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

[EUVL]

This is a power JFET version of the Taylor follower, with the JFET biased in Triode mode. 5 transistors, because we need 2 extra for the triode mode cascode. Great sound, according to the few who have built it.

Some other Source Follower COnfigurations

And you can increase the bias to 1.3A as Nelson Pass did in his Zen V9 for use as power amp. But you probably need to change the values for the Taylor branch to adapt to the increased bias and lower Zout, etc.


Patrick

[Elvee]

Call me a wet blanket if you like, but the improvement over a plain current source isn't that huge: about 2.1ppm against 5ppm.

BTW, a number of HP instruments of the seventies used a similar scheme for buffers stages.

[kenpeter]

Taylor only hides emitter distortion, does nothing to fix the Early effect.
But every time I try to wrap Taylor around a cascode... LTSpice breaks
into wild oscillations.

[Bigun]

Originally Posted by kenpeter

If you can put one of these distortionless A's in parallel with an AB, B, or D....
Either open loop, or as a current dumper... Yeah, then maybe the wastefully
biased Taylor stage wouldn't need so much Iq as to carry the full load by itself.

-------------------

What's not to understand about JLH overbias in the center? The sum of two
base currents is held to a constant under all conditions. But the base currents
needed for full output current swing are much higher than required for normal
class A quiescence. JLH has nowhere to dump this excess of quiescent base
current when it isn't needed, so output transistors idle hotter than necessary...

In my most recent schematic, Q4 is the shunt that bleeds off those excess
drive currents. Is also why it doesn't need a constant current source, drive
currents are shunt regulated instead....

OOpss.. I got it backwards. Q4 is a modulated current source, not a shunt...
Been looking at this thing standing on my head too much.

I don't really understand still - you need to give me a better explanation.

First off, in terms of religious convictions, I don't think of BJTs as current devices. They are voltage controlled devices. The have a base current, but that's a feature of their input impedance. In reality the base currents won't match because the input impedances will be different, the hfe will not be exactly the same and temperatures may not the the same.

If the currents through both devices total up to the same then the voltage drop across base-emitter is the same for both devices (roughly). This isn't a symmetrical circuit so one device is the 'master' and the other is the 'slave'. The 'master' device sets the q-current through the output pair and it's base is driven from the collector of the JLH phase splitting device. The other output device is the 'slave' and it Follows the output, roughly two 'Vbe drops' below the drive signal (one from the JLH phase splitting device, the other the output device it'self).

You'll need to help me here.

Lastly, I don't see why this is bad, why is it inefficient "in the middle", it's Class A afterall

[Elvee]

Originally Posted by kenpeter

Taylor only hides emitter distortion, does nothing to fix the Early effect.
But every time I try to wrap Taylor around a cascode... LTSpice breaks
into wild oscillations.

I don't have an oscillation problem. However, it doesn't bring any improvement either: distortion stays stuck at 2.2ppm.

Bootstrapping the collector of Q1 looks much more promising: distortion now drops to 70ppb (I hope in a month's time, we'll be talking in ppt).

[mfc]

Hi Elvee,

Change R8 from 12 to ~17.5 and it will be in the 10ppb neighborhood.

Mike

[kenpeter]

Originally Posted by Elvee

I don't have an oscillation problem. However, it doesn't bring any improvement either: distortion stays stuck at 2.2ppm.

Bootstrapping the collector of Q1 looks much more promising: distortion now drops to 70ppb (I hope in a month's time, we'll be talking in ppt).

Q1 was the Early I meant to cascode away, the comparator is what matters.
I don't know how much matters having another cascode in the taylor circuit?

I abused a JFET rather than C2 to lift the base of the cascoding transistor in
my attempts that broke into oscillation.

[kenpeter]

Originally Posted by Bigun

I don't really understand still - you need to give me a better explanation. You'll need to help me here.

Lastly, I don't see why this is bad, why is it inefficient "in the middle", it's Class A afterall

Yes, JLH is an A. But lets be clear, its not your "normal" class A.
Normal A holds the output current sum a constant under all conditions.

JLH quiescent is HIGHER than purely linear class A. Because the bases require
a non-linear current input for a linear current output. But total base current
is held constant, at value required to swing peaks. The result is like HyperA,
but JLH nonlinearity bent entirely the wrong way (for idle power efficiency).
The sum of output currents is NOT a constant, but higher in the middle.

As long as you got Watts to burn, and don't mind efficiency worse than A...
Its not a problem.

[Bigun]

I think I'm going to have study it a bit more. I understood only as far as the case that a SE output with a CCS load will keep a constant current through the output (nice for the psu) but the difference between the JLH A and a 'normal' A is still needing some thought.

Revintage - your ideas are the final shove needed to get me going with my own project. I've opened up a separate thread not to abuse this one because it's focus is on Ken's Diamond (like that name).

Looking at Ken's diamond I'm interpreting it differently now. It looks like the JLH follower where the phase splitting device has a modulated emitter load - flavours of Aleph distortion reduction perhaps. I may look at my own JLH from this perspective to see if anything makes sense ?

[tiefbassuebertr]

Originally Posted by HarryHaller

http://www.geocities.com/sonnya5000/basicbuffer.html
Good article. Check it out.
H.H.

now the currently URL:
basicbuffer.page