♫♪ My little cheap Circlophone© ♫♪

[odysseybmx414]

Originally Posted by odysseybmx414

Here ya go:

http://dl.dropbox.com/u/5430178/circ...th%20Drill.pdf

http://dl.dropbox.com/u/5430178/circ...Silkscreen.pdf

Keep in mind I forgot the layer with the jumper wires. But you can look at wakabaki's posted thumbnail for those.

EDIT: Temporarily ignore those. I just noticed what I think is a big error.

Wakabaki, your thumbnail shows several traces that are not seperated at all from the groundplane that surrous them. In the gerber files, they do not show up at all. Am I missing something?

Nevermind, all is good.

[Elvee]

Here is the confirmation of what I said earlier, and also of the versatility of the Circlophone's topology:
I simply dropped two NMOS in place of the darlingtons, in the Ndarlington version, and it worked exactly as expected.

The only things that required adaptation are the G-S resistors that have been increased to 390ohm.

Also, the lower clipping value predicted on the positive side is observed: the clipping occurs ~5V below the positive rail.

If the input voltage is reduced to just under the clipping, the behavior becomes almost identical to that of the BJT version: the THD is a bit higher, probably because of the lower transconductance of the MOS compared to the bipolars.

The bias servo too works in an identical manner.

A reminder:
all this is simulation, and nothing has been built yet.

But I will come up with a version using the same trick as the CircloMOS, build it, and launch a dedicated thread when it's mature and working well enough.

[Tekko]

Why are the later circuits, particulary the last one upside down ? A bit confusing.

[Elvee]

Originally Posted by Tekko

Why are the later circuits, particulary the last one upside down ? A bit confusing.

Because that's the view of the topology I am more familiar with.

But I understand it can be confusing to others.

Here is a more habitual presentation:

[Junm]

Hi elvee,

What is the main purpose of the 2 zeners in series (D3+D4 in last schema), in first schema value was 24v (12v+12v) but in darlington and MOS version it was reduced to 18.2 v (10v+8.2V)? Does the Threshold voltages differences comes to play here?

[odysseybmx414]

Today I will etch a board, and populate most of it. I will just have to get some 2N3019s and the schottky diodes.

[Elvee]

Originally Posted by Junm

Hi elvee,

What is the main purpose of the 2 zeners in series (D3+D4 in last schema), in first schema value was 24v (12v+12v) but in darlington and MOS version it was reduced to 18.2 v (10v+8.2V)? Does the Threshold voltages differences comes to play here?

No, it mainly has to do with the power supply: without D3 and D4, Q5 would see an almost constant voltage equal to 2*Vsupply.

On the other hand, Q6 sees the full output swing, with an average equal to Vsupply.

This means that Q5 would dissipate on average twice as much power as Q6.

Adding ~Vsupply worth of zener decreases the dissipation of Q5 to the level of Q6.
This improves the static balance of the phase-splitter, and avoids the need of a heatsink for Q5.
The first schematic is shown with 25V supplies, whilst the other one uses 20V rails.

This balancing is purely static, and does nothing for the dynamic asymmetries caused by the Early effect.
It might be advantageous to replace the zeners by a resistor, but I haven't (yet) explored that option.

Today I will etch a board, and populate most of it. I will just have to get some 2N3019s and the schottky diodes.

Many other TO5/TO39 and VAS-type transistors are also suitable, the 2N3019 is not at all critical

[odysseybmx414]

Originally Posted by Elvee

No, it mainly has to do with the power supply: without D3 and D4, Q5 would see an almost constant voltage equal to 2*Vsupply.

On the other hand, Q6 sees the full output swing, with an average equal to Vsupply.

This means that Q5 would dissipate on average twice as much power as Q6.

Adding ~Vsupply worth of zener decreases the dissipation of Q5 to the level of Q6.
This improves the static balance of the phase-splitter, and avoids the need of a heatsink for Q5.
The first schematic is shown with 25V supplies, whilst the other one uses 20V rails.

This balancing is purely static, and does nothing for the dynamic asymmetries caused by the Early effect.
It might be advantageous to replace the zeners by a resistor, but I haven't (yet) explored that option.


Many TO5/TO39 and VAS-type transistors are also suitable, the 2N3019 is not at all critical

Already ordered from Mouser. I needed other things as well, and decided I might as well use those.

[Junm]

Originally Posted by Elvee

No, it mainly has to do with the power supply: without D3 and D4, Q5 would see an almost constant voltage equal to 2*Vsupply.

On the other hand, Q6 sees the full output swing, with an average equal to Vsupply.

This means that Q5 would dissipate on average twice as much power as Q6.

Adding ~Vsupply worth of zener decreases the dissipation of Q5 to the level of Q6.
This improves the static balance of the phase-splitter, and avoids the need of a heatsink for Q5.
The first schematic is shown with 25V supplies, whilst the other one uses 20V rails.

This balancing is purely static, and does nothing for the dynamic asymmetries caused by the Early effect.
It might be advantageous to replace the zeners by a resistor, but I haven't (yet) explored that option.

Thanks for the explanation now i got it, so the values are dependent on the desired rail voltages and balance of the 2n3019's.
BTW, what is the maximum rail voltage allowed assuming output transistors are changed to higher Vce/Vds types. Let's say at +/- 50volts what components needs upgrading?

[Elvee]

Originally Posted by Junm

Thanks for the explanation now i got it, so the values are dependent on the desired rail voltages and balance of the 2n3019's.
BTW, what is the maximum rail voltage allowed assuming output transistors are changed to higher Vce/Vds types. Let's say at +/- 50volts what components needs upgrading?

You should find information about the components dimensioning in this post:
♫♪ My little cheap Circlophone© ♫♪

Going much higher than +/-50V would make little sense: even on 8 ohm, this would mean an output power in the region of ~150W, which is clearly unreasonable for a single pair of output transistors.
Even the most rugged types, like the 2N6259 would be pushed to their limits.

I don't recommend paralleling transistors on this design: emitter degeneration resistors would be needed, and they would interfere with the constant transconductance action provided by the bias servo.
This would create crossover artifacts similar to those of a standard class AB output stage.

The optimum range of power for this design is between ~10 to 100W.