four transistor emitter follower (diamond buffer) as power output stage?

The good old four transistor unity-gain buffer which consists of a complementary emitter follower, another emitter follower pair that replaces the bias diodes and two current sources is very popular in complementary integrated op amp designs as well as some discrete designs. For a schemetic, take a look at Sonny's page:

He did some nice simulations about distortion and how to improve it for low impedance loads.

What I wonder about: why is this nice topology apparently not used for power output stages? Is it because it is too difficult to keep the bias constant?




2001-02-04 4:23 am
USA add capacitor to switch to class AB a high power add DC servo to run without feedback
--- Andrea Ciuffoli <[email protected]> wrote:
> Andrea Ciuffoli
> via degli Zingari 50
> 00184 Rome
> thank you
> dennis kleitsch wrote:
> > djk:Sorry,I don't have access to a scanner at
> work.If
> > your library doesn't have Wireless World you can
> send
> > me your postal address and I can mail you a
> > photocopy.It was in October 1976 Wireless World
> page
> > 74
> >
> > --- Andrea Ciuffoli <[email protected]> wrote:
> > > please, send me a san image
> > >
> > > dennis kleitsch wrote:
> > >
> > > > djk: It was published in Wireless World
> magazine
> > > many
> > > > years ago in the Circuit Ideas section.It had
> a
> > > single
> > > > transistor running class A for voltage gain as
> a
> > > front
> > > > end and had no feedback.I'll see if I can find
> and
> > > > copy it for you if you are interested.It would
> be
> > > easy
> > > > to add a DC servo to the current source
> setting
> > > > resistors.
Hi, Wrenchone,

Do you have any idea why this topology never used in commercial power amp output stage? (I never saw one)

From your experience, what is the pro/cons of using this diamond buffer topology compared with EF+VBE multiplier biasing scheme?

How much bias do you use in final stage? ClassA? low biased ClassAB?

And how is the sound?

Is this topology has a "Non-turnoff" properties (like Blomley)?
lumanauw said:
How does a diamond buffer looks like if used for output stage? With many parrarel output transistors? With 0.1ohm RE degeneration? How to make it work practically?

Hi lumanauw,

I've attached a schematic for an amplifier I designed to drive a 1 ohm ribbon speaker where I tried to adapt the diamond buffer topology to a high current bipolar output. Hope this example can give you some ideas for experiments.


  • ribbon_amp.jpg
    66.4 KB · Views: 3,810
Hi, Linesource,

Thanks for the schematic. Very nice design. How is the sound?
With diamond buffer, I wanted to eliminate the need for VBE multiplier, but your example uses both of them :D Without VBE multiplier, how will it looks like?

Hi, PMA,

Nice to see you again. Your example works in full classA. I'm thinking about using it in classAB. Will diamond buffer suitable in class AB or not?
Here's what I used in the output stage of my amp. The output FETs are biased by the voltage drop of the driver pair and the resistors in series with the driver current sources. One can tweak the ratio of the series resistors and the output source resistors to get the desired bias current. The drivers go on the same heat sinks as the output devices for thermal tracking. I generally use driver FETs that are one to two sizes smaller than the output FETs. That way, I can run them with a smaller quiescent current than the output devices and still get sufficient voltage drop for proper bias.

The big disadvantage of this setup is that it lops about 10V off the output voltage swing on either side. I made an amp using a complementary JFET differential folded cascode input stage to drive the diamond buffer. Even with only +/-30V on the rails it sounded like a pretty big amp. This beast is currently in pieces on my work bench getting outfitted with a new toroidal transformer for higher voltage rails (+/-40V), bigger drivers and outputs, a better heatsink and thermal interface, and more voltage across the cascode devices. I'm interested to hear how it will sound when hooked up to my new JFET open loop preamp...

One more thing - Audio magazine reviewed a Mitsubishi amplifier in the late 70's that had what they (Mitsubishi) called a diamond output stage. I'd have to dig up the article (I have no idea at all where it is) to see if it's the same thing we're discussing.


  • diamond.gif
    5.2 KB · Views: 3,562
Hi, Wrenchone,

How much bias do you put in your amp? Will this DB works properly if the bias in the final stage only about 25mA?

when hooked up to my new JFET open loop preamp...
:D You can read my mind. I'm thinking about using this output stage with no-global feedback frontend.

I think of 2 candidates that can do this. One is this DB, and the other is Hawksford EC or NP-PMA by PMA. I think both can be driven without global NFB, but I still don't know which one is better.

Low biased classAB EF+VBEmultiplier cannot perform better than these 2 for this purpose.
This is an interesting variation of diamond buffer (DB) like I attached here. It is submitted by PPL. What is the purpose of R11,12,13,14? Is this 4 resistors making Error Correction or non-turn-off or what do they do?

From that PDF, what happens if I remove R11,12,13,14 but connect C of Q6 to E of Q7 (and C of Q5 to E of Q8) instead to rails. Will this make Q7+Q8 will not turnoff? But will Q5+Q6 saturate because of low VCE?

Hi, PMA,

You've been into these topic for a long time. How is the comparison of DB with your NP-PMA? PPL said something about these here How is the comparison between these 2 according to you?


  • errorcorrction.pdf
    31.6 KB · Views: 1,032
I ran my diamond buffer at about 200ma bias on the output stage, and 20-50ma on the driver stage, depending on device size. If you really want to use 25ma in the output stage, you will need to use smaller MOSFETs or bipolars. Actually, the diamond output stage may be overkill. This afternoon, I simulated a single ended current source loaded source follower using a pair of IRF510s. Bias current was set at about 20ma, supply was +30V. With a 300 ohm load, distortion at 1V out was around 0.016%, and 0.1% for 5V out. The output of such a beast needs to be capacitor coupled, but this may not be a severe restriction. With an open loop amplifier you would probably need a capacitor couipled output anyway.
The diamond buffer was part of a closed loop amplifier. It's the preamp that's open loop.