Having just looked at my DC supply in my LM3886T gainclone via a scope, I noticed it is not flat. In fact, its very wavy, not even close to a flat line. The supply is + and - 35 V, and probably drops to 31 V or so every cycle, even when not under a tough load. When the load is tough it drops even more. And I'm using 8 discrete MUR diodes and low ESR panasonic caps. 1000 uf per rail, separate rails for each channel, so 4 caps total for the stereo amp. Transformer is 500 VA used for both channels.
I have read many discusions here about desired capacitance levels for GC, in terms of sound, but have not seen much in terms of results and measurements. For example, I thought 1000 uf was considered the ideal by many (not all of course, some like more, some like it distributed, some like less....), but the measurement I took was a surprise, I thought it should be flat under a low load. Anyone have pics of the DC waveform that they get? (I can't post any for a few days...but can if you want me to after that).
Just so we are clear, those of you who use 1000 uf or less per rail and who actually have measured your DC supply don't actually get a flat DC supply, do you? Is this something to be worried about?
I have read many discusions here about desired capacitance levels for GC, in terms of sound, but have not seen much in terms of results and measurements. For example, I thought 1000 uf was considered the ideal by many (not all of course, some like more, some like it distributed, some like less....), but the measurement I took was a surprise, I thought it should be flat under a low load. Anyone have pics of the DC waveform that they get? (I can't post any for a few days...but can if you want me to after that).
Just so we are clear, those of you who use 1000 uf or less per rail and who actually have measured your DC supply don't actually get a flat DC supply, do you? Is this something to be worried about?
I've said it before and I'll say it again:
1000 uF is not enough capacitance for a power amp
Somebody saw in the datasheet that each rail needs to have a 1000 uF cap on it, close to the amp. They thought that this capactiance is all that is necessary, but in fact it is intended to be over and above the power supply capacitance. That can be determined using solid engineering theory and methods. The people suggesting that 1000 uF total per rail obviously do not understand those basic electrical theories. The fact is, that for most of each AC half-cycle, the only thing powering the amp is the energy stored capacitors, because the AC voltage has fallen below what is necessary to power the amp. With too little energy storage, bad things happen.
I would suggest 5000 to 10000 uF per rail per 100W of output power.
1000 uF is not enough capacitance for a power amp
Somebody saw in the datasheet that each rail needs to have a 1000 uF cap on it, close to the amp. They thought that this capactiance is all that is necessary, but in fact it is intended to be over and above the power supply capacitance. That can be determined using solid engineering theory and methods. The people suggesting that 1000 uF total per rail obviously do not understand those basic electrical theories. The fact is, that for most of each AC half-cycle, the only thing powering the amp is the energy stored capacitors, because the AC voltage has fallen below what is necessary to power the amp. With too little energy storage, bad things happen.
I would suggest 5000 to 10000 uF per rail per 100W of output power.
Konnichiwa,
Maybe, but in case of Gaincard/clone it is, to listen to music, no matter what the 'scope says. So what?
Making it 1250uF per rail for a 25W Gainclone (original is rated at 25W), NPV is 1,000uF. So are you actually recommending using a 1,000uF Capaictor per rail? You are contradicting your first statement, so it seems...
Sayonara
macboy said:
Maybe, but in case of Gaincard/clone it is, to listen to music, no matter what the 'scope says. So what?
macboy said:
Making it 1250uF per rail for a 25W Gainclone (original is rated at 25W), NPV is 1,000uF. So are you actually recommending using a 1,000uF Capaictor per rail? You are contradicting your first statement, so it seems...
Sayonara
Hmm, that seems like a good point. If a 25W gainclone uses a 1000uf per rail, shouldn't a 50W clone use 2000uf?😕
1000uF is pretty low for supply voltage ripple reduction. W/o going into how it sounds and etc. You are getting about 5.7VRMS of ripple from your supply. This is about what you would expect from only 1000uF of filtering. By comparison, my +24V power supply with 6300uF of filtering has around 2.1VRMS of ripple. If you want to flatten out your power source, you will need more capacitance. Look up bridge rectifier equations in your analog electronics textbooks for the ripple calculations.
I think you'll be better off with atleast 4700uF and probably 10000uF of filtering.
--
Danny
I think you'll be better off with atleast 4700uF and probably 10000uF of filtering.
--
Danny
I'm puzzled by this: some people run gainclones from batteries.
Do battery powered gainclones sound inferior to those with >5V psu ripple?
Do battery powered gainclones sound inferior to those with >5V psu ripple?
I have also found this point interesting. I seem to recall that a very well respected poster to this thread has said that he has not even tried more than 1000 uF per rail, but later stated that a battery supply sounds better! My somewhat limited electronics knowledge suggests that a well filtered power supply will approximate a pure DC (battery) source better than 1000 uF per rail, but what would I know 🙂
lm1875 - 20watts - reccomended near-chip capacitor per rail - 100uF bypassed with 100nF ceramic.
lm3886 - 60 watts - reccomended near-chip capacitor per rail - 470uF bypassed with 100nF ceramic
lm4780 - 120 watt - reccomended near chip capacitors 1000uF bypassed with 10uF bypassed with 100nF ceramic.
these were intended to be close to the IC. ceramic was intended for the better RF performance.
the chips were made with a high PSRR in mind, so supply ripple won't affect musical use by as much.
lm3886 - 60 watts - reccomended near-chip capacitor per rail - 470uF bypassed with 100nF ceramic
lm4780 - 120 watt - reccomended near chip capacitors 1000uF bypassed with 10uF bypassed with 100nF ceramic.
these were intended to be close to the IC. ceramic was intended for the better RF performance.
the chips were made with a high PSRR in mind, so supply ripple won't affect musical use by as much.
there was a perceived need to balance the advantages of a stiff, low-ripple supply against the ill-effects of huge current spikes caused by large power supply capacitances.
since the LM3875 has very high built-in PSRR (power supply ripple rejection), it was felt this should be exploited to sacrifice the former to favor (minimise) the latter.
to have your cake and eat it, i.e. to have both low-ripple supply and low suppy line spikes, there are a few ways, namely
a. battery
b. choke filtering (this is suitable for the large, constant current draw of class A, so less applicable here)
c. active regulation or capacitance multiplier
note all the above three methods then introduce the problem of high power supply output impedance (which can be mitigated somewhat by having large/low-ESR capacitances after them).
since the LM3875 has very high built-in PSRR (power supply ripple rejection), it was felt this should be exploited to sacrifice the former to favor (minimise) the latter.
to have your cake and eat it, i.e. to have both low-ripple supply and low suppy line spikes, there are a few ways, namely
a. battery
b. choke filtering (this is suitable for the large, constant current draw of class A, so less applicable here)
c. active regulation or capacitance multiplier
note all the above three methods then introduce the problem of high power supply output impedance (which can be mitigated somewhat by having large/low-ESR capacitances after them).
Don't flame me this is just MHO and observation.
I am leaning more and more to this way of thinking with GC. Yes the 'scope shows lower ripple with 10,000uF / rail but to my ears it also sounds better. With lower capacitance (1000uF / rail) I hear a gritty sound presentation, over bright but with clear mids. Bass is thinner and overall the sound is less cohesive, I find it fatiguing to listen to for any length of time.
With greater capacitance the bass fills out and I detect no loss of detail or clarity, the grittiness seems absent and the overall presentation is smoother, dare I say it, more valve like, more natural. I have no problem listening for several hours.
Richard C said:
car batteries can crank out a good 20-30 amps at start, and will dip into single digital voltage territory. so that would mean (14-10)/30=0.1ohm (?) of effective internal resistance.
not sure if that's high or low. but it is common to get a 4-5% regulation on a typical transformer. in my case (a 28v 6amp transformer, that means 28*4%/6=0.2ohm from the transformer. additional filter caps will help further reduce that but we probably need a simulation to figure that out.
the point? a battery is likely to have comparable, if not worse, internal resistance to an unregulated power supply.
I've posted one of these days on one thread here (I don't remember what thread...) that, according to my tests, the best TOTAL capacitance for an LM3875/LM3886 GC is around 2000uf.
I've just received a private mail from a member that says:
"You are right about PSU cap for the Gainclone. After reading your post, I change my Pana 1000 FC with 2200 FC and I get better bass, more air in mid and less grainy highs."
😀
I've just received a private mail from a member that says:
"You are right about PSU cap for the Gainclone. After reading your post, I change my Pana 1000 FC with 2200 FC and I get better bass, more air in mid and less grainy highs."
😀
I checked the datasheet for the LM3875 last night. It does have 85dB - 125dB of PSRR. They tested it with 14.1VRMS of ripple 1 rail at a time. I suppose why not take advantage of it.
None-the-less, heavy bass is certainly going to sound better with more PS caps. This is because during heavy current swings (high volumns, low frequencies) the smaller capacitors are going to expend their stored energy faster than a larger cap will. Look at car audio, lots of people put 1F caps with their subwoofers because their bass sounds dead and their lights dim when the bass hits. Same effect.
w/rt to batteries... ideal DC sources have 0 resistance... a real battery will have a pretty low ESR. Certainly not infinite.
None-the-less, heavy bass is certainly going to sound better with more PS caps. This is because during heavy current swings (high volumns, low frequencies) the smaller capacitors are going to expend their stored energy faster than a larger cap will. Look at car audio, lots of people put 1F caps with their subwoofers because their bass sounds dead and their lights dim when the bass hits. Same effect.
w/rt to batteries... ideal DC sources have 0 resistance... a real battery will have a pretty low ESR. Certainly not infinite.
yeah, but in car audio there isn't a diode to prevent the cap from back-charging the battery...
as for ideal battery resistance testing, wouldn't you just take (12.9-2.7)/Icca ? isn't CCA rated as current draw that reduces battery voltage to 2.7V?
and i'd like to note that additional resitance before the caps will increase ripple rejection.
notice for a modest 50 watts and 1000uF you get 10 volts of ripple. this of couse is just an approximation...
as for ideal battery resistance testing, wouldn't you just take (12.9-2.7)/Icca ? isn't CCA rated as current draw that reduces battery voltage to 2.7V?
and i'd like to note that additional resitance before the caps will increase ripple rejection.
notice for a modest 50 watts and 1000uF you get 10 volts of ripple. this of couse is just an approximation...
- Status
- Not open for further replies.