instead, he had an understanding of how the transformer interacted with the circuitry that surrounded it. And that’s where all the problems were.
What he found in those transformers were, “undamped ‘screaming’ high frequency resonances in the ultrasonic region and very inadequate low frequency response,” said Whitlock. “It was bad enough that it would change the timbre of things like kick drums.” It would manifest itself as harshness you couldn’t quite put your finger on, but you could hear something was wrong.
The quest to get rid of all those effects led him to start Jensen Transformers. He soon realised he needed more computing power. So, along with two experts, developed the Comtran Circuit Modelling software. This was the mid-’70s, when state-of-the-art for the day were Hewlett Packard programmable calculators. Whitlock: “The program was written in Rocky Mountain Basic, and Hewlett Packard was so impressed it started selling his software. Deane was so proud that he was the first ever third-party software supplier for Hewlett Packard.
“That was the beginning of the quest, and he gradually discovered how to overcome what was wrong with these transformers. In 1974 the line of Jensen Transformers was born, in fact, this year is Jensen’s 40th anniversary.”
In his research, Jensen found the problems resided in three general areas: the selection of core material, phase distortion, and frequency range. Here’s Whitlock’s beginner’s guide to what was wrong.
SELECTION OF CORE MATERIAL
Whitlock: “Core material selection is critical to both low level and high level harmonic distortion in transformers. The high level part is due to core saturation. Basically, you can wind a coil around a piece of core material and keep increasing the current flow through it and create a stronger and stronger magnetising field, but at some point that core material won’t be able to hold any more magnetic flux. It’s saturated. If you take the coil further than that, all the excess magnetic field radiates out into the air in each direction, because it can’t flow through the material.
“That eventually got solved by using a very high-permeability core material called Permalloy. It’s basically an 80% nickel alloy that’s got traces of molybdenum, iron, and some other things in it. Its permeability is up between 50-100,000 — extremely high, compared to about 300 for ordinary steel. Sometimes it’s called Mu-metal, which is actually a brand name, like calling a facial tissue a Kleenex.
“At the other end, the very low level distortion is caused by what’s called magnetic hysteresis. An analogy is when you use a piece of metal as a spring. If you hold a piece of metal in your hand and push it in one direction it will deflect, and when you let it go it will return to where it started… sort of. And if you push it in the other direction it will come back to its resting position… sort of.
“Hysteresis is that little difference in the resting position after it’s been exercised in one direction and then the other. So if you magnetise the core very strongly in one direction and then turn off the current, the magnetic field in the core doesn’t quite return to zero. The effect is, it has a harmonic distortion, a very small signal.
“Deane found it could be altered with specific annealing processes, because this property is set by the heating and cooling rate that sets the metallurgy of the core material. To this day it’s our trade secret — the cookie recipe for our heat-treated core material that gets us an extremely small distortion level at low-level signals.”
PHASE DISTORTION & FREQUENCY RANGE
The other two points are somewhat related, as Whitlock explains: “The classical specification for an audio system is that the low-frequency response extends to 20Hz. Well, it turns out that to get good replication of frequencies in a waveform down at the 20-30Hz region, the -3dB point has to extend down to less than 1Hz.
“It’s not widely known but it was driven home by a 1989 AES paper, by a Georgia Tech professor called Marshall Leach, that hasn’t received enough attention. He showed, mathematically, the relationship between phase distortion and bandwidth.
“But the popular misconception is that phase shift is the same thing as phase distortion. This really bothered Deane and he wrote a paper about it in 1986 that basically said as long as a phase shift is linear with frequency, it turns into a very benign time delay. It’s no worse to the fidelity of a signal than moving your head an inch further away from the speaker.
“What really alters the shape of the waveform and the timbre of the music is a non-linearity in that phase relationship. That’s why you’ll see on all the Jensen data sheets that we measure phase distortion as a deviation from linear phase. If you can stay linear phase across the audio spectrum, which our transformers typically do, staying within a couple of degrees from 20Hz to 20kHz, you can very accurately reproduce the real waveform of the music, which was Deane’s goal.
“But the really interesting thing happens at the high-frequency end, where bandwidth is related to phase distortion by a factor that depends on the nature of the filters. For a simple one-pole filter, 230kHz of bandwidth at the 3dB points was required to stay within five degrees of phase distortion at 20kHz.
“However, you can make the roll-off much steeper and get that same five degrees of phase distortion criteria in a bandwidth as small as 25kHz, but you have to do it with a filter that’s designed for the purpose, and that’s what Bessel filters are extremely good at.
“All Jensen transformers roll-off as a second-order Bessel low-pass filter. Therefore they have impeccable phase response way up to 20kHz and don’t do any peaking above that. Even today if you do a frequency response sweep of other transformers, you’ll find some very peaky resonances out in the 50-100kHz region. And if you put a square wave through, it will cause ringing on the top of the square wave. This has some rather devastating musical results, which swings us into the topic of spectral contamination that Deane co-wrote a paper about with his friend Gary Sokolich in 1988.
“It basically says if you have resonant peaks that exaggerate ultrasonic frequencies in an audio signal chain, it is almost certain you will cause some very ugly distortions in downstream equipment. Ultrasonic signals come along today from things like DAC clock residue and D/A converter outputs. With phonographs it’s due to ultrasonic mechanical resonances in the stylus, but you can end up with a considerable amount of energy out in the 40-50kHz range.
“If this is allowed to travel through the system, it will end up going through amplifier stages. And because op-amps have the least amount of feedback at these ultrasonic frequencies, they become quite non-linear and behave as mixers.
“If you put two frequencies into any non-linear device like an amplifier, you will get a sum and difference frequencies. And if you put in a complex set they cross-modulate with each other and the harmonic mixing products of these ultrasonic harmonics regurgitate down in the audible spectrum. And they do it at very non-harmonically related frequencies. This has been confirmed with listening tests. It essentially brings down a kind of veil — almost like modulation distortion or modulation noise — that’s very noticeable when it goes away. This can also lead to very blurred imaging in stereo when you have this non-harmonic related background grunge present.
“So transformers can perform as very phase-linear, low-pass filters to keep this ultrasonic garbage from travelling downstream. That’s what Deane wanted to solve, problems caused in ordinary designs by people that did not understand how the transformer interacted with the things around it.”
USING TRANSFORMERS FOR DISTORTION
These days, there’s a lot of talk about harmonic distortion in transformers, without a real appreciation of how it’s generated, or if it’s desirable. While Whitlock and Jensen often turn manufacturers away who are looking for those distortion characteristics, he knows very well why they’re asking: “They’re the people that want colouration. Some people call it ‘warmth’ — the harmonic distortions of the transformer. I’ll grant you, they are very different to electronic distortions because they’re all low order. If you start crowding any magnetic device, including analogue magnetic tape, and start pushing it towards saturation, what you get is second or third harmonic distortion.
“You don’t get a lot of the seventh, 19th and 33rd like you would if you pressed an op-amp — which get real ugly, real quick, because the harmonic distortion mechanism produces not only low order distortion products but extremely high order artefacts that sound really ugly.
“So transformers among the nostalgia crowd, for lack of a better term, are going after that original transformer warmth and overload characteristics which goes hand in hand with the way tubes overload, as opposed to op-amps. They’d rather it gracefully enter their saturation, and older transformers will produce significant amounts of low frequency distortion at normal operating levels.”
Permalloy isn’t a recent invention, it’s been around since 1914. The reason it’s not always used is typically because of cost. Not all transformers on the market are equal, instead we have a wide variety of offerings using cheaper core materials that don’t suit the application; lower windings; and smaller transformers that don’t have high inductance.
This is one of Whitlock’s pet peeves. Because unless the manufacturer is honest, and publishes specs to back it up, it’s hard to know what’s hiding in those coils. Whitlock: “For a given source of signal to keep the impedance of the transformer high as frequency goes down, you have to have a certain amount of inductance. The inductance of a coil is proportional to the number of turns and the permeability of the core material. Since steel has a low permeability, it requires more turns than a high-nickel core material would. On the other hand, steel will support much stronger magnetic fields before it starts to saturate.
“Steel transformers should only be used when the source of the signal has an extremely low impedance, like when you can connect it directly to the output of an op-amp. But in practical equipment, there is a network of some sort between the op-amp and the output pin, so to use the transformer on the output, we rarely, in fact never, recommend using a steel core transformer.
“Most of our competitors will use small steel core transformers because the material is much cheaper, and it will handle a greater signal level in a smaller package. But the distortion characteristics are completely unacceptable. Steel has very high hysteresis distortion and even though that distortion will not increase a lot until it reaches saturation, which is at a higher level, it’s a trade-off we’re not willing to make. They’re often the size of a postage stamp and physics just doesn’t allow a transformer to perform at 20 or 30Hz without being fairly large.”
WHITLOCK’S FINAL INNINGS
Whitlock’s run with Jensen isn’t over, for the next four years at least, Whitlock is staying put at Jensen. But running a company as the legacy of a great but troubled man hasn’t been easy.
Jensen’s father had been a very high-level physicist for Westinghouse, the kind of physicist the government contracted for projects shrouded in secrecy. The company he kept included Albert Einstein, who would occasionally drop around for a cuppa and chat. His father, according to Whitlock, had incredibly high standards and wasn’t the most encouraging of figures. “I think to some extent Deane got the patent [in the 990 op-amp, see sidebar] just to prove to his father how clever he was,” reasoned Whitlock. “One of Deane’s problems was constantly trying to prove himself to his father who was a pious, academic man, and very hard to impress. Deane also was a petit-mal epileptic which he tried to hide all his life, and his identical twin brother was killed as they crossed the street when they were 14 years old, which really left some mental trauma on Deane.”
It’s a lot to carry, and although Jensen was a genius, he gradually became weighed down by problems with the company. The pair had caught up over the years since their first meeting at Quad Eight: Whitlock helped Jensen brainstorm the 990 circuitry; and while he was head of electronic engineering at Capitol/EMI, used Jensen transformers in a stereo widening product he built called The Spatializer. Fed up with Capitol, Whitlock was ready to throw in the towel after seven years, so Jensen suggested they team up and form a consulting team. The consulting business was going well, but Jensen was very close-mouthed about the mounting problems his transformer company was facing.
It all became too much for Jensen, as Whitlock tells it: “He went out for a walk one Saturday morning as he customarily did and didn’t come back. I got worried, and on Monday called the Jensen offices to ask if anybody had seen Deane. They found him in his office with an empty bottle of wine and a revolver in his hands. He’d blown his brains out. In the note there were lots of things including problems with suppliers, a friend that had really taken advantage of him, and he just couldn’t see his way out. In his suicide note he said, ‘Bill is strong, I think he can deal with all this. I leave everything I own to Bill.’”
It turned out Jensen was upside down to the tune of almost half a million dollars, ‘everything he owned’ needed salvaging. At first, Whitlock wanted to smack his friend for not confiding in him. But eventually, against the advice of his lawyer and accountant to close the doors, Whitlock decided to take Jensen on and continue the legacy.
At age 70, he’s had his fill of dealing with the administrative side of the business, and is looking forward to relinquishing day-to-day management to spend more time brainstorming new ideas, writing and lecturing. One of his email taglines reads: ‘Marketing has become the fine art of deception by omission’, so full-disclosure specs is another tradition of Jensen’s he willingly carried on. While he’s seen other companies ruined by cheapening the internals of its product, his 20 years of business with Radial’s President, Peter Janis, gives him confidence in the future. For now, Radial plans to operate Jensen as an independent supplier, of which Radial is a customer. So hopefully we’ll see the Eclipse transformers Radial has been using in its multi-channel variants of the JDI — a move necessitated by their popularity — replaced by the rightful component from Jensen.