[Solved] Best way to amplify a sine wave without distorsion.

[HoracioDos]

Hello.

I'm building a Capacitive Touch Morse Keyer with an Arduino and a MPR121 module.

I managed to make a Wien bridge oscillator with all calculations to generate a quite nice 860-960Hz sine wave. Later I moved the wave above 0V and turn it on and off with a PWM, now I want to amplify it to feed a VHF handheld radio mic with a minimum distorsion. The circuit is attached.

Radio's in/out interface is done. It was a previous project and the arduino programming is also finished.

I thought I could use a class AB amplifier circuit like this.


I have read the pros and cons of A, B and AB amplifiers circuits but as I still don't fully understand transistors and it's calculations, I used a LM386 with a 20 gain config (I guess it's a class A) but I found some problems.

There is always a hiss on the output when the input signal is off and I can't hear the output when the signal is on but there is no hiss and the wave in the scope is very distorted.

Also found that I can hear a commercial radio when the IC input is not wired. I didn't know it was possible.

Should I keep using the IC and make it work or should I use the class AB transistor amplifier circuit? What is best or easier or both?

Thanks in advance!

[FrankBuss]

I don't know the answers to your questions, but your attached Fritzing circuit diagram is difficult to understand. Some obvious problems: You should remove the long bottom wire from -5V to the output of the negative voltage power supply IC, it is sufficient to show that it is connected if you use on both sides the -5V symbol. For C5 you should not connect it to the same GND symbol as used for the neg input of the LM386, place an additional ground symbol, because it has nothing to do with the neg input.

The outputs of the LM358 are not connected, are you sure this is what you built? And would be nice to have standard OpAmp symbols for the two OpAmps in the LM358. I don't know Fritzing, but better PCB CAD programs like KiCad or Eagle don't limit the number of independent symbols per IC packages. You should use one of those programs, if Fritzing doesn't support it.

[FrankBuss]

PS: if you really use a PWM signal to switch your sine wave on an off, it will produce a lot of distortions. Just use a digital out pin of your Arduino and set it to 1 or 0.

[Audioguru]

Like most audio amplifiers, the LM386 is class-AB. It has hiss because it is cheap. Its internal gain is 4500 times and it uses negative feedback to reduce the gain to a useable amount and cancel most of its distortion. With a 5V supply its output power before clipping distortion into an 8 ohm speaker is only 0.14W which is almost nothing. It works best with a 9V supply then its output is 0.5W.

Your transistor "amplifier" has a gain of only a little less than one and it has plenty of distortion. 

[HoracioDos]

Thank you for your comments

You should remove the long bottom wire from -5V to the output of the negative voltage power supply IC, it is sufficient to show that it is connected if you use on both sides the -5V symbol.

I know it's easier to follow it in the way you describe. Fritzing saves the breadboard view, schematic and pcb traces in only one file and creates ratsnets lines in the PCB tab, but if I wire this ways it's easier to convert the netlist to traces in the pcb designer tab later.

For C5 you should not connect it to the same GND symbol as used for the neg input of the LM386, place an additional ground symbol, because it has nothing to do with the neg input.

The circuit has a 20 gain config just like it's shown in the datasheet. I can't find any obvious error on this. Am I missing something?
https://iamzxlee.files.wordpress.com/2014/05/gain-20.jpg
Here is another example.
https://iamzxlee.files.wordpress.com/2014/05/lm-386-main.jpg?w=736&h=424

The outputs of the LM358 are not connected, are you sure this is what you built? And would be nice to have standard OpAmp symbols for the two OpAmps in the LM358. I don't know Fritzing, but better PCB CAD programs like KiCad or Eagle don't limit the number of independent symbols per IC packages. You should use one of those programs, if Fritzing doesn't support it.

Yes, I'm aware of that. Pins 1,7 and 8 are not needed. I'm trying to slowly move from Fritzing to Kicad. I'm sure it's a better EDA tool it's not as friendly as Fritzing and the learning curve is steep.

[HoracioDos]

Hello!

PS: if you really use a PWM signal to switch your sine wave on an off, it will produce a lot of distortions. Just use a digital out pin of your Arduino and set it to 1 or 0.

As far as I understand for a non-PWM pin, analogWrite() with a value below 128 is the same a digitalWrite(LOW) and analogWrite() with a value above 127 is the same a digitalWrite(HIGH). I'll check the program again to verify what i'm doing, but I don't get a distorted signal. Here is the output before it goes to the amplifier stage.

[HoracioDos]

Like most audio amplifiers, the LM386 is class-AB. It has hiss because it is cheap. Its internal gain is 4500 times and it uses negative feedback to reduce the gain to a useable amount and cancel most of its distortion. With a 5V supply its output power before clipping distortion into an 8 ohm speaker is only 0.14W which is almost nothing. It works best with a 9V supply then its output is 0.5W.

That's a problem. I want to power it with a usb port or cell phone charger. I only need the minimum power required to feed a VHF handheld mic.

Your transistor "amplifier" has a gain of only a little less than one and it has plenty of distortion.

I found this circuit that I want to test. Would you say that the tone could not be a clean sine wave before I build it?

[FrankBuss]

For C5 you should not connect it to the same GND symbol as used for the neg input of the LM386, place an additional ground symbol, because it has nothing to do with the neg input.

The circuit has a 20 gain config just like it's shown in the datasheet. I can't find any obvious error on this. Am I missing something?
https://iamzxlee.files.wordpress.com/2014/05/gain-20.jpg

There is nothing wrong with the netlist, only how it is drawn. In your version at first glance it looks like C5 has something to do with the neg input of the LM386, even if that's not the case. That's why I suggested to use 2 GND symbols, one for C5 and one for the neg input of LM386. Sorry when my phrasing is a bit odd, I'm German

[HoracioDos]

Sorry when my phrasing is a bit odd, I'm German

LOL!! No problem. I feel the same many times!!!. English is not my mother language either or mother tongue or whatever..

[Zero999]

Pins 1,7 and 8 are not needed. I'm trying to slowly move from Fritzing to Kicad. I'm sure it's a better EDA tool it's not as friendly as Fritzing and the learning curve is steep.

I think you'll find that pin 1 (the output of the first amplifier) is needed. You must mean pins 5, 6 & 7 aren't needed. The unused op-amp should be correctly terminated, not simply left unconnected, which can cause trouble. Configure the unused op-amp as a unity gain buffer, with the input grounded. Connect the inverting input (pin 6) to the output (pin 7) and the non-inverting (pin 5) input to 0V.

I seriously doubt you actually built the circuit as drawn. What is Q2 supposed to do? Its gate is just connected to the source so it will be off all the time.

As drawn, there's no automatic gain control on the oscillator so the signal will either slowly fade away or clip to form a horribly distorted waveform. Perhaps Q2 is supposed to be the gain control.

Why not AC couple the signal from the Wein bridge oscillator to the LM386 amplifier?

It should also be possible to build a Wein bridge oscillator using a single power supply so you can eliminate U3. Heck, I'd consider removing the U1 and using the LM386 to make the Wein bridge oscillator. There's even an example on the data sheet.
http://www.ee.ic.ac.uk/pcheung/teaching/DE1_EE/Labs/LM386.pdf

[Audioguru]

An LM386 power amplifier has negative feedback so that its gain is only 20 but without negative feedback its gain is about 4500. Its datasheet shows that a capacitor added between pin 1 and pin 8 reduces the negative feedback so that the gain becomes 200.

Your first circuit with many problems drives an 8 ohm speaker.
Your last circuit drives a whopping high signal into a very low level microphone input. Its needs an attenuator.

Your last twin-T filtered "sinewave" oscillator has no automatic-gain-control so its output will probably rise high enough for the transistors to cause clipping that is almost a squarewave.

[Richard Crowley]

If you are only sending out a single tone, I would just make a classic L/C or R/C filter which could take your 5V square wave and make a nice sine-wave out of it.  No fooling around with linear amplification etc.

[HoracioDos]

Pins 1,7 and 8 are not needed. I'm trying to slowly move from Fritzing to Kicad. I'm sure it's a better EDA tool it's not as friendly as Fritzing and the learning curve is steep.

I think you'll find that pin 1 (the output of the first amplifier) is needed. You must mean pins 5, 6 & 7 aren't needed. The unused op-amp should be correctly terminated, not simply left unconnected, which can cause trouble. Configure the unused op-amp as a unity gain buffer, with the input grounded. Connect the inverting input (pin 6) to the output (pin 7) and the non-inverting (pin 5) input to 0V.

I seriously doubt you actually built the circuit as drawn. What is Q2 supposed to do? Its gate is just connected to the source so it will be off all the time.

As drawn, there's no automatic gain control on the oscillator so the signal will either slowly fade away or clip to form a horribly distorted waveform. Perhaps Q2 is supposed to be the gain control.

Why not AC couple the signal from the Wein bridge oscillator to the LM386 amplifier?

It should also be possible to build a Wein bridge oscillator using a single power supply so you can eliminate U3. Heck, I'd consider removing the U1 and using the LM386 to make the Wein bridge oscillator. There's even an example on the data sheet.
http://www.ee.ic.ac.uk/pcheung/teaching/DE1_EE/Labs/LM386.pdf

Hello!

I'm a little bit confused. Let's go by parts...

I wrote that LM386 pins 1,7 and 8 are not needed, not the LM358.

I didn't use the LM386 for a Wien bridge for these reasons:

a) I first began using a LM741 to generate a sinewave and then switched to a LM358.
b) I didn't want anything with a light bulb.
c) LM386 was introduced later in the design as an audio amplifier.

Yes, I'm aware I'm not using the 2nd op-amp in the LM358 based on reason a) and I can't see why it should be a problem. There's an image of the sine wave before going to the LM386 amplifier stage and it's clean.

I decided to use a LM386 to amplify the sine wave because I have read in many websites that a LM358 is not an audio amplifier and It shouldn't be used as one.

Also, I don't fully understand the calculations to build a class ab amplifier with transistors and I didn't want to do it by trial and error. So a low power amplifier should do it easily with less components.

R5 and R6 move the signal above 0V to avoid the transistors could clip the signal. I tested it without the resistors and the sine wave was awful. If I'm wrong I would like to fix that.

Q1 and Q2 was the best way I found to turn on and off the sinewave from the arduino without any distorsion. Arduino pin is always HIGH and switched to LOW when the keyer is touched. If there's a better way, I'll love to learn it.

Yes, You are right, there's no automatic gain control to fade away the signal when it's cut off. I never thought it should be necesary. Perhaps it's a big mistake. I attached an image from the DSO when it happens.

The circuit was built exactly as it's described. I'm sure it can be optimized to be more simple with less components.

First I have to find out how to do the calculations with the proposed circuit and replace the light bulb with something else to get a 800hz sine wave.

In the mean while I would like to find out how to amplify the signal and take advantage of what I've done.  Once it's working I'll try to replace the LM358 with the LM368, drop the ICL7660 and leave only one IC.

[HoracioDos]

An LM386 power amplifier has negative feedback so that its gain is only 20 but without negative feedback its gain is about 4500. Its datasheet shows that a capacitor added between pin 1 and pin 8 reduces the negative feedback so that the gain becomes 200.

Thanks for the explain. I saw the 200 gain config in the datasheet.

Your first circuit with many problems drives an 8 ohm speaker.

Ok, I could try to use 9 volts as you told me previously or try something else.

Your last circuit drives a whopping high signal into a very low level microphone input. Its needs an attenuator.
Your last twin-T filtered "sinewave" oscillator has no automatic-gain-control so its output will probably rise high enough for the transistors to cause clipping that is almost a squarewave.

I thought the output could be too high for the input mic but as It's a published circuit in a Ham Radio Magazine, well I guess I was wrong.

[HoracioDos]

If you are only sending out a single tone, I would just make a classic L/C or R/C filter which could take your 5V square wave and make a nice sine-wave out of it.  No fooling around with linear amplification etc.

LOL!. That was my very first option and I failed miserably. Large story with capacitors, inductors and calculations.
Do you have some circuit at hand?

[HoracioDos]

Your last circuit drives a whopping high signal into a very low level microphone input. Its needs an attenuator.

I was reading your comment again, Do you refer to the circuit that I built not the one published (The third one). If that's the case, yes, it was already in my mind, I thought in a voltage divider with a potentiometer. I don't know if it's the correct way to do an attenuator.

Thanks again!

[Zero999]

Pins 1,7 and 8 are not needed. I'm trying to slowly move from Fritzing to Kicad. I'm sure it's a better EDA tool it's not as friendly as Fritzing and the learning curve is steep.

I think you'll find that pin 1 (the output of the first amplifier) is needed. You must mean pins 5, 6 & 7 aren't needed. The unused op-amp should be correctly terminated, not simply left unconnected, which can cause trouble. Configure the unused op-amp as a unity gain buffer, with the input grounded. Connect the inverting input (pin 6) to the output (pin 7) and the non-inverting (pin 5) input to 0V.

I seriously doubt you actually built the circuit as drawn. What is Q2 supposed to do? Its gate is just connected to the source so it will be off all the time.

As drawn, there's no automatic gain control on the oscillator so the signal will either slowly fade away or clip to form a horribly distorted waveform. Perhaps Q2 is supposed to be the gain control.

Why not AC couple the signal from the Wein bridge oscillator to the LM386 amplifier?

It should also be possible to build a Wein bridge oscillator using a single power supply so you can eliminate U3. Heck, I'd consider removing the U1 and using the LM386 to make the Wein bridge oscillator. There's even an example on the data sheet.
http://www.ee.ic.ac.uk/pcheung/teaching/DE1_EE/Labs/LM386.pdf

Hello!

I'm a little bit confused. Let's go by parts...

I wrote that LM386 pins 1,7 and 8 are not needed, not the LM358.

I apologise for the confusion.

I was indeed talking about the LM358, which doesn't have its output connected to anything, so it can't work. Clearly you are getting a sinewave out of this circuit, so the schematic you've posted here is different to what you've actually built.

The circuit was built exactly as it's described.

You must be mistaken.

Ignore the rest of the circuit for now. Focus on the Wein bridge oscillator. The output (pin 1 on the LM358 goes nowhere) therefore it can't possibly work, as drawn. If it does work, then you've built something different.

[Richard Crowley]

LOL!. That was my very first option and I failed miserably. Large story with capacitors, inductors and calculations.
Do you have some circuit at hand?

Google returned many interesting options when I searched for: 3rd order notch filter
Of course you need the "opposite" of a notch filter, but it is trivial to invert a filter circuit to make a sharp (high-Q) bandpass filter.

[Audioguru]

Your "keyer" changes the DC input on the LM386 audio amplifier so it will make a loud POP sound each time it is keyed and again when the key is turned off.
The LM358 is the dual version of the LM324 quad. They are low power so they are missing output biasing and create crossover distortion.

[FrankBuss]

Your "keyer" changes the DC input on the LM386 audio amplifier so it will make a loud POP sound each time it is keyed and again when the key is turned off.
The LM358 is the dual version of the LM324 quad. They are low power so they are missing output biasing and create crossover distortion.

Interesting, I didn't know that the crossover distortion is so high for such a common OpAmp. Is there any other standard OpAmp with lower distortion? I found the OPA388, for which the datasheet says "zero-crossover", "zero-drift" and it is not too expensive, but max supply voltage is very low.

[sibeen]

Q1 and Q2 was the best way I found to turn on and off the sinewave from the arduino without any distorsion. Arduino pin is always HIGH and switched to LOW when the keyer is touched. If there's a better way, I'll love to learn it.

But how does Q2 do anything? The way you have it configured it is permanently off and cannot possibly be turned on.

[Zero999]

Your "keyer" changes the DC input on the LM386 audio amplifier so it will make a loud POP sound each time it is keyed and again when the key is turned off.
The LM358 is the dual version of the LM324 quad. They are low power so they are missing output biasing and create crossover distortion.

Interesting, I didn't know that the crossover distortion is so high for such a common OpAmp. Is there any other standard OpAmp with lower distortion? I found the OPA388, for which the datasheet says "zero-crossover", "zero-drift" and it is not too expensive, but max supply voltage is very low.

Most op-amps don't have crossover distortion like that. Even the old uA741 didn't have that much crossover distortion.

Fortunately it's fairly straightforward to remove the crossover distortion by ensuring only one of the transistors in the LM358's output stage is active at a time. This can be achieved by connecting a resistor between the output and either supply rail.

[HoracioDos]

You must be mistaken.
Ignore the rest of the circuit for now. Focus on the Wein bridge oscillator. The output (pin 1 on the LM358 goes nowhere) therefore it can't possibly work, as drawn. If it does work, then you've built something different.

I can't stop laughing at myself. How ridiculous I'm! I've learned a lot in this thread.

When I read your first reply, I thought: "Well, he didn't like the circuit even a bit. I'm a noob. I don't have any problems with that. He knows a lot more, that's why I'm here... but ...is he suggesting I didn't build the circuit way I said in front of a lot of people? Is he really saying that? I should post a picture to prove the circuit exists. This circuit have been giving me headaches for so long that I can draw it with my left ear" ... A big avalanche was already going down the hill eating everthing in it's path....

Then I read this reply yesterday and finally the snow swallowed my pride. I was so blind worried about my own interpretation of things that I didn't see what it was so obvious for you about the LM358 connections. I was so sure that the schematic was showing the reality, I would have bet a month salary! You would have loved to see my face when I checked the breadboard! So I must say it " YOU ARE RIGHT!!!"

So, first I must check the schematic very carefully against the breadboard, make corrections in the schematic and review all the previous replies.

Thank you very much!!!

[HoracioDos]

LOL!. That was my very first option and I failed miserably. Large story with capacitors, inductors and calculations.
Do you have some circuit at hand?

Google returned many interesting options when I searched for: 3rd order notch filter
Of course you need the "opposite" of a notch filter, but it is trivial to invert a filter circuit to make a sharp (high-Q) bandpass filter.

These were my first attempts to convert a square wave to a sine wave with transistors and rc network calculations and a 555 also. But the result was a triangle wave or worst.
http://www.learningaboutelectronics.com/Articles/Sine-wave-generator-circuit-with-a-555-timer.php
http://www.learningaboutelectronics.com/Articles/Square-to-sine-wave-converter-circuit.php
It could been easier for me to use an arduino to generate an square wave and find a way to convert it to a sine wave.
It have been a too long journey for such a small project.

[HoracioDos]

Q1 and Q2 was the best way I found to turn on and off the sinewave from the arduino without any distorsion. Arduino pin is always HIGH and switched to LOW when the keyer is touched. If there's a better way, I'll love to learn it.

But how does Q2 do anything? The way you have it configured it is permanently off and cannot possibly be turned on.

Hero999 opened my eyes. I must check if Q1 and Q2 for sure are connected as it's shown. The circuit is obviously working but the schematic may be wrong with this too.