> Cbulk=ipeak*td*dmax/vripple .. =2A*1/250khz*0.95 / 0.1v = 76uf .. cbulk 2 channels = 152uf
Had to stare at that. I guess you have computed ripple at the 250KHz switching frequency. I do not think this matters. The chip designers considered this. The essence of amplifiers, especially feedback amplifiers, is that they "ignore" power supply ripple by making your various-size audio voltages out of the larger available voltage. And I know the class D design community has worked hard to make this problem inaudible.
The issue in AUDIO amplifiers is to keep the supply from "feeling" the audio demand too strongly. Working with simple amps and 8 ohm loads this is similar to "strongly bypassing" a ~~50 Ohm load at the lowest audio frequency. 50 Ohms at 20Hz suggests a 200uFd cap for "slight bypassing". By experience, 1,000uFd is better but not great, 2,000uFd much better, and tens of thousands uFd when cost/space is no object. However your 2-amp BTL outputs "see" 4 ohms each side, so we need 8,000uFd per channel and say 15,000uFd for stereo.
> it has 2 4400uf
OK, this matches my "1,000uFd is ...not great" rule of thumb, when re-levered for the way this amp work. And about what I would expect on a low-cost kit. It works. It's a good deal. Many buyers will be thrilled.
I think the 150uFd you computed should already be ON there as small HF caps. For fine audio you want 100X as much bulk audio capacitance so the amplifier does not have to suck directly against the power brick to handle a large transient.
Caps for this purpose can be plain "power supply" caps. Made in large quantity for millions of uses, and quite cheap. Parallel with a few 10uFd fancy-caps if you wish.
Had to stare at that. I guess you have computed ripple at the 250KHz switching frequency. I do not think this matters. The chip designers considered this. The essence of amplifiers, especially feedback amplifiers, is that they "ignore" power supply ripple by making your various-size audio voltages out of the larger available voltage. And I know the class D design community has worked hard to make this problem inaudible.
The issue in AUDIO amplifiers is to keep the supply from "feeling" the audio demand too strongly. Working with simple amps and 8 ohm loads this is similar to "strongly bypassing" a ~~50 Ohm load at the lowest audio frequency. 50 Ohms at 20Hz suggests a 200uFd cap for "slight bypassing". By experience, 1,000uFd is better but not great, 2,000uFd much better, and tens of thousands uFd when cost/space is no object. However your 2-amp BTL outputs "see" 4 ohms each side, so we need 8,000uFd per channel and say 15,000uFd for stereo.
> it has 2 4400uf
OK, this matches my "1,000uFd is ...not great" rule of thumb, when re-levered for the way this amp work. And about what I would expect on a low-cost kit. It works. It's a good deal. Many buyers will be thrilled.
I think the 150uFd you computed should already be ON there as small HF caps. For fine audio you want 100X as much bulk audio capacitance so the amplifier does not have to suck directly against the power brick to handle a large transient.
Caps for this purpose can be plain "power supply" caps. Made in large quantity for millions of uses, and quite cheap. Parallel with a few 10uFd fancy-caps if you wish.
It's done...
Here is the final result. My friend had some nice caps but were not 25v so we had to use the cheap ones I had, figured the ESR would be low with 3 caps. Also, the TEAPO caps had roughly the same ESR as the ones that were in there so not too bad. I'll wait for my friend to make a big capacitor order so I can get mines and save on shipping but for now they sound better (I'm sure it's placebo) but at least I know they won't blow up with my 19v psu.
Thanks for the help guys, it was a fun first electronics project. Next up: building a Gainclone (eventually...)
Here is the final result. My friend had some nice caps but were not 25v so we had to use the cheap ones I had, figured the ESR would be low with 3 caps. Also, the TEAPO caps had roughly the same ESR as the ones that were in there so not too bad. I'll wait for my friend to make a big capacitor order so I can get mines and save on shipping but for now they sound better (I'm sure it's placebo) but at least I know they won't blow up with my 19v psu.
Thanks for the help guys, it was a fun first electronics project. Next up: building a Gainclone (eventually...)
I could probably swap them out along with the Teapos when I order better ones. It does sound great as it is right now but it's good for soldering practice haha. What should I look for? Low ESR as well?
They are film capacitors, kind of lossless compared to electrolytics. Value, voltage, size and maybe when there is still choice then, MKP (polypropylene). It is between outputs, said to be critical position, when not between outputs 100V seems good voltage, really haven't read about voltage rating between outputs, it should cancel out frequencies, but maybe double the voltage rating then ????
How can I verify what voltage is required? As far as I recall I didn't see any markings on the caps when I opened the amp and the datasheet doesn't specify voltage. Wouldn't I need only around 25 too? Why 100v?
With open outputs or high impedance speakers the outputfilter could oscillate. First couple post in long tpa3116 thread indicate that 100V is needed for each output. The AD modulated tda chip has one capacitor between the +/- outputs, that capacitor sees double the voltage, so maybe 200V is needed ? SMSL98 uses 250V for same position.
I just checked and those to are 100nf and 63v. I'm not sure those 2 are the filter caps, I think they could be named incorrectly. Wouldn't the filter ones be the 2 big and 4 smaller ones by the output? Those are .22uf 63v and .47uf 100v.
In your amp the two just behind the teapo and just before the speakeroutput, in smsl98e the two green just after the speakeroutput, 0.47uF 250V there
(edit: in tpa3116 main 50V (680nF (2 left 2 right) capacitor in filter between positive outputsignal and ground and another 680nF capacitor between negative outputsignal and ground burnt instantly when powered up without a load/speaker and from 19.5V powersupply. Your equivalent capacitors (1 left 1 right) are between positive outputsignal and negative outputsignal, they see twice the voltage the tpa3116 ones see, for 12V psu smsl may use 100V??, for 35V powered smsl98e they use 250V).
(edit: in tpa3116 main 50V (680nF (2 left 2 right) capacitor in filter between positive outputsignal and ground and another 680nF capacitor between negative outputsignal and ground burnt instantly when powered up without a load/speaker and from 19.5V powersupply. Your equivalent capacitors (1 left 1 right) are between positive outputsignal and negative outputsignal, they see twice the voltage the tpa3116 ones see, for 12V psu smsl may use 100V??, for 35V powered smsl98e they use 250V).
Last edited:
Hi all, I have the same board (SMSL 36pro) with one big caps. I have few good Panasonic caps 4700uf 16V with low ESR. I was thinking to substitute the big cap with these two. but I saw the mod that brozium did putting 3 4700uf 25v caps. Is it fine if I do the same mod but using the panasonic caps? Do you think will be worth to do it? Sorry for the noovb question
Hi all, first post for me. I've an SMSL SA-36A Pro that is a few years old now and uses the TPA3118 chip (link includes pictures of the guts of the amp). I've used it with a variety of speakers over the years, but have never really been satisfied with the amount of bass it produces with any of them. As a result, I am considering making some changes...
I thought I might start first with the PSU, as I'm not impressed by the fit or finish of the supplied 12V 3.8A unit. Would using a 24V 5A PSU likely be beneficial? Is a 2.1mm or 2.5mm inner connector needed? I was considering the CWT CAD120241, although it is £30.
Thanks!
I thought I might start first with the PSU, as I'm not impressed by the fit or finish of the supplied 12V 3.8A unit. Would using a 24V 5A PSU likely be beneficial? Is a 2.1mm or 2.5mm inner connector needed? I was considering the CWT CAD120241, although it is £30.
Thanks!