Simple Class A -again

[john_ellis]

Hi
Here is a suggestion for a Class A circuit which uses Allison biasing for the output, a differential pair input, potentiometers to set the output level to zero for open and short input, so that it is fully DC coupled (no "nasty" capacitors or inductors) and exhibits the usual decreasing harmonics you expect from a Class A. Distortion is around .1% at 16W , but proportionally less at lower powers, and there again the main distortion component is only 2nd harmonic. I have use transistors which offer high frequency response and gain.

I hope I have captured the essence of simplicity which John Linsley Hood would have appreciated while maintaining a good response.

Haven't built this yet, so cannot confirm that it would be stable with all loads. May need output inductor/capacitor network with a real speaker since the feedback capacitor (15 pF) is taken directly from the output. May need input capacitor to ground if input impedances are higher than "low".

May need some decoupling between the input stages and power rails for hum reeduction (Class A amps tend to load the power supply and give more background hum than Class B - although in any good circuit the input stages should reject hum. Class B amps sometimes give higher hum with higher output levels which may not be apparent until measured...

John

[Jay]

What are the pros and conts (preferably sound-wise) about this Allison circuit?

[lineup]

Because VAS Q5 BD139 draws so much base current
there is a 5% mismatch in input transistors current.
Should be less than 1%

[alexcp]

You could improve the distortion performance if you make Q5 a darlington - a long tailed pair with a current mirror has high output impedance, and a BD139 with a 1 ohm in its emitter would overload it. Also, you do need decoupling for both the negative rail and for the current sources - your VAS is referenced to the negative rail, and the current through D1-D4 depends on the positive rail voltage. To verify the need for decoupling in your simulation, add equivalent series resistance to your power supply voltage sources and re-measure the distortion.

Oops, lineup was faster!

[lineup]

I was thinking a sziklai darlington for Q5.
For example a 550C at 2 mA + BD140 combo.
This makes equal currents in input.

[lineup]

Here is a fix.
Put one resistor ~3k across Q4.
This will equal currents in input.

[john_ellis]

Hi all

All good points. THe reason I did not use a Darlington VAS was to keep the schematic simple. If you use a D-VAS then the distortion should be very low. But another concern is with stability. The additional transistor may change the phase shift so much that the basic feedback cap no longer works well.

As I indicated this was to try to keep the schematic simple, but I like lineup's solution to balance the currents. ALternatively the LTP currents might be better balanced with, say, a 10 ohm pot.

The 1-ohm emitter resistor in the BD139 was an attempt to linearise the Vbe. Actually, the BD139 does not "overload" the current mirror because current driving it is exactly what is required to eliminate the non-linear Vbe. A low impedance base actually works better. What might occur at higher frequencies is if the LTP current is insufficient to feed the base (Qb or Cbe, whichever you prefer). But this does not seem to be an issue with ~ 4 mA in each of the LTP transistors.

It might be better to use a larger emitter resistor just to be able to use a pair of diodes or a transistor to limit VAS current in the event of a short circuit.
John

John

[lineup]

I like the output. Is it called Allison?
I have set it up in spice. Very clever way for simple Class A output.

The amplifier is FAST. Very fast. I think it is due to the high currents in each stage.
The gain is modest and distortion reasonably low.

It should be easy and possible to build.

[lineup]

You can see my SPICE of this amplifier.

There are some small changes:

- The offset compensation uses higher value potentiometer
not to upset the gain divider.
Also reference diodes only connected to positive current side

- Use of 560C and 550C in input

- Omitted emitter resistors in mirror and BD139

- R13, 2.8K, to equalize currents in input

[john_ellis]

Hi Lineup
Thanks for your comments. Your version is slightly simpler, but shows the good performance that can be achieved with this design.
I would recommend keeping the emitter resistors in the mirror pair because mirrors are very sensitive to temperature differences. They can work well at low currents but in this configuration (in common with many versions of this circuit) the collector voltages are not identical, so the power levels are different.
I agree that you do not need to use two offset balance controls if you only need AC coupling. I can't hear DC either but I thought it useful not to use an electrolytic which might distort.
The main reason for a resistor in the BD139 is for short circuit protection, but this would need another transistor or pair of diodes as well.
Hope to build one soon.
Output transistors need quite a substantial heatsink!

John