It is much better in controlling the idle current. The VBE temperatur drift of the output transistors is now in a DC feedback loop.
About distortion is it hard to say. This output stage goes very well with a current feedback amplifier so i see no difference there.
The questions are just how much lower distortion will we get. And how much do we want to pay for that and the stable idle current.
Before anyone has build it we dont know how it will be in practice.
About distortion is it hard to say. This output stage goes very well with a current feedback amplifier so i see no difference there.
The questions are just how much lower distortion will we get. And how much do we want to pay for that and the stable idle current.
Before anyone has build it we dont know how it will be in practice.
That's almost a call to swap partners at will, isn't it?
Technically and economically, an elaborate design of the last stage makes no sense whatsoever if it does not provide significantly better data in direct comparison with a known simple output stage and this is what is needed.
0.03% THD is my personal benchmark.
If, for example, a KRILL spin-off does not clearly crack this mark, the question of the additional effort is justified. But as we all know, a THD number alone is not worth much in terms of “sound”.
That's a damn good question for the Kendall Castor-Perry proposal to ask itself. Various practical tests can provide an answer, as can extensive metrological investigations.
With Morten's PCB, nothing now stands in the way.
By the way - I do not believe that there will be any significant improvements for daily practice, because where are they supposed to be?
For a large company, however, integration, i.e. an IC that regulates the bias, can be worth a lot - and make a lot of sense. And this is exactly what the design by Kendall C-P shows.
HBt.
Agreed, the feedback keeps the bias current very constant.
The feedback does not reduce distortion by much because the emitter resistors are outside of the feedback loops. Mismatch between the two resistors will generate even-order distortion.
The feedback around the output stage introduces a low-frequency pole. That is not good because the pole takes away from the possible global feedback. Having two identical feedback loops that can see each other is also not good.
The approach is interesting, though.
Ed
True, but the same holds for the boring old output stages almost everyone else uses.
Complementary emitter follower (or source follower):
An emitter follower (or source follower) is nothing but a common-emitter stage (or common-source stage) with local voltage follower feedback. The emitter resistors are outside the feedback loops and the NPN and PNP emitter followers see each other.
Complementary Sziklai pairs:
A Sziklai pair is a two-stage negative-feedback voltage follower or two-stage negative-feedback amplifier with a voltage gain somewhat higher than 1. The collector resistors are outside the feedback loops and the push and pull Sziklai pairs see each other.
The differences between class-i and the conventional solutions are the more complex circuit, the much better control of the quiescent current, the non-switching behaviour and the better controlled transition from one side to the other. The question is how wideband you can make the local voltage followers.
Complementary emitter follower (or source follower):
An emitter follower (or source follower) is nothing but a common-emitter stage (or common-source stage) with local voltage follower feedback. The emitter resistors are outside the feedback loops and the NPN and PNP emitter followers see each other.
Complementary Sziklai pairs:
A Sziklai pair is a two-stage negative-feedback voltage follower or two-stage negative-feedback amplifier with a voltage gain somewhat higher than 1. The collector resistors are outside the feedback loops and the push and pull Sziklai pairs see each other.
The differences between class-i and the conventional solutions are the more complex circuit, the much better control of the quiescent current, the non-switching behaviour and the better controlled transition from one side to the other. The question is how wideband you can make the local voltage followers.
I checked the idle current against the minimum current when the other half is conducting,
With 90 mA idle current it is about 20 mA minimum current. We definitly need a real test to evaluate it not just simulating. Any way in my simulation there was under 0,1% distortion at 10 kHz 20v out with low impedance feed. Some feedback on that will give really low THD.
But I dont trust my TINA much on distortion so a run with another simulator is very interesting.
With 90 mA idle current it is about 20 mA minimum current. We definitly need a real test to evaluate it not just simulating. Any way in my simulation there was under 0,1% distortion at 10 kHz 20v out with low impedance feed. Some feedback on that will give really low THD.
But I dont trust my TINA much on distortion so a run with another simulator is very interesting.
👍 I understand what you are talking right now. I can see how it works without simulation right now.
@hbtaudio, I would like to correct my previous statement. This thing can work beyond Class A. Actually, we can put 2 extra input transistors to MF-A1 and turn MF-A1 into Class i. What's better is it can have gain by it own.
MarcelvdG said:
The class-i feedback is somewhat similar to CFP. EF2/3 does not have a feedback loop that can be opened.
Agree.MarcelvgG said:
That is not very different from an emitter follower output stage.stigigemla said:
Ed
163mA bias.
10Vp, 10KHz, -77dB THD into 8 Ohm. Not bad.
Voltages on the 0.22 resistors. Right, they do not pinch off completely.
PS: I also tried a version with 20x gain, with common source output configuration. That one didn't work. The voltage on the 0.22 resistor is too small to begin with. After dividing that by 20x, it doesn't work anymore.
10Vp, 10KHz, -77dB THD into 8 Ohm. Not bad.
Voltages on the 0.22 resistors. Right, they do not pinch off completely.
PS: I also tried a version with 20x gain, with common source output configuration. That one didn't work. The voltage on the 0.22 resistor is too small to begin with. After dividing that by 20x, it doesn't work anymore.
Sorry. The R20 and R21 needs to be something like 0,3 ohm or more.
Whe we go from idle to output voltage T1 or T3 stops conducting and that gives the double current in T9 or 10. 20 mv more VBE.
20 mv gives 66mA less in the 0,3 ohm. The idle current needs to be more than that. I suggest 90 or 100 mA.
With degradation resistors in series with the emitters the current difference will be bigger than that.
The idle current is set by the difference between R1 and R7 + R4 and R8
Whe we go from idle to output voltage T1 or T3 stops conducting and that gives the double current in T9 or 10. 20 mv more VBE.
20 mv gives 66mA less in the 0,3 ohm. The idle current needs to be more than that. I suggest 90 or 100 mA.
With degradation resistors in series with the emitters the current difference will be bigger than that.
The idle current is set by the difference between R1 and R7 + R4 and R8
Last edited:
The miller-equivalent capacitance plays a bigger role here, not the input capacitance.
The Cgd (known as Reverse Transfer Capacitance) of IRFP140 from Vishay is only about 110pf.
The equivalent that of MJL3281 is about 600pf.
The MOSFET is way easier to drive. Usually, I give about 7mA for the drivers in my past MOSFET designs.
Last edited: