Designing the ultimate monolithic amplifier

Fifteen years ago, my reference amplifier was a current feedback power amp, with 10 W @ 8 Ohm operating in A class, 25 W max, with fairly simple and clean topology (diamond input, another pair of voltage transistors, and output current pair), and no less than 200.000 uF supply capacity. Basically this was the old Japan originating and Jean Hiraga promoted concept from late seventies, but there was no real contender for years. Up to the point I discovered National power opamps. Right from the first moment, even if they missed the scale of previous reference, they were more than seductive. Something about them was just right.

Soon I put more effort in order to bring sonically missing parts of my initial opamp setup, by use of regulated supply and input buffer. I published the sum of these projects – recently kindly hosted by Nick Whetstone’s Decibel Dungeon – and from today’s perspective I can be still proud about the response and interest they brought.  It wasn’t only about the number of successfully built units in DIY circles, which was huge, but also about the way it triggered experience sharing (as opposed to commonly widespread opinions sharing), and about later massive popularity of solutions recommended there.

My later developments stayed offline, however things with this amp did continue to move, and concept was indeed capable to move forward, so I kept tweaking and checking different options. Then, several years later, I started working with Gramofone on capital GF01 amplifier design. It was the first commercial implementation of accumulated experience, and it was real cost-no-object product, both sonically and visually. Audial Model A came as more practical, one box embodiment of the same concept, in a way that fits Audial line of products.

For all this time chip based amplifier remained my main amplifier. And in its current shape I do consider it perfect, in the most practical way.

Now, what is technically plausible explanation on monolithic amplifier qualities?

With monolithic opamps, what we practically deal with is the feedback configuration, impedances, supply, layout, passive parts, and combine them mutually. It is a lot of stuff, and all these parts of design contribute significantly to the final result. Most of traditional amplifier designs were however totally disregarding these parts, narrowing electronic design down to a circuit design, and dealing only partially with passive parts and even more rarely with layout, and applying “that’s what everyone does” solutions elsewhere.

The fun part is that, even the circuits used by huge majority of audio industry were applying actually the same basic opamp scheme: differential input stage, voltage stage, output buffer. Happily, recent generation of power amp chips solved really nicely most of traditional issues of such circuit design. Modern power opamp chips are offering performance that does not only match traditional discrete designs, but exceeds them notably in most major areas, such as distortion and bandwidth, by utilizing advantage of on-chip transistors matching, and of course superiorly small circuit layout. On the other side, thermal distortion, which is known issue with monolithic implementations, is safely kept out of critical zone. Consequently, astute design with such a great integrated circuit will bring today’s state of the art in sonic performance.

As for me personally, back when we were working on GF01 I stopped linking any excuses to the chip. Managing to get its qualities out of it, and to get them all at once, was the real task. Fortunately, and I dare to say that, we managed.

And I can name some things that were important:

●  All supply types were checked for their actual sonic properties, with this amplifier. All the final parts of supply, as well as all the other parts, are tweaked on ear, during very long time.

●  Overall layout experimenting with different physical setups. Opamp layout remains tight, but paths remain well separated where they have to be separated.

●  Material quality is high, and premium parts include also some custom manufacturing, normally seen only in the extremely expensive products. Of course, it takes to know the parts sonic properties.

●  We broke some conventional audio rules, to come with the best available volume control. Low impedance with it is a must, and no compromises are suggested here, as the best things start to happen below 1 kOhm. This does require good driving capable source, but unless it is a tube source, you should ask yourself why does it can’t possibly provide less than 4 mA (!), to drive its 2 V RMS output signal or so into 750 Ohm load (?).

Of course, I can not claim that no part of this design will be ever replaced, or it can not get better, as there is simply no general logical basis for such claims ever. But its thrilling performance is beyond doubt, and I fail to find any shortcoming in it.

Regarding the concept itself, it will be hardly beaten by the other voltage feedback design. To improve upon the concept itself, we might look for monolithic implementation of current feedback power amp, or better, power amp equivalent of AD844, with possibility to use it open loop, when possible. I personally am waiting for it for more than decade, however it is still not likely to appear any time soon, unless we design it and make ourselves. Either way, something like that would be quite an undertake.



Related topics:

Sweet spot of amplifier power

Model A, one year later: Binding posts

Model A, one year later: Power

Model A volume control