Current Dumping topology can achieve Class-A performance with zero biased output stage. It sounds like a fairy tale. Yet, it is true.
The famous Quad 405 amplifier is one of the examples. Please see the attached document if you want to learn more about Quad 405.
How it works, "Current Dumping for Dummies". https://connerlabs.org/current-dumping-for-dummies/
Conceptual Diagram.
What is crazier is that you can actually build it with an opamp. Modern opamps perform much better than than than the Class-A stage of Quad 405 back in 70s.
As the opamp is precompensated internally with a miller capacitor, you don't really need the capacitor shown in the Conceptual Diagram. The 3K3 resistor, together with the miller cap, they determine the "unit loop gain bandwidth", which is also the "Time Constant". Once the "Time Constant" from RC on the left side equals to the "Time Constant" from RL on the right side. The bridge is balanced. (https://en.wikipedia.org/wiki/Time_constant#Time_constants_in_electrical_circuits)
For an opamp, it usually specs the "Gain Bandwidth Product" in the manual. You also know the gain of your circuit. From there, you can get how much loop gain bandwidth in your circuit. You know the bandwidth, you can derive the Time Constant (relation between bandwidth and Time Constant, https://en.wikipedia.org/wiki/Time_constant#Relation_of_time_constant_to_bandwidth). By given the value of the inductor, then you can calculate the feed forward resistor value. The bridge is completed.
Here is the prototype that I come up with. It is almost identical with the conceptual diagram above.
NE5532 is a must, because of its current capability, up to 38mA. Also the max +-22v supply voltage is a plus. You can get 20W with it.
I would suggest to keep the bandwidth (Time Constant) near 3MHz. You want to minimize the current over the feedforward resistor R3, while keep everything stable. You only have 38mA budget from a single NE5532.
Use fast and high beta transistors, MJE15030/31 are good. I use 2SC5200/2SA1943 in my build.
Does it work? Yes, it sounds just same as my other amps. Sorry, I don't have spectrum analyzer at home.
It is just a proof of concept. You can parallel the OPAMP. You can use zero biased CFP to get more current.
It should be your first amp, because of the simplicity. It could be your last amp, because it performs so well why bother other complex designs. Look at this teeny tiny amp. You don’t even need heat sink.
Indices:
Final Version
Feedforward Resistor Value Calculation
Investigate The Current Required to Drive The Feedforward Resistor
Method to Calibrate The Feed Forward Resistor
The famous Quad 405 amplifier is one of the examples. Please see the attached document if you want to learn more about Quad 405.
How it works, "Current Dumping for Dummies". https://connerlabs.org/current-dumping-for-dummies/
Conceptual Diagram.
What is crazier is that you can actually build it with an opamp. Modern opamps perform much better than than than the Class-A stage of Quad 405 back in 70s.
As the opamp is precompensated internally with a miller capacitor, you don't really need the capacitor shown in the Conceptual Diagram. The 3K3 resistor, together with the miller cap, they determine the "unit loop gain bandwidth", which is also the "Time Constant". Once the "Time Constant" from RC on the left side equals to the "Time Constant" from RL on the right side. The bridge is balanced. (https://en.wikipedia.org/wiki/Time_constant#Time_constants_in_electrical_circuits)
For an opamp, it usually specs the "Gain Bandwidth Product" in the manual. You also know the gain of your circuit. From there, you can get how much loop gain bandwidth in your circuit. You know the bandwidth, you can derive the Time Constant (relation between bandwidth and Time Constant, https://en.wikipedia.org/wiki/Time_constant#Relation_of_time_constant_to_bandwidth). By given the value of the inductor, then you can calculate the feed forward resistor value. The bridge is completed.
Here is the prototype that I come up with. It is almost identical with the conceptual diagram above.
NE5532 is a must, because of its current capability, up to 38mA. Also the max +-22v supply voltage is a plus. You can get 20W with it.
I would suggest to keep the bandwidth (Time Constant) near 3MHz. You want to minimize the current over the feedforward resistor R3, while keep everything stable. You only have 38mA budget from a single NE5532.
Use fast and high beta transistors, MJE15030/31 are good. I use 2SC5200/2SA1943 in my build.
Does it work? Yes, it sounds just same as my other amps. Sorry, I don't have spectrum analyzer at home.
It is just a proof of concept. You can parallel the OPAMP. You can use zero biased CFP to get more current.
It should be your first amp, because of the simplicity. It could be your last amp, because it performs so well why bother other complex designs. Look at this teeny tiny amp. You don’t even need heat sink.
Indices:
Final Version
Feedforward Resistor Value Calculation
Investigate The Current Required to Drive The Feedforward Resistor
Method to Calibrate The Feed Forward Resistor
Last edited:
You are one of the very few people here who dare to admit that they don't hear any difference.
This for sure is working fine in a limited range. A nice, but not new idea. You may go a step further on that way to simple amp design and even combine the op-amp with a chip amp. "Done right" is it called by a developer we know. You take the comfort of the chip amp with the precision of the op-amp.
On the other hand, the NE5532 is not really a single part, but a quite well thought out amp with a number of components. Take all these components and include them into your schematic. You will end up with something looking like a "normal" amp.
On the other hand, the NE5532 is not really a single part, but a quite well thought out amp with a number of components. Take all these components and include them into your schematic. You will end up with something looking like a "normal" amp.
I used current dumping with an op-amp when converting a cheap satellite loudspeaker into a doorbell, see the attachment. The loudspeaker is connected to P2, the switch to pins 3 and 4 of P1. P5 is for muting; it is meant for use in a radio studio, so it should keep silent when a microphone is open. I didn't dare to try it without C9.
It actually is a single part, but it isn't a discrete part per se, as the op said, big difference. He also hasn't stated it is a new idea, literally in the first post. He just made a proto.
Recommend a chipamp? It will be severely limited in drive capability though, when put in class A... More limited than what his proto actually can achieve.
Maybe try the lm3886. 200 Watt are no huge problem. Ask tomchr for details, but I'm sure you know better anyway.
In the OP schematic there is high input voltage and very low gain in the circuit.
Could this be fixed?
Could this be fixed?
Nothing in post #1 works in class A. The same holds for post #4.
In the real world and not in simulation, the original class A driver, or in this case the opamp, has to supply the load current until the dumper transistors conduct. The 405 MK1 had a large dead zone, one reason it didn't like 4 Ohm loads This is why later Quad models biased the dumpers closer but not quite, to class B
I thought it would be much better if I put some bias on the output dump. Nope, they aren’t any better. The beauty of the current dumping is that the performance mainly depends on the error correction amplifier(the opamp in my case).
The condition of the balanced the bridge is:
R1 C = L / R2
When you realize R1 C is the Time Constant and it can be derived from the loop gain bandwidth of your circuit, you can just balance the bridge without the C and R1.
It is just a proof of the concept.
The lower the gain it is, the higher bandwidth you get from the opamp. 3MHz bandwidth ensures the performance. Also the inverting configuration resolves the common mode input limit.
NE5532 comes with dual package, you can add another gain stage at the front. It is highly recommended to do so.
Last edited:
You get a more complicated feedback loop around the op-amp than you normally have: extra emitter followers when the dumpers are active, an extra voltage divider consisting of the small resistor in the bridge and the load and an extra inductor when the dumpers are off. You have to ensure that doesn't eat up too much of the phase margin. As in my case it was only for a doorbell anyway, I didn't bother and just used C9, with a phase correction resistor in series.
The unit loop gain bandwidth of the class A stage of Quad 405 is about 2.4MHz. It is pretty aggressive for 5MHz fT output transistor.
In the simulation, once you run out of fT, you get a Zero(in the transfer function) naturally. 2N3055/2N2955 in my simulation, can do 10MHz inside the NFB. As there is a coil that separates the load at high frequency, it seems very stable. I don’t have the guts to stretch it to 10MHz. I think 3MHz is about right with modern 30MHz fT output stages.
Last edited:
2 things affect the square wave. The slew rate of the opamp, and the size of the inductor.
Please keep R1, R2, R3, L1 as the same values as mine. They are part of distortion cancelling bridge. The values below only work with NE5532.