As to ESR, ESR= tan delta/2piifC
Now please explain how could tan delta be improved by much more than a factor of roughly 2 perhaps? 10xsmaller caps against 1x big cap will giv you (neglecting interconnections) roughly half the ESR, add to this the interconnections and you are back at roughly the same ESR that 1 single cap has.
Now please explain how could tan delta be improved by much more than a factor of roughly 2 perhaps? 10xsmaller caps against 1x big cap will giv you (neglecting interconnections) roughly half the ESR, add to this the interconnections and you are back at roughly the same ESR that 1 single cap has.
Last edited:
No, it proves that he didn't have a thin-enough board on hand, at the time, as he clearly stated. If you read what he said, he calculated that it would be LESS that 0.5 nH.
The equations for paralleling already guarantee diminishing returns. You don't need to claim to be trying to make that point.
Sorry. Your paragraph above is nonsense. You need to learn more, and improve your logic, precision of language, and etiquette, before you again too-casually attempt to cast a pall over other the work of others. It is too fatiguing, for those who might feel obligated, to rebut so many ill-formed arguments at one time.
No. You explain. Hint: There's more than one capacitor. That ESR(DF) equation is only valid for one cap at a time.
Please show how 10x 1R resistors in parallel have an equivalent resistance of 0.5R, as you claim.
Andrew
Would you agree that the lowest low-frequency inductance of two very thin conducting planes and having similar but opposing currents flowing forth and being spaced at the minimum distance applied voltages allow, and being at least as wide than long cannot by made any smaller?
If you can agree to that than I suggest you also do your homework and calculate what is the theoretical lowest possible lf-inductance, it will be an eyeopener.
Would you agree that the lowest low-frequency inductance of two very thin conducting planes and having similar but opposing currents flowing forth and being spaced at the minimum distance applied voltages allow, and being at least as wide than long cannot by made any smaller?
If you can agree to that than I suggest you also do your homework and calculate what is the theoretical lowest possible lf-inductance, it will be an eyeopener.
This is not what i claim.
For example, 10xRifaPEG124, 1000uf/63V, 100HzESR 90mOhm = 9mOhm,
1xRifaPEH169,10000uF/63V, 100HzESR = 14mOhm.
Smallcap total ESR 9/14= 0,64. If we use a different big cap with ESR of 18mOhm then we get 9/18=0,5. This is what I talked about, I hope you get the picture, not much can be gained in respect to ESR especially since the above neglected interconnection resistances. Add connection R of a dubbelsided
35umCu....I think you get the picture....
As I explained allready, the ESR of a cap is made up of several resistances and the liquid makes up the mainpart of it (as you should be able to see when you look at ESR curves of wet caps). Dissipationfactor for 1 small wet cap does not differ much from the DF of a big wet cap, it is the same liquid.
Anyway, I thinck worth to mention is also the load impedance, the lower the load impence the more you can gain by paralleling. But at 8ohm speaker load
propable not worth the effort as long as you are able to have low RL connections. I do that by using 2 sandwiched 75x 0,2mm copperstripes with
dubblesided gluetape between them. Is flexible and lower R and L than you possible can get otherwise. Next best flexible connection can be made by twisting 2 or more pairs of magnetwire and connect them so currents go back-forth- back a.s.o.
gotee, i hope there is no need to be rude, please go into more detail because your outburst does not make much sense to me either
1. Please explain in detail what you think is nonsense
2. where i lack precision and have to improve my logic (always room for improvement for any of us)
3. etiquette ( i have really no idea what this could refere too)
4. many ill-formed arguments at one time (just lets go through them 1 by 1 please)
Last edited:
No, there was no need to be rude. Sorry. I was aiming at "curt" but not "rude". I was posting from work, which I very-rarely do, and was in too much of a hurry. The reason I was not being as nice as I could have been was that it was somewhat frustrating to see such casual public disagreement with what I considered to be well-established theories, especially since the statements contained what seemed to be errors of fact and logic. I do not usually see the word "proves" used so loosely, for example, in engineering contexts.
At any rate, I think I should ask you to allow me to mostly retract even "curt", for now. I believe that there might have been a bit of misunderstanding on my part, possibly mostly due to communication style and language differences.
Here is your paragraph I was referencing:
I still disagree, and also believe that most of that is just wrong (which I sometimes call "nonsense").
Sorry but the first sentence instantly struck me as nonsense (especially since my trigger-level was already lowered):
Looking at the Nichicon UHE series electrolytics, in my old Mouser catalog, for example, every voltage rating, from 6.3V to 100V has a 5mm-diameter (or 6.3mm) x 11mm cap. And the 100V ones have a 5mm diameter but the 6.3V ones only go down to 6.3mm diameter.
Maybe you meant to say that C*V had to be low, rather than C, if the diameter was 12mm.
The UHE series only has 12.5mm, not 12mm. The largest (uF) of the 12.5mm diameter models at each voltage rating are 330uF 100V, 680 uF 63V, 1000uF 50V, 1800uF 35V, 2200uF 25V, 3300uF 16V, 4700uF 10V, and 3300uF 6.3V. We would typically be using 35V or 50V caps, maybe 63V, so those values would be 1800uF, 1000uF, and 680uF, which don't seem particularly low.
Also, in every case, where there are two with the same uF and voltage rating, the one with the lower ESR is the one with the lower diameter and greater height. I think you claimed the opposite of that in one or two posts.
Also, as board area increases, inductance would decrease, not increase.
At any rate, the Nichicon UHE series I am looking at only goes up to 1200 uF, in 63V rating. 1200uF has 7.5mm lead spacing. So we could guess its inductance at 7.5L nH (where L is some factor). The 120 uF caps hav 3.5mm LS. So their inductance would probably be close to 3.5L. 10x of them in parallel would give 0.35L. That's 1/21 as much inductance. (I actually have no idea what the inductance would really come out to, in this case. I downloaded an Archambeault paper from the IEEE library, today, at work, but didn't get a chance to look at it, well-enough. It covers the calculation method for exactly this case, and compares calculations to measurements for some examples. But I would need to measure or look up the inductances of the caps, too.)
But even if it was only a 1/4 reduction for a 10/1 increase in number of caps, that would only mean that 10x was not enough caps. We would most likely be in a design situation where we had a target impedance that should not be exceeded over some frequency range. If we needed 57 caps, then that's what we would use (assuming no better way (than caps) was found).
OK. I can now understand what you were saying, about the 0.5.
But we aren't limited to any particular number of caps, like 10. So apparently we could make the total ESR as low as we wanted, if it was needed and worth doing.
I wasn't considering any particular load impedance. But these caps would be upstream from a power amplifier's power supply inputs. They would need to be able to support the power output devices' full current slew-rate capabilities. And we might want 1% maximum ripple at maximum output power. So pretty soon you could be wanting at least a few hundred kHz where you need no more than 10 mOhms of impedance, as seen by the device power pins. (Please don't try to say that the caps need to be able to respond at only up to audio frequencies.)
Paralleling also buys you the ability to do a better layout, with at least some of the multiple much-smaller caps closer to the active device for decoupling, for minimum inductance connections.
Now THAT'S interesting! These are the connections from amp to speakers, correct? 75mm wide x 0.2mm thick copper planes, with thin tape between them! Have you calculated or measured their inductance and resistance (and capacitance) per unit of length? (I bet you don't have too much problem with them acting like an antenna, either, with almost no gap between them.)
I think that there is also available some two-sided copper foil ribbon or tape. I was once looking for some one-sided, but very narrow (and in aluminum not copper), to use for (attempting) making a planar magnetic speaker (like Magnepan, sort of), and THINK I might have seen some two-sided material.
Anyway, I like extremely-low-impedance ideas like that. Thanks so much for sharing it.
AlphaCore publishes (or at least used to publish) RLC parameters for their Goertz cables and interconnects, though their inductance measurements seem high by a factor of two. It's easy to do the RLCG parameter extraction analytically with a parallel plate approximation. Broadside coupled stripline impedance calculators work too.
Writing a 2D EFIE solver for this stuff isn't that hard either; it's generally a single week's homework assignment in a 500 level electromagnetics class.
Correct. It'd be an interesting way of making twisted pair, though in my experience working with Goertz MI2 it's kind of a b*tch to keep it from shorting. If you open the cuts enough to be able to get in there to round over the edges so they don't cut through the dielectric between the conductors the stuff tends to bend and not want to go back into a planar, low inductance geometry. Speaking as someone who's done both it's can actually be faster to build and EQ a triamp dipole (and thereby pretty much avoid the need for low impedance cabling to control back waves) than do the planar cable assembly, termination, and installation for a box speaker.
Nice that we at least can talk now and can exchange ideas, because I was really disappointed and felt to be intentional attacked and misunderstood.
I felt a bit sad since I remember we had a nice exchange of thougths some time back when. May be you remember I mentined I developed a new type of powersupply wich can, under certain load conditions, offer advantages over currently known types.
No back to the cap-bank, we want them caps physically as far away from our load as necessary but electrically as close as we can get. When I said for higher voltages you need bigger caps (I really dont want 1000uF/63V caps
with 12mm diam, even if I could get them) and that this will increase boardwidth and inductance i was not very accurate, i should have said it will increase boardwidth and length but most of all it will increase the distance and inductance from the bigger caps to the load. Anyway I thougth about ways on how to calculate the inductance of the bank.
It would be easy to calculate if the load would be connected with evenly spread out current fro/to source and load. But it aint. What I came up with may be sound strange to you and I will tell should anyone show interrest.
Last edited:
If you really need big currents and super low inductances you can use hard eloxated aluminium (25um good up to 500V) and/or stack multiple striplines.
I have not measured them. If antennas are a prob you just gotta make the outside ground, plug the holes on both sides with thin coppertape a bit of solderin (If you need to solder you can use 0,13mm glassfiber/teflontape) and you got yourself a very low Z rectangular coax. Easy to calculate (if needed) you just need to know epsilon rel of the dielectricum. The lowest of the low inductances you can make with paralleled and stacked striplines made of hard eloxated aluminium (25um is good up to 500V, thicker than 25um can be made but it gets brittle and therefore becomes unreliable).
From amp to speaker i prefer as short as possible multipairs of twisted ca.0.5mm dubble isolated dynamowires (0.5 is still flexible, easy to connect, has adequate isolation, and skin is yet not to serious as long as you dont overdue it by adding to many pairs). The first time i used copperstripes with thin isolation between as connections from the rectifiers to the resevoircaps and on to the load was decades ago. Back then I used a wider triple copper (stripline) with solid 0,5mm glassfiber isolation all the way from the clamped on rectifier coolingblocks over the resevoircaps to the decoplingcaps/load. I used it in many variations mostly in induction heating applications. But principally quite similar but not as low inductance as the 3-strip line is the 2 strip/tape-thingy i told you about. It is easier to build and can be more easely bent at 90deg (to minimize the inductive coupling before and after the caps) and you can even twist it if needed. I used it in a 600W/2ohm amp mainly to get the caps properly coupled/decoupled and have a low loss/low inductance/ low stray connection to the amp. Btw, some time ago i stumbled over something similar used as speaker cables and I thougth about its usefulnes in transformers, but it had a unsuitable and far to thick PVC-isolation and was unsuitable for the task i had in mind. For the same transformer-thingy i then bougth some big Mundorf speaker crossover chokes made out of thin copper (just to get what i thougth,thin pp isolated copperfoil) but it turned out to be unsuitable for my purpose. The thin isolation was wound together with the copper and then molten together at the edges. I had to cut away the edges but then all i got was a roll of copperfoil and unusable tape. It turned out to be not the most clever design otherwise too, because you can not just roll a stripe any way you like and expect it to be a high-Q chocke. The designer clearly did not know what he was doing.
Last edited:
Just asking for more troubles i guess
But caps could be improved, esl could be reduced by 4 point connection and properly positoned stabs,
also putting connections closer together would help, and thicker wires...and..and...and
But even real RF-power caps have often much room for improvement when it comes to connections
Last edited:
Hi gootee
If you want me to respond to the rest of your lately posts (wich I have not responded too yet) please let me know,
I made the mistake of not relating a real single cap of 100000uF to 100 real caps of 1000uF. I was not even aware someone would use such a cap. The esr and esl of those overcapaciticed caps is not any better than one of the better caps with 1/10 of that capacity. Offcourse the cap-bank wins such caps with easy. But at other values things are not as drast. And offcourse, if for whatever reason you need the smallest esr or esl then you choose the best suited way to get there and build your cap-bank.
In audio i usually ( tube freak) dont need lower esr/esl than 1 or 2 decent caps can provide. In professional applications electrolytic caps mostly would not cut it and would be bypassed with something much more suitable for the task.
If you want me to respond to the rest of your lately posts (wich I have not responded too yet) please let me know,
I made the mistake of not relating a real single cap of 100000uF to 100 real caps of 1000uF. I was not even aware someone would use such a cap. The esr and esl of those overcapaciticed caps is not any better than one of the better caps with 1/10 of that capacity. Offcourse the cap-bank wins such caps with easy. But at other values things are not as drast. And offcourse, if for whatever reason you need the smallest esr or esl then you choose the best suited way to get there and build your cap-bank.
In audio i usually ( tube freak) dont need lower esr/esl than 1 or 2 decent caps can provide. In professional applications electrolytic caps mostly would not cut it and would be bypassed with something much more suitable for the task.
Last edited:
This is not so much for you gootee, i thinck you dont have a problem with it anymore, but just to make sure Andrew T does not have either
Df=0.08 is what you can find in textbooks, is not exactly what it is in real life
1 cap, 1000uF 100hz ESR = Df/2pii*f*C = 127,32mOhm
a real cap of 1000u/63 - 100V would have somewhat lower ESR
100 cap 1000uF in parallel 127,32/100 = 1,2732mOhm
1 cap, 100.000uF 100hz ESR = Df/2pii*f*C = 1,2732mOhm
a real 100.000uF/63-100V cap would have up to 4 x higher ESR
caps with ca 10.000uF would have more favourable ESR values as the 100.000uF cap has
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.