Graham Slee Ultra-Linear Amplifier

How Ultra Linear makes better music...

Ultra-Linear: better music reproduction

Better music reproduction depends on wide open-loop bandwidth with stability but today's solid state op-amp and discrete transistor amplifiers don't have much open-loop bandwidth and rely purely on negative feedback to provide sufficient bandwidth for the musical signal.

Take a look back: if you're old enough to remember the Japanese (and some British) made amps of the 70s and 80s which boasted "flat: DC to light" but didn't sound very musical: their open loop bandwidth wasn't all that great or it wasn't very stable and they relied on negative feedback to rectify the matter.

The valve amp has made a come-back. The reason why is that (most) valve amps had much wider open-loop bandwidth before negative feedback was applied.

This works because of reduced phase modulation. An amplifier whose open-loop bandwidth is small will give rise to more phase modulation on application of negative feedback than one with a higher open-loop bandwidth.

Courtesy "The Art of Electronics" (2nd edition), Horowitz and Hill, Cambridge University Press, 1989

Harold S Black of Bell Laboratories invented negative feedback to reduce distortion in amplifiers in 1927. Valve amplifiers having transformer output stages could not use much negative feedback before transformer phase displacement caused them to oscillate (a term called motor-boating because of the resulting noise), and therefore only enough negative feedback is applied to reduce distortion to an acceptable level.

But with the introduction of solid state amplifiers which can directly drive heavier loads (pre-amplifiers included) transformers were not required and a lot more negative feedback could be applied. In the 60's and 70's specifications sold amplifiers, especially distortion specifications, with figures better than 0.01% often being quoted.

This led designers of audio op-amps as well as discrete stages to go all out for the maximum open-loop gain they could obtain, such that on the application of gross amounts of negative feedback the distortion would be so low as to be able to claim superior audio performance.

However, on listening, few people agree that such designs sound very good. But that has not stopped the relentless pursuit of ever increasing open-loop gain, and audio op-amp manufacturers continue to make great claims for this "solution".

A practice called "Chip Rolling" is quite common in DIY circles where such op-amps are tried over and over again, which should tell you that satisfaction is seldom achieved using such amplifiers. The result on a number of DIY forums has been much venting of frustration with little or no progress.

The f3dB point marked 0.707 on the above plot (the lower graph) shows bandwidth.

The further to the right it can go before rolling down indicates a wider bandwidth and the further to the left a narrower bandwidth. As open-loop gain goes up the further to the left goes the bandwidth - it decreases. Application of negative feedback increases bandwidth so specification wise you'd expect the sound to be OK, but the negative feedback is working against the phase shift shown in the upper of the two graphs.

All commercially available audio op-amp data sheets I have seen (and I have seen multitudes) have a very limited open-loop bandwidth, often just a few tens of hertz. Please also realise that most discrete stages are simply "copies" of audio op-amps done with separate transistors - or I should say that op-amps are actually "discrete circuits" that have been integrated onto one piece of silicon. In either respect the goal has been high open-loop gain which results in an open-loop bandwidth of only a few tens of hertz.

On the application of negative feedback to bring the gain to a usable level, any frequency above a few tens of hertz has to be "converted" from being 90 degrees out of phase back to being nearly in phase - and it does achieve this. But if you consider each frequency at a time, the minus 180 degree negative feedback converts a minus 90 degree signal - a signal that has already undergone phase shift - there is going to be phase modulation - a skipping and catch up scenario those who understand modems will be able to comprehend - in a nutshell, distortion. A kind of distortion that's very difficult to measure but exhibits itself as utter confusion in complex musical works - it is easily and often heard. If all you listen to is simple low level music then audio op-amps and similar discrete circuits will work for you, but large scale orchestral works and rock music don't stand a chance.

Now consider the open-loop bandwidth being a decade wider. Instead of being just a few tens of hertz it is now a few hundreds of hertz. This cannot be done at the same open-loop gain because nature has given us something called parasitic capacitance: the f3dB point can only go toward the right if the open-loop gain comes down proportionally. For the same resulting usable gain a smaller amount of negative feedback is being used, but the main point is that any phase modulation has been shifted higher in frequency (the phase shift depicted by the upper graph has moved a decade to the right). The mid frequencies where our hearing is most sensitive are now much less distorted and it can be clearly heard in a more organised - more musical portrayal of massed instruments and complex music.

The Ultra-Linear technology does exactly this. Depending on application the open-loop bandwidth can be taken even further to the right, especially where a lower gain is required in such as headphone amplifiers, line stages and power amplifiers.