Class A amplifier with op-amp input stage is refusing to work

[Plecto]

I refreshed this site every 30. minutes or so hoping for a reply, but I got too impatient so I etched the circuit. It does not work of course, steady odd looking 150khz or so oscillations. I added caps in parallel with the feedback resistors, but that didn't do anything (got up to 470nF at the end there).

 

[betwixt]

May I ask what your design creiteria are. You seem to be fixated with the idea of using class A amplifiers which are the least efficient and using DC coupling throughout. While DC coupling *may* have some advantages, if removing 'pop' at switch on is your intention there are easier ways to do it which do not use relays. Although a relay can disconnect the output for a while after powering up, it might only delay the pop if all it did was prevent charging currents flowing until its contacts close.

Without predjudice to your intentions, you would get far better results using a small class AB amplifier with either split supplies or a virtual ground for the output signal. I'm thinking along the lines of a three transistor complimentary amplifier with the output ground being the mid point of two large capacitors across a single supply. This would DC stabilize without frequency or phase skewing networks and could give a flat response from one or two Hz up to maybe 30KHz or more and may even use fewer components.

Brian.

 

[Plecto]

Well, class A is a criteria. While I know the audio quality gained from class A as opposed to a proper class AB might not be audible, it's more of a marketing thing. I'm planning on selling these and the letters "class A" really makes an audiophools eyes to shine.

These are the criteria: Class A, minimum of 100mW into a 62Ohm load, 12-15V single ended supply (wall-wart) and simplicity. The reason for using an op-amp as a input stage was to make it simpler as well as being able to calculate the gain easily. I guess I got so fixated at the design I first linked that I didn't care to ask if there perhaps was a better design I could use

About the popping noise. Using a relay will remove this, i just need to have a resistor connected in parallel with the load so that the cap charges while the output is disconnected.

 

[betwixt]

As you are talking about low power amplifiers, why not use a simple class AB design but increase it's quiescent current so it stays in the 'A' conduction cycle up to at least the 100mW target. It would still be more efficient than using a load resistor and single ended output.

Brian.

 

[Plecto]

I guess I could do that, but that would strictly speaking be AB, wouldn't it? Perhaps it's fine though, but how do I bias it so neatly? I'm not sure if biasing with resistors will do. With BJTs I will have their Hfe's increase as temperature increases and with fets I can't be exactly sure of what the threshold voltage is, right? So how would I do this so that the current stays stable regardless of the type of fet or bjt that's used?

 

[FvM]

The idea of "perfect" class A amplifier is ignoring some problems of the discussed real amplifier designs. Besides loop stability, which can be managed with suitable compensation methods, there are also issues like OP non-linearity (really bad: LM358), or supply voltage hum. Means, you can design class A amplifiers with excellent THD numbers, but I don't expect it for the discussed circuits,

 

[Plecto]

I was hoping that a regulator at the input would decrease humming to a level of inaudibility, but that's a different matter. I don't worry that much about THD values, I care more about output power and the simplicity of the circuit. What if I remove the op-amp and go for a design like the Zen amp? **broken link removed**. Instead of the current source I'll go with a drain resistor and of course I won't be having a transformer. I've made this circuit before and I got all kinds of humming and hissing. I replaced R5 with a pot for volume control, but it was all a failure. The amp worked and the sound was pretty good, but the humming and hissing changed in intensity when turning the pot so I just had to turn it almost all the way up. I also have the feeling that the bias point changed with time as I think the amp has started clipping now (which shouldn't be possible without blowing up my headphones). This was the reason I wanted to go with an op-amp input stage. Easily to set gain, easy to set the volume and a stable bias point.

I'm really up for suggestions as I'm itching to etch a new board and get something working

 

[betwixt]

You only need 2.5V of of signal to drive 100mW into 62 Ohms.
The Zen design is basically a constant current generator as a load to a FET voltage amplifier. It will work but look at the efficiency - about 5W in for 0.1W out = 2% !
I think you will find the most important factor when driving headphones is the amplifier output impedance. The headphone impedance will vary greatly over frequency and even on how tight the phones are to the listeners ears, particularly if they are closed cup types. If the amp can't damp the signal enough the frequency response will be horrible.

I've attached a sketch (sorry for the quality, my workbench is so cluttered I had no flat surface to draw it!) showing a simple class AB amplifier with bias stabilization. I guessed the values but they are probably good starting points for a simulation. If you set the bias current to about 100mA (just over twice the output current) it should work in class A mode up to around 100mW of output then start to move into class B mode. It would be worth experimenting with small emitter resistors in the output transistors and possibly bootstrapping the top bias resistor. I'm in the middle of a big assembly language program at the moment so I don't want the distraction of entering it in a simulator, I leave that to you.

Brian.

 

[Plecto]

Thanks for the circuit I've been simulating a little bit and I got a working result: https://www.circuitlab.com/circuit/ge873g/class-a-biased-ab-amp/

I feel a little unsure though. First off I have no idea how to calculate the currents and voltages in this circuit. I have learned about transistors at uni, but everything is just so connected together here After some trial and error I got the bias point to about 1/2 the supply to 120mA, but in a real life circuit, even if the diodes are mounted next to the transistors, can I really trust that this will stay that way? Giving that there's no output cap, a change in the bias voltage would be pretty bad. Also, how do I calculate the gain as it seems pretty darn high? I put in two voltage sources instead of a virtual ground just for the sake of simulation. It became to tedious otherwise.

About the zen amp and efficiency. I can't see how you are getting 2% efficiency. If I have a 12V source and replace the current source with a 50Ohm resistor I would get 6^2/50*2=1.44W of heat. The maximum peak output voltage into 62Ohm would be 3.32V. That would equate to 88mW output power and a efficiency of 6%, but if a current source was used instead it would be able to output a voltage pretty close to the rails (depends on the current source I guess).

Edit: How about this: https://www.circuitlab.com/circuit/863645/class-ab-op-amp-input/

 

[betwixt]

I can't run the simulation here because my intenet connection is too slow. It takes around 2 minutes just to load the opening page. Despite being bombarded with offers of 20 or 50Mb internet speeds, my rural location can only support about 100K at best and is usually much slower. It used to be faster with dial-up but that's now been withdrawn.

The input ground side should be connected to the emitter of the first transistor, not the output common point. It will DC stabilize because any rise in output voltage will increase the bias current in the first transistor and restore equilibrium. It should also be thermally stable if the diodes are coupled to the transistors because the quiescent current through the output transistors is decided by the voltage dropped across the diodes and as they heat up, their Vf will reduce and drop the available bias voltage. It will work in class B with two diodes in series, adding a third should increase the bias too much, the resistor across one of the diodes is to control the increased current so the 'A' to 'B' changeover point can be set.

The idea of series output capacitors is that as you switch on, they charge equally and their mid point assumes half supply voltage and hence no pop is heard. I added the resistors across them to help them maintain Vs/2 in case they were not well matched.

Brian.

 

[Plecto]

I see. Are you talking about the 1000uF caps? I'm not worried about them though, I'm worried about the amp outputting DC. I can only fine tune the bias point so much, but I'm doomed to get some DC output. This might not be that bad, but can I trust the bias point to stay that way after lets say a year of use?

What about the circuit I made though. As you can see I'm fond of op-amp input stages The fets was perhaps a bad idea though. If I were to find fets with a Vth of perhaps 4V, I would only get 0.5V maximum peak at the output So I would have to go with BJTs. The op-amp will also make sure that the bias point is kept steady as well as perhaps improving the sound quality a bit by compensating for the transistors unlinearity? I won't have any problem calculating the gain at least, but the question is if it will be stable?

 

[Audioguru]

The threshold voltage of a Mosfet is when it is almost turned off (0.25mA), not when it is turned on.

When BJTs are used as emitter-followers at the output then they have no voltage gain so the circuit is stable.

You show a TL082. Why not use an TL072 dual audio opamp instead that is a TL082 selected for low noise? It costs the same.

 

[Plecto]

Yes I know. I mean that the difference between the output voltage and the rail voltage can never be less than the threshold voltage, if it was then the fet would be turned off. So using a fet with a high threshold voltage would sincerely limit the output voltage capabilities.

I'll try building my class A biased class AB amp (is it called that? :S ) with an op-amp input then and see how it goes.

You show a TL082. Why not use an TL072 dual audio opamp instead that is a TL082 selected for low noise? It costs the same.

It's just the default op-amp chosen by circuitlab. I'm going to use a NE5532.

 

[Plecto]

I've now completed the amplifer. I clearly underestimated the power I needed, 100mW is not enough for my power hunger AKG K-701. I was thinking of using a 9V supply, but the amp clipped too soon. I then went for a 12V supply and I can now get a volume that I'm more satisfied with, but I'm sure some of the tops are still clipped a little bit. I also underestimated how dynamic music is and how load some portions of music can be, right now it can output 163mW, but I think I wan't to double this just be absolutely safe that the signal is at no point clipping. The whole idea of having it being class A isn't quite working out. If I'm going to use 15V or perhaps 24V, theres absolutely no way I can have a bias current of 100mA without a more proper heatsinking solution. I have biased it some, but I'm unsure about how high the bias current is, at least it should be an okay class AB amp (too bad I don't have equipment to measure distortion). Here's how my prototype turned out:

One last very off-topic question, does anyone know what that cone-shaped drill bit in the upper right corner is called? I got one for christmas, was wondering if I could get them in other dimensions (they are absolutely brilliant )

 

[keith1200rs]

It's a "step drill".

Keith

 

[betwixt]

To put your problem in perspective, I know someone who insisted on class A power amplifiers for their home stereo. They were, in my estimation, expert design engineers so I have no doubt they followed alll the rules and calculated everything precisely. Each output transistor had it's own heat sink, from memory they were black anodized blocks with fins on both sides, measuring about 15 x 15 x 6cm (about 6 inches square with 2.5 inches of fins) and the whole amplifier had a large cooling fan blowng through it all the time. I can remember the heat from it kept the room warm, I would guess it was equivalent to a 1KW heater.

It was powerful enough to drive Quad electrostatic loudspeakers to comfortable volume so I would guess it's output power was around 100W peak.

Brian.

 

[keith1200rs]

I have a class A amplifier which can be switched from class A to class AB. In class A mode (20W per channel) it consumes around 1kW, I seem to remember. I tend to run it in class AB! I cannot remember how much it weighs but I think it is around 15kg. It has a conventional transformer - no switching circuits whatsoever.

Keith.

 

[Plecto]

Yeah, it seems that a class A headphone amplifier that can drive even the heaviest headsets to max would require some heatsinking. Even if we are just talking milliwatts, the extreme variation in output impedance (headsets can range from maybe 16Ohms to 600Ohms I believe) makes it extremely inefficient.

I'm a little curious to why I'm not getting more output voltage though. Right now I have a 12.44V supply and I'm getting 4.5Vp output. I know because of the threshold voltage of the bjt's I will have a theoretical output peak voltage of 5.52V, can't the op-amp output anything more than Vs-1V? I believed it was closer to the rails than that. Another reason might be that the op-amp is running out of power. I'm using TO-220 BJT's with a Hfe(min) of 20 so I guess it's possible.