rdf said:
rdf said:
Guilty as charged. I'll take that as a complement.Naturally as such I didn't think about large currents but it should work well for preamps, low power amps, etc. If the target is lowering HF PS impedance for reasons of stability no reason a small 'RF' bypass won't work with a large main cap jumps out. I wouldn't build a glass circuit without using this technique, to my ....perception it always brings clear benefits.
This approach works fine for more current hungry applications as welol. The only detail I can add is that when crossing the 500uF mark, bypass ratios up to 1:200 is just fine.
I have recently made a few experiments with a 100.000uF > 680uF > 50uF (film from this point and down) > 10uF >1uF > 0.22uF.
This layout showed no ill effects. I started out with a 10.000uF cap in between the 100.000uF and the 680uF, removing it made no difference.
Magura
If you play around with the technique in Spice it quickly becomes obvious significantly larger values for the inductors work better. The example numbers were min values, Andrew's right they're in the range of isolated (non-groundplane) traces, as you'ld expect since the inductance of modern film caps is in the range of their physical length.
Though I'm a big LTSpice user, I'm very wary of what it tells me regarding caps when using a frequency scan. As far as I can tell, it models losses with a simple fixed resistor, and that's only good at a single frequency. Though I've had a couple suggestions on how to deal with this, I haven't really grasped how to do it. BTW, there was a circuit presented for comparing caps in audio equipment, I think by Jung. IMO, bridge type circuits have the same problem- when you null the DF between two caps it's with a fixed resistor, good at a single frequency, but insufficient for a listening test over the audio range, depending on the characteristics of the caps in question. So we have two levels of insufficiency, simulation and actual comparison! I can do a series of measurements that should be illuminating, but tying it all together is another story.
I have been reading through this and other threads about power supplies, caps, wire, etc.
The way I see it is that an amplifier is really nothing more than a modulated power supply. This "power supply" is supposed to be Impedance-matched to the load (speakers) the best as possibly can be within a passband of frequencies. All the impedances between the stages of the amplifier (modulated power supply) all must follow this same idea.
With all of that in mind, it does not matter much what type of parts are used if everything is impedance matched within a particular passband.
It is a balancing act.
The hard part is gathering data to figure out all those impedances.
The way I see it is that an amplifier is really nothing more than a modulated power supply. This "power supply" is supposed to be Impedance-matched to the load (speakers) the best as possibly can be within a passband of frequencies. All the impedances between the stages of the amplifier (modulated power supply) all must follow this same idea.
With all of that in mind, it does not matter much what type of parts are used if everything is impedance matched within a particular passband.
It is a balancing act.
The hard part is gathering data to figure out all those impedances.
Impedance matching is important for RF, where you have reflections to deal with, and can create matches with baluns and ununs. For audio, it's neither desirable or a useful concept. What we generally want is a "stiff" supply, impervious to sudden transient loads. To me, that means an energy storage capacitor that holds its properties over a wide range is more desirable than one which doesn't.
Eva- just to clarify your beliefs, you don't think large electrolytics should generally be bypassed at the cap (I agree to a large extent, but still want to create a perfect cap), but surely you believe local bypassing, say for a high bandwidth op-amp, using small low ESR caps is desirable? Or is a big electrolytic enough?
FWIW, I recently changed the supply and bypassing for my phono preamp to consist of 47uF OS-CONs, but no smaller value bypasses. I'm not very happy with it now, and intend to try a variety of other schemes. Given the PSRR and all, I don't even know why it matters, but it seems to.
Eva- just to clarify your beliefs, you don't think large electrolytics should generally be bypassed at the cap (I agree to a large extent, but still want to create a perfect cap), but surely you believe local bypassing, say for a high bandwidth op-amp, using small low ESR caps is desirable? Or is a big electrolytic enough?
FWIW, I recently changed the supply and bypassing for my phono preamp to consist of 47uF OS-CONs, but no smaller value bypasses. I'm not very happy with it now, and intend to try a variety of other schemes. Given the PSRR and all, I don't even know why it matters, but it seems to.
Hi Conrad, no doubt about LTSpice's limitations but unless I'm way off it really doesn't matter. In general the aggregate topology doesn't appear to be that sensitive to wide variances in the cap's self-inductance if the preceeding inductor is large enough. LT should do a good enough job modelling small air core inductors.
Can't do the tests I want to do right now, but here's something I was able to throw together with what's on the bench right now. This may not be important, but it's another way of looking at things that might be useful. Let's say at high frequencies a capacitor's phase shift is closer to zero than -90 degrees. That means it's closer to a resistor, albeit a low value one. Last time I checked, a resistor isn't a very good energy storage device. So let's say I want two things. I want to use enough capacitance so the reactance (X) stays under some arbitrary value, like 1 ohm. Let's say I also dictate that the phase shift (loss tangent) shall stay between -80 and -90 degrees. It turns out that this is pretty easy below about 3khz. It's also pretty easy above 10 or 20khz, where you can use reasonably obtainable film caps. The middle region is surprisingly difficult. Take a look at the following Excel spreadsheet. Check my math, of course, as I've been known to make pretty big mistakes! Most important, tell me why the phase shift, loss tangent, DF, or however you want to describe it, doesn't matter. Also, I'm now thinking the poly motor capacitors might be a great answer, but the one I have is too small for the frequency range where I need it.
Cap spreadsheet
Cap spreadsheet
Doesnt matter if you dont have considerable voltage swing accross your caps. And we dont want that if its about power supply caps.Conrad Hoffman said:
Lets take another look at large 1000uF 50v electrolytic capacitor not able give and take energy at 50khz swithing frequency. Lets say that supply voltage is 5.0025v and 0.025ohms ESR. we charge our DUT with 500mA for 10us and discharce 500mA for 10us.
1000uF cap voltage rises 5mV during charge perioid and ESR drop is 12.5mV. Energy input to cap: 0.5*0.001*5.005^2 0.5*0.001*5.0v^2 =12.525mJ-12.5mJ =25uJ
Energy loss in ESR: 0.5A*0.025*10us=62.5nJ
So our elko stores 99.7% of the energy.
mzzj- As usual, looking at things from an energy standpoint clears them up quickly! I'm probably better off ditching the AC model and just looking at the thing as a battery and a small resistor- you're right, there shouldn't be a significant AC voltage across the cap anyway. Still, any power amp I've ever looked at (with unregulated supplies) does have easily visible "music" signal on the supply rails. It would be interesting to see if a spectral analysis of this "music" showed LF content due to insufficient capacitance, and MF content where it's hardest to get low DF, that 3-10khz range.
Conrad Hoffman said:
I'm afraid it's not that simple. You can't just account for the DC figures as with a battery and make it all add up. As you stated youself, there is AC in the shape of the music on the supply rails, and hence we are forced to look at the AC portion of the problem.
I personally tend to lean towards your "big film cap" solution, as from what I can measure a 50uF PP film cap, as such seems to keep the phase shift down to a worst case scenario of 80 degrees at 7kHz.
Magura
Conrad Hoffman said:
I am affraid there is nothing "intresting" to see in that spectral content.We(at least me) dont want low DF for power supply, its nothing but trouble because it creates ugly LC-resonances with every wire inductace. Low DF is great if you want to build oscillator...
What I want is stiff voltage source, NOT a building part for LC-resonator.
Imagine you have perfect, indutanceless 1000F capacitor with superb low DF. That would be real nice for power supply right? Now add 0.01milli-ohm resistance in series and DF goes trough the roof and it behaves purely resistive trough entire audio band. In your theory its not a good supply because its DF is high, for me its near-ideal voltage source with 0.01mohm impedance in entire audio band.
Forget DF or start building induction heater LC-resonators
Conrad Hoffman said:
does the aligent work or operate under complex harmonic draw/ loading?Or is the test signal capacity basic in nature? This is key in getting further down the road in understanding what is going on...
After all the dynamic draw/charge/direction/etc pattern of a DC to AC situation that a capacitor faces in either a DC rail support situation or a AC coupling situation is far more complex than a basic test signal.
for example: Strip the can off of a 20kuf 75V capacitor. Slide a high impedance lead between various layers, in order. Measure those points under dynamic loading. Compare the loading characteristics of the various points, in comparison to the complex loading (first do simple signals) results. Also compare to the original signal, both complex and simple signals.
I'll bet you get completely different results, or some data which will lead to a slightly different understanding than you have now. If you want to take it further, I'd say that's the next step.
The polystyrol caps from russia are likely an attempt to deal with these issues. If you do a search under my handle, and on 'polystyrene', you'll find the internal data on their construction.
My thinking behind putting small air core inductors between caps is precisely to control those resonances and assure the bypass doesn't see the preceeding cap's parasitic inductance. Lead and trace inductances/resistances are inevitable, isn't better to control their effect by swamping them with known values and designing appropriately?
It is always interesting to remember that a large value inductor that is connected open ended (one end connected to ...nothing...) is actually a high speed capacitor that will dump it's stored energy at it's propogation speed limits..into a collapsing dc rail. Below that speed (length of the wire is also a component that measurement) it becomes ineffective, obviously. It can create more problems than it solves, but it is a interesting phenomenon that must be tackled if one really wishes to understand what capacitors are doing.
At audio and even low bypass frequencies, I'd hope I can treat everything as lumped parameters. If not, I probably need a new hobby. Unless the foils or metallization of a cap have dreadfully bad conductivity, probing any internal location had darn well better display the same signal. My understanding is that the Agilent LCR meters use the conventional method of sine waves and phase detection, not music signals, to measure caps, as was originally patented by General Radio Corp back when Henry Hall developed the Digibridge. But, I digress. IMO, this is a good dialog. I know all the conventional reasons why none of this should matter and everything sounds the same (I subscribe to a lot of them myself), but too many people seem to hear differences in supplies, and I'm leaning that way myself, though I hesitate to admit it in public. Oops. I have an extra hole in the front of my preamp, and I'm contemplating installing a switch so I can actually change filter components or even the entire supply, while listening. I suppose I could have a friend solder the final wires so I wouldn't know which position was connected to what.
While somewhat tangential to the original thread, I have been thinking about building a simple remote switching unit to A/B test input caps as well as cap filter banks. What are the implications of switching caps "on the fly"? I have doubts about my ability to discern minor changes with several intervening minutes
7/10
That the human hear only hears the leading edge of the given audio signal and largely ingores the rest (90% ignored). 100% of our hearing is associated with decoding that 10%. So normal methods of measurement involving the whole signal as a 'distortion measurement' or the like, are factually speaking...grossly ignorant and wholly inaccurate. Base all measurments of distortions or otherwise on those leading edges ONLY and you will finally begin to get numbers which agree with what the human ear hears. This places the human ear squarely in the 10,000th to millionth particle level of capacity for noise, harmoinoc and temporal undrestandings of the distortion aspects of a given 'standard' measurement methodology. Get it?
I bought a 'brown's gas' generator a while back and have been delving deeply into understandings of EKD (totally new words Here:"ElectroKinetiDynamics") and MHD. That browns gas has contributed to a better understanding of capacitors and materials, well, that's something that I'd have to explain, so I will. Hopefully some of you might get it.
Gravitation: the MHD version of this EKD:
http://amasci.com/weird/unusual/e-wall.html
Read the above bit, then consider:
That the gas is the result of an electrical expansion of the molecular considerations. This also occurs in the minute sense within the charge/discharge and all dynamic considerations of the gap between the foils and films, etc. This also is very noticable within the aspects of the sound of a given paper and oil capacitor, for the very same reasons. Think of the separate atomic structures not just charging but their little vortexes polarizing and unpolarizing, or twisting and then untwisting.
Depending on the method by which the given capacitor is built (dry or wet) the sound will change, due to these hydrodynamic considerations of either the film/foil..or the foil/film/electrolytic or oil. And the conditions between one fraction of a second and the next, is that the micro-harmonic loading-unloading characteristics of the capacitor will be different under any given different load delta. Add to that, that the human ear only hears the changes that I speak of and ignores the rest of the cap's loading characteristics and standard measurements...then the numbers suddenly DO correlate with what is heard. If those measurments are weighted in the way the human hear hears it.
To see it otherwise is to tell me that a given loudspeaker's driver response characteristics wil NOT be different..depending on how far the voice coil is in or out of the gap-and what the pressue laoding is and what the mehcanical loading is. We know that such is not true. Very large subsets of the driver's distortions originate from each of these considerations. And how does the given driver respond to a given signal input when it is in one of these given dynamic conditions? It is repeatable - but the distortion considerations are wholly dynamic and changing.
Same for the capacitor.
LCR considerations are merely gross numbers to throw at a wall to see if they stick, a basic measurment, if you will... they are not cause and effect. They are each (L, C, And R) merely lumped parameters that have minimal bearing or relation to the origin of the numbers or effects they allude to representing.
On the microdynamic-hydrodynamic frequential quanta and vortex level, each parameter becomes considerably more complex. One can even say that inductance also comprises capacitance as one of it's aspects, or visa-versa. (capacitance also partially containing the nature of inductance within the truth of what it actually is, for example)
I bought a 'brown's gas' generator a while back and have been delving deeply into understandings of EKD (totally new words Here:"ElectroKinetiDynamics") and MHD. That browns gas has contributed to a better understanding of capacitors and materials, well, that's something that I'd have to explain, so I will. Hopefully some of you might get it.
Gravitation: the MHD version of this EKD:
http://amasci.com/weird/unusual/e-wall.html
Read the above bit, then consider:
That the gas is the result of an electrical expansion of the molecular considerations. This also occurs in the minute sense within the charge/discharge and all dynamic considerations of the gap between the foils and films, etc. This also is very noticable within the aspects of the sound of a given paper and oil capacitor, for the very same reasons. Think of the separate atomic structures not just charging but their little vortexes polarizing and unpolarizing, or twisting and then untwisting.
Depending on the method by which the given capacitor is built (dry or wet) the sound will change, due to these hydrodynamic considerations of either the film/foil..or the foil/film/electrolytic or oil. And the conditions between one fraction of a second and the next, is that the micro-harmonic loading-unloading characteristics of the capacitor will be different under any given different load delta. Add to that, that the human ear only hears the changes that I speak of and ignores the rest of the cap's loading characteristics and standard measurements...then the numbers suddenly DO correlate with what is heard. If those measurments are weighted in the way the human hear hears it.
To see it otherwise is to tell me that a given loudspeaker's driver response characteristics wil NOT be different..depending on how far the voice coil is in or out of the gap-and what the pressue laoding is and what the mehcanical loading is. We know that such is not true. Very large subsets of the driver's distortions originate from each of these considerations. And how does the given driver respond to a given signal input when it is in one of these given dynamic conditions? It is repeatable - but the distortion considerations are wholly dynamic and changing.
Same for the capacitor.
LCR considerations are merely gross numbers to throw at a wall to see if they stick, a basic measurment, if you will... they are not cause and effect. They are each (L, C, And R) merely lumped parameters that have minimal bearing or relation to the origin of the numbers or effects they allude to representing.
On the microdynamic-hydrodynamic frequential quanta and vortex level, each parameter becomes considerably more complex. One can even say that inductance also comprises capacitance as one of it's aspects, or visa-versa. (capacitance also partially containing the nature of inductance within the truth of what it actually is, for example)
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.