I dread starting a capacitor discussion as much as anyone, but I've been looking at datasheets on various brands and types of capacitors to use for the +/- power rails on a SS power amplifier. I work mostly on tube amps, so I have years of experience knowing which type and brand to use for a given application. With SS rail caps, I'm not as sure which way to go.
For example, there are the well-known Nichicon Audio Grade lines like LKG and others. But if you look at the datasheet specs by other companies, you see that in recent years, they have managed to bring ESR down in some cases to half that of the Nichicon Audio Grade capacitors and have increased ripple current figures. Companies like TDK/Epcos and United Chemi-con seem to be leading in this area.
Reading sites like the Elliot Sound Projects pages, there's an emphasis on low ESR for SS power supplies, and I've seen some techs using arrays of smaller capacitors to replace single larger capacitors to lower ESR by putting them in parallel. In other words, I see emphasis on this parameter from multiple sources.
On the other hand, Nichicon specifically markets several lines of capacitors specifically for audio, and I've used several of them with good results in signal path applications like KZ and ES Muse types. For the most part, however, other companies simply don't market their capacitors in this way for specific applications. (Elna comes to mind as another company that does, but they are mainly for signal path.) Nichicon doesn't state exactly what it is about their audio grade capacitors that makes them better for audio.
Making comparisons even more difficult, you have to do a pretty deep dive into the datasheets to compare apples to apples since the various companies don't all report their specs in the same way. Math is involved.
So if I'm looking for a 3300uF capacitor, I could go with a Nichicon LKG "Type II" with 120mΩ ESR. But there is also a Vishay/BC Components capacitor from the MAL2056 series listed as having 67mΩ ESR @ 100Hz and 40mΩ at 10kHz. It's also listed as a 12,000 hour part compared to 1,000 for the Nichicon. Ripple current is about the same as Nichicon, but for a longer life.
Then there's a TDK/Epcos cap from the B41505 series that claims 56mΩ ESR @ 100Hz and 55mΩ ESR @ 10kHz. It also claims around 60% greater ripple current than Nichicon or BC Components and a life of 5,000 hours @ 105C, 12,000 hours @ 85C, and 25,000 hours @ 40C. Looks like this line has been around since 2016.
On paper, it looks like the TDK capacitor would win the specifications contest, but does that lead to better sonic performance specifically for a power supply rail capacitor? Would Nichicon's audio grade sound better even though Nichicon is pretty vague on why that would be the case? Or is it simply the case that Nichicon pursues the audio market in a way their competitors do not? I've used Panasonic and Chemi-con capacitors in some power supplies in the past, as well as Nichicon, but I can't claim one sounded better than the other. They were all very good. And is that the answer? Has performance of all modern electrolytics by top companies gotten so good that it's just splitting hairs as to which is "best"?
For example, there are the well-known Nichicon Audio Grade lines like LKG and others. But if you look at the datasheet specs by other companies, you see that in recent years, they have managed to bring ESR down in some cases to half that of the Nichicon Audio Grade capacitors and have increased ripple current figures. Companies like TDK/Epcos and United Chemi-con seem to be leading in this area.
Reading sites like the Elliot Sound Projects pages, there's an emphasis on low ESR for SS power supplies, and I've seen some techs using arrays of smaller capacitors to replace single larger capacitors to lower ESR by putting them in parallel. In other words, I see emphasis on this parameter from multiple sources.
On the other hand, Nichicon specifically markets several lines of capacitors specifically for audio, and I've used several of them with good results in signal path applications like KZ and ES Muse types. For the most part, however, other companies simply don't market their capacitors in this way for specific applications. (Elna comes to mind as another company that does, but they are mainly for signal path.) Nichicon doesn't state exactly what it is about their audio grade capacitors that makes them better for audio.
Making comparisons even more difficult, you have to do a pretty deep dive into the datasheets to compare apples to apples since the various companies don't all report their specs in the same way. Math is involved.
So if I'm looking for a 3300uF capacitor, I could go with a Nichicon LKG "Type II" with 120mΩ ESR. But there is also a Vishay/BC Components capacitor from the MAL2056 series listed as having 67mΩ ESR @ 100Hz and 40mΩ at 10kHz. It's also listed as a 12,000 hour part compared to 1,000 for the Nichicon. Ripple current is about the same as Nichicon, but for a longer life.
Then there's a TDK/Epcos cap from the B41505 series that claims 56mΩ ESR @ 100Hz and 55mΩ ESR @ 10kHz. It also claims around 60% greater ripple current than Nichicon or BC Components and a life of 5,000 hours @ 105C, 12,000 hours @ 85C, and 25,000 hours @ 40C. Looks like this line has been around since 2016.
On paper, it looks like the TDK capacitor would win the specifications contest, but does that lead to better sonic performance specifically for a power supply rail capacitor? Would Nichicon's audio grade sound better even though Nichicon is pretty vague on why that would be the case? Or is it simply the case that Nichicon pursues the audio market in a way their competitors do not? I've used Panasonic and Chemi-con capacitors in some power supplies in the past, as well as Nichicon, but I can't claim one sounded better than the other. They were all very good. And is that the answer? Has performance of all modern electrolytics by top companies gotten so good that it's just splitting hairs as to which is "best"?
I tend to use banks of Panasonics. Nichicon have been reported to be a little noisier.
What sounds better? Well, in my experience, MORE sounds better. Maybe with some headphone amps and really good youthful hearing could one hear any other sonic differences. Through speakers, not me. But going from 6,000 uF single cans to 27,000 uF array was clearly audible in "liveliness". One of those subjective hard to define terms. I am sure one could measure rail droop dynamically and make some objective measures.
On noise, again, you could easily measure any difference in noise. But as I tend to have additional bypass caps of low ESR and low inductance at the points of power consumption, not sure the cap differences are meaningful there either.
I believe "audio grade" is a label put on a cap to justify a higher price than the same cap as a generic electro.
What sounds better? Well, in my experience, MORE sounds better. Maybe with some headphone amps and really good youthful hearing could one hear any other sonic differences. Through speakers, not me. But going from 6,000 uF single cans to 27,000 uF array was clearly audible in "liveliness". One of those subjective hard to define terms. I am sure one could measure rail droop dynamically and make some objective measures.
On noise, again, you could easily measure any difference in noise. But as I tend to have additional bypass caps of low ESR and low inductance at the points of power consumption, not sure the cap differences are meaningful there either.
I believe "audio grade" is a label put on a cap to justify a higher price than the same cap as a generic electro.
I bought a reasonably expensive LCR meter (ET4510) with 4-point Kelvin probes, so I could measure these things for myself and not rely upon datasheets. If you'll pay the shipping expenses for round-trip shipping, I'll measure your capacitors' ESR at several different frequencies.
(Sales page on Amazon)
(Sales page on eBay)
(Users Manual)
(Sales page on Amazon)
(Sales page on eBay)
(Users Manual)
I completely understand the purpose of capacitor arrays and how that puts capacitor ESR in parallel, but it's not always an option due to clearance/practicality issues. I'm asking here about SINGLE capacitors.
Though I don't have as sophisticated a capacitor tester as the one mentioned above, I have an impedance bridge that has internal 1kHz test signals and the option of supplying external signals. I also have the Anatek Blue ESR meter that tests at 100kHz.
As I said in my post, it may be the case that no one really knows the answer, and I can understand why. Swapping caps for testing is a PITA, and if you're like me, once you hear something that sounds good, your enthusiasm for disassembly, parts replacement, and reassembly wanes a bit.
Though I don't have as sophisticated a capacitor tester as the one mentioned above, I have an impedance bridge that has internal 1kHz test signals and the option of supplying external signals. I also have the Anatek Blue ESR meter that tests at 100kHz.
As I said in my post, it may be the case that no one really knows the answer, and I can understand why. Swapping caps for testing is a PITA, and if you're like me, once you hear something that sounds good, your enthusiasm for disassembly, parts replacement, and reassembly wanes a bit.
Don't know if Nichicon LKG "for audio" is a significant breakthrough or snake oil. ESR is one critical dimension. Another is self inductance. High self inductance of wound caps can lead to lack of response to high frequency signals like drum hits, tinkly bells, top octave wood piano, etc.
Some amps I repair have a pair of 0.22 uf film caps paralleled right on the feed line to the output transistors. Whereas the 10000 uf wound caps are 8" away by wire harness. I would say the sharp edges the output transistors are fed by the film cap until the current from the wound cap has time to get there. Some people parallel wound caps at the source with a smaller film cap. Some use a smaller electrolytic cap mounted on top of the big wound rail cap. Experts warn this arrangement can lead to oscillation. As the designers of my PV-1.3k included the bypass caps in the original design, I don't have to worry about the oscillation bugaboo.
I've used nichicon & panasonic industrial grade long service life caps in all my upgrade of geriatric class AB amps. I've used a few CDE or multicomp when I couldn't get the first two. All caps service life >2000 hours. No cost cutter parts. No problems so far. I play a top octave piano track through SP2(2004) speakers as part of my QA check. My ears go to 14 khz, and I have a Steinway console piano for calibration of proper sound.
I've used "low ESR" caps only for front of switcher power supplies. Seems to be required there. A recent QSC CX302 from early switcher era has 10 parallel 470 caps, instead of one 4700. Probably additional board space for 10 caps was cheaper than low ESR caps 25 years ago.
Some amps I repair have a pair of 0.22 uf film caps paralleled right on the feed line to the output transistors. Whereas the 10000 uf wound caps are 8" away by wire harness. I would say the sharp edges the output transistors are fed by the film cap until the current from the wound cap has time to get there. Some people parallel wound caps at the source with a smaller film cap. Some use a smaller electrolytic cap mounted on top of the big wound rail cap. Experts warn this arrangement can lead to oscillation. As the designers of my PV-1.3k included the bypass caps in the original design, I don't have to worry about the oscillation bugaboo.
I've used nichicon & panasonic industrial grade long service life caps in all my upgrade of geriatric class AB amps. I've used a few CDE or multicomp when I couldn't get the first two. All caps service life >2000 hours. No cost cutter parts. No problems so far. I play a top octave piano track through SP2(2004) speakers as part of my QA check. My ears go to 14 khz, and I have a Steinway console piano for calibration of proper sound.
I've used "low ESR" caps only for front of switcher power supplies. Seems to be required there. A recent QSC CX302 from early switcher era has 10 parallel 470 caps, instead of one 4700. Probably additional board space for 10 caps was cheaper than low ESR caps 25 years ago.
Hi indianajo, We were wondering what had happened to you the other day on OF. Someone was discussing one of your old posts. Glad to know you're still up and running. I just signed up for this site.
Good point about self-inductance. Some of the caps I listed, the Vishay and TDK, give impedance values at different frequencies. Nichicon does not list anything other than ESR and allowable ripple current at 120Hz.
Good point about self-inductance. Some of the caps I listed, the Vishay and TDK, give impedance values at different frequencies. Nichicon does not list anything other than ESR and allowable ripple current at 120Hz.
I tend to put a small, like 10uf or less film cap in parallel to drop the esr to help the array of paralleled electrolytics. lower esr does give more slam for lack of a better term. The film cap gives that leading edge of power delivery until the electrolytics are up to speed. Also one big cap is under more stress than a few in parallel. I like 105C caps because they are not sweating at 85C and last longer.
I stoped using LKG years ago after comparing ESR. Now use UCC KMH or SMH for main caps. Nichicon UPW on unregulated rails sometimes on regulated where I try to squeeze in an audio cap like UKW, UFG or...
Is ripple/ESR that important on the regulated rails, sure still target low ESR but does it need to be the lowest...
Is ripple/ESR that important on the regulated rails, sure still target low ESR but does it need to be the lowest...
The rail caps on an SS amp are for stability mainly - the bulk decoupling comes from the filter caps on the power supply, and that's going to predominate at audio frequencies, since the thick wires to the bulk caps are typically less resistance than the dinky rail caps' ESR. At RF frequencies the inductance starts to matter and that's where the rail caps do the work of stabilizing by preventing inadvertent HF feedback paths via the rails - the filter caps may have too much ESL and there may be too much wiring inductance between amp board and PSU to work well at higher frequencies.
You might see 10mF bulk caps perhaps 10 milliohms away from the amp, and 100µF caps on the rails.
At 5kHz for instance a 100µF cap has a reactance of 0.3 ohms or so. The main filter caps at 10mF are 3 milliohms of reactance at the same frequency. Guess who's doing most of the work at 5kHz - I suspect the filter caps 🙂
So the important parameters of the rail caps is the ESL and high frequency performance, though adding ceramic decoupling can relax these requirements considerably (which is why its normally done).
You might see 10mF bulk caps perhaps 10 milliohms away from the amp, and 100µF caps on the rails.
At 5kHz for instance a 100µF cap has a reactance of 0.3 ohms or so. The main filter caps at 10mF are 3 milliohms of reactance at the same frequency. Guess who's doing most of the work at 5kHz - I suspect the filter caps 🙂
So the important parameters of the rail caps is the ESL and high frequency performance, though adding ceramic decoupling can relax these requirements considerably (which is why its normally done).
Mark, As someone who's primarily a tube amp tech used to dealing with B+ supplies only, maybe I'm using the wrong terminology, but this is a weird amp. It's a Carver M-4.0T, so it has three B+/B- supplies operating from secondaries on the same transformer. Each set of +/- rails has its own rectifier bridge and filter caps.
So it's the filter caps I'm asking about, but there are three pairs of them: 2 x 9,000uF for the high voltage rails, 2 x 2200uf for the mid voltage rails, and 2 x 3300uF for the low voltage rails. And really I'm only asking about the mid and low voltage rails because the only practical way these days to replace the high voltage rail caps is to use an array. I don't think anyone makes a 9-10,000uF cap rated at least 125V that will fit in place of the originals, which were long and thin.
I inherited this amp along with tons of other gear from my late brother, so I'm making my way through it.
I have used the nichicon kg type 1 and type 2 and then the super through caps on my aksa nirvana amp. The clear winner was the type 3 super through and yes they are the most expensive caps.worth the money for better sound definitely.
Well then you need to figure out the droop due to ESR at maximum current peaks, the droop due to the capacitance discharging over a mains half-cycle (if a linear power supply), and figure out if you need more capacitance and/or less ESR.
And decide on the temperature/lifetime spec.
And then check the ripple current rating is adequate.
Then pick a suitable capacitor to match these specifications (though one might not exist and you'll have to relax your requirements or consider parallel array).
Just like any other component, figure out the desired performance, then do a parametric search (and avoid dodgy won-hung-lo brands of course!).
Yes before trying the nichicons i used some computor rated capps with the screw type connection. My impression was they tend to sound shut in and rather steally in the midrange and treble. The bass to my ears sounds disjointed.
I used Nichicon KG series in the last 10 years. Now changed to BC Components 056 series, and feel it much better.
There is some reports about low ESR capacitors on forum. It sounded worst, caused ringing on rails etc. and saying that let stay away from them. I have used low ESR Epcos Sikorel B41505 for few years, after careful comparison testing, i changed to use Epcos general purpose B41231. The Vishay BC 056 is also planned but their cost high, then i change to Nichicon LKG. However, the LKGs sound quite bad on midrange, not neutral, not open. On this side, the Kemet ALC10S sound much better.
I have some of the b41505 in my DAC, they are great parts. Also use some of the 22,000uf Epcos B41550 in my solid state amp, along with 1,500uf b41505, closer to the circuit.
The Mundorf AG are nice also.
It seems to matter less the farther awake from the amplifier the parts are. The audio labeled parts tend to have lower noise that isn’t as noticeable in less sensitive applications.
The Mundorf AG are nice also.
It seems to matter less the farther awake from the amplifier the parts are. The audio labeled parts tend to have lower noise that isn’t as noticeable in less sensitive applications.
Hi Michael C
Interesting.. as these have been highly rated in other threads in the forum.
Would you consider sharing what make and size they were?
Hi Walkalone
Also interesting.
Is there any indication to what is considered low ESR and what is considered acceptable ESR?