Thank you! Way too complicated for me, I'm afraid! 😎
WHat I would really like is an OPAMP front end for the KCP OPS. As simple as possible to make an operating amplifier...
Honestly, why? Your 50 watt KRILL is also a non-switching EF. Steve's D. implementation is much very simpler. Very good OP amp-based front-end circuits have been on the market for at least 50 years ...
Basically, it's a great thing to have a non-switching module, the power amplifier (or simply put, the last stage, the impedance converter) in the drawer, but the most important question is how much power you /we /one need.
But you should never forget the recipe (the rules for dimensioning and implementation) of Douglas Self's Blameless approach. Unfortunately, the rest must always compete against this standard when it comes to elegance and simple beauty of design in the audiophile home sector. So let's say 5 to 50 watts and a pleasant room volume.
The rest is academic posturing or simply fun and the joy of being different.
#
Do not switch or switch as quickly as possible, i.e. infinitely short.
But you should never forget the recipe (the rules for dimensioning and implementation) of Douglas Self's Blameless approach. Unfortunately, the rest must always compete against this standard when it comes to elegance and simple beauty of design in the audiophile home sector. So let's say 5 to 50 watts and a pleasant room volume.
The rest is academic posturing or simply fun and the joy of being different.
#
Do not switch or switch as quickly as possible, i.e. infinitely short.
The description of the circuit that I attached in my previous post is as follows:-
The amplifier is based on the Class i output stage described in LinearAudio - Volume 2 by Kendall Castor-Perry. The output stage has a smooth transition between the two halves of a Class B output stage without any high power switching and without requiring any adjustments and is mostly free of cross over distortion.
The amplifier is split into 2 sections. The first section provides the voltage gain and the second Class 'i' section provides the current gain. Overall / global feedback is required to lower the output resistance because the Class 'i' stage has an output resistance set by the current sense resistors R36 and R37 to 0.3 ohm. The global feedback is by R38 and R5 which sets the voltage gain to 19.5.The input resistance is set by R5 to 2 k ohm.
The voltage amplifier bias current is set by the current source U1 to about 1 mA. The current mirrors formed by Q1A, Q3A and Q3B then set the currents in Q5A and Q6A to the same 1mA. Because the base emitter voltage of Q5B is similar to that of Q5A the bias current is similar. The collector and emitter currents of Q6A and Q6B are similar for the same reason.
R8, R9, R10 and R11 reduce the effect of different temperatures and imperfect transistors matching ensuring that the currents through the transistors Q5A to Q8B are similar. They also linearize and reduce the transconductance.
The bias current of 1 mA flows through R14 which with R15 and R16 set the bias current in the output buffer stage formed by the emitter followers Q12 and Q14 to about 5 mA. Q11 and Q13 provide some temperature compensation of the bias current through Q12 and Q14.
C3 and C4 set the the voltage amplifier open loop gain to be unity at about 4 MHz as measured on the prototype.
The MOSFET Q38 sources the load current is controlled by the differential amplifier formed by Q19A, Q19B and Q20A.
When the output is required to source current to the load, RL, Q19A is on, Q19B is off. The differential amplifier Q19A and Q20A ensures that the voltage at the source of Q38 follows the voltage at the Class i input voltage (label on the schematic).
When the output isn't sourcing current Q19A is off and Q19B is on. The current through R36 is held to about 100 mA by the differential amplifier Q19B and Q20A.
The Class i amplifier bias is set by the current source formed by D11, R29, Q31A, Q31B, D10, Q30, Q27A, Q27B, Q28 and Q29 to about 10 mA.
Q21, Q22, Q23A, Q23B, R23, R24 form a Wilson current mirror to set the tail current of the differential amplifier formed by Q19 and Q20 to 10 mA.
R19, R20, Q15A, Q15B, Q16 and Q17 form another Wilson current mirror forcing the currents through Q19A, Q19B and Q20A to be close to equal. The current is required to be reasonably accurate because the voltage difference between the the emitters of Q19A , Q19B along with the difference in resistors R21 and R22 set the bias current through the output stage when it isn't sourcing current to the output. The difference in resistance between R21 and R22 = 39 - 33 = 6 ohms. With a current in each half of the differential amplifier = 5 mA the offset = 6 ohms x 5 mA = 30 mV. This sets the voltage across R36 to 30 mV hence the current = 100 mA.
Q18 ensures that Q19A and Q19B collector voltage is similar to that of Q20A minimising the difference in base emitter voltages of Q19B and Q20A for an equal collector current (Early effect).
D4 and D5 prevent excessive reverse bias to either Q19 or Q20 that could otherwise cause the Class i output to latch up.
The amplifier is based on the Class i output stage described in LinearAudio - Volume 2 by Kendall Castor-Perry. The output stage has a smooth transition between the two halves of a Class B output stage without any high power switching and without requiring any adjustments and is mostly free of cross over distortion.
The amplifier is split into 2 sections. The first section provides the voltage gain and the second Class 'i' section provides the current gain. Overall / global feedback is required to lower the output resistance because the Class 'i' stage has an output resistance set by the current sense resistors R36 and R37 to 0.3 ohm. The global feedback is by R38 and R5 which sets the voltage gain to 19.5.The input resistance is set by R5 to 2 k ohm.
The voltage amplifier bias current is set by the current source U1 to about 1 mA. The current mirrors formed by Q1A, Q3A and Q3B then set the currents in Q5A and Q6A to the same 1mA. Because the base emitter voltage of Q5B is similar to that of Q5A the bias current is similar. The collector and emitter currents of Q6A and Q6B are similar for the same reason.
R8, R9, R10 and R11 reduce the effect of different temperatures and imperfect transistors matching ensuring that the currents through the transistors Q5A to Q8B are similar. They also linearize and reduce the transconductance.
The bias current of 1 mA flows through R14 which with R15 and R16 set the bias current in the output buffer stage formed by the emitter followers Q12 and Q14 to about 5 mA. Q11 and Q13 provide some temperature compensation of the bias current through Q12 and Q14.
C3 and C4 set the the voltage amplifier open loop gain to be unity at about 4 MHz as measured on the prototype.
The MOSFET Q38 sources the load current is controlled by the differential amplifier formed by Q19A, Q19B and Q20A.
When the output is required to source current to the load, RL, Q19A is on, Q19B is off. The differential amplifier Q19A and Q20A ensures that the voltage at the source of Q38 follows the voltage at the Class i input voltage (label on the schematic).
When the output isn't sourcing current Q19A is off and Q19B is on. The current through R36 is held to about 100 mA by the differential amplifier Q19B and Q20A.
The Class i amplifier bias is set by the current source formed by D11, R29, Q31A, Q31B, D10, Q30, Q27A, Q27B, Q28 and Q29 to about 10 mA.
Q21, Q22, Q23A, Q23B, R23, R24 form a Wilson current mirror to set the tail current of the differential amplifier formed by Q19 and Q20 to 10 mA.
R19, R20, Q15A, Q15B, Q16 and Q17 form another Wilson current mirror forcing the currents through Q19A, Q19B and Q20A to be close to equal. The current is required to be reasonably accurate because the voltage difference between the the emitters of Q19A , Q19B along with the difference in resistors R21 and R22 set the bias current through the output stage when it isn't sourcing current to the output. The difference in resistance between R21 and R22 = 39 - 33 = 6 ohms. With a current in each half of the differential amplifier = 5 mA the offset = 6 ohms x 5 mA = 30 mV. This sets the voltage across R36 to 30 mV hence the current = 100 mA.
Q18 ensures that Q19A and Q19B collector voltage is similar to that of Q20A minimising the difference in base emitter voltages of Q19B and Q20A for an equal collector current (Early effect).
D4 and D5 prevent excessive reverse bias to either Q19 or Q20 that could otherwise cause the Class i output to latch up.
Thank you for this detailed explanation!
It looks like the overall amplifier is inverting, is that right? Or is it the IPS that makes it look so?
(From the question you know I'm no designer... huh!)
🙂 morten
I want to add some of my thoughts.
1. We have to admit that there is NFB going on. Not just one, there are 2 NFB, one for upper half, the other for the bottom half. We need to be caution for anything more than one NFB. If there is any mismatch between 2 halvies, the bias point won't be constant under dynamic condition. A cap is necessary between the gates of N-mosfet and the gate of P-mosfet, to smooth out the ripples.
2. As said, there is NFB going on. Stability has to be inspected. I lean toward the method used in the OP. Please see C3, C4, R17, R18 in OP.
3. It has the same culprit as MF-A1. It is a dumb idea to bias the amp by sensing the current over the emitter/source resistors. The sum of voltage drops of these 2 resistors only stays constant under class A. Under class B, the symmetry is broken and the performance of the whole thing starts to fall apart. MF-A1 dodges this problem by keeping the amp working in class A.
1. We have to admit that there is NFB going on. Not just one, there are 2 NFB, one for upper half, the other for the bottom half. We need to be caution for anything more than one NFB. If there is any mismatch between 2 halvies, the bias point won't be constant under dynamic condition. A cap is necessary between the gates of N-mosfet and the gate of P-mosfet, to smooth out the ripples.
2. As said, there is NFB going on. Stability has to be inspected. I lean toward the method used in the OP. Please see C3, C4, R17, R18 in OP.
3. It has the same culprit as MF-A1. It is a dumb idea to bias the amp by sensing the current over the emitter/source resistors. The sum of voltage drops of these 2 resistors only stays constant under class A. Under class B, the symmetry is broken and the performance of the whole thing starts to fall apart. MF-A1 dodges this problem by keeping the amp working in class A.
Hello m0rten,
Yes you are correct it is overall inverting. The intention is to add a balanced to unbalanced input stage with some voltage gain to drive the rather low input impedance of 2 k ohm.
I did look at using an operational amplifier for the voltage gain but they are few parts that support the supply voltage. Also because the class I stage is within the overall feedback loop (to keep the output impedance low) it would be desirable to have some control of the operational amplifier frequency compensation to keep the loop stable. I am interested in any part that could be used.
I have attached an Oscilloscope picture showing the voltage across R36 showing the flat constant current part (about 100 mA) and the sourcing current to the load (about 3.3 A). It shows the relatively smoth transistions bwteen the two modes. The differential probe is home made and not very accurate so the numbers are not to be trusted. The waveform does degrade somewhat at 20 kHz.
Yes you are correct it is overall inverting. The intention is to add a balanced to unbalanced input stage with some voltage gain to drive the rather low input impedance of 2 k ohm.
I did look at using an operational amplifier for the voltage gain but they are few parts that support the supply voltage. Also because the class I stage is within the overall feedback loop (to keep the output impedance low) it would be desirable to have some control of the operational amplifier frequency compensation to keep the loop stable. I am interested in any part that could be used.
I have attached an Oscilloscope picture showing the voltage across R36 showing the flat constant current part (about 100 mA) and the sourcing current to the load (about 3.3 A). It shows the relatively smoth transistions bwteen the two modes. The differential probe is home made and not very accurate so the numbers are not to be trusted. The waveform does degrade somewhat at 20 kHz.
I have tried to make a similar concept but with OP:s as drivers. It failed while when an output side is off the OP has maximum opposite voltage out.
And the time to get the output to conducting the missed output voltage was crossover distortion.
I guess this switching time will be shorter with these transistor stages but i fear it will be some switching time anyway.
And the time to get the output to conducting the missed output voltage was crossover distortion.
I guess this switching time will be shorter with these transistor stages but i fear it will be some switching time anyway.