Larger caps are needed because you are trying to deliver power to a low impedance load. 100-1000uF is not sufficient, either, unless you want to be one of the mindless horde that strictly adheres to the "original" gainclown "design" and then complains about anemic bass.
Try 10,00-20,000 uF on each rail for each amplifier. Many will disagree, but many do not understand the function of the caps and the meaning of a PSRR spec. Use 100-1000uF to bypass the larger caps, and use 1uF-10uF film caps to bypass those.
I know, that's not how Sakura Labs (or whoever originated this nonsense) makes them, but judging by their assembly skills from what I have seen in photos of their stuff, they aren't exactly experts.
If you want to make an amp that works reliably with whatever load you connect to it, I suggest read and follow the info in the data sheet for the chip. I know it's hard to believe, but the guys who designed the parts actually know what they are doing.
I_F
Try 10,00-20,000 uF on each rail for each amplifier. Many will disagree, but many do not understand the function of the caps and the meaning of a PSRR spec. Use 100-1000uF to bypass the larger caps, and use 1uF-10uF film caps to bypass those.
I know, that's not how Sakura Labs (or whoever originated this nonsense) makes them, but judging by their assembly skills from what I have seen in photos of their stuff, they aren't exactly experts.
If you want to make an amp that works reliably with whatever load you connect to it, I suggest read and follow the info in the data sheet for the chip. I know it's hard to believe, but the guys who designed the parts actually know what they are doing.
I_F
Moreover, with gainclones, there is not a lot of smoothing on the rails before it hits the large "decoupling" caps. In all actuality, these are the primary smoothing caps for the gainclones. If you were to insert another small 0.1uf cap to the chip's power pins, these would truly be decoupling caps.
As it stands, the incoming power to the chipamp.com and audiosector.com kits is rectified, but with no cap bank to get rid of the "humps" in the rectified power. This is why they use large caps to get rid of the ripple right before it hits the chip.
[disclaimer] At least this is my understanding of it. I haven't ever built one of the chips, but looking at the schematics, this is how it should work. [/disclaimer]
As it stands, the incoming power to the chipamp.com and audiosector.com kits is rectified, but with no cap bank to get rid of the "humps" in the rectified power. This is why they use large caps to get rid of the ripple right before it hits the chip.
[disclaimer] At least this is my understanding of it. I haven't ever built one of the chips, but looking at the schematics, this is how it should work. [/disclaimer]
0.1uF decoupling is only suitable when the chip and the big supply capacitors are in the same PCB and quite close together, which is the usual situation in consumer products.
However, never ever use 0.1uF decoupling capacitors alone in DIY projects. That's because in DIY the supply rails usually come to the amplifier boards through long and poorly laid out wires, whose inudctance will resonate like crazy together with the 0.1uF capacitors. In these circumstances, 10uF should be considered as a minium, being higher values like 1000uF advantageous because they reduce the amount of "chopped" class B "trebble" and "midrange" currents flowing through the supply wires.
However, never ever use 0.1uF decoupling capacitors alone in DIY projects. That's because in DIY the supply rails usually come to the amplifier boards through long and poorly laid out wires, whose inudctance will resonate like crazy together with the 0.1uF capacitors. In these circumstances, 10uF should be considered as a minium, being higher values like 1000uF advantageous because they reduce the amount of "chopped" class B "trebble" and "midrange" currents flowing through the supply wires.
I_Forgot said:
YOu posted while I was typing! I'll agree about liking more capatance on the rails, but I have found 4400uf per rail per chip was my sweet spot (just because I had a bunch of 2200 uf caps around!). This puts me close to the 10,000uf (why don't we ever change units on capatance????) IF recommends. Of course more won't hurt at all, at least in my experience. It definitely fills out the bass response better!
Keep in mind ALWAYS, that when you see a schematic, an example circuit in a datasheet, that example is NOT the entire "thing" you are building.
If you see (for example), "+35V" somewhere and only a 0.1uF cap in the subcircuit, that does NOT mean you only need a 0.1uF cap in the entire "thing" you are building, it means you need ALL of the parts necessary to have a relatively clean power supply of 35V getting TO that subcircuit the datasheet pictures.
Some parts are more tolerant of noisey power than others, have good PSRR spec for example, but until you know for certain this is the case, you have to assume you are supplying the rest of the design around the subcircuit provided.
This extends to most circuits, certainly not just Gainclones.
If you see (for example), "+35V" somewhere and only a 0.1uF cap in the subcircuit, that does NOT mean you only need a 0.1uF cap in the entire "thing" you are building, it means you need ALL of the parts necessary to have a relatively clean power supply of 35V getting TO that subcircuit the datasheet pictures.
Some parts are more tolerant of noisey power than others, have good PSRR spec for example, but until you know for certain this is the case, you have to assume you are supplying the rest of the design around the subcircuit provided.
This extends to most circuits, certainly not just Gainclones.
Eva said:
Even on commercial designs such as by Naim, and Cyrus audio with external PSU's it was a head scratcher!
So It's better to think of these designs as a capacitor with 100cm leads...
Is there any difference between decoupling caps, and caps to provide current to the load ?
An externally hosted image should be here but it was not working when we last tested it.
http://www.6moons.com/audioreviews/47labs5/gaincard.html
Hi all,
The whole point of decoupling caps is to negate (as much as possible) the effects of wire/track inductance. Think of it this way: the wires carrying the current to the load have some inductance (and resistance, but you can't do much about that). Due to the AC components of the current drawn by the load, there will be reactance in the wires from the inductance, which will cause the voltage rail to sag. However, by using decoupling capacitors, the higher frequency AC components in the signal will be shunted through this capacitor (instead of travelling through the wires), just leaving low frequency AC and DC that is travelling through the inductive wires. Hence there will be less loss on the wires.
Note that at audio frequencies decoupling doesn't have a massive effect (although side effects like reduced chip oscillation etc mandate decoupling), high speed digital switching is where it really matters. The value and type of capacitor is also important - you should check datasheets for the resonant frequency of the capacitor - this is where it decouples most effectively. However, when people are saying that you should use 10,000uF caps - this is not decoupling. This provides smoothing of the rectified AC, quite distinct from decoupling which aims to reduce AC components of the load current such that wire inductance doesn't have a significant effect.
Chris.
The whole point of decoupling caps is to negate (as much as possible) the effects of wire/track inductance. Think of it this way: the wires carrying the current to the load have some inductance (and resistance, but you can't do much about that). Due to the AC components of the current drawn by the load, there will be reactance in the wires from the inductance, which will cause the voltage rail to sag. However, by using decoupling capacitors, the higher frequency AC components in the signal will be shunted through this capacitor (instead of travelling through the wires), just leaving low frequency AC and DC that is travelling through the inductive wires. Hence there will be less loss on the wires.
Note that at audio frequencies decoupling doesn't have a massive effect (although side effects like reduced chip oscillation etc mandate decoupling), high speed digital switching is where it really matters. The value and type of capacitor is also important - you should check datasheets for the resonant frequency of the capacitor - this is where it decouples most effectively. However, when people are saying that you should use 10,000uF caps - this is not decoupling. This provides smoothing of the rectified AC, quite distinct from decoupling which aims to reduce AC components of the load current such that wire inductance doesn't have a significant effect.
Chris.
Hi,
I have been looking at the PSU schematic in the link below.
http://www.eetchina.com/ARTICLES/2003OCT/PDF/2003OCT03_POW_AN.PDF
Small cap's after a bridge recifier with a load resistor, and then capacitance.
I'm not sure about using the cap's like that..
Has anyone feed ac directly into a LM38XX chip power pins ?
I have been looking at the PSU schematic in the link below.
http://www.eetchina.com/ARTICLES/2003OCT/PDF/2003OCT03_POW_AN.PDF
Small cap's after a bridge recifier with a load resistor, and then capacitance.
I'm not sure about using the cap's like that..
Has anyone feed ac directly into a LM38XX chip power pins ?
one note worthy mention is, the bigger the bank, the less it is noticed when it gets doubled.
so going from 10,000 from 5,000 uF does indeed sound different, but going for 100,000 uf from 50,000 uf makes little difference if any at all.
i made my own b/s rule of 1,000 uF / watt / rail for dual supply, and 2,000 uF for single supply application after the rectifier.
the rest is a nother question.
and allso worth mentioning that if you have say.. 20 volt rails, then don't use 25V rated caps. go higher. the physical size will grow, and as far as i know you get better ESL and ESR ratings.
specially important for single supply stuff like my most loved amp, the STA540.
even if a stabilised psu is used, a beefy bank of caps is never going to hurt. ever.
some say smaller caps takes less to recharge and then you get tighter and faster response. well i say the current drawn in case of a big and a small cap would be identical. the large cap has less ESR and ESL, the rail sag will be less, and therefore, it can be recharged faster.
so you get faster response from big caps.
(notice, both is ********.)
the big cap bank is there to make sure that a high current pulse does not collapse your rails. and thats it.
what we want from it is high capacity, ability to discharge a huge lump of current in a short time and after wards soak up a lot of current swiftly.
since there are no pure caps out there, just reactances with inductive and capacitiv and resistive parts inside, a cap is a mix of the 3 where capacitance is the major one. we want something, that has big capacitance, low inductance and low resistance.
this is all what you are looking for.
electrolitic caps are mostly okay. they are big, and affordable.
even the exact values have quite a lot of spread, but its not that interesting to be bothered with at all.
some can even spread out to +-30 %, but the ratio of ESR to capacity and ESL to capacity is far more interesting.
as ESL and ESR define the maximum current that can charge the cap ( if the xformer can supply the needed current that is) and allso limit the maximum discharge current.
allso worth mentioning that if the rails do collapse, but do not hit near the output voltage +few volts, then you won't notice a thing.
say.. you have 20 volt rails, and they collapse to 15 volts, and your amplifier does not need to supply more than 10 volts on its output then no amount of caps will change anything.
however, filtering is a nother thing those big caps do, but not too well.
that is why small caps near the ic (preferably as close to the ic package as possible, and really good quality ones) are required.
its worth more to inest into a better small cap, than to have a far too big bank of big reservoir caps.
and the best is to have properly high sensitivity speakers, so you don't have to have a lot of voltage gain, and don't have to pump a lot of current into the speakers. A well behaving crossover is one that- when the drivers are connected to it, and they are in the desired enclosure- elliminates as mutch as reasonably possible from the impedance peaks and dips.
its less painfull to drive a speaker that is not too reactive. music is dynamic alone , don't need a hard to drive load there too.
so going from 10,000 from 5,000 uF does indeed sound different, but going for 100,000 uf from 50,000 uf makes little difference if any at all.
i made my own b/s rule of 1,000 uF / watt / rail for dual supply, and 2,000 uF for single supply application after the rectifier.
the rest is a nother question.
and allso worth mentioning that if you have say.. 20 volt rails, then don't use 25V rated caps. go higher. the physical size will grow, and as far as i know you get better ESL and ESR ratings.
specially important for single supply stuff like my most loved amp, the STA540.
even if a stabilised psu is used, a beefy bank of caps is never going to hurt. ever.
some say smaller caps takes less to recharge and then you get tighter and faster response. well i say the current drawn in case of a big and a small cap would be identical. the large cap has less ESR and ESL, the rail sag will be less, and therefore, it can be recharged faster.
so you get faster response from big caps.
(notice, both is ********.)
the big cap bank is there to make sure that a high current pulse does not collapse your rails. and thats it.
what we want from it is high capacity, ability to discharge a huge lump of current in a short time and after wards soak up a lot of current swiftly.
since there are no pure caps out there, just reactances with inductive and capacitiv and resistive parts inside, a cap is a mix of the 3 where capacitance is the major one. we want something, that has big capacitance, low inductance and low resistance.
this is all what you are looking for.
electrolitic caps are mostly okay. they are big, and affordable.
even the exact values have quite a lot of spread, but its not that interesting to be bothered with at all.
some can even spread out to +-30 %, but the ratio of ESR to capacity and ESL to capacity is far more interesting.
as ESL and ESR define the maximum current that can charge the cap ( if the xformer can supply the needed current that is) and allso limit the maximum discharge current.
allso worth mentioning that if the rails do collapse, but do not hit near the output voltage +few volts, then you won't notice a thing.
say.. you have 20 volt rails, and they collapse to 15 volts, and your amplifier does not need to supply more than 10 volts on its output then no amount of caps will change anything.
however, filtering is a nother thing those big caps do, but not too well.
that is why small caps near the ic (preferably as close to the ic package as possible, and really good quality ones) are required.
its worth more to inest into a better small cap, than to have a far too big bank of big reservoir caps.
and the best is to have properly high sensitivity speakers, so you don't have to have a lot of voltage gain, and don't have to pump a lot of current into the speakers. A well behaving crossover is one that- when the drivers are connected to it, and they are in the desired enclosure- elliminates as mutch as reasonably possible from the impedance peaks and dips.
its less painfull to drive a speaker that is not too reactive. music is dynamic alone , don't need a hard to drive load there too.
i have been using both 3875 and 3886 with from 2200uf per rail to 20 000 uf per rail ,regulated, normal...., in both cases transformers were overkill (400VA for stereo 3886 and 250VA for stereo 3875) i havent heard any difference, hell i could connect 3875 with 2200uf on one channel and 3886 with 20 000uf on other i wouldnt hear difference ,not sure how is this possible though but it sounded same to me.
given an xformer with enough VA to overcome the surges,
the caps are only needed for stability.
if you run a pair of lm3886 chips on 400 VA then you may even be lucky and run it without any reservoir caps at all.
only a small one needed close to the ic pins.
specially if you are not pushing the limits with high gain settings, and try to go near maximum poweroutput.
the caps are only needed for stability.
if you run a pair of lm3886 chips on 400 VA then you may even be lucky and run it without any reservoir caps at all.
only a small one needed close to the ic pins.
specially if you are not pushing the limits with high gain settings, and try to go near maximum poweroutput.
ok seriously ,here is for my ear best combination for 3875, it was connected to lt1083 regulator. 4400uf/rail,44V before regulation 2200uf/rail,33v after.
for 3886 is similar , just unregulated 40vdc(when mains high peak) with 20 000/rail. both have same bass response for my ears
for 3886 is similar , just unregulated 40vdc(when mains high peak) with 20 000/rail. both have same bass response for my ears
Attachments
Last edited:
Very interesting. I'm trying to determine if there's an upgrade for the 1500uf 50v Panasonic FCs in my LM3875 Gainclone. OK, so a large cap has less esr and esl than a small cap. Question, if you wire 2 - 2000uf in parallel (esr,esl then being half a single 2k), would the resultant 4k cap have less esr, esl than a single 4k cap and therefore be a better approach ? Or am I oversimplifying it.
Seeing as you can't get BGs any more, is there a successor that gives the best SQ in a gainclone? Was thinking of Elna Silmic ii, Nichicon KZ, etc anyone familiar with Kendeil caps, they look great. Also, I've heard it's better to use caps with higher voltage ratings, so, would a 4700uf 200v be significantly better than a 50v?
I've read posts alluding to super beefy PSUs for the GC allowing great improvements in SQ but can find no details, anyone have any info?
Seeing as you can't get BGs any more, is there a successor that gives the best SQ in a gainclone? Was thinking of Elna Silmic ii, Nichicon KZ, etc anyone familiar with Kendeil caps, they look great. Also, I've heard it's better to use caps with higher voltage ratings, so, would a 4700uf 200v be significantly better than a 50v?
I've read posts alluding to super beefy PSUs for the GC allowing great improvements in SQ but can find no details, anyone have any info?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.