I wan't to make a headphone amplifier that can drive any type of headphones

Will the dual supply with a virtual ground (as I linked) work?

Click to expand...

I thought my statement was clear so far: As long as the said DC path is missing, it won't work. Using a DP switch and ground connection of battery center point can easily solve the issue...

Im not following. Why cant a virtual ground work? I can't get a real dual sided supply from a wall-wart

A virtual ground can work, if it has sufficient low DC impedance. A pure capacitor divider doesn't. As said, the amplifier output must be expected to source at least a small DC current. It must be absorbed by the virtual ground circuit.

Okay. I've now gotten my TA6120. I see that the space between the pins are quite smaller than the DIP standard, but I managed to solder everything on and test it (allthough the result could have been prettier). The amp worked for like two seconds, before it (I assume) started oscillate (the sound got low and distorted). I powered it with 2x9V batteries in series and used 2x220uF caps and 2x3.6kOhm resistors to make a dual sided supply with a virtual ground. I chose a reverse feedback resistor of 10kOhm (making the gain 10, right?) and I follows the circuit in the datasheet with a 50Ohm resistor on the input and a 10Ohm on the output. The load was one of the channels of a 32Ohm headset.

I have a couple of questions. How do I know which input and output resistor values to choose? Will the 50Ohm and 10Ohm be standard for every application or are there other considerations to be taking into account?

I am not etching PCB's so using SMT chips like this is a little pain. Is there any technique or equipment that can be used? Like a SMT to DIP socket converter or something? And will my solution (soldering the resistors directly to the pins) be less favorable than a propper PCB solution thinking of chance of oscillation?

Here is the datasheet by the way: https://www.ti.com/lit/ds/symlink/tpa6120a2.pdf

The problem is, that according to it's bandwidth, TPA6120 can be easily used as a VHF transmitter. This has some impact on layout requirements. Placing the IC over a ground plane, may be a RF vero board, with ceramic bypass caps placed near the supply pins is absolutely necessary. The feedback resistors should be dimensioned at least roughly near to the datasheet suggestions. Somewhat hidden in the datasheet, the device is a current-feedback amplifier, thus the level of the feedback resistor Rf sets the loop bandwidth.

I already told by doubts regarding virtual ground operation.

I will make the circuit again and try to get a better result, I will also add 2x0.1uF caps from the power pins of the chip to ground.

I don't understand why the virtual ground is the problem. I understand it if the circuit is drawing alot of power, that would mean smaller dividing resistors leading to more current loss, but the power isn't big at all in this circuit. How will the chip know if the supply is a "real" dual sided supply or a supply with a virtual ground?

Perhaps i misunderstand your virtual ground configuration, but I fear, that the DC output current will pull the virtual ground out of balance.

I tried it again, but it didn't work. I used 2x9V batteries to make a real dual sided supply. I've dessided to put this chip to rest for a while. I found it hard working with SMT without using PCB's and even if I would get it to work, I wouldn't be very comfortable with the result. I think it's better to use this chip when I've learned enough about SMT and how to etch pcb's.

I wan't to give the LM386-4 another go. Even though it won't blow up any 600Ohm headsets, it will do a far better job than any other equipment (like computers, cellphone etc.), I'm betting that you hardly won't get any sound at all using a 600Ohm headset on a cellphone

There are two problems though. Even though the amp itself works, there are some minor issues. There is a constant hissing sound which is quite loud when nothing else is on the input. I was told that this was due to the amp itself and that putting a resistor with a value of 10x the headphone impedance value in series with the loads would fix this. So I put 620Ohm resistors in series with the output and it did indeed work, the problem is that my headphones are now at avarage only receiving 10% of the voltage it did before, so it starts clipping at a really low volume. Is there anything else that could fix the hissing without dropping the voltage? Lowering the gain might fix it, but the datasheet explains that the gain is set internally to 20, and with a combination of a cap and a resistor I can vary the gain from 20 to 200, but not lower then 20 It said that by putting a cap between pin 1 and 8, it would bypass the internal resistor thus making it a gain of 200. I don't quite get this, if the internal resistor is bypassed then the gain should be 1, right? A higher feedback resistor leads to a higher gain, right?

The second issue is with these new wall-warts I just bought. I've previously used a rather big psu with an adjustable output (I'm betting that it's of decent quality) and that worked fine. With these new wall-warts it will produce a constant buzzing noise, like the amp is beeing fed a 50hz sinewave through the input. I've been told that replacing my current 330uF decoupling cap with a 4700uF cap will reduce the noise drastically. I'm assuming that the buzzing comes from the wall-wart having a 50hz sinewave mixed into the DC and that a big cap would filter out this AC. A 4700uF cap sounds very drastic with such a small application. These caps are big and expensive, isn't there another solution to the problem? And why isn't my 330uF cap doing the job? I understand how the decoupling cap work, the cap is constantly charged to the supply voltage, and then discharged through the amp when/if the supply voltage falls below the cap voltage. I understand that if my amp is drawing alot of current, the cap's voltage will be lessened by quite a bit when the supply voltage falls, but how can the two LM386 (with no input) cause my 330uF cap's voltage to fall that drastically that the buzzing sound is made?

I wan't to open up the thread again.

I've come to the conclusion that the LM386 isn't doing what I wan't it to, one of the main reasons is the set gain of 20, the noise coming from this amp is just too great.

I'm expecting some SOIC to DIP converters to land in my mailbox in a couple of days so I wan't to give the TPA6120 another chance. I've abandoned the idea with battery operation. It's not worth the fuzz and I don't think it would be that much of a use for me anyway. I wan't to go with a center tapped transformer instead (no virtual ground operation). The transformer will take up abit of space, but it doesn't matter, the extra weight will make the amp feel more solid aswell

My worry is all this talk about oscilliation. I read the datasheet once more and saw that if just one of the amps was to be used (as I did the last time I tried), a feedback resistor along with a resistor from the output to ground had to be put on the amp not beeing used in order to prevent it from oscillating. I did not do this and might explain why it didn't work. I also read about propper testing, scoping the output looking for oscillation in the Mhz region, oscillation that might not even be that easy not notice by just listening to the sound. This is what's worrying me, why can't the amp be eigher HIGH or LOW, work or not work? And why can't a standard design be made, a design that ensures that the amp isn't oscillating? I get the feeling that theres some devine beeing that kickstarts oscillation at random when he feels like it.

Here's the design I was thinking about:



Let's say I make every resistor aswell as the 0.1uF caps kiss the pins of the chip, what else can I do to prevent it from oscillating?

Usually the best thing is to check the evaluation kit document for the ic, to glean some ideas from the manufacturer's implementation. See here. If you can find them, then the etched SMD prototyping boards are a good way, you could place it over a ground plane. See here for an example pic of a different device soldered. If you look closely, you'll see there are SMD resistors and capacitors in the
pic too; SMT looks intimidating but actually is not too hard at this size (practice with some cheap devices first ; )

Usually the best thing is to check the evaluation kit document for the ic, to glean some ideas from the manufacturer's implementation. See here. If you can find them, then the etched SMD prototyping boards are a good way, you could place it over a ground plane. See here for an example pic of a different device soldered. If you look closely, you'll see there are SMD resistors and capacitors in the
pic too; SMT looks intimidating but actually is not too hard at this size (practice with some cheap devices first ; )

Click to expand...

I read the user guide. I'm glad there so much easy to understand documentation on this device

SMT doesn't look that intimidating, it seemed easy enough when I watched this video: https://www.youtube.com/watch?v=ihoX7x0RBz8&t=4m0s. I've tried searching on ebay for those SMT prototyping boards, but I can't seem to find it. I also can't see how a result like that can be pretty, I would expect it to look very like the board in the picture you sent.

I'm sure the result would have been absolutely wonderful if I were to make a pcb, but I'm not quite there yet. Me and a friend of mine have talked about making pcb's, even bought in alot of what we need, but I feel we need some experience before etching something for a chip like this.

I also wonder why you wan't me to go with sm devices? I can't avoid jumper cables with a smt prototyping board anyway, bessides, I can get the caps and resistor pretty close to the chips pins regardless. As for paracitics, isn't it less of them in a decent jumper cable rather than a tiny pcb trace?

Ordinarily you'd want decoupling caps within a few millimetres of the IC, which is sometimes difficult with a SMD->DIL adapter, which
ends up being 1-2cm of track and pin length, depending on the exact type. However, it may be ok.
**broken link removed** is the type of thing which can be used, it is quite cheap. Agree, it is not pretty, but it's for
prototyping. Alternatively, a PCB is not too expensive, e.g. **broken link removed** offers 5 boards for $55, and they are double-sided
so you could have a nice ground plane. Etching yourself is even cheaper, but agree it takes a bit experience before you get good results.
As long as you've got the pre-sensitised boards and very sharp and opaque track lines, then you're virtually guaranteed a good result (at least
for single-sided). Hot water and ferric chloride work ok, as long as you gently stir it regularly.
However getting the sharp/opaqueness is quite hard. There are companies who can do photoplotting onto a sheet for you, but they charge a lot.
The results are perfect with that. I've not tried with a printer, but I guess experimentation is the key. If you can hold the dense areas in front of
a lamp and see no faint light through it, only then proceed : )

EDIT: I suppose that even if the printer was not completely dense you could touch-up with a pen marker, since the SOIC outline is still
pretty large for SMD these days. (i.e. "go over" the printed tracks with a pen marker, or maybe try to print again if you could align the
transparent sheet 100% accurately.. anyway, this is digressing from the point of your original post.

Ordinarily you'd want decoupling caps within a few millimetres of the IC, which is sometimes difficult with a SMD->DIL adapter, which
ends up being 1-2cm of track and pin length, depending on the exact type. However, it may be ok.
This is the type of thing which can be used, it is quite cheap. Agree, it is not pretty, but it's for
prototyping. Alternatively, a PCB is not too expensive, e.g. this company offers 5 boards for $55, and they are double-sided
so you could have a nice ground plane. Etching yourself is even cheaper, but agree it takes a bit experience before you get good results.
As long as you've got the pre-sensitised boards and very sharp and opaque track lines, then you're virtually guaranteed a good result (at least
for single-sided). Hot water and ferric chloride work ok, as long as you gently stir it regularly.
However getting the sharp/opaqueness is quite hard. There are companies who can do photoplotting onto a sheet for you, but they charge a lot.
The results are perfect with that. I've not tried with a printer, but I guess experimentation is the key. If you can hold the dense areas in front of
a lamp and see no faint light through it, only then proceed : )

EDIT: I suppose that even if the printer was not completely dense you could touch-up with a pen marker, since the SOIC outline is still
pretty large for SMD these days. (i.e. "go over" the printed tracks with a pen marker, or maybe try to print again if you could align the
transparent sheet 100% accurately.. anyway, this is digressing from the point of your original post.

Click to expand...

The adapter you linked is exactly what I have ordered. I'll give it a shot, if it doesn't work I can try going for a smt solution.

One thing I really wan't with this amp is a bass boost, or maybe a treble dampener. The treble can be rather intense at high volumes. I read that a capacitance in the feedback path would be a bad idea, maybe I can just put a low-pass filter on the input?

I guess you could use a low-pass at the input, but it may not be ideal. There has been a lot written about different types of tone controls, but unfortunately I'm not really super-knowledgeable on audio : ( Maybe search for "baxendall tone control". Or you might be able to adjust at the source (e.g. ipod allows you to adjust the tone from within it's menus).

I finally got my SMT-dip adapters. I soldered the 4 0.1uF caps on top of the adapter (to minimize the space between the chip's pins and the caps) so that the distance is maybe 6-8mm from the negative supply pin, and maybe 4mm from the positive supply pin. I found out that my 12-0-12V tranformer is output a too high voltage so I had to settle with 2x9V batteries for now. I also used 2x4700uF caps in parallel with the 0.1uF caps (Didn't have any other electrolytic caps), but the amp is oscillating I get one of the channels to work pretty well, allthough it seems that I'm not getting the gain of 5 that I set (the sound is pure and bass-heavy though), but this channel seems to drift into and out of oscillation after some time The other channel is just making some crackling noises allthough it seems to start working a short second after I disconnect one of the batteries (can't begin to emagine how).

This really frustrates me How am I supposed to be sure that my result is stable if it's this easy to make it oscillate? Also, the difference in the paracitic inductance and resistance has to be absolutely miniscule between 1mm and 6mm of a conducting path? Even if I did turn to all SMT parts (resistors and caps aswell), what guarantee can I have that it won't oscillate then?

You might need to post a photo of your layout, so that it is possible to see what can be done to fix this.
Also, you may wish to replace the 4700uF caps with the recommended 100uF and 10uF ones as shown in the data sheet
application circuit, and make sure they are 'low ESR' ones (they tend to cost a bit more than normal electrolytics,
a few dollars or maybe £1.50 each). Just 47000uF ones will not help for this circuit possibly. What kind of 0.1uF
capacitors did you use? Were they ceramic?
Which circuit are you following, the one on the application circuit? The components it mentions are all fairly important,
including the series 10-100 ohm resistors for the audio output.

Well, the 4700uF electrolytic caps are not low-esr (as far as I know), they are as cheap as they come. The 0.1uF caps are regular ceramic disc caps.

I used the circuit mentioned earler in the thread. I use 5.1k resistors as reverse feedback resistors. 1k resistor from inverting input to ground, 50ohm input resistor and 10 ohm output resistor. I'm testing it with a 30ohm headset (I have a 62Ohm headset aswell, I just don't want to try that one before I know the amps working, this headset is kind of expensive).

Replacing the 4700uF caps with 10 or 100uF caps is no problem, but does ESR increase with Farads when talking about electrolytic caps?

The datasheet mentions that Rf needs to be extremely close, and usually that would mean a surface mount device within a couple of mm
of the package, or on the underside of the board. It also says something similar about Ro. It may be worth trying to solder Rf directly
over the top of your ic, rather than on the stripboard. I suspect it may be the culprit. Also, nowadays there are very low ESR caps,
and possibly I'd be worried that the 4700uF ones are not low ESR if they were very cheap. Better to use the 100uF and 10uF low-ESR
ones which will also be smaller than 4700uF, and will allow you to place them much closer like the ceramic ones in your photo. I think in this
case, the distance of the caps may also be having an effect.

Hmm. I guess I have to buy some low ESR electrolytic caps and then turn to smt

One thing though. The datasheet sais that adjusting the output or input resistor might be wise in order to make the amp more stable. I guess I could put a pot in there, but I guess it's easier to use a constant smt resistor, bessides, I wan't this amp to be used with every common source and most types of headset and it wouldn't be ideal if the input and output resistor had to be adjusted every time the source or headset has changed. So what input and output resistor is the best in my scenario?