Truly impressed by the tpa6120

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I soldered together this afternoon a small preamp/headphones amplifier based on the TPA6120 and the opa134 and I've to say I'm very impressed by the sound of it. I guess it's old news for many of you but it's really a nice sounding chip, even using cheap leads and a 36V printer wallwart with 2 1K resistors to provide the +/-18V power input. It doesn't have quite the magic of my behemoth headphones tube amp but I'm willing to admit it's probably more faithful to the source than the tube amp is.

The schematic is quite simple (see below). However, laying out a two-layers board for it was quite nightmarish: keeping the feedback loops short, decoupling the tpa6120 close enough, keeping the power lines close to one another... I finally brought the +/-12V lines by jumpers running flat around the bottom of the board.

I've added 22r at the output as it will be mostly used as a preamp. I've not tested it yet for stability without that output resistor. It seems fine with it (but I've only a 20mhz scope)

Schematic, pcb and pic of the assembled unit attached.
 

Attachments

  • preamp schem.PNG
    preamp schem.PNG
    13.8 KB · Views: 808
  • preamp board.PNG
    preamp board.PNG
    105 KB · Views: 770
  • 3.JPG
    3.JPG
    103.2 KB · Views: 754
wrong design of Vcc & Vdd connection. the trace should be first connected to the bypass caps and THEN you should run shortest possible separate trace from that caps to the IC pins. it's not good to place the bypass caps parallel to existing supply wires (the way it looks like on your PCB). lookup TI's geniue evaluation board - they've done it right. you should also connect the GND pins of the tpa6120 (#6-#15).

but still - TPA6120 is truly great opamp. I have built headphone amp with design very similar to Per-Ander's QRV07 and like it a lot.
 
Thanks for the comments.

wrong design of Vcc & Vdd connection. the trace should be first connected to the bypass caps and THEN you should run shortest possible separate trace from that caps to the IC pins. it's not good to place the bypass caps parallel to existing supply wires (the way it looks like on your PCB). lookup TI's geniue evaluation board - they've done it right. you should also connect the GND pins of the tpa6120 (#6-#15).
Actually, V- is wired pretty much the same as the evaluation board and the way you suggest. V+ is a bit more problematic, with indeed the incoming jumper connecting in between the bypass cap and the ic pin, which adds a few nh in serie with the bypass caps. It should be quite easy to fix (if I ever get new boards done). I might have to switch from 1206 to 0805 caps though.

Fixing the orientation of the decoupling caps could be more problematic. I probably should also get rid of the thermals... :rolleyes:

Is there a reason to connect pins 6 to 15 to gnd ? According to datasheet, they're not connected to anything ? :scratch:
 
Last edited:
0805's are easy to hand solder, there's no reason not to use them.)

the thermal vias are important if you (as I see) not solder the board by reflow. another solution for hand soldering is one big (1,5 mm / 60 mils) hole under the thermal pad - it's easy to hand solder and has a better thermal properties when filled with tin. tpa6120 can overheat if it's pwr_gnd isn't connected to suficient gnd plane.
pins 6-15 should be connected to gnd plane to further improve the thermal dissipation.

you should not be afraid of vias, small parts and thin tracks, they're inevitable in designs like this :)

in attachment you can see a detail of my own tpa6120 board - the red side is top, blue is bottom. on bottom there's one big gnd plane under the whole board (excerpt for the power transformers), signal + power tracks are on top (if possible).
it's not ideal (some parts are pretty too much close so it's bit harder to solder) but works somehow and it's quiet (I have it sitting on the top of my work-table and there's no detectable hum/hiss of any kind).
 

Attachments

  • icfha_detail.jpg
    icfha_detail.jpg
    63.3 KB · Views: 686
Small misunderstanding for the "thermals", what I meant are the narrow connection in between the ground pads of the bypass caps and the top groundplane, which facilitates soldering but adds some more inductance. The thermalpad of the tpa6120 is soldered through the vias under it.

As far as small tracks and parts, I'm not "affraid" but using Olimex to get reasonnable prices on small runs of proto, I've some limitations (their smallest standard via is quite big and clearances are 8mil at best).

And, btw, my pcb as shown is dead silent into low impedance, sensitive headphones and the tpa6120 is barely lukewarm after hours of listening at rather high volume (with high impedance headphones however). I'll try to run some measurements later on but I've no access to a good soundcard for now. Maybe next week.

Still, I'll probably do a few changes to the board under eagle, for a possible (but not likely to happen anytime soon) re-run.
 
I point out some output load decoupling inductor alternatives that give near 0 Ohm audio output impedance and still keep the output from seeing destabilizing (cable) C load
http://www.diyaudio.com/forums/head...0-reference-design-project-8.html#post2528662 and a few more of my posts over the next couple of pages

and there is good reason not to use 0805 in feedback R - the power rating is so low that self heating can become a distortion source - use bigger parts and/or lower TC - NiCr or TaN only add a few $ to the project cost, bulk foil are closer to $10 ea
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.