Itsy Bitsy TPA6120 Amp

Well, someone just had to make the suggestion that I try this chip out, actually as part of my PGA2311 preamp project. I've got to hear it first before I commit to the boards for the other project, so I've done up a little board to try it out. Could potentially make a decent portable amp, but battery life will not be fantastic (TPA6120 eats up 23mA, OPA551 another 8.5mA). Should sound great though, assuming I haven't made any design errors...

I'd also like to get at least one successful fab job under my belt before going all out on the preamp boards.

It's set up for a gain of 5, with buffered virtual ground by OPA551. The plan is to fit a Hammond 1455C802 (50mm x 80mm) with room to squeeze 2x9V batteries in. I'm not totally certain this will be possible; the current layout is 32mm in length, and 9Vs are nominally 48mm long - leaving little room for the connectors. It may be possible to squeeze it in though. Unfortunately the TPA6120 isn't specified to run at +/-4V, but may work anyway, for single 9V operation. I don't think I can squeeze any more space in that dimension due to the necessary connectors. The ground plane is unbroken except for one pesky V- trace I couldn't find a way to reroute. It doesn't split things too badly though.

No coupling caps are onboard, and 10R output resistors are placed as suggested in the TPA6120 datasheet. Because these are for opamp isolation, I don't think they serve their purpose if placed inside the feedback loop, so they are outside the loop. They shouldn't impact available current when driving low impedance phones.

This chip is supposed to be one of the best sounding single-chip amps around, I have high hopes for this simple design. I'll be placing an order with Olimex this afternoon for one eurocard, which should fit 8 boards. If anyone wants boards, I'll eat the shipping and send to US/Canada for $6.

As everything I do, this design is open. If anyone wants gerbers or the Altium files, let me know.

Comments before I go to fab?

The main info post is down the thread a bit and I'll be updating it as changes are made.

Yea, misread the dimensions of course. Not a chance of getting two 9V in there, and I'd rather not go to the next longer case, the 1455C1202 (though this should definitely work). Some of the component values are overly optimistic too, 680uF and 120uF are both too large. 330uF and 100uF will fit.

A DigiKey BOM: http://spreadsheets.google.com/pub?k...j8iJOBIDEoN9Mg totals to $21.77 CAD, including stepping up to min qtys. I adjusted 200K voltage divider resistors to 100K to reduce line item count (uA of current don't really matter when Iq is so high), and subbed cap values that will actually fit the board.

I want to go ahead with this, but I'm niggled by the fact that I can't find any appropriate enclosure that can accommodate 2x9V without being way too large for a portable amp. I could potentially switch to AA's plus a DC-DC converter, and I was looking at the Serpac H-45 or H-65 with battery holders, but the technical drawings are awful and I couldn't deduce the available board size from them. The Hammond might squeeze a 3xAAA holder with the current board, but it would be tricky to fit the necessary DC-DC in the minimal available room.

Any suggestions on portable enclosures?

Yes, this is basically a CMoy with a buffered virtual ground; though CMoy is really just 'non-inverting amplifier based on a single opamp', even an LM3886-based amp looks like a CMoy on paper . Given the specs on this chip I don't really feel that separate input and output stages are necessary. I plan to supply 18-20VDC, though I think I will now add a linear regulator onboard and increase board size slightly (32mm -> 40mm), since I can't find a way to make this portable (and power draw is unreasonable to do so anyway). I took a look at going to 3xAAA in the larger Hammond enclosure, but the power supply got too complicated for my liking, and added too many variables for me to 'hear' the 6120, which was the point of this exercise at the start. Final boards will probably accept DC input of ~20V, regulate it down to ~18V via an LDO and have additional capacitance onboard. I'll modify the board accordingly this evening, and then I'll probably be satisfied going to fab.

Thanks for the advice, gain is set for 5 but if I do increase board size that will give more room for capacitance as well. I've followed the instructions in the datasheet regarding layout of the ground plane for stability, I've placed V+ and V- decoupling for both sides of the TPA6120 (and for OPA551), and the layout is nice and tight around the chip so I think it will be okay stability-wise. I'm hoping that it will be possible to solder the power pad to the ground plane by using the vias I placed underneath it to heat sink it.

I'm not sure about the buffered ground though, I've never built one before and I'm wondering if input/output caps are required in this situation. I'm thinking that input caps may be required, but I don't fully understand the implications. I think OPA551 current should be enough, and I don't know any way to build a buffer that's anywhere near as good with a small parts count. A full-on diamond buffer would take as much room as the rest of the amp itself. At full 80mW output of TPA6120 it would be asked to sink/source 50mA RMS, so I think I will leave this chip.

I've made some changes and I think I can consider this final:

• Board widened from 32mm to 40mm
• 940uF FM rail capacitance (not 100% sure this will fit, may end up being 660uF or 16V caps instead)
• Virtual ground with TLE2426 (SO-8 with Cnr) isolated and buffered for output by OPA551
• Onboard 3-terminal fixed regulator (78L15)

I plan to supply approximately 18V from a cheap switching wallwart. My simulations on the power supply indicate that the 180mV of ripple will be easily quenched by the simple supply I implemented (the sims show a ruler-flat output for the passives I've used), and this will not require finicky unregulated supplies to be used. For critical listening I'll probably hook my σ22 directly to the provided test points, but really even this should not be necessary.

I'm about to put in a fab order at Olimex for just a single eurocard. Assuming there's nothing heinous wrong with my output I should have boards in a couple weeks.

The BOM at DigiKey comes out to just over $40 USD, complete including case and wall wart.

Some files you might like too:
Fab Outputs (gerbers) (these files are not up to date, PM me if you want current ones)
Nice PDF schematic/layout




Edit: Okay, scratch the order for now. Olimex's annular ring requirements make it pretty much impossible to mount TO-92 parts straight on. How bad is that?! Getting a quote from another fab house while I see if I want to bother offsetting them and redoing most of the pads.

Update Oct 05 2007
See post #22, I've got this working, which required that I change the 100K resistors to 5K due to input bias on the TPA6120 causing excessive DC offset. Other than that, everything is working fine for me with OPA551 on ground and it sounds great! Not as good as my full-fledged mini3, but still drives all my cans quite respectably. I'd say great for its diminuitive size, this would be great to integrate into something like a preamp where a headphone out is desirable, but not the main purpose.

I don't think I will be using it in my preamp though, due to how finicky it seems to be to work with. Definitely don't want to be plagued by DC offset and the like in that application. Nothing against the sound of it though.

If you want a board, I have 3 left (Oct 15/07) that I'll part with for $9 a piece, shipped. If you build this, please verify the parts against the schematic, as I'm not positive the silkscreen matches the BOM.

Good catch, updated the qty but not the designators.

Ground is the bottom layer. It's broken for a couple of traces, any through-holes not connected to ground, and underneath the TPA6120 and OPA690 to reduce parasitic capacitance (per the TPA6120 datasheet). Also, the output ground is buffered and is therefore a trace. The bottom layer is the blue fill in the images, top is red.
Size indeed. There was room for more capacitance, and I thought the existing 330uF was definitely on the low side, so I increased the diameter of the caps and doubled them up. It also reduces ESR (R || R = 0.5R), though larger caps would have less ESR as well.

Yes, I thought of using Panasonic EVJ series, but then a power switch would be required. It's a wash, and I prefer the switched pot. Vertical jacks aren't a bad idea, but I don't feel the need for space. There's probably enough room between jacks to fit an LED, but I dislike having them on everything, and if I were to place one I'd probably airwire it at the back of the case where the DC comes in.

The 1455C802 is very small, 50mm x 80mm x 20mm. 2x9V won't fit either way, nor will even 2xAAA with holder. All the other enclosures I looked at I didn't like or were too large, so I've given up on making this portable. If you want to take the design and put it in a huge case with 10xAAA, be my guest . If you jerry-rigged it, you might squeeze a small DC-DC and this board in a Serpac H-65, but I couldn't read their technical diagrams (they're hideous) to actually design around it.

PS. I couldn't help myself and fixed the board to fulfill Olimex's requirements, and sent off a request for a quote. DigiKey order placed for a whole ton of stuff. Wee, winter's here!

PCBs have been ordered from CustomPCB. I've ordered 12 pcs, at a rather high cost of $8 each, but they took my Protel files directly and I didn't have to fiddle around. As a bonus they fab and ship much faster so I should have boards in a week or less.

I usually use Protel's built in simulation since it has models for a lot more components, but I also often use SwCAD III since it's easier to put together a simple one-off sim. In this case I used SwCAD.