doobooloo's TPA6120-based balanced headphone amp

Hello everyone!

So I am extremely bored over summer, and I was reminded of my old TPA6120-based balanced headamp project. I was excited as I found something to kill time with, but I realized that when I reformatted my computer a while back all my Eagle files and libraries that I had built up had perished.

But guess what. That just gave me another reason to design the same amp from scratch, just much better this time.

So after a few days of playing around, here is my design on a 2.5" by 4" board (fits in the smallest Lansing case):

Everything:


Just pads and traces:


Top layer:


Bottom layer:


Click here for higher-resolution pictures.

So what makes this amp exciting? To list a few features that I really wanted when I was designing this amp...

1. It is a fully balanced amp
2. It is extremely high-performance (TPA6120)
3. But, it takes single-ended inputs, allowing the use of numerous (read: simple, cheap, small, proven, good) sources
4. Outputs are in the form of two 1/4" jacks - one inverted and the other non-inverted - which allows the use of conventional headphones and balanced headphones with the use of a simple adapter cable
5. It employs the DS1802 digital potentiometer coupled with a rotary pulse switch allowing for easy and precise volume adjustment as well as a higher sound quality, better channel matching, and long-term durability compared to analog pots
6. All necessary jacks can be mounted onboard - at the same time standard 0.10" pads are provided for external wiring and the use of 3-pin Molex KK connectors if desired
7. A desired sound signature can be achieved by opamp selection (since it's SMD, no easy opamp rolling, though)

I used Eagle Light to design this board, so no groundplane fill function is available. Also, board space is limited by the Light version, and the only reasonable board size within these limitations with a widely available case was 2.5" x 4" using Lansing MicroPak cases.

Anyway... I have a breadboard prototype amp employing a very similar amp circuitry running and it sounds damn good. With a more optimized board design the amp section can only sound better... and this is very exciting.

The part that I am a bit nervous about is the DS1802 digital pot section which I took a lot of hints from Armond Chen's design and the accompanying Headwize thread. I replaced the LM336Z-2.5 with an LM4040 since the latter has higher accuracy and is available in a smaller surface-mount package. Other than that, I am pretty much following Armond's design for this board.

Anyway... any comments?

Oh, forgot to add... The amp requires a bipolar power supply, so something like the Elpac WM71 (+15V/-15V) with a male DIN-5 plug. Different PSU manufacturers, including Elpac, seemed to be using the same DIN-5 standard for dual and triple output power supplies so I thought that would be nice in terms of flexibility.

Of course, my opamp of choice is the AD8066 which can only take up to +12/-12V... I am probably going to build myself a simple dual-TREAD bipolar supply to go with this unit or make another small matching PSU PCB so that the two cases can neatly be stacked or something...

Another thing I forgot to add again... All the surface mount resistors and capacitors are in 0603 package.

How are you splitting the single ended signal to balanced?
DRV134s? those arn't that great last I checked.. the high performance amp circuit may be for nothing if you're using line drivers as splitters..

Quote:



Two dual opamps take in signal, acts as buffer into TPA6120 providing low impedance output. One TPA6120 is set at positive gain and the other at negative gain. Gain magnitude should both be equal, of course, just inversed.

True the DRV134 isn't great but it's not a bad rail splitter either. I found it better then discrete solutions when I tried it but it did fail to outperform Transformers.

My guess looking by the layout, one of the opamps will flip the signal.

The 2.5V -2.5V option for running the DS1802 works like a treat. The problem I had when integrating it in my balanced amp was that the input signal is often overpowering. I think 10V p-p are some of the balanced output signal swings, so I had to put a 10k resistor in series with the input to bring the level down slightly. A studer cdplayer that becomethemould dropped off here one day did however still clip the input.

The other thing is what rotary pulse switch are you using? Double check the datasheets to ensure the pulse time of the switch is long enough. I used a quadrature rotary encoder and the decoder for it didn't pulse long enough for the chip to register a click. In the end Daki][er sent up a uC with a new decoder programmed in.

Quote:



Thanks for the comments! The opamps are both configured identically, and it is one of the TPA6120 chips that does the flipping.

Regarding the input clipping... maybe I should make space for that option on the board. Two resistors to form a voltage divider would be better than a simple resistor though, right?

The rotary switch was simply taken from the Headwize thread. I have no clue if it will work well or not. I'll just have to try it - but for now that's really the only option I have since I don't want to set up a uC... If the prototype board fails with the rotary switch then I'm going to have to redesign with two pushbutton switches or something...

Also... regarding aos's comments on the Headwize thread:

Quote:



Is this necessary? I think I've heard some arguments both for and against this "dividing" of analog and digital grounds with an inductor or ferrite, so I am not sure what to do. I omitted it in this design but it wouldn't be too hard to add space for it.

Actually, maybe I can just make space for it so I can play around and simply jumper across it if it's not necessary?

Also, any suggestions for a PCB house for this small board?

sparkfun.com used to be ok for small boards.. might want to try them

thoughts:

1. not fully balanced, fully balanced would have balanced inputs and balanced outputs.

2. TPA chip works fine on +12/-12v and with balanced config the extra 6v swing in running 15/15 will make very little difference.

3. Amps in this configuration tend to have the sonic signature of the TPA chip rather than smd opamps, provided the smd opamp is fast enough. Very little difference in sound signature with the input opamp selection.

4. Wouldn't it make more sense to just have 4 electrolytics bypassed with 4 polyprop or polyester caps (possibly even 4 silver mica as well if you want to go extreme) to filter the power??

5. it looks like you have input ground connected to power ground which could cause a ground loop. Suggest maybe using a ground loop breaker or sticking to one or the other.

6. considering you're happy to mount resistors/caps on the opposite sides of the board you might want to think about mounting the resistors/caps on the tpa right under the pins or under the chip.

7. nothing stopping you having the input opamps socketable.

dooob. i think what AOS says is a good idea.. putting a ferrite there would help a fair bit with the hf noise (if any)..

4pcb is by far one of the nicest boardhouses evar. $33 / 60 square inches.. if you're a college student that is.. its a very nice deal to say the least.

unlike ******* in other sections, i line by line because you roxor.
Quote:



find a portable balanced source. this amp looks VERY well sized for portable use.

Quote:



it is actually closer to 12V of swing because it is balanced. FWIW the swing with a 12V sup-ly would be DOUBLE the swing of a single chip with a +-12V input probably about 40vP-P without getting anywhere near the rails yea, you dont need this.... drop suply to +12V/-12V. alternately, a 2X9V system could be used... you would still have about 32vP-P swing. can anyone say "uberportable"

Quote:



the tpa chip has the sound signature that is to sound like the source... it is VERY uncolored. i would wory more about the input opamps than the tpa chips.

Quote:



the chip used is SOO insensative to power suply "quality" that it is really unnecessary.
Quote:



you may run into space/heat issues from the power-pad solder point.
Quote:



+1, DO IT.

Thanks for all your replies! Regarding some comments...

Quote:



What impedance and DCR values should I look for when choosing a ferrite for that position? I'm looking at the Vishay-Dale SMD multilayer ferrite beads on Mouser here: http://www.mouser.com/catalog/626/710.pdf and I'm not quite sure which will work best.

Quote:



Yup, +12/-12V would be ideal, since many nice opamps will not tolerate higher anyway. 2x9V is very interesting but battery life probably will suffer...

Quote:



Why find a portable balanced source if the amp takes unbalanced? I was thinking of mating this with the Alien DAC or something that's small which is much easier to find unbalanced than balanced... All I care about is the balanced output, really. Anything balanced that's cheap and small is likely to have balancing mechanism that's crappier (using lesser parts) than what I've done on this board anyway.

Quote:



Some SMD resistors and caps (related to power and ground) are mounted under the board. I tried to minimize the use of vias in the signal path so all the resistors in the signal path are on the top layer.

Quote:



Heat issues? The TPA that I'm running now doesn't get that hot - with a small "groundplane" attached to the powerpad solder point, will heat be such an issue with what I've done?

Quote:



I personally didn't like the DIP socket idea - potential stability problems with using higher speed opamps on adapters and sockets and space issues as well. Also, more of my favorite opamps only come in SOIC anyway so why bother with DIP?

Anyway, thank you so much for all your comments so far. I'll try to get a revised version up soon with your feedback applied!

With the amount of board space you're dedicating to the caps, any reason you could just use an AC input (either voltage doubler with single AC, or CT Xformer) and just go with an on-board regulated bipolar PSU?

I personally don't like the idea of inductors in HF or switching applications, which is what incidently a lot of people recomend them for. The main problem is implimentation. I may say at this point too that while testing the DS1802 for my amp I ran them from a very basic shunt based on the amp's powersupplies. A simple capacitor on the outside was all the filtering I put in and there is precisely nada difference from going to a separate supply. So that said I wouldn't bother with an inductor.

If you do pick an inductor you should pick one that minimally reacts with the capacitor beforehand. Putting an Ls, Rs, and Cs in the same circuit is a recepie for ringing. Info on how to impliment this correctly: http://www.hagtech.com/pdf/snubber.pdf, http://www.designers-guide.org/Design/bypassing.pdf.

Quote:



Don't know if this has been covered, but why not? Select the polygon tool, type in the name of you ground net and hit enter and you can draw a ground plane. When you're done, use the ratsnest tool to render it

/U.

A ferrite will create no problem (at higher frequencies, its inductance is about zero) and they're dirt cheap. Pick a ferrite with enough current capacity and a middle of the road resistance.