If you're calling out the leakage current of the decoupling capacitors as a bad thing, you're doing it wrong 🙂 Polymer caps like OSCON's can have well over a mA of leakage current, potentially 100x the leakage of the electrolytic capacitors you're replacing.
No, you're fine with this. If you want to use ceramics, you may go for C0G (Type1) or X7R (Type2) ceramics. If you need a tight tolerance fit, you might just measure the real capacitance of your caps.
What i meant for the -3dB lower cut-off is, when you change your gain, you'll have to change the input caps as well to maintain your desired cut-off. (So whats your desired -3dB lower cut-off? 2Hz, 10Hz, 20Hz) The higher the capacitance your chances to have pop-on-issues increase if you're using a pot in front with rather high resistance (>=10kOhm) due to inbalance at the differential inputs while charging to the bias-voltage.
Well, i am not saying "i know it all", don't get me wrong. I just find it curious that alot people here say, that changing caps will always improve "things". At first you should define "things" or what you want to improve. So is there something you miss in the actual configuration?
But lets see what we can "improve" here by changing the bulk caps.
As i cant find any informations about "GSK Taiwan" Caps, i had a look for something "similar/cheap" -> CapXon Caps.
As seen on the picture, these are 2200uF 25V, considered to be low ESR.
Having a look into the CapXon catalogue
http://capxon-europe.com/downloads/catalogue2012.pdf
at page 140 "GH series" gives us:
2200uF, 25V, ~2600mArms max ripple, 50mOhms ESR.
or page 142 "FH series":
2200uF, 25V, ~2880mArms, 22mOhms
The following calculations are derived from:
http://www.ti.com/lit/an/sloa031/sloa031.pdf
Lets assume the following parameters per Channel:
PVCC=20V
RLoad=4Ohm
Pout=50W (Which is 50Wpk -> 25W continous output power)
fs=400kHz
dutycycle_max = D = 90% = 0.9
Td = 1/fs = 1/400kHz = 2.5us
V_peak = SQRT(P_peak * R_L) = SQRT(50W*4Ohm) = 14.14V
I_peak = V_peak/R_L = 14.14V/4Ohm=3.54A
Cbulk_min = (I_peak * Td * Dmax)/V_ripple
Lets assume an allowed max ripple voltage of 1% = 0.2V at PVCC=20V:
Cbulk_min = (3.54A * 2.5us * 0.9)/0.2V = 39.825uF
Next bigger value would be 47uF to maintain 1% V_ripple.
Lets go further:
ESR_max = V_ripple/I_peak = 0.2V/3.54A = 0.056Ohm = 56mOhms
Regarding to the CapXON FH series we could use any cap with >=47uF and <=56ohms to meet this requirement which leads into: 680uF 25V 50mOhms from the datasheet (but only 1400mArms is a bit on the low end).
So having this lets calculate our final maximum ripple voltage by:
V_maxripple = I_peak * (((Td*Dmax/C) + ESR + R_DS(on))
with R_DS(on) = 2x*120mOhms (from TI Datasheet) = 240mOhm
At first, with the calculated values from above:
V_maxripple = 3.54A * (((2.5us*0.9/39.825uF) + 56mOhms + 240mOhms) = 1.24V
With the values of the CapXON FH cap:
2200uF, 25V, ~2880mArms, 22mOhms
V_maxripple = 3.54A * (((2.5us*0.9/2200uF) + 22mOhms + 240mOhms) = 0.93V
If you consider unlimited capacitance with zero ESR, the best you can achieve is:
3.54A * (0 + 0 + 240ohms) = 0.8496V
For the ease, we would say that 0.8496V = 100%, so 0.93V = 109.5%, and 1.24V = 145%.
Lets change these caps for an Panasonic OS-Con:
Panasonic OS-CON Conductive Polymer SEPF "25SEPF330M" 330uF 25V 14mohm 5000mArms
V_maxripple = 3.54A * (((2.5us*0.9/330uF) + 14mOhms + 240mOhms) = 0.8499V
Which is nearly our "100%" optimum. Compared just to ripple, you'll get ~10% better, but compared to our Vpeak = 14.14V:
0.93V = 6.577%
0.8499V = 6.01%
Our "gain on improvement" is 0.57%.
Finally lets see how things going when are at moderate listening levels, so our D would be 0.1:
Cbulk_min = (3.54A * 2.5us * 0.1)/0.2V = 4.425uF
.. 😱
With the values of the CapXON FH cap:
2200uF, 25V, ~2880mArms, 22mOhms
V_maxripple = 3.54A * (((2.5us*0.1/2200uF) + 22mOhms + 240mOhms) = 0.927V
Panasonic OS-CON Conductive Polymer SEPF "25SEPF330M" 330uF 25V 14mohm 5000mArms
V_maxripple = 3.54A * (((2.5us*0.1/330uF) + 14mOhms + 240mOhms) = 0.902V
compared to our Vpeak = 14.14V:
0.927V = 6.56%
0.902V = 6.38%
So 0.18% improvement.
So if you can hear <=1% improvement in ripple vs. V_peak, yes, there is an improvement.
Last edited:
You've just turned science on its head.
We have an observation here - that 'bass articulation improves' - presumably that's a testable (i.e. falsifiable) observation. Therefore the theory needs to rise to meet the observation. Theory that doesn't meet observations is to be discarded in favour of theory that does. Theory isn't to be used as a reason to dismiss observations - rather when they don't match up its time to test assumptions made in formulating the theory.
Then it appears to me that you've not understood scientific methodology - which in my understanding is where observations precede and inform theorizing. If I've got that wrong, do please explain how you see it.
Thank you very much for taking the time to explain. Much appreciated. I learn something from it.
I agree with you that quite often there are some "myth" or "assertions" circulating without much scientific support, leading to a lot of misconceptions and false ideas. Personally I prefer using "better" parts for my build if I can afford the price. Not necessarily because I think I will get better performance, but more for peace of mind that there are less chance of getting unexpected nasty surprises that "lesser quality" parts might bring. More for insurance purpose.
Using the same cap leakage as an example, the leakage itself is probably not doing any harm to the circuit. However, the leakage could be a sign of poor quality of the material used and the component might blow at the most unfortunate moment, costing a lot of damages. Since it is difficult to tell what is the root cause of the leakage, the only way to mitigate is to get better parts, risk management. Beside, apart from cost, there is no downside in using "better" parts.
I do apologize if my previous comments are overly harsh.
Regards,
No problem mate. 🙂 As gmarsh already said, the oscons leaks generally more than an aluminum cap.
Thanks guys for the lesson!!!...
now my doubts...
I wanted to mod the tpa3116 + pt2313 board I got last summer, it happens that one of the boards came with a ripped of inductor and I got a new main board, so now I have two main boards with the pt2313 and Tpa3116, Irrebeo made a good review of it, not the best tpa3116 amp, but zero noise and with a few mods it could be a nice amp for a little setup, with diferent inputs, one extra output and tone control and remote control...
So, at the signal inputs of the pt2313 the board comes with some elna 4.7uf electrolitic cap, the recomendation as per datasheet its 2.2uf, so I thought of changing them for the correct value, it seems that was designed to use 7.5x5mm film caps, I been looking at digikey and the I found this EPCOS alternative. but they are bigger, so I should put three on the top and 3 on the bottom. But I was wondering if it could be better to go for this WIMAS from mouser, which they fit perfectly in the gap, which do you think it will do a better job!?.
The other matter its bothering me its that at the output of the pt2313 the board comes with the same elnas as at the input, per datasheet its recommended 10uf, so I thought about some MUSE ES from nichicon, but then right after the output signal from the pt2313 the via goes directly to the input cap at the tpa 3116, its has a smd cap, which I have no idea which could be the value, so I thought either take it off and bypass the signal with some wire or changing them with some tdk input caps that I already have from a TI evm bom list. What do you think I should do!?
Other mods that I have in mind is the Oscons, the snubber, that with a little of surgery it can be done... Also maybe the inductors, but I still havent decided which ones, any recommendation!?.
I just thought, I have a Little bear P5 valve preamp and some JAN 5654W for it, I could put it between the pt2313 and the tpa3116... maybe it worths a try...
now my doubts...
I wanted to mod the tpa3116 + pt2313 board I got last summer, it happens that one of the boards came with a ripped of inductor and I got a new main board, so now I have two main boards with the pt2313 and Tpa3116, Irrebeo made a good review of it, not the best tpa3116 amp, but zero noise and with a few mods it could be a nice amp for a little setup, with diferent inputs, one extra output and tone control and remote control...
So, at the signal inputs of the pt2313 the board comes with some elna 4.7uf electrolitic cap, the recomendation as per datasheet its 2.2uf, so I thought of changing them for the correct value, it seems that was designed to use 7.5x5mm film caps, I been looking at digikey and the I found this EPCOS alternative. but they are bigger, so I should put three on the top and 3 on the bottom. But I was wondering if it could be better to go for this WIMAS from mouser, which they fit perfectly in the gap, which do you think it will do a better job!?.
The other matter its bothering me its that at the output of the pt2313 the board comes with the same elnas as at the input, per datasheet its recommended 10uf, so I thought about some MUSE ES from nichicon, but then right after the output signal from the pt2313 the via goes directly to the input cap at the tpa 3116, its has a smd cap, which I have no idea which could be the value, so I thought either take it off and bypass the signal with some wire or changing them with some tdk input caps that I already have from a TI evm bom list. What do you think I should do!?
Other mods that I have in mind is the Oscons, the snubber, that with a little of surgery it can be done... Also maybe the inductors, but I still havent decided which ones, any recommendation!?.
I just thought, I have a Little bear P5 valve preamp and some JAN 5654W for it, I could put it between the pt2313 and the tpa3116... maybe it worths a try...
How I'd hot-rod an amp:
Input capacitors definitely should be film, as they have very low leakage current, important here since the input capacitors have DC across them in the steady state. Value here isn't too critical, I think 1uF is fine for most applications, higher values just raise the highpass frequency of the amp. Note that the input resistance of the amp changes with the gain setting when you're calculating.
Decoupling capacitors should be low Z up into the MHz, and rated for a few amps minimum for reliability reasons. OSCON recommendation is a good one. You'll also want some MLCC ceramics close to the chip to handle the high frequency currents drawn by the TPA's power stage. The TPA is essentially a multichannel synchronous buck converter capable of switching >5A per channel, and it should be decoupled with the same care and attention that you'd put into a FPGA core voltage regulator or similar.
I think the output filter is probably the most effective change you can make..
Change the output inductors to a part that has a flat inductance versus DC curve up to 6+ amps, more is better especially if you're pushing the amp hard which will heat up the inductors. Coilcraft MSS1210 will drop in place of most 12x12mm SMT inductors and has a beautifully flat L versus DC curve.
If they're not already, change output filter capacitors to film capacitors, because of their negligible capacitance change with voltage. Both this effect, and inductance change with current in the inductor, will cause the output filter's response to change with the audio envelope and create intermodulation.
I'd match the output filter to the speaker impedance to minimize rolloff or peaking - I'd go with 10uH/1.5uF for 4 ohms, 10uH/0.68uF for 6 ohms, and 15uH/0.47uF for 8 ohms. Make sure the speaker impedance is flat to the high end of the audio band, add a zobel network if your speaker doesn't have one.
I don't really have an opinion on OUTxx snubbers or isolating AVCC, I'd have to spend some time in front of an Audio Precision set to know how effective these things are.
Input capacitors definitely should be film, as they have very low leakage current, important here since the input capacitors have DC across them in the steady state. Value here isn't too critical, I think 1uF is fine for most applications, higher values just raise the highpass frequency of the amp. Note that the input resistance of the amp changes with the gain setting when you're calculating.
Decoupling capacitors should be low Z up into the MHz, and rated for a few amps minimum for reliability reasons. OSCON recommendation is a good one. You'll also want some MLCC ceramics close to the chip to handle the high frequency currents drawn by the TPA's power stage. The TPA is essentially a multichannel synchronous buck converter capable of switching >5A per channel, and it should be decoupled with the same care and attention that you'd put into a FPGA core voltage regulator or similar.
I think the output filter is probably the most effective change you can make..
Change the output inductors to a part that has a flat inductance versus DC curve up to 6+ amps, more is better especially if you're pushing the amp hard which will heat up the inductors. Coilcraft MSS1210 will drop in place of most 12x12mm SMT inductors and has a beautifully flat L versus DC curve.
If they're not already, change output filter capacitors to film capacitors, because of their negligible capacitance change with voltage. Both this effect, and inductance change with current in the inductor, will cause the output filter's response to change with the audio envelope and create intermodulation.
I'd match the output filter to the speaker impedance to minimize rolloff or peaking - I'd go with 10uH/1.5uF for 4 ohms, 10uH/0.68uF for 6 ohms, and 15uH/0.47uF for 8 ohms. Make sure the speaker impedance is flat to the high end of the audio band, add a zobel network if your speaker doesn't have one.
I don't really have an opinion on OUTxx snubbers or isolating AVCC, I'd have to spend some time in front of an Audio Precision set to know how effective these things are.
My board just arrived.
It is slightly different from yours, not by much though.
I hooked it up quickly, no pop, no hiss.
Attachments
Xiongzhi's GSK JZ is supposed to be low esr, that is the green one in some pics. Boards seem produced with the black Xiongzhi GSK JH, who knows the chinese sellers might find that low esr too, but could also be cheapest general purpose capacitor they could find and that does make sense too. Then we assume Taiwan GSK are equal to Xiongzhi's GSK, both we don't have any data on LOL and China could even produce fake Xiongzhi GSK capacitors just because they produced the wrap foil and operator let it run for 1 minute longer to sell foil to his neighbour who produces GSK's in his living room, picking up capacitors from dumpcontainers near productionfacilities. It is like making an assumption not having any data on Feastrex speaker that it will be constant 4 ohms. Or having an unmarked "Panasonic" cap and saying lets pick FM datasheet to have data to use. It will tell you FM capacitor is indeed ok, like the CapXon could be ok, CapXon isn't mentioned too often on bad cap posts on the web.
Hi,
here is my first black/blue board with just a little mods: bootstrap caps, Panasonics 330uF/25V and Coilcraft inductors. With a 50k pot.
I biefly tested it with a 12V SMPS and I have a problem:
Right channel is working apparently perfectly, but the left one is almost mute: only from half potentiometer up I start hearing sound.
I don't know where to check, since everything went very easily with the mods and all the connections seem ok.
Any idea, please..?
here is my first black/blue board with just a little mods: bootstrap caps, Panasonics 330uF/25V and Coilcraft inductors. With a 50k pot.
An externally hosted image should be here but it was not working when we last tested it.
I biefly tested it with a 12V SMPS and I have a problem:
Right channel is working apparently perfectly, but the left one is almost mute: only from half potentiometer up I start hearing sound.
I don't know where to check, since everything went very easily with the mods and all the connections seem ok.
Any idea, please..?
Irribeo, feel free to measure and evaluate your parts, the math is there.
But wait, as from some postings of you before, you prefer fake caps instead of genue stuff.
As I stated in the post above, the GSK data is an assumption, of course, but that doesn't makes the math wrong.
Do you have any useful hints beside your LOLing about "not your opinion?
It would be an ease to just measure the specs of those parts and evaluate the rest of the system. Did you ever considered that? Would be much more useful regarding this discussion than just blaming others for being a believer in "heard-about, must do, is better".
Dont get it, really.
Regards.
But wait, as from some postings of you before, you prefer fake caps instead of genue stuff.
As I stated in the post above, the GSK data is an assumption, of course, but that doesn't makes the math wrong.
Do you have any useful hints beside your LOLing about "not your opinion?
It would be an ease to just measure the specs of those parts and evaluate the rest of the system. Did you ever considered that? Would be much more useful regarding this discussion than just blaming others for being a believer in "heard-about, must do, is better".
Dont get it, really.
Regards.
Last edited:
In all your post doctormord you assume the designers know what they are doing. Better assume the cheapest capacitor available is used then on the Chinese ampboards, because that is what is found on the ampboards. I admire your math but your goal is to proof replacing the cheapest the chinese could find isn't going to matter. Forget about that and your post could be valued better, or buy yourself some chinese ampboards and find out that replacing the electrolytics changes quite a lot.
Does anyone know the quality of this board:
XH M190 2 100W Digital Amplifier Board 12V 24V TPA3116 Dual Channel Amplifier | eBay
I cannot find much information about it.
XH M190 2 100W Digital Amplifier Board 12V 24V TPA3116 Dual Channel Amplifier | eBay
I cannot find much information about it.
And the only proof you have at this are your ears? Did you ever took a measure on this? Ripple current, impedance, DC resistance? Let's say you are right and they are cheap, does this also mean they are bad/crap? I mean I can buy some decent Nichicons at Mouser that are cheaper than Panasonic and are within the same specs. Does the price make them more worse than the Panasonics? As from the past, there is a post where you say, that fake OS-CONs are better than genue Panasonics/OS-CONS from Mouser. Aren't this fake OS-CONs cheaper than the originals but perform better to your saying? Having really low ESR within 2200uF is much easier than doing so in 220/330uF. Just sayin. Do you have some of those "crappy cheap" Chinese caps at your office? If so, pick them up and do some measurements or get some valid data so we can discuss further and do some math and/or simulations to really see what's doing on. 😉
Thanks for the info gmarsh
I have read in some other forum, I dont recall which one now, some opinion about this amp, having a lack of bass.
Talking from the ignorance.
At the input of the PT2313 IC it has a 4,7uf, then at the output it has the same 4,7uf and then the signal goes to the input cap of the TPA3116 that it suppose to be 1uf, that would make a total capacitance of 0,70uf at the input of the TPA3116. Of course all this calculation its being done supposing the the input cap of the TPA3116 its 1uf. So according to this, the value of the total capacitance using the correct values respecting the PT2313 recommendations it would be of 0,64uf. So now it make sense to me that the producer of this amp had chosen those values for this design. Then I don't see here a reason of having a lack of bass, unless the fact of using electrolytic instead of film, which I dont know.
So, now to the point, I have clear now, that I should change the PT2313 input caps per film ones, But... Should I use film caps at the output of the PT2313?, I would say yes, and if I use film for both, I would take off the smd input cap of the TPA 3116 and connect the pads directly with a little piece of thin wire. What do you think!?
I will go for OSCONS then.
I will upgrade the inductor for the coildcraft using your recommendation for 8hms.
As I already have from some other project the smd caps and resistor for this, and after reading that its a really improvement. Once I have to take the inductors away, It will need a little of surgery, but not a very complicated one. I will do it.
The best of all, that as I have to main boards, I can make A/B tests with both boards and see which mods has the best improvement.
The first I will do cause it has to be done as this board has a ripped of inductor it would be changing the inductors and at the same time, as it will be easier to work without the inductor on the board, it would be bosttrap snnuber and the output cap.
I will report whenever I have time for this.
I have read in some other forum, I dont recall which one now, some opinion about this amp, having a lack of bass.
Talking from the ignorance.
At the input of the PT2313 IC it has a 4,7uf, then at the output it has the same 4,7uf and then the signal goes to the input cap of the TPA3116 that it suppose to be 1uf, that would make a total capacitance of 0,70uf at the input of the TPA3116. Of course all this calculation its being done supposing the the input cap of the TPA3116 its 1uf. So according to this, the value of the total capacitance using the correct values respecting the PT2313 recommendations it would be of 0,64uf. So now it make sense to me that the producer of this amp had chosen those values for this design. Then I don't see here a reason of having a lack of bass, unless the fact of using electrolytic instead of film, which I dont know.
So, now to the point, I have clear now, that I should change the PT2313 input caps per film ones, But... Should I use film caps at the output of the PT2313?, I would say yes, and if I use film for both, I would take off the smd input cap of the TPA 3116 and connect the pads directly with a little piece of thin wire. What do you think!?
I will go for OSCONS then.
I will upgrade the inductor for the coildcraft using your recommendation for 8hms.
As I already have from some other project the smd caps and resistor for this, and after reading that its a really improvement. Once I have to take the inductors away, It will need a little of surgery, but not a very complicated one. I will do it.
The best of all, that as I have to main boards, I can make A/B tests with both boards and see which mods has the best improvement.
The first I will do cause it has to be done as this board has a ripped of inductor it would be changing the inductors and at the same time, as it will be easier to work without the inductor on the board, it would be bosttrap snnuber and the output cap.
I will report whenever I have time for this.
I'm using my modified Blue/black board from yj. Ordered another one. I'm planning to use them in pbtl mode with differential input. I have a spare lambda dual 24v output smps at hand. Can I drive both mono boards with this one? According to smps specs max current output is 4.5A. If this approach is not logical I'll build a linear supply with dual secondary toroid xformer.
DJ le Roi; just from looking at the picture the high value inductors would indicate that the amp is designed for 8ohm+ speakers. Also I dont like that the power caps are that far from the amp chips and that close to the audio input side either. I would guess that the amp has a lot of hiss/white noise because of that. I guess you could use it of <87db speakers and for far field listening perhaps.. Sorry for my not so positive opinions cause it really is cheap!