Hello, I am a student tasked to build a working amplifier, I do not know much of what I am doing. I've got a three-stage configuration that consists of a single ended differential input stage, a common-emitter VAS stage, and a Sziklai pair output stage. It's supposed to be a 15W amplifier to a 15W 8ohms speaker in single supply. Output should be capable up to 15.5Vpeak in order to deliver ~11Vrms for the 15W rating and 15W speaker. Current on the other hand, about 2A peak. I was able to get good results on the output using LTSpice simulation, but I am wondering if I would even get such result in real life. Any advice? improvement? things I should change?


My reference on this design is from three books, Bob Cordell's, Douglas Self, as well as Horowitz and Hill's Art of Electronics in which I got the simplistic design of this one from. Most of the time, they got extra things like current sources, current mirrors, in which I have removed them for simplicity, but I am afraid it is no longer feasible in real life for it to function.

1. Coupling Capacitors:
Here's what I came up with through trial and error of the values of each component that I have. I'd probably be asked as to why my C2 and C3. I got these values because using lower ones such as 22u-33u-47u-100u-1000u values get my output attenuated, I also noticed that around 33u-47u, there would be oscillations in my output. It's the same case with my C4, using lower than 470p leads to oscillations. I did not do any calculations on how to get that value, I just sat and adjusted the values of those components until I got a satisfactory output.

2. INPUT STAGE - Single ended differential pair:
I have this kind of design because I am familiar with how to setup a feedback on a operational amplifier. I like how easy it is to just sample the output voltage via a voltage divider and feeding it back to the input which is just the base of the other input. The 3.3kohm allows about 6mA emitter current which would separate into 3mA current for each of the input transistors. I'd like to know if this kind of simple differential design would be able to work properly in real life because the ones from the books have these current mirrors as well as current sources. The Q1 is then fed into the VAS stage transistor Q3.

3. VAS STAGE - Common-emitter amplifier:
This one is just a common-emitter amplifier that's going to amplify the output of the differential stage. Diodes D1, D2, D3 are voltage bias for the Sziklai pair output stage. I believe it'll give me around 2.1V bias resulting to about 0.6V-0.7V impressed on R9 and R10. The 3.3kohm, R8, is just there to set the current for the stage to be about 6mA, which I think is already adequate current for the output stage to pull from at peak voltages. I would like to know if this design is also good or would even work. I simplified the VAS stage I see on Self's and Cordell's book in which they use a current source instead of like my R8.

4. OUTPUT STAGE - Sziklai Pair
I chose this output stage design mainly because of the bias voltage. I did read why such stage is good because of stability or what, I could not keep up with the technical terms. But all I know is that Q4 and Q5 are pre-drivers and that adding those 150ohm resistors would result to a better response of the output stage at higher frequencies. R8 and R9's values are also acquired from plugging in values and praying for good output. Their values lead to a ~670mA quiescent current on the output stage.

5. Overall Gain
As gleaned, with R6 and R7, I'd be getting about 31 voltage gain. I tried 11 as well 21 but I was having oscillations, distortions on the output, strangely I should have gotten better output with lower gain but, like before, I came up with that number through experimenting with what values gives the best.

That's pretty much it. I'd like to know if there are things I can improve on, maybe start on what should I critically consider first. Although I am able to get it working with satisfactory output in two simulators, Multisim and LTSpice, still I'd like to know if this design would do well or even work in real life.

 

The frequency response looks quite good. Stable.

 

The frequency response looks quite good. Stable.

Thanks, that's good to hear. I am even quite surprised I was able to get it flat towards ~400kHz.

 

The output stage quiescent current is big, so your output stage will work in class A for the output power lower than Pout < (2*670mA* 8Ω)^2/(2 *8Ω ) ≈ 7.2W
I'm not saying that it is a bad idea.
Also, R4 is not needed, you can replace it with a short. And R5 looks big for 3mA of quiescent current. I would expect to see something in this range of 0.7V/3mA = 240Ω
And R6 should be equal to R2||R3 = 50kΩ. And you should replace R8 with a bootstrap circuit.
https://forum.allaboutcircuits.com/...acitor-in-power-amplifier.114339/#post-890349 (Rx = Ry = R8/2)
And if you would like to build this circuit do not forget to use a "big heatsink" and provide a thermal "connection" (thermal coupling) between D1 and Q4. Or used Vbe multiplier instead of D1, D2, and D3.

 

I removed one diode and the output did not change (no crossover distortion. With 3 diodes the amplifier heats with 40V x 0.67A= 26.8W all the time (even when not playing sounds). The idle current and heating are much less with only 2 diodes. The diodes should be mounted on the same heatsink as Q4 and Q5 to track their heating.

 

You have an excellent choice of reading material. I would add another author: J. Linsley-Hood. His "The Art of Linear Electronics" and "Audio Electronics" will tell you a lot about power amplifier design. "Audio Electronics" has a section which tells you the advantage of a current source over a tail resistor, and the advantages of a current mirror over simple load resistors.
Your coupling capacitors form a high-pass filter with a cut-off frequency of 1/(2πRC). It's not always perfectly obvious which R to use. C3 is the most obvious - it's the load resistance.
For C2 it's R7
For C1 it's R2 in parallel with R3, but it is also in parallel with the input resistance of the long-tailed pair, which is significant.
Don't just set them all to 20Hz otherwise each will contribute a 3dB attenuation at 20Hz, and your output will be 9dB down at 20Hz.

 

I corrected the trans command so you can see the waveform:

 

The output stage quiescent current is big, so your output stage will work in class A for the output power lower than Pout < (2*670mA* 8Ω)^2/(2 *8Ω ) ≈ 7.2W
I'm not saying that it is a bad idea.
Also, R4 is not needed, you can replace it with a short. And R5 looks big for 3mA of quiescent current. I would expect to see something in this range of 0.7V/3mA = 240Ω
And R6 should be equal to R2||R3 = 50kΩ. And you should replace R8 with a bootstrap circuit.
https://forum.allaboutcircuits.com/...acitor-in-power-amplifier.114339/#post-890349 (Rx = Ry = R8/2)
And if you would like to build this circuit do not forget to use a "big heatsink" and provide a thermal "connection" (thermal coupling) between D1 and Q4. Or used Vbe multiplier instead of D1, D2, and D3.

Thanks for this! indeed, I changed R5 to a 220Ω ohm one. I also changed R6 for 50k, I readjusted the gain to about ~21. I have to get my input now to about 750mV to achieve the required 15.5Vpeak swing for 15W power output.

I have yet to apply the bootstrap circuit. If I understand, I can probably use this to lower my single supply of 40V but still get the rated power output?

I am thinking of indeed thinking of using a Vbe multiplier instead of the diodes, but I can't wrap my head around how they would work in single supply setup like the one I have. So, I am still researching on it.

 

I removed one diode and the output did not change (no crossover distortion. With 3 diodes the amplifier heats with 40V x 0.67A= 26.8W all the time (even when not playing sounds). The idle current and heating are much less with only 2 diodes. The diodes should be mounted on the same heatsink as Q4 and Q5 to track their heating.

Thanks for this one! Woah, my god you are right, my output stage current went down to mirco amps. I suppose it is now working in class B mode? Although noticed some kind of oscillation in my output with the diode removed.


But, when I replaced R1 to a tail emitter current source, the oscillation in the output went away, back to a clean sine wave.

hmmm...... I apologize for the file names haha. I have created multiple versions of the initial one so that I can track the changes in the circuit when I am following the suggested changes.

 

Thanks for this one! Woah, my god you are right, my output stage current went down to mirco amps. I suppose it is now working in class B mode? Although noticed some kind of oscillation in my output with the diode removed.

Also might be worth a try to replace ONE of the 1N4148 diodes with a Schottky instead of deleting it altogether.

 

You have an excellent choice of reading material. I would add another author: J. Linsley-Hood. His "The Art of Linear Electronics" and "Audio Electronics" will tell you a lot about power amplifier design. "Audio Electronics" has a section which tells you the advantage of a current source over a tail resistor, and the advantages of a current mirror over simple load resistors.
Your coupling capacitors form a high-pass filter with a cut-off frequency of 1/(2πRC). It's not always perfectly obvious which R to use. C3 is the most obvious - it's the load resistance.
For C2 it's R7
For C1 it's R2 in parallel with R3, but it is also in parallel with the input resistance of the long-tailed pair, which is significant.
Don't just set them all to 20Hz otherwise each will contribute a 3dB attenuation at 20Hz, and your output will be 9dB down at 20Hz.

Oh boy, thanks for the book! I had skim on The Art of Linear Electronics as well as Audio electronics. I like the way he writes! It seems simple and tries to avoid complex and technical terms, I also like the way he started from tubes to transistors. Really thanks a lot for this one. I think I would have myself occupied by these two the following days .

Your words on the coupling capacitors helped a lot in understanding in what values I was plugging in. So it is a high pass filter, I see. I computed their individual cut-off frequencies. I think I'm all right with their values so far. With C1 set to 1uF, I got its cut-off to ~3.18Hz. For C3, it is at 9.04Hz.

I am not sure about C2, I do not know if I have to set it at a certain value like the other two, all I know is that it should be large in order to keep the DC level in the negative feedback.

 

Also might be worth a try to replace ONE of the 1N4148 diodes with a Schottky instead of deleting it altogether.

Ooh, I think I got this one. That way, I can still get the output stage operate in AB and also maintain the output stage quiescent to a lower value than my starting 600mA+. I now understand the Vbe multipliers significance. Because when you suggested the Schottky diode, I immediately started hunting for ones with voltage drop lower than 600mV, even thinking if it is available in shops near me which is a hassle already unless I go to my local shop's inventory and just see what they have X_X. So yea, thanks a again for this one. I have to start working on how to get a Vbe multipler setup instead.


Edit, not only voltage drop, I have to consider their forward current as well to make sure they have the full drop. Which means altering my VAS stage quiescent current.

 

You have Q8 upside down.

It's often said that you should never use a time constant involving an electrolytic capacitor to determine the frequency response. Without C2, the DC gain is R6/R7 and it will amplify any DC offset from the input transistors, so it is advisable to roll off the gain to unity at low frequencies. Most designs choose a -3dB of about 3Hz.
Most designs also have positive and negative power supplies and are DC coupled on the output. That doesn't eliminate a large electrolytic, it just moves one from the output to the power supply, but it moves it from where its distortion and series resistance are noticeable to somewhere where they are not.

 

Ooh, I think I got this one. That way, I can still get the output stage operate in AB and also maintain the output stage quiescent to a lower value than my starting 600mA+. I now understand the Vbe multipliers significance. Because when you suggested the Schottky diode, I immediately started hunting for ones with voltage drop lower than 600mV, even thinking if it is available in shops near me which is a hassle already unless I go to my local shop's inventory and just see what they have X_X. So yea, thanks a again for this one. I have to start working on how to get a Vbe multipler setup instead.


Edit, not only voltage drop, I have to consider their forward current as well to make sure they have the full drop. Which means altering my VAS stage quiescent current.

The common ones have a part number starting with "BAT" and drop about 0.25V to 0.35V at the current you are using.
Vbe multipliers allow you to adjust the bias. They also allow someone else to adjust it wrongly. They also involve presets which can fail. They also can compensate for temperature changes if they are thermally coupled to the output transistors.The Holy Grail is an amplifier that sets the perfect bias current without needing a method of adjustment!

 

The common ones have a part number starting with "BAT" and drop about 0.25V to 0.35V at the current you are using.
Vbe multipliers allow you to adjust the bias. They also allow someone else to adjust it wrongly. They also involve presets which can fail. They also can compensate for temperature changes if they are thermally coupled to the output transistors.The Holy Grail is an amplifier that sets the perfect bias current without needing a method of adjustment!

I was hunting for the Schottkey diodes a while ago, I landed on seeing BAT41 as well as BAT46 which are 450mV, one is at 1mA forward current while the other is 10mA. Other parts are surface mounts in which I think looking at their names, I doubt my local store has stock.
I can get my VAS stage quiescent to 10mA instead of my 6mA. Small price to pay I think, so that I can lower my output stage quiescent current. I'll just see what I can do with BAT46.


With BAT46 and 1ohm emitter resistors on the output, I was able to bring down the large 600mA+ current down to 150mA. I am still somewhat unsatisfied as going by what John Linsley Hood Said about class AB output stage, it is typically 100mA quiescent current, I do not know if that is good. No distortion or clipping in the output so far. I've attached the file with the Schottkey diode in.

 

I am glad to see that you fixed Q8:

 

They also allow someone else to adjust it wrongly. They also involve presets which can fail. They also can compensate for temperature changes if they are thermally coupled to the output transistors. The Holy Grail is an amplifier that sets the perfect bias current without needing a method of adjustment!

About the bias of the output stage, I am thinking of going with this design.


R14 as well as R15 are able to give me a voltage drop of about 1.5-1.6V which is able to set the quiescent current of the output stage, measured at R9 and R10, to be 95mA-125mA. I think having a transistor instead of diodes, I can easily thermally couple it to Q4 and Q5. I have not ditched the diode and schottky yet though.

 

Make R15 a 100Ω preset in series with a 270Ω resistor.
Q10 should ideally be thermally coupled to Q4 and Q5.

 

Dore the new circuit simulate producing 18W with low distortion?

 

Dore the new circuit simulate producing 18W with low distortion?

Although I am not going for 18W, it seems capable of doing so without distortion, and actually, 18W seems to be the limit, trying to crank it higher, I am getting clipped outputs, but since I am only aiming for 15W, I think that is good?

The sine wave output seems clean at ~18W. I experimented between operating the output stage in AB and B. I found that as long as my output stage quiescent current (I measure it on R9 and R10) is about 100mA, I am getting satisfactory output. I've attached the circuit with the new changes.