Hello all,
I'd like to buy a TPA3116 2.1 amp, but due to the size of this thread I couldn't really find a recent comparison/consensus on different versions of these boards. Due to the Chinese new year and me being impatient, I want to receive the board quickly, but unfortunately there seems to be only a single seller within EU to be selling this board at a reasonable price here; is this version of the board good enough or should I really not bother with it and get the 'upgraded' version here?
I'd like to buy a TPA3116 2.1 amp, but due to the size of this thread I couldn't really find a recent comparison/consensus on different versions of these boards. Due to the Chinese new year and me being impatient, I want to receive the board quickly, but unfortunately there seems to be only a single seller within EU to be selling this board at a reasonable price here; is this version of the board good enough or should I really not bother with it and get the 'upgraded' version here?
Hi,
In my spare time im working on a budget amplifier. I have chosen the tpa3116d2, because of many nice reviews. My implementation is based on what i have learned from the ti docs and here.
But there is one question, that comes often to my mind. There is an avcc input. What if i use an analog low noise psu based on a torroid for the avcc and a smps for the other vccs ?
Will this work and perhaps sound nicer than smps on avcc ? My experiments with dacs showed me, that a torroid is always better for analog supply, than switching mode. Is this true for class d amps too ?
A torroid for such a purpose is not cheap, so i wanted to ask, before buying.
What about torroid psu for the whole amp ? Could this sound nicer ? I know, ti suggests smps, but i think only because on efficency and price or is my assumption wrong ?
Thanks guys
Tom
In my spare time im working on a budget amplifier. I have chosen the tpa3116d2, because of many nice reviews. My implementation is based on what i have learned from the ti docs and here.
But there is one question, that comes often to my mind. There is an avcc input. What if i use an analog low noise psu based on a torroid for the avcc and a smps for the other vccs ?
Will this work and perhaps sound nicer than smps on avcc ? My experiments with dacs showed me, that a torroid is always better for analog supply, than switching mode. Is this true for class d amps too ?
A torroid for such a purpose is not cheap, so i wanted to ask, before buying.
What about torroid psu for the whole amp ? Could this sound nicer ? I know, ti suggests smps, but i think only because on efficency and price or is my assumption wrong ?
Thanks guys
Tom
I recently got this:
....
......
Does anybody know what might be causing this phenomenon?
Also while measuring as Doctormord advised, I advise you to remove the heatsink and check to see if the TPA3116 output pins actually have a wide enough trace on the PCB to carry current from the outputs >> LC filter and out. Also flip the board over and check the PVDD traces, particularly where PVDD vias to the top of the PCB. I'm guessing you'll find those to be terribly insufficient.
I bought the dual chip version (red, MH-M190 V2) of that board and was pretty disgusted with it. It was capable of only a few watts as the PCB traces were measured at only 0.34 mm wide exiting the chip outputs. A single PVDD via on the back is capable of passing ~1 amp of current or so (if that!) to power the amp. At PVDD 12 volts the dual chip boards were straining as you explained and had reached max SPL. The PCBs were so useless I simply burned them with the trash.
If your boards are sufficient Doctormords advice is spot on.
Last edited:
The TPA3116D2 requires only a single DC power supply, max 24 volts to PVDD. The chip has internal regulators to provide power for its function circuits. Choice of supply design is yours... low noise if you can get it.
It might well work, equally though it might not. TI say the voltage difference between these two pins (pin17 and pin18/19) should be small, but on power up and power down it'll be tricky to keep to that rule.
That said, it was the single biggest improvement to my TPA3116 amp when I installed a filter on the AVCC pin, so the effort of a separately regulated supply for the AVCC may well not be in vain.
You'd want a toroid plus a linear regulator methinks. It might sound nicer as SMPSUs do generate common-mode hash which is hard to filter out.
You can find some details in this post http://www.diyaudio.com/forums/class-d/237086-tpa3116d2-amp-63.html#post3936825 and there's more clarification in several subsequent ones. Your question has been asked more than once on this thread.
Using just 470uF isn't right, because usually the ESR is to high to have adequate attenuation. ESR is the dominant factor here to get good filtering. 10uF MLCC is the way to go but keep DC-bias in mind. I also would recommend to increase 3R3 to 10R.
Its not simply about attenuation though, its also about how much load-induced noise there is. There's a trade-off between those - increasing 3R3 to 10R raises the level of load-induced noise (whilst also improving attenuation).
Right to mention DC bias - TDK caps (my first choice) are probably below 3uF at full bias. Others are often even lower.
Right to mention DC bias - TDK caps (my first choice) are probably below 3uF at full bias. Others are often even lower.
Right but having to much capacitance will eventually violate TI's rule of having PVCC and AVCC tight together. (+-0.3V) I'd suspect that otherwise the protection/clamping diodes will conduct.
Indeed. I tried 'too much capacitance' and it made no audible improvement to my ears, hence why I recommend 470uF. Add more if you wish but I can't myself see any point.
If I remove the heatsink, what's a good way to stick it back on. Is a dab of plumber's goop okay? Or I have some double stick tape...
Regarding adequacy of traces, I'm not sure I'd know if I was looking at them. Is there a minimum width for the output traces that is sufficient? For the vias, is there a minimum diameter? Not sure what I'm looking for there.
I was looking at the diagram for the chip and I see "PVCC" on pins 18, 19, 31, and 32. This is the supply, correct? Is this different from "PVDD?" Please forgive my ignorance. I'm learning as I go.
Thanks for the input! I'm hoping it's not a hopeless case.
I don't have a particularly nice meter. Looks like minimum resistance range is 200ohm. Doesn't seem like that'll do the trick. I'll check supply voltage the next time I have a little time to tinker.
I see 680nF output capacitors and 33uH output inductors in the output filter. With an 8 ohm load, you get this electrical response.I recently got this:
version of the TPA3116 board and put it together with a pair of 4ohm 100W bookshelf speakers to use in place of an old pair of computer speakers. I'm using a LiShin 19V 3.42A laptop power supply.
I've been doing a little bit of listening since I got it running and it sounds nice for the most part. My complaint is that it doesn't seem to handle dynamics well. By this I mean, when instrumentation is sparse ( for instance bass, guitar, vocals, light percussion) it sounds very nice. The sound is rich, balanced, lots of details. When the instrumentation gets more complex and the overall dynamic increases (especially when there is a heavier percussion, lots of symbols, distorted guitar etc.) it seems like the overall volume stays roughly the same, but the low and mids get robbed and I'm left with a more thin brittle sound and some really unpleasant highs.
This is with volume set well below the maximum, though so it doesn't seem to me like it should be a problem with the amount of power available. Don't have much experience here though so I don't really know.
Does anybody know what might be causing this phenomenon?
A great big treble peak. This explains exactly what you're hearing. Take the 33uH inductors off and put down some 10uH ones.
OK, I screwed that up. Here's the new response.
MATLAB code (should work in Octave too):
j = sqrt(-1);
c = 680e-9;
l = 33e-6;
rload = 4;
f = logspace(log10(1),log10(100000),1000);
zload = rload/2;
zc = -j./(2*pi*f*c);
zshunt = 1./(1./zload + 1./zc)
zl = j*2*pi*f*l;
resp = zshunt./(zshunt+zl);
subplot(2,1,1);
semilogx(f,20*log10(abs(resp)));
grid on;
subplot(2,1,2);
semilogx(f,angle(resp)*180/pi);
grid on;
MATLAB code (should work in Octave too):
j = sqrt(-1);
c = 680e-9;
l = 33e-6;
rload = 4;
f = logspace(log10(1),log10(100000),1000);
zload = rload/2;
zc = -j./(2*pi*f*c);
zshunt = 1./(1./zload + 1./zc)
zl = j*2*pi*f*l;
resp = zshunt./(zshunt+zl);
subplot(2,1,1);
semilogx(f,20*log10(abs(resp)));
grid on;
subplot(2,1,2);
semilogx(f,angle(resp)*180/pi);
grid on;
I didn't know what I was looking at before, but now that you mention it the inductors on the board I received don't match those in the picture on the ebay ad. If I understand the nomenclature, they are indeed 10uH inductors. Here's a picture:
An externally hosted image should be here but it was not working when we last tested it.
I think the caps are the same "684J100" though it's difficult to read in my photo and I don't have the board in front of me right now.