133 Dalmeny Drive, Prestons NSW 2170, Australia +612-96074650 & 0412-203382

  JLTi Oppo Player        CD & DVD Player Upgrades        JLTi Tube Amps   

TUBES MEETS GAINCLONE TECHNOLOGY

AN ESSAY

Introduction

First, some years ago an English Hi-Fi magazine reviewed a locally (UK) made low power amplifier (I think it was 15 Watts) powered by batteries. This was not a cheap design, it sold for £1500. Internally there were only two devices, one monolithic power amp chip per channel (from here on will be called the Gainclone IC). Now there has been amplifiers, especially integrated types from oriental sources, that have used IC power amp modules. But what made this different was the rave review, it was to all who heard it, evidently a real surprise packet.

Again, in more recent times, an unusual small scale specialist Japanese manufacturer also brought out a product using the exact same IC power chip. But this time it was presented as a one only input integrated amplifier. This was done by adding a 12 position attenuator ( 24 position would have been better). The asking price for this combo including external power supply (not battery operated) was then US$2700, but now US$3300.

It is apparent these products were aimed at the hard core audiophile fraternity, especially for those who may also be interested (even intrigued) by possible alternatives to SE Triodes. These products were priced accordingly.

The pricing of these products were not based on the basic costs of parts, but rather that the sound quality made it possible to exploit such a situation. Some have called it a rip-off. Those are harsh words.

Could someone bring out a reasonably priced version, not just a copy but an attempt to further the art?

Gainclone Circuitry

The story does not end here. This essay is also about tubes and how does this fit into this scheme? Bear with me.

Here is a basic recommended circuit of the Gainclone chip:

I would now like to analyse the above. The circuit uses conventional shunt feedback. This has some tricky drawbacks. For one, notice C2, this has to be an electrolytic yucky cap, and ideally a larger value than shown. Why is C2 used? This is part of keeping the DC Off-Set into the speaker load down to a few mV. C1 also decouples non-inverted input (+ pin 1) to DC. There are good reason why R2 and R4 has similar values, the input current on both inputs are almost the same (I measured it) and consequently the difference is not amplified by the gain set by R4 & R3. So in this way DC Off-Set is indeed very low.

How can we get rid of C2? We can by using series feedback!

Let us look at the alternative, the inverted variation:

Looks a wee bit different (thanks to Thorsten Loesch for this circuit).  Now our signal is going to the inverted input (- pin 2) but it itself has also become part of the feedback path. We have also eliminated the Electrolytic cap. The 2.2uF input film cap is the minimum value and can even be made larger. Again notice both 220K caps, the fact that they are identical,  along with our input cap, ensures very low DC-Offset. The 0.1uF is there to ensure our non-inverted input (+) is AC signal grounded.

So what are the pros and cons?

It has already been established that series feedback (inverted) has the potential to sound better than equivalent shunt version (non-inverted) - the reasons for this was established by Lindsay Hood and others. Yet, for its superior qualities, it is  rarely used, because the low value input resistor defines the input impedance. So by nature the input impedance is low. Hence feedback values needs to be scaled up much higher, typically ten times as much. This makes for pour noise figures in low level circuits and that is why series feedback is rarely used in circuits for mics or phono cartridges. But in power amps, this is not problematic where signal to noise is good to start with, due much higher signal levels.

There is also the question of the source (preamp or CD in some cases). This now becomes, arguably I know, part of the series feedback loop, so it has to be Lo-Z. This means that using 1 Meter interconnects the feedback loop is now more than 1 Meter long... more about this later.

The other con is that, in practice, the series feedback amplifier is far more tricky, the layout, the earth arrangement and wiring become even more critical.

So what is it about the Gainclone monolithic chip that makes it special. Let us take a look at its internal circuitry. This is the simplified circuit:

Do we note anything? Starting at the inputs, emitter followers with individual current sources, then a differential stage also current sourced, indeed everything up to the final output stage is current sourced! In fact you can count no less than six of them. For those not in the know, this generally implies good things and also that all stages up to the output stage are in Class A.

This helps to explain why the Power Supply Rejection (PSR) is typically 120dB. That is a ratio of a million to one! Finally the output stage itself is not fully Complimentary, but Quasi! This avoids the use of PNP bi-polar devices. There are still those that say NPN/PNP complimentary 'power pairs' are never truly such. They say that the NPN will always be superior. No such problem here. So our internal view of the Gainclone IC shows some significant reasons why it has the potential to sound as good as it can. Many discrete circuits are less interesting than what is on show here.

So What About Tubes & Gainclones Together?

Now we come to the crux of the matter.

This is the current JLTi Mark 2, simplified circuit.

Our series feedback situation needs a defined source impedance (Lo-Z), what better than a Tube Buffer which has a stable 200 Ohm output impedance.

Take a closer look at this basic schematic, the right half should look familiar. This part is very similar to our earlier circuit but with added features. The series feedback is actually split after what is clearly the coupling cap (on the output of Tube Buffer). The Gainclone Philosophy is to keep things very compact, especially the power supply reservoir caps to be straight on the plus and minus of the Gainclone chip. This also applies to the feedback loop. This is easy when it comes to shunt feedback but flies out of ball park when applying series feedback. So we need a buffer for the source signal (our glorious music) so that we can at least keep it down to inches and not feet or metres.

What better than a Tube Buffer, and not just any kind of buffer, but a Unity Gain Constant Voltage and Constant Current Tube Buffer (remember what we said about current sources, here the same good things apply) .

But we need two other things accomplished as well. By splitting the feedback path into two, we can keep the series feedback extremely short as per original Gainclone Philosophy. It is above the normal audio bandwidth that we can accomplish this while at the same introduce a passive single order low pass filter function to prevent potential slew rate problems. The above circuit accomplishes these functions.

The sharp eye will also observe that the input to the Tube Buffer is direct coupled. This keeps the basic circuit simple as it is basically self biasing. The complexity then comes from the fact that the circuit is powered by split rails, as the Gainclone chip is as well. There are no high tension applied and thus no real danger of electric shocks.

Because of the split rails, the DC Off-set of the Tube Buffer is only around one volt. The coupling cap does not need high voltage rating. I use a parallel 100V rated non-inductive Polystyrene Foil by-pass cap. These caps would be horrendously expensive if required at high voltages normally used in tube circuits.

Something else may have caught your eye as well: What is that bi-polar transistor doing on top of what is a Triode functioning as a Cathode Follower? This is a bootstrapped device, this applies the same voltage swing to the Anode as the Cathode. It has been noted by a number of Triode aficionados that Cathode Followers do not sound as good as they should. Yes, they have potential low output impedance but they miss out of some of that Triode magic. The reason is that Cathode Followers expends their gain by one hundred percent local feedback - this total feedback situation is not desirable. But what if we can keep the voltage constant across the Triode itself, and also keep the current constant as well, then we accomplish another very neat trick: The Triode does not see the AC signal either in the form of voltage or current, hence the hundred percent local feedback situation is avoided, and yes indeed, it sounds that way. We still get the Lo-Z output impedance but without the sonic penalty. Not only does it sound great but also SPICE simulations show extremely low distortion. Sometimes the two can, despite the cynicism of some, correlate nicely.

The acronym we use to describe this follower is SLCF - short for Super-Linear Cathode Follower... and it is a Vacuum State Electronics exclusive.

Current Sources Galore

The JLTi Integrated amplifier uses no less than sixteen Current Sources, two in the Tube Buffer, two in the split rail power supplies  for the Tube Buffer, finally 12 (six per channel) inside the Gainclone ICs. A record? Not exactly but for a supposedly budget product (but hard  a budget performance), perhaps yes. For a more detailed discussion re Current Sources, read Allen Wright's The Tube Preamp CookBook.

FOR MORE INFORMATION AND MEASUREMENTS, GOTO:

Gainclone & Sound Quality - Gazing Into The Dark

Home Page