“Hello John,
I just finished watching the Hegel Video. I thoroughly enjoyed it. You make it look so easy…but I assume it isn’t; making it look easy is very hard as far as my experience goes.
I very much appreciate your non-audiophile take on things and the gentle winks and nods. I hope your audience can enjoy many more of these very well made longer form films in the future.
Your approach brings a very much needed breath of fresh air to audio gear coverage.
So a big thank you and kind regards,
Jacques V.”
“John – great Hegel H590 video. You have a real talent of making the technical aspects of hi-fi more relatable and understandable and wrapping your thoughts in an attractive, entertaining package.
One query: the internally-linked video to Bent Holter explaining Hegel’s Sound Engine left me a little confused. I know you’re also a words guy. Any chance you might try to explain Bent’s work to those of us who flunked Math in high school and/or to those of us who can’t stand to watch a 30-minute iPhone video?
Yours – Hanson T.”
I’ve sat in front of Bent Holter’s SoundEngine2 presentation twice this year. The first time was a solo visit in February to get a feel for what Hegel do, how they work, what makes them tick and what makes them different. The second was a German press event held in August to formally launch both the KEF’s 2018 R Series and the H590, which we learn began its life as a power amplifier before morphing into a powerhouse Class A/B integrated with internal DAC and streamer.
Holter had already given us a brief outline of the H590 from the floor of Munich High-End in May but four months later at the Hegel HQ presser in Oslo – and a short distance from the man’s quad of desktop monitors (see above) – he once again detailed the SoundEngine recipe with a Flipboard and a marker pen. This time, I was ready with smartphone and arm stamina to film the whole thing. (I agree, the video quality is basic at best).
Holter begins his explanation by asking: what’s ‘wrong’ with normal amplifiers? The answer according to this Norwegian is the negative feedback used to correct distortions introduced by each ‘local’ amplification stage or to the entire ‘global’ circuit — a feedback loop in which an output is fed back to its input for differencing and the resulting distortion wiped away.
If this sounds gloriously simple in theory, Holter is quick to burst our bubble with a dose of reality: transistors don’t work instantaneously. Their operation introduces a small time delay. By the time the output is fed back to the input, the input has already moved on, if ever so slightly, with this feedback error correction introducing phase errors. The more stages in the amplification circuit, the longer the time delay and the greater the potential for phase errors.
Furthermore, according to Holter, a negative feedback-loaded circuit demands high-frequency gain be reduced so that transistor time delay doesn’t cause those high frequencies to oscillate.
And if speeding up a feedBACK loop isn’t easy, how about slowing down a feedFORWARD? That’s the premise of Hegel’s Soundengine circuitry, now in its second generation. A copy of each amplification stage’s input is peeled off to an “analogue computer” where it is continually subtracted from the output to isolate a real-time distortion signal. This distortion signal is then fed into a second “analogue computer” which inverts it before returning it to the amplification stage’s original output to cancel out distortion. But not always. Only when the distortion crosses a specific threshold.
But what about transistor time delay? As the music signal and its associated distortion fluctuates, Holter’s Soundengine 2 correction circuit is itself time delayed to precisely align with the amplifier’s own output delay and is, therefore, according to its designer, reliable all the way up to 300kHz and will reduce dynamic distortion by a factor of 30 to push damping factor into the thousands. Such are the benefits of feeding forward.
The Soundengine 2 can also be applied to an amplifier’s final output stage as it talks to a loudspeaker. In its sights is crossover distortion: the shortfall or overrun of transistor A, charged with handling the upper half of the music signal (the push), handing off to transistor B, charged with the lower half of the music signal (the pull). Holter asserts that crossover distortion is what lends Class A/B amplifiers their ‘transistor sound’.
None too dissimilar to the baton handoff in an athletics relay, if transistor A hands-off too soon we get a tiny horizontal pause in the output signal before B starts its run. Turn off too late and transistor A will overrun vertically into transistor B’s domain. B can also turn off/on too early or too late to inject small vertical and horizontal lines into the (music) signal.
According to Holter, these signal ‘flatlines’ generate high odd-order harmonic distortion — well known to be less agreeable to the human ear than even order.
Varying bias current run through the transistors regulates the possibility of over- or under-runs with an optimal bias point minimising the associated crossover distortion. But that optimal bias point becomes a moving target once we move from a symmetrical sine wave to a music signal that constantly calls for different amounts of power to keep transistor temperature in motion. As the transistor’s operating temperature varies, so too does its optimal bias point.
In other words, crossover distortion, like amplification stage distortion, is dynamic. It changes with the musical input and can be corrected for by the Soundengine 2’s real-time monitoring and adjustment.
Bent Holter might not be the only amplifier designer to use feed forward correction in an amplifier but he tells us that he has had his SoundEngine design patented since the early noughties.
Further information: Hegel Music Systems