<--my face when I saw over one hundrer measurements)
There was a question proposed: "Why would you use a polyester cap?"
I would use a polyester cap when I want an effect that isn't abrupt. That can make an amp zobel that doesn't "chop" into the tweeter's response like MKP.
Is there any use for polyester in power supplies?
I am planning to make this:
(one rail shown)
AC
input
Diodes (MUR4XX)
10k
2200uF ecap (Mallory/Cornell 50v SEK)
2200uF ecap . .
0.22uF polyester (14ESR)
output
The figures are 100% imagination, but what does this power supply do?
I would use a polyester cap when I want an effect that isn't abrupt. That can make an amp zobel that doesn't "chop" into the tweeter's response like MKP.
Is there any use for polyester in power supplies?
I am planning to make this:
(one rail shown)
AC
input
Diodes (MUR4XX)
10k
2200uF ecap (Mallory/Cornell 50v SEK)
2200uF ecap . .
0.22uF polyester (14ESR)
output
The figures are 100% imagination, but what does this power supply do?
danielwritesbac said:
Hi Daniel,
The first thing I noticed is the 10K resistor.
You didn't mention the AC amplitude, but, a 10K resistor, if in SERIES, would severely-limit the current. I have used resistors (and inductors), there, and even another one between the elcos, but the resistances have all been extremely small, maybe typically from 0.1 Ohm to probably something less than 1 Ohm, probably mainly so the voltage is not dropped too significantly for large-ish currents.
With a small resistance in series, before a large capacitor, you get a beneficial low-pass filter, with -3dB frequency at 1/2.Pi.R.C . For example, for 0.5 Ohms with 4400 uF, the f(-3dB) would be at about 72 Hz, which would even help attenuate the 120 Hz or 100 Hz rectifier ripple.
But, note that you DO have to worry about the resistor's power dissipation. You can use i-squared times R or v-squared over R, to calculate that. Example: 6 Amps through 0.5 Ohm gives 6 x 6 x 0.5 = 18 Watts, i.e. probably at least a 30W-capable resistor needed, or paralleled smaller ones. And it's not a steady DC current. So you'd actually need to use the RMS value of the ripple current, to calculate the dissipation more-accurately.
On the other hand, if your 10K is in parallel with the caps, then it would be fine, and would discharge them after power-off.
Regarding the 0.22 uF polyester film cap in parallel with the 2200 uF elcos: According to what 'looks like' the concensus reached in this thread, it's probably not a good idea, or, at least, might be way too small. (By the way: What did you mean by "14ESR", for that cap? Is that 14 milliOhms?)
10k is parallel (instead of 2.2k parallel that burns up)
14 ESR is roughly equivilent to:
7 ohms--perfect cap--7 ohms
However, that's not the whole picture because you'd need a lot of resistors and ceramic caps to duplicate what's going on inside a high-esr model like polyester.
I just didn't want to use polypro and a 2r resistor there because that typical application will eventually burn up the resistor.
What value capacitor would you suggest in order to help complete the frequency response of the 2200uF ?
Thanks!
14 ESR is roughly equivilent to:
7 ohms--perfect cap--7 ohms
However, that's not the whole picture because you'd need a lot of resistors and ceramic caps to duplicate what's going on inside a high-esr model like polyester.
I just didn't want to use polypro and a 2r resistor there because that typical application will eventually burn up the resistor.
What value capacitor would you suggest in order to help complete the frequency response of the 2200uF ?
Thanks!
danielwritesbac said:
I didn't think that any polyester caps had high ESR. Those kinds of numbers are at the level of the ESR of a small electrolytic, on a bad day.
As far as calculating (or estimating) a good value for a film cap, to parallel your 2X 2200 uF caps, there are some formulas (and some 'rules of thumb') that have been proposed, in this very thread. But, to use them, you would need to know, or estimate well, the inductances of the caps and the interconnects and what's upstream (according to what I think I understand from what I've read here, so far, anyway).
I am usually more inclined to put a resistor in series with the film cap, and use the equations for calculating snubber network values. And I'm not usually in possession of good-enough numbers, for that. So I usually end up estimating most of them and then tweaking the calculated R and C snubber values in a working prototype. (To not "eventually burn up the resistor", you just use one that's rated for high-enough power dissipation.)
There is a good paper about snubber design on the Cornell Dubilier website, at http://www.cde.com , in the technical papers section:
http://www.cornell-dubilier.com/design.pdf
There are some other good ones on line, too:
http://www.hagtech.com/pdf/snubber.pdf
http://www.ridleyengineering.com/snubber.htm
http://www.maxim-ic.com/appnotes.cfm/appnote_number/3835
http://archive.chipcenter.com/circuitcellar/november00/c1100rp58.htm
The one at maxim-ic.com, above, has a nifty procedure for determining the R and C values for a snubber, from empirical measurements, in a circuit where there's ringing.
Perhaps I underestimated what the resistor is doing here in this sequence
diode bridge
resistor (2.2k)
caps (large)
cap (very small and highly efficient)
RC
Doesn't it create a load from which to calibrate (arrange) the shunting of unwanted signals? In other words, is that a baseline (reference point) or is it my imagination?
If its a reference point, then the plan is to make a zero-sound situation, so the little bypass cap and/or RC would need to be really efficient, like polypro or ceramic caps.
I'd have to guess that you'd need 1 really efficient cap (very high pitches) and 1 RC (fill in the gap)?
What would you suggest as the most generally likely estimate for values at RC?
I can read most of the documents above and get some sense, but I can't apply them, so i need a bit of help.
Thanks!!
diode bridge
resistor (2.2k)
caps (large)
cap (very small and highly efficient)
RC
Doesn't it create a load from which to calibrate (arrange) the shunting of unwanted signals? In other words, is that a baseline (reference point) or is it my imagination?
If its a reference point, then the plan is to make a zero-sound situation, so the little bypass cap and/or RC would need to be really efficient, like polypro or ceramic caps.
I'd have to guess that you'd need 1 really efficient cap (very high pitches) and 1 RC (fill in the gap)?
What would you suggest as the most generally likely estimate for values at RC?
I can read most of the documents above and get some sense, but I can't apply them, so i need a bit of help.
Thanks!!danielwritesbac said:Perhaps I underestimated what the resistor is doing here in this sequence
diode bridge
resistor (2.2k)
caps (large)
cap (very small and highly efficient)
RC
Doesn't it create a load from which to calibrate (arrange) the shunting of unwanted signals? In other words, is that a baseline (reference point) or is it my imagination?
If its a reference point, then the plan is to make a zero-sound situation, so the little bypass cap and/or RC would need to be really efficient, like polypro or ceramic caps.
I'd have to guess that you'd need 1 really efficient cap (very high pitches) and 1 RC (fill in the gap)?
What would you suggest as the most generally likely estimate for values at RC?
I can read most of the documents above and get some sense, but I can't apply them, so i need a bit of help.
Hi Daniel,
I am not capable of thinking of it in those terms. I'm sorry.
But I have to say that I really don't see a need for any smaller caps, or even much need for any high-frequency snubbing, at the power supply reservoir. Maybe upstream, a snubber might be appropriate, to try to cancel the inductance of the transformer (if that's even necessary or desirable; just guessing), and maybe to try to quell any tendencies for the rectifier diodes to ring when they abruptly turn off.
You might initially reasonably guess that it could be beneficial to have a snubber in parallel with a large power supply reservoir capacitor to try to cancel its inductance, or a small cap there to try to 'fill out' the capacitance versus frequency. BUT, the leads from the reservoir caps to the load are probably already *much* more inductive than the caps, which (i.e. a power supply lead's or trace's inductance) will probably also negate, and cause to be merely a localized benefit, any benefit from the smaller parallel caps (with nothing local, there, to actually benefit). Plus, as explained rather thoroughly in this thread, you run a significant risk of generating a lot of high-frequency noise or even oscillation, due to unintended resonances caused by inserting a small parallel capacitor.
Therefore, I think that the smaller caps, and maybe a snubber (if there's enough need for a snubber, or if you're just gung-ho-enough to try for perfection etc), would be MUCH more useful and effective if they are placed AFTER the final power supply leads or PCB traces, i.e. directly AT the point of load.
If there is to be a three-terminal or other voltage regulator near the reservoir caps, that might change the story, a bit. But, again, I believe in placing any regulator (or maybe a second one) as close to the load as is physically possible. So I still tend to believe that most of any effort toward cleaning up, and optimizing the frequency response of, the power supply rails, will be most-effective if done as closely as possible to the point of load (e.g. at a chipamp's power supply pins), so that the good effects are not spoiled by any intervening parasitic inductances, etc.
If you really want to improve the power supply's effects, you might want to first worry about the possibly really-significant ones. Besides major things like worrying about keeping the load's power ground quiet (I've seen suggestions like separating the two big caps' grounds, etc), you will (eventually at least) want to look into using super-regulators. It's said that there is 'no comparison', regarding the actual sound that's produced, when using them.
I know that you're a hands-on type of guy, and have been building and testing chipamps quite prolifically, lately. And that, my friend, appears to be a huge understatement, judging from what I've been reading. Kudos and attaboys to you, for all of that! And I know that you are or at least were some type of computer networking engineer. But, if you don't have at least an oscilloscope, and ideally other test equipment as well, it might be difficult to properly investigate or evaluate the effects of things like snubbers and the paralleling of large and small capacitors. Maybe you should check out http://www.govliquidation.com , when their monthly test-equipment sale comes around again, or ebay.com , and do some searches for Tek* (* is a wildcard, at both sites). At the very least, I recommend doing some google searches for thing like 'sound card oscilloscope download'. Keep in mind that their bandwidth is far too low for many important tasks. But they would definitely be far better than nothing, IMO.
However, if you can't immediately get a scope (and even if you DO have one!), there is a lot that can be learned by using the LTspice circuit simulator. It gives you a virtual scope display of the voltage or current of any points in a circuit to which you touch the probe (mouse pointer is a probe symbol), and can do the same for frequency response plotting. And so much more is possible, with it. The ability to zoom-in to see small parts of waveforms, alone, is wonderful. If you also insert parasitic inductances and resistances of conductors and components, you can actually see the nasties and come up with pretty-good snubber designs to kill them. It's extremely fascinating, actually, and could be a great tool for you, especially after a short learning period. (If your computer is fast-enough, note that you can ALSO use real WAV files, for both inputs and outputs! So, theoretically, one can actually LISTEN to a SIMULATED amplifier, for example. I have done that, as a matter of fact, while evaluating DC Servo designs.) Your technical background could only help, too. If you want to start with power supplies, I have some READY-MADE ones that you can download and run *immediately*, and apply 'tweaks' and see the results in minutes, etc, etc. Everything you need to download to get started is at http://www.fullnet.com/~tomg/gooteesp.htm .
OH!! I just slapped my forehead! Of course!!
Thank you SO much for highlighting so many of the involved variables all at once.
Maybe the reason that I can't quite seem to get a handle on the snubber plan is that I haven't left much opportunity for it to be useful? Do you think that I should finally buy a toroid transfo and find out?
I can't use modeling software. It seems to be some sort of "personal difficulty" whereby the end results are always worst-case scenerios. Okay, maybe I do need to create a situation where I can more-easily observe the power supply at work.
I do like the measuring software, though.
Yes, I need some measuring stuff. . . something more definitive than grabbing a stopwatch and playing Enya to see how low it takes the cats to tump over.
Thanks for your help and encouragement!
P.S. Regulated best? Yes, agreed.
Thank you SO much for highlighting so many of the involved variables all at once.
Maybe the reason that I can't quite seem to get a handle on the snubber plan is that I haven't left much opportunity for it to be useful? Do you think that I should finally buy a toroid transfo and find out?
I can't use modeling software. It seems to be some sort of "personal difficulty" whereby the end results are always worst-case scenerios. Okay, maybe I do need to create a situation where I can more-easily observe the power supply at work.
I do like the measuring software, though.
Yes, I need some measuring stuff. . . something more definitive than grabbing a stopwatch and playing Enya to see how low it takes the cats to tump over.
Thanks for your help and encouragement!
P.S. Regulated best? Yes, agreed.
Inductance is what causes resonance. Pure ohmic resistance is required to "neutralize" inductance.
Capacitor combinations may be easily analized directly on the target circuit by applying a square wave with very fast rise and fall times through a resistor. The resulting waveform will show interesting data like inductance, ESR and all the ringing modes (if any). I'm currently using this method to check and adjust supply impedance in class D and SMPS circuits.
Capacitor combinations may be easily analized directly on the target circuit by applying a square wave with very fast rise and fall times through a resistor. The resulting waveform will show interesting data like inductance, ESR and all the ringing modes (if any). I'm currently using this method to check and adjust supply impedance in class D and SMPS circuits.
Relating it to a speaker?
Apparantly, you're right.
Although my abilities on this topic are limited, I did experiment with several scenerios.
Here's what I understood so far.
First was the bypass cap typical of audio power supplies. Okay, so this serves to increase the high frequency abilities of a big cap. However, the typical approach uses a far-too-small bypass cap, so that its effects are quite uneven.
Main point: Either way makes a noise, just exactly like the noise you can make if you try to make a full range speaker do woofer duties via a 2nd order crossover.
Where that cap is, is also a great big sound. Put a better cap, and you get a bigger sound. This is also true of bypass cap practice.
I take it that most of the point of the power supply filtering is to get rid of sounds.
Well, for a 1.5R speaker (about like a power supply), I'd never think that a 2 ohm + 0.1uF zobel would muffle it. That's far too small.
Don't we want to muffle it?
Given that example, then there's this:
(1 rail)
3x 2200uF caps (6800uF)
Okay, then the zobel is 2200uF + 5R
Next,
680uF
A zobel with 220uF
Next,
68uF
a zobel with 22uF
Next,
6.8uF
a zobel with 2.2uF
Next,
.68uF
a zobel with 0.22uF
Next,
.06uF
a zobel with 0.02uF
Output (if any) is DC. Is there an easier way?
Eva said:
Apparantly, you're right.
Although my abilities on this topic are limited, I did experiment with several scenerios.
Here's what I understood so far.
First was the bypass cap typical of audio power supplies. Okay, so this serves to increase the high frequency abilities of a big cap. However, the typical approach uses a far-too-small bypass cap, so that its effects are quite uneven.
Main point: Either way makes a noise, just exactly like the noise you can make if you try to make a full range speaker do woofer duties via a 2nd order crossover.
Where that cap is, is also a great big sound. Put a better cap, and you get a bigger sound. This is also true of bypass cap practice.
I take it that most of the point of the power supply filtering is to get rid of sounds.
Well, for a 1.5R speaker (about like a power supply), I'd never think that a 2 ohm + 0.1uF zobel would muffle it. That's far too small.
Don't we want to muffle it?
Given that example, then there's this:
(1 rail)
3x 2200uF caps (6800uF)
Okay, then the zobel is 2200uF + 5R
Next,
680uF
A zobel with 220uF
Next,
68uF
a zobel with 22uF
Next,
6.8uF
a zobel with 2.2uF
Next,
.68uF
a zobel with 0.22uF
Next,
.06uF
a zobel with 0.02uF
Output (if any) is DC.
Is there an easier way?small paralel filmcaps on power supplys elcaps...to do it properly and finding the right cap and size of it you need measurements...otherwise you risk to make things worse
Leach, and others has written about it
Skilled people dedicated to tubeamps making LCL supply filter (?) will also know the importanse of that
Leach, and others has written about it
Skilled people dedicated to tubeamps making LCL supply filter (?) will also know the importanse of that
sorry about my ignorance, or maybe the topic is out of my league to absorb. has read the entire threat for 1 1/2 time but still no clue on the conclusion as i do not follow all the links.
whats the conclusion of the discussion? is paralleling films caps to electrolytic caps good to go? if good, whats the rule of thump or scientifically proven beneficial? 1/10? 1/100 1/1000?
thanks in adv
whats the conclusion of the discussion? is paralleling films caps to electrolytic caps good to go? if good, whats the rule of thump or scientifically proven beneficial? 1/10? 1/100 1/1000?
thanks in adv
Hmmmm....
Well i have to build a supply soon for my lil "monster" that employs LME49810 as a driver and 8 pairs of thermal tracks as output devices. the LME will be fed separatley by a regulated supply where the output section will be fed by the main supply which will be a high current using a 1.5KVA toroid
Here's the schematic...
T1 - 1.5KVA 80VAC CT @ 18.8A
D1 - D4 -> RURG5060
C1 - C4 & C14, C15 -> decoupling caps 0.1uF 250V - Ceramic
C5 - C14 -> 10000uF 100V CHEMI-CON kMH Caps
C17 -> 1000uF 250V Chemi-Con SMQ
L1, L2 -> 5uH
R1, R2 -> 2.2Ohms 10W
Still thinking sneaking in 6 x 100uF 100V Nichicon KZ caps
2 after the chokes, 2 after the bank of 4 caps ( C5 - C8 ) prior to the resistors and then right before the output before the Ceramics...
Input Please...
Well i have to build a supply soon for my lil "monster" that employs LME49810 as a driver and 8 pairs of thermal tracks as output devices. the LME will be fed separatley by a regulated supply where the output section will be fed by the main supply which will be a high current using a 1.5KVA toroid
Here's the schematic...
An externally hosted image should be here but it was not working when we last tested it.
T1 - 1.5KVA 80VAC CT @ 18.8A
D1 - D4 -> RURG5060
C1 - C4 & C14, C15 -> decoupling caps 0.1uF 250V - Ceramic
C5 - C14 -> 10000uF 100V CHEMI-CON kMH Caps
C17 -> 1000uF 250V Chemi-Con SMQ
L1, L2 -> 5uH
R1, R2 -> 2.2Ohms 10W
Still thinking sneaking in 6 x 100uF 100V Nichicon KZ caps
2 after the chokes, 2 after the bank of 4 caps ( C5 - C8 ) prior to the resistors and then right before the output before the Ceramics...
Input Please...
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.