Oh THAT ONE !? Claude you know the saying, what is not claimed for a year and a day...... Anyway, now that you bought good LS for it, i'll enjoy to listen to it in
Gilles.
I tried this home made bank of 12 parallel 3300uF 50 V capacitors between the 32 V 5 A power supply sold by AIYIMA and the A04. Using top of the line Accuphase CD player and preamplifier, and JBL studio loudspeakers, the capacitor bank certainly makes a difference. The soundstage is more solid with more impact. My A04 also has OPA1656 OPAMPs. I now believe that when using SMPS for audio applications, it is a good choice to use a filter capacitance like the one used for linear transformer based power supply.
Attachments
If I am not in error, the open loop audio gain of a class D amplifier is proportional to the power supply voltage of the power MOS even if we have ideal MOS. I do not know the open loop gain of the TPA3251, but if it is not much higher than the closed loop gain of 20 dB: see https://www.ti.com/lit/gpn/tpa3251 the effect of dynamical changes of the open loop gain due to power supply voltage dynamical changes might not be negligible. These changes may follow the power envelope of real music.
It certainly helps in many cases.
However, it won't help the getting the HF garbage out, so it could be worthwhile to make also provision for a SMPS filter.
I strongly believe that "SMPS + SMPS Filter + additional PS capacity + additional bypass caps" can make in many audio cases an as good PS as the best top of the line linear ones. Bare perhaps for some applications around some Class AB amps
Claude
However, it won't help the getting the HF garbage out, so it could be worthwhile to make also provision for a SMPS filter.
I strongly believe that "SMPS + SMPS Filter + additional PS capacity + additional bypass caps" can make in many audio cases an as good PS as the best top of the line linear ones. Bare perhaps for some applications around some Class AB amps
Claude
It might be interesting to analyze the design of a capacitor bank....
Suppose that we want to provide 5 amps for one second. The electric charge is Q=current * time = 5 * 1 = 5 Coulomb.
Then we want that our capacitor discharges by 1 volt (delta V) during that period of time. The equation is charge Q= C * delta V = C * 1, therefore C=Q/delta V = 5/1 = 5 Farad = 5,000,000 uF, five million microfarads.
We may now say that our capacitor is indeed fully charged every 1/100 second, it follows that the required capacitance for the same voltage drop of 1 V is 1/100 of 5,000,000 uf, that is 50000 uF.
A04 has about 6000 uF inside, with the external 40000 uF we have 46000 uF
With 50 Hz mains frequency, a well designed diode bridge (and transformer) does indeed provide the 1/100 s periodic charging time. A SMPS should make a better job because it is actively regulated (in the 1/100 s interval and also in average), but because it does its job that way, it has an ACTIVE role in sound quality, something that we do not want !!
The conclusion is that if we accept one volt fluctuations at full load (5A) (say from 32 to 31 V, and let the audio amplifier compensate for that) and we do not want an active role for any kind of power supply (for the 1/100 s mains fluctuations), 40000 uF is the minimum capacitance to be added to this amplifier.
NOTE: the calulations are simplified, Charging a DC loaded capacitor with diodes and common transformers is a problem that can be solved with simulation software, no exact formula exists.
Suppose that we want to provide 5 amps for one second. The electric charge is Q=current * time = 5 * 1 = 5 Coulomb.
Then we want that our capacitor discharges by 1 volt (delta V) during that period of time. The equation is charge Q= C * delta V = C * 1, therefore C=Q/delta V = 5/1 = 5 Farad = 5,000,000 uF, five million microfarads.
We may now say that our capacitor is indeed fully charged every 1/100 second, it follows that the required capacitance for the same voltage drop of 1 V is 1/100 of 5,000,000 uf, that is 50000 uF.
A04 has about 6000 uF inside, with the external 40000 uF we have 46000 uF
With 50 Hz mains frequency, a well designed diode bridge (and transformer) does indeed provide the 1/100 s periodic charging time. A SMPS should make a better job because it is actively regulated (in the 1/100 s interval and also in average), but because it does its job that way, it has an ACTIVE role in sound quality, something that we do not want !!
The conclusion is that if we accept one volt fluctuations at full load (5A) (say from 32 to 31 V, and let the audio amplifier compensate for that) and we do not want an active role for any kind of power supply (for the 1/100 s mains fluctuations), 40000 uF is the minimum capacitance to be added to this amplifier.
NOTE: the calulations are simplified, Charging a DC loaded capacitor with diodes and common transformers is a problem that can be solved with simulation software, no exact formula exists.
Last edited:
Hmm... No time at all to elaborate here, but I suggest you indeed run a simulation... or use formula that are closer to the way a linear PS is charging and discharging (far away from a square form, the tranny conducts quite a lot etc.). BTW, when the tranny conducts you hear it aswell, so not just cap's sound acting as buffer all the time, and indeed active compensation has also its downside (I always found SMPS benefitted from a bit more buffering, which has also its downsides and is sometimes to wise dpending on the application, but say for power amps OK).
As for 1V 'fluctuation', that can be quite high already, it depends a lot on your amp... and its ability to reject PS fluctuation (PSRR). Key words are noise floor, hum etc., in relation to your signal... in fact you need to take the noise floor you want and substract the amp's PSRR to dertermine what "further max voltage fluctuation in dB (relative to your signal)" you need... etc.
Just as a sideline I believe the PSRR of the TI chip in our application is roughly 60dB, so 1V can be quite high already, depending on what you seek. But on the other hand it depends a lot if you use a linear PS or SMPS etc., in our case SMPS is doing already an excellent job re regulation and PS fluctuation under load (with some deficits of course, and that real life ripple is higher than advertised but can be found in most test sheets (Meanwell etc.).
Anyway... I realy have no time for this indeed intersting topic, but digging a bit you may find that to enable 1V fluctuation for a 5A load with a linear PS (you post)... 8500uF are adequate.
Of course nothing wrong going higher as long as you can / the rest of the PS copes at start up (but not sure you would here a difference*), nor of course that bigger / more / different caps wouldn't affect the sound in a different way (read stiff PS, ESR across frequencies, especialy as a Class D pumps at high frequency), but we already tried to address that with bypass caps. That's why IME caps with as low ESR as possible across a wide frequency range are important.
* we did't need more than 1 Watt in our application, so difficult to say were the limit lies if needing full power (not sure this amp would cope anyway), but we ended up with a total of over 19200uF AFTER a low ripple SMPS and adding more (we tried +6600uF) didn't bring any sonic difference... in our case.
All this ti say that the 4400uf used in the application sheet and that can be found inside the unit make indeed sense. With your aim and a linear PS, a total of 8500uF would be fine... but then I would use a SMPS with a HF good filter and some moderate additional caps (nothing close to 50 000uF).
Sorry, I won't elaborate more as time consuming, but i am sure others can and if not internet will provide you with these basics
Have fun
Claude
As for 1V 'fluctuation', that can be quite high already, it depends a lot on your amp... and its ability to reject PS fluctuation (PSRR). Key words are noise floor, hum etc., in relation to your signal... in fact you need to take the noise floor you want and substract the amp's PSRR to dertermine what "further max voltage fluctuation in dB (relative to your signal)" you need... etc.
Just as a sideline I believe the PSRR of the TI chip in our application is roughly 60dB, so 1V can be quite high already, depending on what you seek. But on the other hand it depends a lot if you use a linear PS or SMPS etc., in our case SMPS is doing already an excellent job re regulation and PS fluctuation under load (with some deficits of course, and that real life ripple is higher than advertised but can be found in most test sheets (Meanwell etc.).
Anyway... I realy have no time for this indeed intersting topic, but digging a bit you may find that to enable 1V fluctuation for a 5A load with a linear PS (you post)... 8500uF are adequate.
Of course nothing wrong going higher as long as you can / the rest of the PS copes at start up (but not sure you would here a difference*), nor of course that bigger / more / different caps wouldn't affect the sound in a different way (read stiff PS, ESR across frequencies, especialy as a Class D pumps at high frequency), but we already tried to address that with bypass caps. That's why IME caps with as low ESR as possible across a wide frequency range are important.
* we did't need more than 1 Watt in our application, so difficult to say were the limit lies if needing full power (not sure this amp would cope anyway), but we ended up with a total of over 19200uF AFTER a low ripple SMPS and adding more (we tried +6600uF) didn't bring any sonic difference... in our case.
All this ti say that the 4400uf used in the application sheet and that can be found inside the unit make indeed sense. With your aim and a linear PS, a total of 8500uF would be fine... but then I would use a SMPS with a HF good filter and some moderate additional caps (nothing close to 50 000uF).
Sorry, I won't elaborate more as time consuming, but i am sure others can and if not internet will provide you with these basics
Have fun
Claude
Hi,
Thanks all for sharing very insightful info, especially Claude!!
I did mod my A04 based on the recommendation, result is simply say great.
But now I did volume change for the DACT type small attenuator.
It is not difficult to remove the old one actually and easy to install the new one. Simply say, just fit!!
sound is clearly improved. I highly recommend this if you accept 24 steps volume control (for me, no problem)
Thanks all for sharing very insightful info, especially Claude!!
I did mod my A04 based on the recommendation, result is simply say great.
But now I did volume change for the DACT type small attenuator.
It is not difficult to remove the old one actually and easy to install the new one. Simply say, just fit!!
sound is clearly improved. I highly recommend this if you accept 24 steps volume control (for me, no problem)
Hi All, In all of the talk regarding the op-amps and capacitors.. Has anyone thought of using surface mount bypass capacitors? They have no leads. No leads = no lead inductance. An EIA0805 capacitor in many cases neatly fits between the legs of through hole parts. Just solder them to the pads. Another comment regarding op-amps. A trick I learned once upon a time from an audio design legend. Add a resistor from the output of the op-amp to ground. (Or the negative rail) The output of the op-amp is at around 6VDC in the case of this amp. If you add a 1.5k resistor to ground, (you can try low resistance values too for even more current. Sometimes distortion goes up... but so does musicality) the resistor is pulling ~4ma through the op-amp to ground. What does the op-amp have to do to compensate? Pull a DC bias in the opposite direction. This puts the op-amp into a state where it isn't dancing around the reference point of 6V. it's akin to forcing it into class A. This simple trick can turn even mundane op-amps into much better performers.
Hi Soundblaster
Thanks for your input
As reply to your question, yep, been there, done that.
There are of course some very fine SMD caps and inded sometimes their size come very handy between larger component's legs. And you get rid of quite a big portion of the parasitic inductance indeed thanks to the absence of legs.
There are of course some limitations due to the form factor and technology, eventhough SMDs without legs and housing etc. are quite good in relative terms. The prob is you can't get quality items in larger size (PPP etc.) obviously. More rleated to this project, it is probably easier for beginners to fit and replace / play with bypasses that are larger size - eventhough in fact SMDs are easy to remove. Another negative is that SMD caps are quite sensitive, if you mess with temp you can degrade them and unless you cook them it is difficult to find out you degraded them, so less good when it comes to experimenting and undoing several times.
As of me, at the end I had no idea where that would lead, and Gilles was doing most of the fitting due to COVID lockout, so it was just handier to work with good old components I knew well... and that were handy for some.
But there is nothing against SMD and once you know what values to go for / sound best, it is even easier to implement in series...
Have fun experimenting
Claude
Thanks for your input
As reply to your question, yep, been there, done that.
There are of course some very fine SMD caps and inded sometimes their size come very handy between larger component's legs. And you get rid of quite a big portion of the parasitic inductance indeed thanks to the absence of legs.
There are of course some limitations due to the form factor and technology, eventhough SMDs without legs and housing etc. are quite good in relative terms. The prob is you can't get quality items in larger size (PPP etc.) obviously. More rleated to this project, it is probably easier for beginners to fit and replace / play with bypasses that are larger size - eventhough in fact SMDs are easy to remove. Another negative is that SMD caps are quite sensitive, if you mess with temp you can degrade them and unless you cook them it is difficult to find out you degraded them, so less good when it comes to experimenting and undoing several times.
As of me, at the end I had no idea where that would lead, and Gilles was doing most of the fitting due to COVID lockout, so it was just handier to work with good old components I knew well... and that were handy for some.
But there is nothing against SMD and once you know what values to go for / sound best, it is even easier to implement in series...
Have fun experimenting
Claude
regarding the op amp trick, I played with it over 2 decades ago and still have my ART DI/O with AD825 amps fitted like this... browse and google, my name hasn't changed ;-)
However, a few bits:
Bottom line, all nice tricks and worth trying, but you need to know what you are doing and understand why it may work or not. One can not copy paste and also some old tricks do not apply to modern parts as the tech has evolved meantime. I used to do some very complicated bypasses on my old power amps. Today's PS caps need far less, but on the other hand today's caps, due to some build tech no more being used, don't last as long.
Trade off I guess, and we need to adapt, that's DIY
Have fun
Claude
However, a few bits:
- The resistor is Okish, but has some negative to bias in Class A... much better is to use a transistor and a resistor to do that. They are even more complex ways to bias / force in Class A an op amp, you can google, but overkill for us IMHO
- That trick works great on some OPA, and does nothing on some... and degrades some OPAs. If you search on this forum you will find that when OPA1656 was released some time ago (got some prototypes / early releases) I tried to force it into Class A with a resistor and transistor. It didn't work. I posted wondering why and the designer of the chip replied, it is in the OPA1656 thread (hint: it has already some similar trick built-in by design).
Bottom line, all nice tricks and worth trying, but you need to know what you are doing and understand why it may work or not. One can not copy paste and also some old tricks do not apply to modern parts as the tech has evolved meantime. I used to do some very complicated bypasses on my old power amps. Today's PS caps need far less, but on the other hand today's caps, due to some build tech no more being used, don't last as long.
Trade off I guess, and we need to adapt, that's DIY
Have fun
Claude
For anyone interested, I can confirm swappping those four 10 uf caps between OPAMP and the chip with 1uF MKP types doesn't lessen bass in any capacity! I have verified this using Frequency spectrum sweep test. Though I am doing this on A07, I hope the results should be consistent with A04 as well. I will be posting my entire mod here once i have done with my tweaks.
This makes sense. If one were to read the data sheet for the IC. The input resistance for the input pins is 24k. A 1uf cap will still give you a corner frequency of 6.63Hz. Going to 0.47uf will move it up to 14Hz. Of course phase shift starts creeping in but it will be easier on the power supply.
Hi, Great find ! Which out of 5 caps in the input stage are DC blocking caps for OPamps ? I am in mid of tweaking my unit .
Though this is for A07, i think the layout is pretty much the same as A04. Capacitor marked 1 is responsible for OPAmp power supply. Capacitor 2 shows continuity with Capacitor 1. So i changed them to Panasonic FM 220uF 35V (fits fine). For this position i have tried Nichicon KZ, Pana FM, Nichicon FW and few other but Pana FM was clearly the best audibly even with the stock NE 5532 (might be due to high ripple current as opposed to Samwha WF it replaced). I have the Cap 1 bypassed with Vishay MKT 370 0.22uF as well. Under the board the OPAamp VCC & GND is further bypassed with Wima FKS2 1500pF x 2 each, so 3000 pF in total. The OPA 1656 is in order & am looking forward to bypass the top terminals with a small size Pana FM electro &/or Wima 0.033uF FKP2 (largest size available).
The 5 caps before OPA has been changed to Nichicon FW (can't say if they made any significant difference). The 1st & 4th cap from the bottom right has been bypassed with wima 0.01 uf (just because i had them laying around). The middle of the 5 caps near the OPAmp has been changed to 100uF 35V. Previously i had completely removed the cap in that position and replaced it with Wee WME 2.2uF film cap. I saw some weird behaviour, the sound quality changed for very good initially then after some time the sound changed to mono like (like am listening to bluetooth speaker) with no seperation. And occasionaly right channel will stop working. I almost got a mini stroke. Changed back to NIchicon FW 100uf 35V. Everything back to normal ! Big Relief !!!
Based on your suggestion & mine initial experience as well i bypassed them with Wima but felt the soundstage collapsed & became somehow narrow, don't know if this was real or my bad experience with full film cap replacement tricking my mind but i finally removed the bypass. So, no bypass in this position now.
The four caps on the output side of OPA i am contemplating to change but am confused between Nichicon ES, Nichicon FG or Elna Silmic II. But thinking of doing this once i have my OPA 1656 in the place since that would define a better route in which would i like to go depending on their sonic signature (also as per your report no significant benefits). No bypass on these four caps as of yet.
Prior to these I changed the coils to Coilcraft 10uH (since i had them with me & they made significant upgrade in my TPA 3116 amp when i did it). Its pretty neat & is very firm. I am not able to compare the improvements made since most of the mods were done in a go & it's hard to imagine how the amp sounded in stock. but change must be very positive. The DC caps has been changed to Nichicon KZ 220uF 50V x2 (since these were my fav caps on other TPA 3116 board, based on various choices i tried) on the board & bypassed with extended wire to Nichicon FW 1000uf X 2. I didn't noticed any change with the increase in capacitance (might be due to my less demanding needs for wattage). Nichicon KZ is bypassed below the board with 0.22 uF Vishay MKT 370. I don't want to go crazy on bypass in this position since i am very happy with KZ sound signature & heavy bypass will alter that sparkly HF am getting. I am contemplating the idea to bypass those 4 small ceramics close to the chip itself with some FKS2 (or FKP2 if slim enough) since they fit inside the cavity of heatsink just fine so will be much beneficial than heavy bypassing on the power caps lead out. Also FKP2 or FKS2 have a much lower DF compared to MKT or MKP types being a foil cap and are much suited for pulse duty. So even a smaller value can outperform a larger value MKT or MKP.
After all these mods i really have no idea how much the sound have changed since i have no memory of what the amp sounded in stock (i almost jumped on modding it, considering how much difference it made on my other TPA 3116 amp) but irrespective of that, it just sounds very very good. Very much close to my reference DIY amp with even better low end. Am just hoping that OPA 1656 chip brings further sonic improvements.
Attachments
Last edited:
Cap 1 & 2(possibly) are responsible for VCC & GND supply line on OPamp. This is on A07 but most likely will be similar on A04 as well.
DIMM is your friend.
Attachments
Based on A07, but A04 might be very similar. Changing the cap in the middle of those 5 caps to film type collapses the soundstage to almost mono like & some weird behaviour of right channel cutting out. Done that, but changed to Nichi FW 100 uf 35V now. I tried bypassing them with Vishay 0.22 uF as well but felt the sound stage becoming narrower ! Weird !! I felt that bipolars (film or electrolytic) wont be favourable in this position , so removed the bypass. May be bcoz you are using a smaller bypass (0.1uF) you are not detecting this behaviour or may be my setup is more revealing.
Infact i doubt the middle of the 5 caps are the PS cap for the OPamp. The middle cap shows continuity with GND & INP pin on both the OPamps while the VCC & GND are in continuity with cap marked 1& 2. I have changed Cap 1 & 2 with low esr Pana FM & there is already a noticeable sound improvement with even the stock NE 5532. I have tried bunch of my fav caps in this position but Panasonic was clearly better sonically.
Attachments
Last edited:
It sounds like a lot of work, well done.
Without pix and numbers on it it is difficult to follow what has been done sadly. Well, must confess that since I don't own such a unit and it has been ages it would probably make no difference to mu understanding anyway, so don't waste time on this. Further, it appears this unit had quite a few variations along its production time, so what was true on the 3 units we tweaked might not necessarly be on other units. But these productions variations were quite small and after all this is DIY.
Without pix and numbers on it it is difficult to follow what has been done sadly. Well, must confess that since I don't own such a unit and it has been ages it would probably make no difference to mu understanding anyway, so don't waste time on this. Further, it appears this unit had quite a few variations along its production time, so what was true on the 3 units we tweaked might not necessarly be on other units. But these productions variations were quite small and after all this is DIY.