Very simple class B amplifier

[ingenieus]

Here is my suggestion for an amplifier that is as simple as possible, but no more than that. The main design objective was to get a low-priced amplifier with as few components as possible, while still getting good performance.

To this end, I went for integrated circuits and Darlington transistors. I even went as far as using the LM334 current source. A conventional current source has at least five components, maybe more. The design might not be the last word in state of the art and high performance, but it is as simple (and cheap) as I can make it.

LTspice has this amp giving at 1kHz:

• 16W into 4 ohms with a THD of 0.045%
• 32W into 8 ohms with a THD of 0.027%

The limiting factor is the power supply voltage of the NE5534 op-amp, which is + 22V. If more power is required two of these can be bridged. The generously dimensioned Darlington output transistors will handle the additional current required, but only into an 8 ohm load.

A 4 ohm load might be okay in a bridged configuration but each output transistor then has to source 7A of current. They have an absolute maximum rating of 10A, so we are sailing a little close to the wind here.

At these power levels, we might be able to do away with the Zobel network, or maybe just bits of it. Any thoughts on this would be appreciated.

There are various ways of getting more power out of this design, but then the cost and/or number of components goes up. The first way is to use a higher supply voltage and to drop the supply to the op-amp with Zener diodes. You then need an output stage with voltage gain. This can be achieved with complimentary feedback pairs and local feedback, However, all of the simplicity is gone.

Another way is use a high voltage opamp like the OPA551, which can handle a supply of + 40V. This will give about 70W of output power. However, these opamps are fairly expensive, which defeats the design objective. And for just a little more money you can use the LME49830 input stage from National Semiconductor.

Using an op-amp in a power amplifier seems to be a bit of a technological dead end, for the reasons mentioned above. But if you don't want a lot of power, it serves its purpose.

[Bigun]

I'd keep the Zobel, it's not about power but about avoiding parasitic oscillations of the output pair and it prevents r.f. entering the amp via the negative feedback loop.

Don't you want a capacitor in series with R1 to ensure gain rolls off at dc (better dc-offset stability) ?

[nigelwright7557]

Its not class B, its class AB due to the Vbe multiplier.

[Tekko]

You could add a VAS which would reduce the drive voltage needed from the opamp.

[Bigun]

Originally Posted by Tekko

You could add a VAS which would reduce the drive voltage needed from the opamp.

not allowed - you have to reduce parts count not increase

How about you ditch the constant current source with a resistor, replace the Vbe multiplier with a TL431 to maintain a constant voltage ?

[ingenieus]

Originally Posted by Bigun

not allowed - you have to reduce parts count not increase

How about you ditch the constant current source with a resistor, replace the Vbe multiplier with a to maintain a constant voltage ?

That's what I'm talking about - (nearly) ruthless simplification. It will reduce the parts count by one. The picture below (from the TL431 datasheet) shows how it is done. I will investigate further.

[ingenieus]

Originally Posted by Bigun

I'd keep the Zobel, it's not about power but about avoiding parasitic oscillations of the output pair and it prevents r.f. entering the amp via the negative feedback loop.

Don't you want a capacitor in series with R1 to ensure gain rolls off at dc (better dc-offset stability) ?

That additional capacitor will really have justify its existence. When I get round to building one, I will check it out. LTspice does not provide answers on this. It is certainly needed in a discrete long tail pair with not-quite-matched transistors, but my thoughts at this stage is that the NE5534 will have the DC offset thing well under control.

[lineup]

Nice to see you use NE5534 (compensated) and not NE5532.
Because NE5534 is the better one.
And ridiculously lowcost vs. performance.

[ingenieus]

Replacing the Vbe multiplier with a TL431 or LT1009 might not be a good idea. There will be no thermal coupling to the output transistors, as these voltage references are designed to maintain a constant voltage no matter what the temperature (see graph from LT431 datasheet below).

In a simulation it does work. I ran it with the LT1009 model included in LTspice. What happens when things get hot is another question, though. It seems like a good idea that will not work. The comprehensive TL431 application notes from National Semiconductor does not mention this seemingly obvious idea. Perhaps there is a message there. Any thoughts?

[wahab]

Once made tries with the 5534 as base for low power amps,
and did build some , using as well 5532s when factor form
is in the low size range.
In the simulators , assuming the model is accurate enough ,
it s amazing that ppm class THD is not out of reach , despite
the simplicity of the schematics.