Guitar Headphone Amplifier

[tech_builder]

Alright, time for the next episode. I’ve changed the clipping diode’s reference to be two diode drops from the positive supply, rather than being referenced from ground, as suggested by Hero999. This means that as the battery dies and the voltage begins to drop, the clip point will drop with it.

I made the changes Hero999 suggested with the bypassing capacitors, and what a change! I am really impressed with how well it worked. I tried a few different size capacitors and could not hear/see a difference between a pair of 1000uF caps and a pair of 220uF caps, so I stuck with the latter.

I made a few other small changes. The output resistors have gone from 5ohms to 1ohm as per Audioguru’s suggestion, and the 10n capacitor across the output feedback loop is now gone. I also used an OPA4171 quad opamp package, mostly because I had a smd to dip convertor for it, while I didn’t have one for the OPA1664 that I was originally going to try. This opamp worked great as far as I could tell, it was better than the TL082 I was using before at least.

Calexanian, I am curious about the other circuits you have used/come across for this purpose as you mentioned there are other circuits, what are they?

Thanks again, I think it's finally coming together. Next is the circuit board layout and enclosure making.

[Zero999]

I'm glad it worked. On reflection 1000µF was a bit overkill, 220µF is more than enough.

R10 is not needed, since the base current into the transistors is limited by the voltage change on the emitters. For example, if the voltage on T2's base increases, the voltage on the emitter will too, reducing the voltage between the base and emitter, thus limiting the base current. All R10 will do is increase the voltage drop.

[calexanian]

Going To have to dig for those. Have not used them in many years. I may have to take a picture of a schematic out of an old data book or re draw it or something.

[technix]

Maybe you should take a look on the LM386 chip as it packs the op amp and the power amplifier in one chip. Maybe you still need one op amp input stage for impedance matching but the second stage op amp and power amp can be merged.

Also when designing the PCB socket the chips. Different input stage op amps can have different performances so you can swap them in and out. Dual op amp chips almost always have the same pinout, and SO-8 adapter boards result in the same footprint as DIP-8, so you can safely use the TL082 DIP-8 footprint. LM386 when driven too hard for too long (like a big session of death metal) can fry if insufficiently cooled so socket them for easier replacement.

[akis]

I have looked quickly and maybe I am making a mistake, but can I just say that an electric guitar requires a very specific equalisation with serious chopping down the mids and amplifying the highs (5KHz) just to get the usual, clean sound. As was noted right at the start, a 47K bias resistor would steal all the high frequencies - it would sound *exactly* like when you plug your guitar into your stereo system, it really does not sound like a guitar at all. I believe the 1M resistor is too high and just asking for noise and offsets. I would use a 500K instead, it is very common.

But then guitar pickup amps have a ridiculous ceiling, and any amp will clip, unless you have equally ridiculous voltage rails. You cannot do it with 5V, or with 30V (2x15V). You have to account for clipping and make the clipping as symmetrical as possible and as smooth as possible.

Typical pre-amps found in quality valve amps have rails of 400V and two cascade input stages of Av=50 followed by another Av=50, letting the valves do the smooth clipping. Look out for the very common Fender tone stack to see how it aggressively cuts down on the mids allowing the highs to go through. That is more or less the sound you are looking for.

In terms of smooth clipping produced by a transistorised circuit, my best experiments have been with series of BAT48s.  I remember dozens of them, anti-parallel, and even then clipping can happen at any place (even inside the op-amp) so the design has to be tested on breadboard and check the scope to see the smooth curves (not square clips).

You are looking for around 15% distortion for the "crunch" sound and maybe 20-25% for the "metal" sound. Technically there cannot be "clean" sound (called "crystal" on some amps) because you do need valves for that or to turn down the pots on the guitar maybe.

[tech_builder]

Hero999: So I tried removing R10 and I got a lot of noise and waveform deformation, so I have decided to keep it in 

Calexanian: That would be great if you could post them here. Any format you can document it with is fine by me!

Technix: I know the LM386 is an easy way to make this thing, but I’m more interested in the design and build of a transistor amplifier. I don’t get to do much analog design at my job and I have forgotten most of it due to neglect, so I am reviving it with a fun project. Hopefully it will be one of many that I do, and having a great community like this one to help me along is fantastic! So bear with me if I take the hard way, it’s for a good cause.

Akis: Thanks for looking at the schematic. This first iteration of the guitar headphone amp is not going to have any designed in distortion. I am cutting it off pretty much where it is so that I can at least get something for my son to play through before he grows up and moves out! I’m interested in your comment about the input resistor R17. You mention that a 500k resistor to ground is common, where is this configuration typically used?

Thanks!

[Zero999]

Hero999: So I tried removing R10 and I got a lot of noise and waveform deformation, so I have decided to keep it in 

That's interesting. Don't you want to know why?

When you mean noise, does it produce a waveform, when there's nothing attached to the input? If you put a square wave in, does it ring?

Could it be that it's better with R10 because it reduces the open loop gain slightly, therefore stabilising the amplifier?

[akis]

R10 is a last ditch effort to save the current limiting of the op-amp. Get rid of those caps. Maybe reduce the biasing resistors to 3K each and remove the 220Rs.  I think headphones are 64R (are they?) so make the emitter resistors, say 5R each, even 10R each. If you do all that change R10 to something just for lip service, 10R, or remove it.

[akis]

I am not sure your ground is a good ground, trying to split the 9V supply into two. The op-amps draw a lot of current, as do the buffer transistors, and the 10K split supply resistors will allow what, .45mA, whereas each op-amp draws a few mA itself. In order for the split supply to work you'd need an impedance 10x, or 20x of the largest current draw, so for example if all the op-amps and all the buffer transistors drew, say, 12mA in total, you'd need a potentiometer of 1.2mA (10x) or 0.6mA (20x) to make some half-way house ground and even then. You could try to make a better ground by using a spare op-amp if there is any. Or better, do away with trying to split the supply. Or use 2 x 9V batteries, more headroom too!

[Zero999]

R10 is a last ditch effort to save the current limiting of the op-amp.

No, R10 shouldn't be relied upon for current limiting. If it's shorted the transistors would still have enough base current to allow a damaging collector current, made higher by the fact the current gain has a positive temperature coefficient.

Get rid of those caps. Maybe reduce the biasing resistors to 3K each and remove the 220Rs. I think headphones are 64R (are they?) so make the emitter resistors, say 5R each, even 10R each. If you do all that change R10 to something just for lip service, 10R, or remove it.

Perhaps you should read the thread more thoroughly? This is not a criticism, since I don't blame you for not having the time to read through a such a long thread. A similar configuration to what you've described has already been tried and found unsatisfactory. Here's a link to the schematic.
https://www.eevblog.com/forum/projects/guitar-headphone-amplifier/msg1214040/#msg1214040

Here's why:

If you remove the bypass capacitors then you'll need to also reduce the value of the biasing resistors to the point, where the quiescent current becomes very high to get the desired power output. Removing the 220R resistors may help reduce the quiescent current but it would increase distortion dramatically.

The headphones have an impedance of 32R per channel and are connected in parallel making 16R, so increasing the emitter resistors is a bad idea.

I am not sure your ground is a good ground, trying to split the 9V supply into two. The op-amps draw a lot of current, as do the buffer transistors, and the 10K split supply resistors will allow what, .45mA, whereas each op-amp draws a few mA itself. In order for the split supply to work you'd need an impedance 10x, or 20x of the largest current draw, so for example if all the op-amps and all the buffer transistors drew, say, 12mA in total, you'd need a potentiometer of 1.2mA (10x) or 0.6mA (20x) to make some half-way house ground and even then. You could try to make a better ground by using a spare op-amp if there is any. Or better, do away with trying to split the supply. Or use 2 x 9V batteries, more headroom too!

There is some truth in this but it's grossly exaggerated. The current taken by the op-amp and the buffer transistors makes absolutely no difference to the split power supply because it just passes from +V to -V and doesn't flow into or out of the 0V node. The biasing currents going in/out of the 0V node will also be insignificant. As long as the current from the +V and -V rails is equal, then no current flows in or out of the 0V rail.

The 0V rail is bypassed by two 220µF capacitors, which effectively make 440µF, which will present a low impedance path to 0V for the headphones and the input voltages, so that isn't a problem.

The places where the positive and negative currents are not balanced is what will cause the 0V rail to rise/fall. The most glaringly obvious one is R11, which is connected to +V, via a couple of diodes. Fortunately this is easy to fix: R11 could easily be connected to -V, without much change to the voltage across the diodes. R13 is more of an issue, although perhaps as it's a clipping detector, one would say it doesn't matter, since when the LED lights and starts to pull the 0V rail up, then the user should adjust the volume any way. That then leaves R27, which will leak a small current into the 0V point, more so at higher output levels.

[Zero999]

Looking at this again, DC can be avoiding in the 0V rail by referencing the peak detector circuit to 0V.

I would also consider upping the value of the bypass capacitors, to make a lower impedance path at AC.

[tech_builder]

Hero999 I had actually just made similar changes to the schematic. I had not increased the power supply caps though. I also increased R13 to account for the larger voltage drop across the LED. I haven’t specified an LED yet though, but I do want a low current one.

I am curious about the reason the removal of R10 caused the opamp to be so wonky. The output of the circuit when I remove R10 is shown in the first image below. There is quite a bit of noise in the output signal. I tried various resistors from 100 to 270ohm and found that 220ohm was the lowest value resistor that didn’t show any distortion; at least that I could see on the scope trace.

I remember classes talking about open loop gain, but I’ll have to dig a bit to pull that out of my memory. I do recall phase being important because I believe that as you get closer to 180 degrees out of phase your feedback becomes more positive instead of negative and no longer functions correctly.

Akis, as Hero999 already mentioned that was my original design. However, since I am using a fairly limited voltage I was running up against the rails and getting a lot of clipping during my testing. When I put the bypass caps on the output it really opened it up. I’m still not too sure how it is working though, perhaps someone can elaborate? I’m guessing it might be similar to how putting a bypass cap across a common emitter’s emitter resistor would increase its gain at signal frequencies? However, the capacitor is now across the base emitter...sorta. I have to think about it more.

Thanks again!

Edit: Yellow is input, Green is output.

[akis]

Hello

In my opinion

Supposing the load is 16R (someone said so above I think). My "pro" headphones are 64R but anyway. The op-amp cannot drive such a low impedance load, so we add a buffer stage. The low E is 80Hz. We need approx 15mW to drive good quality headphones. With that in mind and on simulation (rather than on breadboard)

1) set the frequency to 83Hz
2) try to get 15mW on the 16R load (around 500mV RMS, 1.42V p2p)
3) on my simulation it does not work

I believe it is because you are returning the power output from the speakers into your fake ground.

If we accept that the fake ground is unnecessary and does not work for returning loads to it, then I can suggest:

1) We can use a single supply design which basically only attempts to set the Vin+ of the op-amp to half the supply voltage , you can have huge resistors for that, eg 200K each.
2) We have to return the speaker to the V- and power the speaker through a large capacitor
3) Need a complementary pair to give us a lot of power without distortion, but we can get rid of the extra two transistors for slightly higher distortion
4) I attach a schematic for ideas

[BrianHG]

Hero999 I had actually just made similar changes to the schematic. I had not increased the power supply caps though. I also increased R13 to account for the larger voltage drop across the LED. I haven’t specified an LED yet though, but I do want a low current one.

I am curious about the reason the removal of R10 caused the opamp to be so wonky. The output of the circuit when I remove R10 is shown in the first image below. There is quite a bit of noise in the output signal. I tried various resistors from 100 to 270ohm and found that 220ohm was the lowest value resistor that didn’t show any distortion; at least that I could see on the scope trace.

I remember classes talking about open loop gain, but I’ll have to dig a bit to pull that out of my memory. I do recall phase being important because I believe that as you get closer to 180 degrees out of phase your feedback becomes more positive instead of negative and no longer functions correctly.

Akis, as Hero999 already mentioned that was my original design. However, since I am using a fairly limited voltage I was running up against the rails and getting a lot of clipping during my testing. When I put the bypass caps on the output it really opened it up. I’m still not too sure how it is working though, perhaps someone can elaborate? I’m guessing it might be similar to how putting a bypass cap across a common emitter’s emitter resistor would increase its gain at signal frequencies? However, the capacitor is now across the base emitter...sorta. I have to think about it more.

Thanks again!

Thant distortion is due to cross-over distortion + influenced oscillation.  Choose a different op-amp.

[Zero999]

Hero999 I had actually just made similar changes to the schematic. I had not increased the power supply caps though. I also increased R13 to account for the larger voltage drop across the LED. I haven’t specified an LED yet though, but I do want a low current one.

I am curious about the reason the removal of R10 caused the opamp to be so wonky. The output of the circuit when I remove R10 is shown in the first image below. There is quite a bit of noise in the output signal. I tried various resistors from 100 to 270ohm and found that 220ohm was the lowest value resistor that didn’t show any distortion; at least that I could see on the scope trace.

I remember classes talking about open loop gain, but I’ll have to dig a bit to pull that out of my memory. I do recall phase being important because I believe that as you get closer to 180 degrees out of phase your feedback becomes more positive instead of negative and no longer functions correctly.

Akis, as Hero999 already mentioned that was my original design. However, since I am using a fairly limited voltage I was running up against the rails and getting a lot of clipping during my testing. When I put the bypass caps on the output it really opened it up. I’m still not too sure how it is working though, perhaps someone can elaborate? I’m guessing it might be similar to how putting a bypass cap across a common emitter’s emitter resistor would increase its gain at signal frequencies? However, the capacitor is now across the base emitter...sorta. I have to think about it more.

Thanks again!

Yes, I think you're right about the resistor reducing the loop gain, thus increasing the stability. It's interesting how the oscillation only occurs on the negative going side of the waveform.

The AC coupling capacitors work by allowing the op-amp to the transistors both on as well as turn off, by allowing audio frequencies to bypass the diodes. Without them, the op-amp can only turn the transistors off, as the diodes prevent current flow in the other direction, so the drive current for the transistors has to come from the 10k resistors.

Hello

In my opinion

Supposing the load is 16R (someone said so above I think). My "pro" headphones are 64R but anyway. The op-amp cannot drive such a low impedance load, so we add a buffer stage. The low E is 80Hz. We need approx 15mW to drive good quality headphones. With that in mind and on simulation (rather than on breadboard)

1) set the frequency to 83Hz
2) try to get 15mW on the 16R load (around 500mV RMS, 1.42V p2p)
3) on my simulation it does not work

I believe it is because you are returning the power output from the speakers into your fake ground.

If we accept that the fake ground is unnecessary and does not work for returning loads to it, then I can suggest:

1) We can use a single supply design which basically only attempts to set the Vin+ of the op-amp to half the supply voltage , you can have huge resistors for that, eg 200K each.
2) We have to return the speaker to the V- and power the speaker through a large capacitor
3) Need a complementary pair to give us a lot of power without distortion, but we can get rid of the extra two transistors for slightly higher distortion
4) I attach a schematic for ideas

I agree about not using the false ground for returning power from the headphones but compound transistors on the output are a bad idea. They increase the voltage loss, clipping and could make the amplifier less stable, as they have more voltage gain than ordinary emitter followers. It's better to only have one set of emitter followers, which provide more than enough current gain to drive the headphones and AC bypass the biasing diodes with capacitors.

Thant distortion is due to cross-over distortion + influenced oscillation.  Choose a different op-amp.

I think you have a point regarding crossover distortion but why blame the op-amp when it doesn't have a class B output stage? That also doesn't explain why it disappears when the loop gain is reduced.

If there's crossover distortion, it will be down to the buffer and can be reduced by increasing the value of the 220R resistors in between the diodes slightly, which would increase the quiescent current drawn too. The voltage across the 1R resistors could be monitored, whilst adjusting the bias, to reach a compromise between quiescent current and crossover distortion, but since reducing the loop gain already fixes this, there's little point.

The OPA1664 is a 4 opamp package so I only need to power all of them once. I’m not too sure what to do with the unused opamp, any ideas?

You could use it to reduce the impedance of the 0V rail.

[Audioguru]

The application note for the LM380 little amplifier (similar to the LM386 little amplifier) says that the output produces high frequency oscillation on the negative going swing, exactly like we have here when R10 is removed. They fixed it by adding an external zobel series RC network from the output to ground.

[BrianHG]

Thant distortion is due to cross-over distortion + influenced oscillation.  Choose a different op-amp.

I think you have a point regarding crossover distortion but why blame the op-amp when it doesn't have a class B output stage? That also doesn't explain why it disappears when the loop gain is reduced.

If there's crossover distortion, it will be down to the buffer and can be reduced by increasing the value of the 220R resistors in between the diodes slightly, which would increase the quiescent current drawn too. The voltage across the 1R resistors could be monitored, whilst adjusting the bias, to reach a compromise between quiescent current and crossover distortion, but since reducing the loop gain already fixes this, there's little point.

My recommendation to try a different op-amp, is not a description of an exclusive problem being in that op-amp, it's just that some op-amps have a better time when feeding such a class AB buffer without going as haywire pronouncing that NPN/PNP switchover problem.

Another thing which might help is to tie a 100 ohm resistor between the output of the op-amp and the center of the 2 resistors inbetween the emitters of the 2 transistors.  This helps conduct a bit of signal inbetween the switch-on and off of the PNP and NPN transistors removing that distortion without sacrificing base drive current.

[tech_builder]

Thanks Akis. That is a neat idea. I’ve built it up in a very simple simulator (Everycircuit) and it seems to work pretty well. I guess it would still need another stage at the front to moderate higher level input signals since the pot wasn’t able to attenuate it to zero as a volume control (it just stopped it from clipping).

I like the idea of using the extra opamp to lower the ground rail impedance, it’s certainly better than doing nothing!

I’m not too sure what a zobel network is, but it doesn’t sound simple. Maybe I’ll take a look if I find some time.

I do have another opamp to try for this circuit BrianHG. It’s an OPA1664, which they state on the datasheet as being designed for audio. Not too sure if that’s going to make a difference here. I didn’t have a way to mount it into a breadboard since it only comes in surface mount packages. I did recently acquire the means to mount it into the breadboard, but now I just have to get motivated enough to rewire the spaghetti mess for a new opamp.

[Audioguru]

I like the idea of using the extra opamp to lower the ground rail impedance, it’s certainly better than doing nothing!

Yes, it is much better.

I’m not too sure what a zobel network is, but it doesn’t sound simple. Maybe I’ll take a look if I find some time.

A zobel network is a low value resistorat the output of an amplifier in series with a capacitor to ground. It provides a load at high frequencies where the inductance of a speaker or headphones is not a load.
Maybe your resistor should be 16 ohms and the capacitor can be 33nF. Here is an LM386 little amplifier with its zobel network:

"into a breadboard', "rewire the spaghetti mess"

A solderless breadboard has capacitance between its rows of contacts and the spaghetti mess of wires all over the place that guarantees a circuit like this to oscillate at a high frequency as you show. Use a compact pcb or a compact stripboard layout with the parts soldered on it instead.

[tech_builder]

Thanks Audioguru, I put the Zobel network on to the output of the amplifier and it cleaned up the signal as seen in the image below. I am currently going through the Learning the Art of Electronics book, which is a lab based learning tool that compliments the well known Art of Electronics text. I found an example of a simple speaker driver that they show as having a similar zobel network on the output, but they call it a “snubber”. It also has a 1k resistor to ground in parallel, which I haven't tried yet. Green is output, yellow is input.

You can see the spaghetti mess below. I have been wondering what the implications of such a setup would have on performance and how much I can trust the results I am getting. I am planning on making a PCB out of the design I have this far. Hopefully the performance will only get better!

Attached is a schematic of the state of the guitar headphone amplifier as it stands.

[Audioguru]

Like all solderless breadboard circuits, yours is a mess. There is stray capacitance and antenna wires that pickup interference all over the place.

The value of R1 and R2 that provide base current to the output transistors is much too high so the base current is almost nothing.
The 2N3904 and 2N3906 transistors cannot provide enough current even if R1 and R2 have low values.

Work it out:
1) The output voltage might need to be +1.6V or -1.6V. Then the current in an output transistor will be 1.6V/16 ohms= 100mA.
2) Look at the datasheet for a 2N3904 and 2N3906. The current gain drops sharply above 50mA. At 100mA the text says the minimum current gain at 100mA is 30.
3) The base current must be 100mA/30= 3.3mA.
4) The base voltage will be about 1.6V + 0.9V= 2.5V. Then the base resistor needs to be (4.5V - 2.5V)/3.3mA= 606 ohms, much less than your 10k.
You might need an output higher than 1.6V then the 2N3904 and 2N3906 will burn out.

2N4401 and 2N4403 little transistors have a minimum current gain of about 120 at 100mA so the base resistors can be about 606 x 4= 2424 ohms. Their maximum allowed current is 600mA.

You show a little 9V battery. It is too small to provide these high currents.
   

[Zero999]

The value of R1 and R2 that provide base current to the output transistors is much too high so the base current is almost nothing.

No. You've forgotten that R1 and R2 only need to provide enough current to bias the transistors into conduction when there's no signal to avoid crossover distortion. C7 & C8 bypass the high value biasing resistors at AC, thus providing a low impedance path for the base current from the op-amp's output stage.

[Audioguru]

No. You've forgotten that R1 and R2 only need to provide enough current to bias the transistors into conduction when there's no signal to avoid crossover distortion. C7 & C8 bypass the high value biasing resistors at AC, thus providing a low impedance path for the base current from the op-amp's output stage.

Yes, I forgot that the opamp drives plenty of base current into the transistors through the capacitors you added.

[alexanderbrevig]

I've been a long time reader of this thread but decided to reply mostly to get it in my "Show new replies to your post" section.
Also, to thank everyone! I've learned things reading this thread. 

[Zero999]

I thought I'd revisit this topic.

I've simulated both the compound transistor and emitter follower with bypass capacitor output stages. With both circuits, I adjusted the biasing to get rid of most of the crossover distortion, without too much bias current.

Both circuits have a gain of about 0.9, which is to be expected, as emitter followers always have a gain of less than one and some voltage is dropped across the emitter resistors.

The compound transistor circuit causes clipping at 1.2V less than the power supply rails, around +/-3.8V.

The emitter follower circuit doesn't clip. The output will be able to reach close to the power supply rails, minus the voltage drop on the emitter resistors and transistor saturation voltage.