TPA3116D2 Amp

Member
Joined 2008
Paid Member
Mention of six figure capacitances is a red rag to a bull :D However I did the search, no results found.

Later I found the original link on AudioCircle - 'does not ship to China'. Methinks a bit of a waste of low ESR caps mounting them so far distant from any load.... Also based on my listening, the bass performance of the TPA3116 is its weak point so I doubt there's much to be gained by going so extreme on the supply decoupling. To do justice to all those caps, get an NXP amp chip such as TDA8932....

Very strange the bass as weak point statement. I get all the bass I can handle with a MTM two way 90db speakers. Especially if the input cap values are increased. Even 3.3uf at 32/36db gain is too much in my space. I think a lot disclosure about the speakers being used is important before calling the TPA weak on bass.
 
I'm not talking about bass quantity here, I'm talking how distinct the bass is. ISTM that the TPA3116 is adding some LF noise which muddies the bass somewhat. Adding supply caps does precisely zero for the bass quantity on any amp you care to name. The speaker types are taken out of the equation because I'm using the same speakers and comparing with two other amps here - a class AB chip amp and the NXP chip I already mentioned.
 
Member
Joined 2008
Paid Member
I'm not talking about bass quantity here, I'm talking how distinct the bass is. ISTM that the TPA3116 is adding some LF noise which muddies the bass somewhat. Adding supply caps does precisely zero for the bass quantity on any amp you care to name. The speaker types are taken out of the equation because I'm using the same speakers and comparing with two other amps here - a class AB chip amp and the NXP chip I already mentioned.

Wasn't talking supply caps but the input caps and values as recommended by TI, but yes I agree the TPA does not have the same bass tightness and control as say the typical Class AB. Can't have it all! (for $25 anyway)
 
Sure, it makes sense to have more than the bare minimum value for the input coupling caps - this can't fix up the quality of the bass, just the quantity.

The NXP chip doesn't do the HF nearly as well as the TPA3116, however its bass tightness is considerably better than my reference classAB chip amp. So yeah seems that we can't have it all but I'm thinking about bi-amping possibilities to get a little bit closer to the whole enchilada....:p
 
mmm, tpa curves for low frequency THD+N for 12V look the same as NXP BTL 22V, for 24V tpa vs 22V NXP BTL our TPA has ~0.02% more THD+N in low frequencies? And that is the muddy low? (SE figures NXP are worse for low frequencies)


What I don't understand is that a weaker input-signal(external) is equally affected by that noise (and distortion) as a stronger input-signal. The noise is there with no input, with a little input and same noise it seems to me input is more affected then with stronger input and same noise??
 
I'm fairly sure that looking at distortion plots won't reveal the subjective difference in bass performance. That's because if you try the experiment of adding supply caps, it won't vary the measured THD at all, but it sure makes a difference to how the bass sounds.

The noise I'm talking about isn't there with no input, its noise that varies as the signal does.
 
I'm fairly sure that looking at distortion plots won't reveal the subjective difference in bass performance. That's because if you try the experiment of adding supply caps, it won't vary the measured THD at all, but it sure makes a difference to how the bass sounds.

The noise I'm talking about isn't there with no input, its noise that varies as the signal does.

What objective measurement/test can one use to verify the "LF noise" that you are referring to then? We cannot rely subjective listening alone since you, me, and everyone hear things differently? Any suggestions? The fact that everyone has their own favourite speakers make this even more complicated? May be what you heard has something to do with the way the TPA3116 amp interacts with your speakers as compare to your class AB amps? Have you tried the amp with other speakers? I am always curious how the output filter interact with the crossover of the speakers.

Regards,
 
What objective measurement/test can one use to verify the "LF noise" that you are referring to then?

It would be nice if someone were to come up with a measurement. For the time being though I'm not aware of any.

We cannot rely subjective listening alone since you, me, and everyone hear things differently?

Who is 'we' here? I rely on my own subjective listening seeing as measurements don't yet cut it. Even when they eventually do, they're likely to be embedded within expensive audio test gear with little appeal to DIYers. No-one else is under any kind of obligation to rely on what I hear - I just post it up here that someone might find it useful/interesting.

We already have an example of both wushuliu and myself hearing the bass performance of TPA3116 similarly. So whence this 'everyone hears differently'? When people hear differently perhaps its because their setups are different.

Any suggestions? The fact that everyone has their own favourite speakers make this even more complicated?

I don't see 'complicated'. If you're happy with your TPA3116, just ignore those of us who are interested in pushing the limits.

May be what you heard has something to do with the way the TPA3116 amp interacts with your speakers as compare to your class AB amps?

Could be, yep. The output filter is something of a wild card with a classD amp and different speakers will damp the response of that differently.

Have you tried the amp with other speakers? I am always curious how the output filter interact with the crossover of the speakers.

No, I've already cannibalized my TPA3116 as I wanted the output inductors for the NXP chip. The great bass of the NXP carried across to my latest set of speakers though.
 
I haven't tried a different powersupply but its something I probably should try before changing more components. I just realized that I haven't changed the bootstrap capacitors.. Could they be the problem?
As a side note, I use 10uF silmic 2 as input signal dc de-coupling caps and got lots of bass. Its true what they say about them:) With the original ceramic caps it sounded hollow and weak.. I've also experimented with higher value caps further away from the chip and that didnt help at all! If the cap legs extruded 5mm up from the board it added more noise. From that experience I tinned up the +wires into the chip to reduce the resistance. As dc power caps I now use 220uf 35v Panasonic hybrid electrolytics from a motherboard I had. They removed even more noise and no change in bass:) From my experience long legged large caps doesn't help as dc power caps! The most important factor is emediate placement to the chip as earlier stated here on this thread!
I have also soldered the input directly onto the pot pins and cut of the extra pcb leads.. That also removed some noise! When I got this board I thought it was very noisy, borderline unstable.. White noise comparable to a bad radio at times.. Could my extra noise come from the high voltage? I've read in earlier threads that some diy boards can't handle full design voltage because of uncontrolled RF. I've gotten rid of the noise now even at 25.3V now it sounds like a good large amplifier! It has no problem driving my XTZ - 99.26(6ohm, 87db) almost to max:) I just need to get rid of that coarseness!
 
I just tested with different psu solutions now; 14.8V lipo, 19,1 laptop brick and back to the toroid with PI filter at 26,2V!? I must say that they all sounded just about the same in tonality! But there was less punch and dynamics from the 14.8V battery. I belive I might have found the cause to the noise I've been having.. a bad\loose rca cable at the DAC! Now I feel pretty dumb.. Anyways, I will be upgrading the bootstrap ceramics just to see if I can hear a difference:) Those suckers are expensive on eBay:(
 
Member
Joined 2008
Paid Member
There are two differential inputs "+" and "-" for each channel. With a single ended source, the "-" inputs are connected to Ground while the "+" inputs are connected to the the source. For the proper summing of the inputs, they must have the same impedance.

What is the purpose of those 1uF caps (C7 and C10 on danzz's schematic) if they are connected to common GND on the inputs? For single-ended/unbalanced/RCA input, with the YJ blue/black circuit being what it is, I don't understand why those caps can't be replaced with just a wire.

Now, a related but separate question: what if one wanted to use balanced (XLR) inputs with this board? Obviously that three-pin terminal block is out. (And in this mode, I understand why you need equal caps on the + and - of each input, as per Rhing's statement above.) So, again referring to danzz's schematic, R+ goes to C6, and L+ goes to C32 (i.e. same as stock/unbalanced). Now, R- should go to C7 and L- goes to C10.

Where does the GND for the input signal go in this case? Is it OK to connect to any convenient GND spot on the board? Or is there an ideal location?
 
impuls60:
-two series bipolar elco's have a lower distortion then most mkt/mkp's I believe I read somewhere:) 1 has a significant higher distortion. I did experience some "altered" sound when trying 1 green nichicon after listening very long to wima mks, I say altered, I don't know how to put into English words. I didn't wait very long to switch back to wima, didn't try double bipolars.

- what is the current outputfilter combination you have? inductor/capacitor? seems from testreport highfrequency respons of your speaker has a decreasing impedance, to ~4 ohm at 20khz and maybe dropping little further.
 
datasheet TI and also NXP's btw advise to connect the c7/c10 when input is SE to ground at the audio source, preamp or whatever source is, saying that will give better noise performance then connection to ground near chip.


TI and NXP and I think all chipproducers have same values for these caps in SE as inputcaps.

And NXP also states about ground for differential inputs:
4.2 Supply GND connection​
The best practice to avoid any common ground path with the power supply is to leave the​
supply floating. The power supply should be attached to GND at the amplifier side. The​
differential input should be grounded at the sound processor and not at the amplifier side.
 
What is the purpose of those 1uF caps (C7 and C10 on danzz's schematic) if they are connected to common GND on the inputs? For single-ended/unbalanced/RCA input, with the YJ blue/black circuit being what it is, I don't understand why those caps can't be replaced with just a wire.

They can't be replaced with just a wire because the input pins to the chip don't sit at 0V, they're biassed internally to some voltage above ground so as to be able to handle bipolar audio signals. So if you replace them with a wire you'll upset the internal biassing and have an amp which can only handle the positive going parts of an audio waveform, not the negative.

Now, a related but separate question: what if one wanted to use balanced (XLR) inputs with this board? Obviously that three-pin terminal block is out. (And in this mode, I understand why you need equal caps on the + and - of each input, as per Rhing's statement above.) So, again referring to danzz's schematic, R+ goes to C6, and L+ goes to C32 (i.e. same as stock/unbalanced). Now, R- should go to C7 and L- goes to C10.

Where does the GND for the input signal go in this case? Is it OK to connect to any convenient GND spot on the board? Or is there an ideal location?
When using balanced inputs, the GND for the input signal should go to the chassis earth directly. The GND on the PCB should be a wire from the same point on the chassis. If you're just using bare boards and have no chassis then connect the balanced in GND to the place where the GND enters the PCB from the power supply.
 
Last edited:
Bought Feixiang 2.1 amp. I'm very disappointed about the sound quality of the amp.
Previously I used Red YJ 2.0 board with a ne5532 preamp tone control board. The sound quality was extremely great in that setup.

There's no way to boost high frequency in Feixiang 2.1 board (which is possible in ne5532 preamp tone control board). Also adjusting the subwoofer frequency doesn't work as it should (I used a ne5532 low pass filter board and it worked very well with my L15d amp)

So in my point of view it is best to avoid these kind of boards (all in one).

Also I bought LM4562 opamps to replace ne5532 opamps on both Feixiang 2.1 board and my tone control board. I think those LM4562s are fake because I didn't hear an audible difference after replacing them.
 
Bought Feixiang 2.1 amp. I'm very disappointed about the sound quality of the amp.
Previously I used Red YJ 2.0 board with a ne5532 preamp tone control board. The sound quality was extremely great in that setup.

There's no way to boost high frequency in Feixiang 2.1 board (which is possible in ne5532 preamp tone control board). Also adjusting the subwoofer frequency doesn't work as it should (I used a ne5532 low pass filter board and it worked very well with my L15d amp)

So in my point of view it is best to avoid these kind of boards (all in one).

Also I bought LM4562 opamps to replace ne5532 opamps on both Feixiang 2.1 board and my tone control board. I think those LM4562s are fake because I didn't hear an audible difference after replacing them.

Well, it has been pointed out a number of times in this forum (please search the posts/thread) that "rolling opamps" do not necessary brings improvement in sound. A lot depends on the circuit design and the operation parameters that the opamps are subjected to. For example, in a recent thread, someone reported that the load impedance that the opamp is working with has significant impact on the sound.

As such, I do not think you can say the opamp is fake because you did not hear any "improvements in sound". You do not want to throw out a good piece by mistake. By the way, where did you buy the LM4562?

Regards