Home Amp/Preamp Asylum

Looking for a new Amp or Preamp? If you're after tubes, post over here.

RE: Ayre's Diamond output stage - STUPIDLY LONG!!!

The Diamond circuit has been around for a long time. Specifically, it was developed and patented in the US in the 1960s, but for use in digital switching circuits. I could never figure out why it was called a "Diamond" circuit, as it really looks nothing at all like a diamond the way it is normally drawn. For example, here is how they show it in an app note from Burr-Brown on a now discontinued part:





There are some extra parts shown in here, with the current sources and some other bits, but there isn't anything particularly "diamond-like" about it. I include this schematic, because the datasheet for this part (OPA660) clearly describes the operation of the circuit and repeatedly refers to it as a "diamond" buffer. If you take away all of the extra parts and strip the circuit down to its bare essentials, it looks like this:





Now it is much easier to understand. It is simply a pair of complementary cross-coupled emitter-followers. But while now it is much easier to understand the operation, it still makes no sense to me why in the world it would be called a "diamond" buffer. But I am an amateur historian of audio technology and audio companies. So I kept digging...

Then I found the original patent for the circuit. It was invented by a professor at MIT, Richard H. Baker, who is best known for the "Baker clamp", a circuit that prevents problems when a transistor amplifier clips. (Although Baker patented it and his name is associated with it, the circuit was actually known for several years before he applied for the patent.)

Back to the "Diamond" buffer. Professor Baker developed another circuit which he called "The "Diamond Circuit" in an internal publication for MIT, but on the patent he called it "A Gateable Bridge Network Having Power Gain", with some properties that mad it sort of a hybrid between a digital circuit and an analog circuit. In the patent application, Baker drew it rather differently than it is normally drawn and one can now *easily* see why he called it a "Diamond Circuit", especially in Figure 2:





Figure 2 is simply a rearranged version of the simplified circuit above, but when drawn in this manner it is obvious to see why Baker called it the "Diamond Circuit". In actual practice it was most useful as a "buffer" -- that is, a circuit with no voltage gain but a high current gain, which is precisely what is needed to drive a low impedance load. Hence it has become popularly known as a "Diamond buffer".

It is commonly used in the output stage of "buffer" IC's that have unity voltage gain, but boost the current to drive heavier loads. Sometimes it is also used with discrete circuits, where it requires careful matching of the opposite polarities of parts (NPN and PNP) to minimize the offset voltages.

It is typically only used in low power stages, such as preamplifiers outputs or things of that nature. After playing around with it for quite a while I came to understand why this is so. To work properly, the driver stage needs to use an identical transistor to the output stage.


An output transistor is a big beast that typically draws a lot of current at idle and also has large capacitances between the terminals. So this means that the driver stage also has to use a "big beast" and this stage also needs to draw a lot of current and is fairly difficult to drive due to the large capacitances.

So using a diamond buffer as an output stage has two fairly significant disadvantages:

1) It draws a lot of current and this adds to the total power consumption (from the wall outlet) of the amplifier, as well as requires additional heatsinking to dissipate the power drawn by the driver stage.

2) Since the input transistors of the diamond buffer must be identical to the output transistors of the diamond buffer, they are large and rather difficult to drive. This means that one cannot drive them directly with the typical voltage amplifier used to provide the gain in a power amplifier circuit.

To address the first problem means that the amplifier needs to be larger and heavier, with a more powerful power supply. To address the second problem means that one needs to use a more powerful driving source than the typical amplifier uses.

The large majority of transistor power amplifiers use a "front end" circuit that provides the voltage gain, and then an "output stage" with no voltage gain, but that provides the current gain required to drive the low impedance of the loudspeakers. (One version of the Wilson WATT/Puppy dipped to around 0.3 ohms in the upper midrange. While this is unusual, it is important that a high-quality power amp won't be bothered by loads below 2 ohms or even 1 ohm. Most electrostatic loudspeakers drop below 1 ohm at frequencies above 20 kHz.)

The typical solid-state power amplifier only uses two transistors signal path of the output stage. (There may be more in parallel to provide more current *handling* capability, but this does NOT increase the current gain.) A really good output transistor will have a current gain of around 100x, and the same is true of a driver transistor. This means that a two-transistor output stage will have a current gain of 100 x 100 = 10,000x. So if you connect a 4 ohm loudspeaker, the load on the "front end" (voltage gain stage) will be 40,000 ohms.

While this sounds like a lot, it really isn't. It is a low enough load that it puts a strain on the front end and increases the distortion significantly. The solution that 99.9% of all designer use is to add loop negative feedback. Works great on paper, but doesn't sound so great.

At Ayre, we have always used *three* transistors in series in our output stage. So if the third transistor also has a current gain of 100x, now the total current gain is 100 x 100 x 100 = 1,000,000x. Now if you connect the 4 ohm speaker, it only puts a load of 4 megohms on the front end and the distortion drops dramatically without having to use negative feedback. These extra transistors only add a dollar or so to the Bill of Material (BOM) so unless we are talking ultra-cheap junk audio equipment, this is completely negligible.

Back to the Diamond circuit. If it has these disadvantages, why in the world would anyone use it? The short answer is it sounds better.

The long answer is that I really don't know why, which doesn't help much as that's not much of a story to tell, and writers like to tell stories and people like to listen to stories. At this point the best that I can do is to speculate. So here goes nothing....

~~~~~~~~~~~

Before transistors came along (in the vacuum tube era) there were two types of power amplifier circuits in common use. One was the single-ended circuit and the other was the push-pull circuit. In the single-ended circuit, there is one tube (or possibly more, but they are simply connected in parallel to act as one larger tube), and this tube must be "on" all the time or the signal will be horribly distorted with clipping on either the top or the bottom (depending on the particulars of the design.

This limits the power to a few watts ("flea-power") unless the designer resorts to huge power tubes originally designed for radio station transmitting tubes. Even then it is hard to get more than 20 (or *maybe* 25) watts out of the amp and it draws a ton of juice from the wall, which is pumped out as heat int your listening room.

So when the push-pull output stage was invented, it allowed for MUCH more powerful amplifiers to be built very easily. Instead of a single 6L6 tube putting out perhaps 5 watts, a pair in push-pull could easily put out 30 watts. And when they developed bigger tubes like the KT-88 (6550 in the US), instead of a single tube putting out maybe 8 watts, a pair could put out 60 watts! This was just in time for stereo, when all of a sudden one needed TWO speakers and the small "acoustic suspension" designs took over the market place. They needed all that power, as they were so inefficient.

With a push-pull amplifier, the input signal is split into two halves with opposite polarities with a stage called a "phase-splitter". Then each output tube amplifies one half of the signal, and these two "half-signals" are recombined by the output transformer.

When solid-state came out, they were able (eventually) to make two devices that were opposite and (approximately) equal. So a designer could pull the same trick of splitting the signal into two parts, one for each output transistor, and then the output of each output transistor was combined back into a single output containing the entire signal without the need for an output transformer (which is big, heavy, and very expensive). This type of design is referred to as a "complementary output stage".

One can also do the same trick with transistors of the same polarity by adding an extra transistor to invert the drive signal to one of the output transistors, and this is called a "quasi-complementary output stage", but it doesn't change anything fundamental about the situation. Specifically, the input signal on ANY modern power amplifier is split into two parts, sent through an output stage that has a separate device for each half of the signal, and then the two half-output signals are recombined to make one final output signal. This is true whether it is tube or transistor, uses transformers or is or not, is truly complementary or quasi-complementary.

What is different about the diamond buffer from a typical output stage is that after the signal has been split into two halves to send to each output device, is that the two half signals are rejoined at ONE single point. There is NOTHING in between these two half signals. A conventional output stage always has *something* in between the two inputs of the two half-output stages before they are recombined. Typically it is a bias circuit of some sort that sets the idling current in the output devices. but there is *always* something there.

The diamond circuit is the only is commonly output stage in use that feeds the two half-output-stages from the same exact point. This can be most easily seen in the second of the three figures in the schematic diagrams above. In contrary, here is a typical two-transistor NON-diamond output stage. (Two transistors for each half of the signal, for a total of four transistors):





As you can see, there is *something* between the two half signals in the traditional output stage. Invariably it is a bias circuit that sets the amount of output current in the output transistors. And so the two halves of the signal going through the two halves of the output stage are slightly different somehow and when the recombine to recreate the entire signal out the output of the output stage, it simply doesn't sound as good as the Diamond circuit.

In a way, this entire argument makes no sense, because if you look at the simplified Diamond circuit in the second figure (near the top) again, there is no connection whatsoever between to two halves of the signal at the half-way point through the output stage. That is, at the input, the signals are identical as they are hard-wired together, and at the output the signals are identical as they are hard-wired together. But *in-between* the two transistors in series, there is absolutely no connection between the top half of the circuit and the bottom half of the circuit.

But that is the beauty, the mystery, and the art of audio circuit design. If we knew all of the answers to everything, then everything would sound identical and everything would sound perfect. There is still clearly more awaiting discovery here...

~~~~~~~~~~

The only other power amplifier that I have ever seen that uses the Diamond buffer for its output stage is the original DartZeel amplifier. But it works in a very different way than does that Ayre AX-5. Remember how we needed to have *three*stages in the output stage to properly isolate the loudspeaker from the front-end circuit, whereas the Diamond circuit has only two?

In the DartZeel, they solve this problem by giving some extra "heft" to the front end of the amplifier circuit by using three transistors with a feedback loop around them. This makes them "strong" enough to handle the load presented by only using two transistors in series in each half of the output stage.





In this simplified schematic from the DartZeel website, I have drawn a red box around the Diamond output stage (identified as the "Third Stage" at the bottom of the diagram) and two green boxes around the two half-stages that amplify the "bottom" and "top" halves of the signal, using a feedback loop around the three transistors in each state (identified as the "Second Stage" at the bottom of the diagram).

Of course at Ayre, we absolutely avoid the use of feedback in everything that we do. Feedback can only attempt to correct an error *after* it has happened. It works great on paper and can be mathematically shown to be sound (as long as certain rules or guidelines are followed). But to my ear, there is something more natural and musical about an amplifier that does not use any feedback whatsoever.

So we developed a special buffer stage to go in between the front end circuitry and the Diamond output stage that has a directly coupled input *and* a directly coupled output.


This post is made possible by the generous support of people like you and our sponsors:
  Atma-Sphere Music Systems, Inc.  


Follow Ups Full Thread
Follow Ups

FAQ

Post a Message!

Forgot Password?
Moniker (Username):
Password (Optional):
  Remember my Moniker & Password  (What's this?)    Eat Me
E-Mail (Optional):
Subject:
Message:   (Posts are subject to Content Rules)
Optional Link URL:
Optional Link Title:
Optional Image URL:
Upload Image:
E-mail Replies:  Automagically notify you when someone responds.