Took a few hours and banged out my take on a 'proper' TPA PCB.
Size is 96mm W, 50mm H. Yeah, the silkscreen's half wrong, I'm not finished yet 🙂
I gave up trying to find a catalog heatsink that would fit the TPA3116 and clear nearby decoupling, and threw the TPA3118 on there instead. The 3118 is only rated for 8 ohm loads in the datasheet, but it uses the same die as the TPA3116 and has a comparable Rjc - and I guarantee the ground plane heatsinking done here works better than the tiny aluminum heatsinks you find on half the eBay TPA3116 cards out there. There's 7 inches of largely unbroken ground plane on the bottom layer, and lots of vias connecting top/bottom layers directly under the TPA.
Input uses film coupling caps, 5mm lead spacing/7.5mm square, typical part would probably be 2.2uF/50V. Spots are provided to put down 10K input resistors between +/-, and a 100 ohm resistor from - to ground for unbalanced input, I might futz with this some more.
Inductors are ICE Components 1D14A, caps for the output filter are 0.68uF/100V film. I'm not using diodes to clamp the post-filter output, thought being if you're pushing the output close to clipping and the output filter is peaking into your speaker load, you could end up with unwanted conduction. Using 100V caps should hopefully keep things from catching on fire with no output load - if not, hole sizes for the film caps are generous so they should be easy to change if they burn up 🙂
5 decoupling cap locations are provided per TPA channel - two 0603 caps, a 1210 cap, and two 10mm diameter caps. An example configuration providing low Z over a wide bandwidth could include a 1000pF NPO, 0.1uF/50V X7R, 4.7uF/50V X7R, a 100uF/35V polymer and a 470uF/35V wet electrolytic. Bootstrap caps (0603 X7R's) are placed tight against the TPA. Immediately next to them are locations for OUTxx snubbers, aka "bootstrap snubbers", with a short return path back to the TPA.
A 4-position DIP switch is provided for adjusting switching frequency and selecting BD/1SPW mode. Gain is set with a pair of 0805 resistors. PLIMIT is currently tied to GVDD, I've contemplated stuffing a pot on there to allow manually setting it like the TI evaluation board provides though I dunno if anyone will use the feature.
I'm using an ATTiny13A microcontroller to provide popless power-on/off and fault handling - originally I had a dual comparator and several discretes handling power sequencing, but decided this was easier. There's a couple pins provided on the input terminal block (/EN and ground) - shorting these two pins turns on the amp, leaving these open turns the amp off with a standby current of ~50uA dominated by the TPA itself, handy for battery powered apps.
Terminal blocks are nice Phoenix rising cage ones that can take 12AWG to 28AWG, and handle 16 amps or something ridiculous.
All parts required to build the card can be ordered from Mouser.
So yeah, probably a bit overengineered compared to everything else out there, it'll be definitely more expensive to build compared to an eBay special, but it should outperform them and allow more customization.
Any change suggestions, thoughts, criticisms, etc... are welcome.
Excellent!! Look like the "ultimate board" is arriving. Good work!!
Regards,
Hi guys,
I just clued into this thread. I never did fab my TPA3116 design but I can offer you what I came up with, if it is of any use.
Find attached the schematic and layout as it stands uncompleted. The way i look at it, you might as well put as many options in as you can to cover as many bases as you can.
I have made the mistake of not considering or thinking of a package to enclose the amp. I suggest that you do the same unless you have done so already.
I was planning on using a Hammond 1455J1201.
Let me know if I can help further.
Cheers
Rick
I just clued into this thread. I never did fab my TPA3116 design but I can offer you what I came up with, if it is of any use.
Find attached the schematic and layout as it stands uncompleted. The way i look at it, you might as well put as many options in as you can to cover as many bases as you can.
I have made the mistake of not considering or thinking of a package to enclose the amp. I suggest that you do the same unless you have done so already.
I was planning on using a Hammond 1455J1201.
Let me know if I can help further.
Cheers
Rick
Awesome, I'll throw the gain/voltage adjustment bits in.
I'm still not sold on having the input power routed from cap to cap to cap, having a fat power plane provides a lower impedance between the caps and the TPA. I s'pose it also gives a lower impedance path for crap to get in off the power supply... I'll throw a ferrite on the input to take care of that.
I'll have a peek, thanks for posting the design.
I think I'm going to keep this board as a rectangle with terminal blocks as it's the most flexible thing to use. You could buy a 1455 case, screw the board down inside, mount some panel mount connectors/switches to the end plates and get something that looks much the same.
If you mount this board in a closed metal case like a 1455, I'd suggest buying standoffs and poking a berquist gap-pad between the case and PCB. This is what I'm doing in the cAMP design.
For the right ferrite you definitely need to know what to filter. Just having a ferrite there wont help in any way. For sure wider tracks are lower impedance but they bypass the current away from the caps. For the pure DC component, they will take the path off lowest resistance. For the higher AC component, they will form a "beam" between the poles and you will have the caps terminals to be that poles. (Well okay, it's just 400kHz) 🙂
Last edited:
+1. You should always "force" the current to go to your cap pins, otherwise it takes the lowest impedance path which is the shortest, and in this case it will be partially bypassing your capacitors.
Oh and also, I would recommend, if possible, to keep the chip-inductor connection on one layer. The inductance of the grouped vias might pose an issue IMO.
Perhaps rotate the outer inductors 90 degrees, then you can route the power polygon between the pins and connect the inductor right to the chip? Looks nice apart from that, good job on the ATTiny control, it's a good idea.
Perhaps rotate the outer inductors 90 degrees, then you can route the power polygon between the pins and connect the inductor right to the chip? Looks nice apart from that, good job on the ATTiny control, it's a good idea.
The decoupling here serves two purposes, #1 keeping the supply rail stable through load demand, and #2 keeping crap from the external supply from disturbing the supply to the load.
To solve #1 and only #1, the solution is to make the lowest impedance path possible between the capacitors and the load. A plane is the best you can do here, jumping from cap to cap to cap increases the inductance between the far caps and the load.
Now that parasitic inductance and the small value local decoupling does provide a low pass filter, which helps #2, this is the primary reason to go cap-cap-cap. But rather than depending on that inductance at the expense of #1, I'd sooner throw down a filter further upstream. At frequencies where the difference between plane and cap-cap-cap starts making a difference, a typical ferrite will have tons of resistance and work much better.
That parasitic inductance has a 10uH inductor in series with it, I think it's fine.
So you say the groundplane is prefered because of lower impedance but then you'll put an inductance within the path to rise that impedance/inductance again, no?
Beside this, you want to have your current coming from your bulk capacitance, not mainly from your supply, as the connection between the psu and your ampboard forms a nice di/dt limiter.
But hey, i just can give you this advice from my knowledge. 😉
Just sayin.
Beside this, you want to have your current coming from your bulk capacitance, not mainly from your supply, as the connection between the psu and your ampboard forms a nice di/dt limiter.
But hey, i just can give you this advice from my knowledge. 😉
Just sayin.
I think we're misunderstanding here..
I want the lowest impedance possible between the decoupling cap array (including the bulk capacitance) and the TPA, which is why I put it on a plane. I won't bury ferrites or anything in the middle of that, I'd put a single ferrite back at the terminal block where external power comes in to keep HF trash out. Or probably just leave it as-is.
I've got a bit of knowledge here too, several years back I beat my head against a wall decoupling a >600MSPS RF DAC for FM radio broadcast. Had to achieve a >80dB SFDR to meet CE specs, over a 88-108MHz output frequency range, over a varying output level, without modulation, with endless different mechanisms for signals to mix together... ugh. Proud of the work but can't say I loved doing that part of the job.
I want the lowest impedance possible between the decoupling cap array (including the bulk capacitance) and the TPA, which is why I put it on a plane. I won't bury ferrites or anything in the middle of that, I'd put a single ferrite back at the terminal block where external power comes in to keep HF trash out. Or probably just leave it as-is.
I've got a bit of knowledge here too, several years back I beat my head against a wall decoupling a >600MSPS RF DAC for FM radio broadcast. Had to achieve a >80dB SFDR to meet CE specs, over a 88-108MHz output frequency range, over a varying output level, without modulation, with endless different mechanisms for signals to mix together... ugh. Proud of the work but can't say I loved doing that part of the job.
My idea with the Hammond enclosure was that the pcb slides and sits/rides along the Al extrusions designed for such. The pcb only needs attachment at the end plates. One can make a simple aluminum bracket to some how use the Hammond case as the heat sink for the TPA3116, saving $ on a fancy HS or extending the one that is chosen.
The Hammond 1455J1201 requires a pcb that is 75(w)x120(l)mm, what are your current dimensions?
Good luck with your design.
Last edited:
...
I'd leave it at first. HF mainly dont care at first.
Nice one, you doin what you do!
Last edited:
Decided to use dual-comparator route for the shutdown circuit instead of the AVR, so you should be able to solder it together and go.
Soldering supplies:
- Temperature controlled soldering station. You'll need a heavy tip and high temperature to heat the thermal pad from underneath and solder the TPA3118 in place, I use a knife tip on my Hakko 936 to do the job.
- Small diameter 63/37 solder
- Liquid flux. I use a MG Chemicals rosin flux pen.
- Alcohol/flux remover/Q-tips/whatever for cleaning up.
- Tweezers
- Good surface to work, ideally a static mat. So you won't lose small parts when you're taking them out of cut tape, or scratch up or burn your kitchen table and **** off your significant other 🙂
The 65 mil pitch on the TPA is the finest SMT on there, followed by some 0603 resistors/capacitors and a SOD323 diode.
If there's demand, I'll get a local shop to build assembled cards. I'll warn that it'll cost a few times more than a cheap TPA3116, because I'll be manufacturing them here with greater-than-chinese labor rates at lower quantities.
If you changed the design to pbtl/mono, I would assume you could have an even smaller board, and still more compact layout? Also, given the same speaker load, I would assume pbtl puts half the current through the IC, making cooling easier?
I'm just guessing here, these ideas could be completely silly.
I'm just guessing here, these ideas could be completely silly.
- Status
- Not open for further replies.