Talk:Decoupling capacitor

From Wikipedia, the free encyclopedia
Jump to navigation Jump to search
WikiProject Electronics (Rated Start-class)
WikiProject iconThis article is part of WikiProject Electronics, an attempt to provide a standard approach to writing articles about electronics on Wikipedia. If you would like to participate, you can choose to edit the article attached to this page, or visit the project page, where you can join the project and see a list of open tasks. Leave messages at the project talk page
Start-Class article Start  This article has been rated as Start-Class on the project's quality scale.
 ???  This article has not yet received a rating on the project's importance scale.
 

Mention of saturating inductances[edit]

The "Switching subcircuits" section mentions that the sagging supply voltage to an IC is only temporary since the line inductances and so on will eventually "saturate". Is this accurate? The line inductances are parasitic inductances, they are not like inductor components which will generally have some kind of ferromagnetic core. There shouldn't be any (significant) saturation effects. Even if an inductor does saturate, its inductance never goes to zero, it just becomes much smaller. I think whether the parasitic inductances saturate or not is besides the point. The inductances make it hard for the chip to change its supply current quickly when it switches every now and again. By their very nature, these load changes are temporary and transient, so inductor saturation is not a necessary explanation for why the sagging supply is transient, nor does it seem accurate. Correct me if I'm wrong. --Hddharvey (talk) 03:34, 8 August 2018 (UTC)

Yes, they do not saturate.Constant314 (talk) 04:05, 8 August 2018 (UTC)
The whole idea behind parasitic inductance is that there is more inductance than you would like. In some cases, though, a series inductor is used as part of a decoupling circuit. They used to be, though rarely are now, in power supply filter circuits, and those were usually iron core inductors. Lossy ferrite cores are now often used for decoupling, especially for RF signals that travel on the outside of cables, or common mode signals on cables that aren't supposed to have common mode. In that case, they could saturate, and that would reduce the desired effect. I don't know that this needs to be included here. I suppose there is no article on decoupling inductors. Gah4 (talk) 04:18, 12 August 2018 (UTC)
Yes, I've seen instances where a sub-circuit does have a small series ferrite inductor, however, the inductor is selected such that it does not saturate when the circuit functions correctly as that would defeat its purpose. But the article was using saturate incorrectly to mean that the inductor current reaches a peak and then returns to normal. Constant314 (talk) 19:09, 12 August 2018 (UTC)
The usual case of unwanted saturation is running transformers on too low line frequency (60Hz transformers on 50Hz). Most should have enough margin, but not all. The inductance limits the current, and if it goes down, the current goes up. LC power supply filters were common in vacuum tube amplifiers, possibly because capacitors weren't as good as now. More iron makes transformers and inductors more expensive, so they tend not to overdo it. Decoupling gets harder when you have mixed analog and digital circuitry in the same device. Gah4 (talk) 19:56, 12 August 2018 (UTC)

Misc[edit]

Do you need any more capacitance than the inductance of the supply dictates?--SpectrumAnalyser 18:53, 12 August 2007 (UTC)

Page title[edit]

Should this page be called electrical decoupling to include other forms of decoupling as well as capacitors?--SpectrumAnalyser 19:35, 12 August 2007 (UTC)

This page need major corrections to the language used to describe circuits; it is too casual. Circuits do not "see" current or signals, nor do they have "intentions" of behaving a certain way. —Preceding unsigned comment added by 70.239.100.6 (talk) 01:12, 3 January 2008 (UTC)

Please feel free to edit to your satisfaction.--TreeSmiler (talk) 01:31, 3 January 2008 (UTC)

No schematics!?! —Preceding unsigned comment added by 208.22.104.18 (talk) 14:01, 5 June 2009 (UTC)

Yea I agree, would be nice to see some schematicsDiscmon (talk) 13:44, 10 March 2013 (UTC)

That's what I thought too when reading this article, so I added an example schematic Dalva24 (talk) 04:52, 22 September 2017 (UTC)
Now reading it again, and if I undertand it correctly, I only added an example for analogue "smoothing" capacitor. Maybe someone can also add an example for digital logic decoupling Dalva24 (talk) 05:08, 22 September 2017 (UTC)
Started a subtopic called: Examples. Added two schematics along with their simulated oscilloscopes. Plus added descriptive text to help focus the reader's attention on some key points.Vinyasi (talk) 02:03, 1 January 2018 (UTC)
From another discussion on this page, there are sometimes series inductors, forming LC filters, used in decoupling. I don't believe that is enough for a whole page on decoupling inductors, though. For digital circuits, the current pulse due to fast changing signals is (all?) that is important. For analog circuits, it is more complicated. In that case, you might need to decouple a wide range of frequencies. You need to keep 60Hz (or 50Hz) out of the signal path. Even just keep the two halves of a stereo amplifier apart. I know of ones that have separate power supplies for each half. We could have a redirect for decoupling inductor if that is useful. Gah4 (talk) 20:01, 12 August 2018 (UTC)

Why are decoupling capacitors not directly integrated?[edit]

Surface-mount components on a flash drive's circuit board. The small rectangular chips with numbers are resistors, while the unmarked small rectangular chips are capacitors.

Correct me if I'm wrong: The capacitors on most circuits with IC chips fulfill the function of being decoupling capacitors. Books such as Designing Embedded Hardware suggest placing decoupling capacitors literally everywhere. If decoupling is so essential and ubiquitous, then why are these decoupling capacitors not directly integrated within the ICs? This should reduce the PCB's size and the systems BOM. Awaiting your ideas - Abdull (talk) 20:14, 10 February 2010 (UTC)

It would also simplify designs when IC's could do without separate, nearby placed decoupling capacitor(s), and reduce board space a bit. And make it less of an issue when sub-optimal wiring / sockets are used. Perhaps it's a technology / cost issue: to integrate a capacitor, you'd either have to make some sort of hybrid IC, or make more process steps (=increased cost) to integrate a capacitor as part of the silicon. I'm guessing external capacitor is less elegant, but overall cheaper solution? --RetroTechie 24 April 2010, 18:14 CEST
Some current high speed ICs do include decoupling caps in the package, but external decoupling caps are still needed, since the amount of decoupling capacitance needed will vary with the specific PCB design and frequency of operation.
Including decoupling caps in the die area is prohibitively expensive, too much area is needed to create even a very small cap, so nobody (that I know of) does this at this time. —Preceding unsigned comment added by 201.163.84.60 (talk) 03:14, 19 May 2010 (UTC)
The initial discussion would benefit from being split into paragraphs, and maybe with a better opening. I will have a go myself, but I make no promises regarding quality. — Preceding unsigned comment added by Evildoctorbluetooth (talkcontribs) 09:07, 23 June 2015 (UTC)
I think the answer is "the amount of silicon area required to create a useful capacitance in silicon takes up a lot of silicon area and far more expensive than external capacitors which are created out of other cheaper materials. • SbmeirowTalk • 07:36, 24 June 2015 (UTC)
Large capacitances on chip need a complete different integration technique, the Bosch Deep reactive-ion etching, you can reach capacitances of 250 nF/mm2, much lower than MLCCs, which use the same height as the IC. As silicon capacitors (Passive integration) [[1]] this types you can get from IPDIA [[2]] --Elcap (talk) 08:41, 24 June 2015 (UTC)
There are some IC packages with capacitors on top. Bigger packages with more pins, means longer distance pin to chip, so more inductance. But otherwise, keeping the inductance down is the main problem in design of IC packages. Gah4 (talk) 04:40, 4 January 2018 (UTC)

Decoupling for digital logic vs in analogue circuits[edit]

(comment moved from top of page by User:JulesH) The title is misleading, as it describes the use of a capacitor as a "smoothing" device - a crude regulator of the output voltage of a power supply. "Decoupling" capacitors are connected between two circuits to allow a.c. signals to pass from one to the other, while at the same time "decoupling" any d.c. levels. The source might have a d.c. level of 2.5V with an a.c. signal superimposed on it (say 50mV rms). If this is input to an amplifier circuit whose input is biased to 1.5V and designed to accept a 50mV rms signal, the d.c. difference will upset the biasing of the amplifier and lead to distortion or no signal transfer. A "decoupling" capacitor will block the d.c level, and allow the 50mV rms signal to pass, without upsetting the biasing. — Preceding unsigned comment added by 80.254.147.236 (talk) 10:28, 29 November 2011 (UTC)

At the very least this page needs a hatnote. While power supply bypass capacitors are commonly called "decoupling" capacitors in digital design, the phrase is used with the very different meaning the anonymous commenter above describes in analogue design, particularly audio amplification. I came here looking for an article on the latter meaning, unfortunately I have not yet found one. JulesH (talk) 15:39, 5 February 2012 (UTC)
Found it. JulesH (talk) 15:50, 5 February 2012 (UTC)
The latter should not be decoupling capacitors, though they might be coupling capacitors. Gah4 (talk) 23:03, 1 January 2018 (UTC)

comment[edit]

This phrase in discussion doesn't seem correct to me, "so small and large capacitors are usually placed together in parallel to fully cover circuit bandwidth". This is because putting 2 caps in parallel merely sums the individual capacitances. So, you effectively have one capacitor. On top of that I've never seen a circuit or a schematic showing 2 capacitors in parallel. This info about capacitors is so basic it is taught in physics before electrical engineering courses are started. — Preceding unsigned comment added by 74.77.148.238 (talk) 06:28, 22 February 2017 (UTC)

Yes it is common to have more than one capacitor, as larger capacitors turn into inductors at higher frequencies. There are usually a few large capactors for a board, and smaller ones near each IC. Physics mostly ignores this effect. Gah4 (talk) 23:01, 1 January 2018 (UTC)
Discussion of SPICE simulation of decoupling capacitors that drifted into WP:FRINGE
The following is a closed debate. Please do not modify it.

I don't see the connection between images and WP:OR.[edit]

Regarding your deletion, I don't understand the reference. I thought images are not original research and merely aid the reader to focus on the text? What am I missing? How does this apply?Vinyasi (talk) 03:04, 1 January 2018 (UTC)

You need a reliable source to say that circuit is properly implemented and that it demonstrates the effect. If a manufacturer gives a circuit on a datasheet and you "paraphrase it" (redraw it yourself), that would not be OR and you could cite the manufacturer's data sheet. If your SPICE implementation were obvious enough, we might accept that it is an acceptable representation of the manufacturer's circuit. Your circuit, as drawn, is not obvious on these counts: the action of the source and the switch require expert knowledge of a particular SPICE implementation. For example, is the switch on or off when the voltage is zero? Finally, unfortunately, SPICE lies, especially on the transient analysis. But before we invest a lot of energy in this debate, lets see what others say. Also- Happy New Year! Constant314 (talk) 03:24, 1 January 2018 (UTC)
Constant314 Switch is OFF when voltage is measured zero at Vin. Switch is ON when voltage is measured zero at V(1). The exorbitant positive and negative spikes measured at V(1) are the release of back EMF originating at the inductor. Good question. I had to sit and stare at it for a few moments. Thanks for asking. Maybe I should edit the diagram/s by adding text to highlight your concern for vagueness? We can always wait to hear from others.... As for a simulator lying, well, I don't know. I tried the simulation using more strict parsing methods which LTSpice gives the user the option to try out, and I found no variation for both images used in my Decoupling capacitor edit. Vinyasi (talk) 04:48, 1 January 2018 (UTC)
Having stared at your circuit, I think what you have depicted is not a common application of a decoupling capacitor. In fact, it probably would not be considered an example of decoupling at all. Constant314 (talk) 05:24, 1 January 2018 (UTC)
It might be that I don't understand what Vinyasi is doing, but the common need for decoupling is lead inductance and di/dt. Lead inductance is small, but di/dt is big. Gah4 (talk) 22:57, 1 January 2018 (UTC)
This is the reason why I attempted to write articles on shielded inductors and decoupling capacitors.
Vinyasi (talk) 03:51, 5 January 2018 (UTC)
Version two of Joseph Newman's Energy Machine with slightly faster processing time by the simulator at the expense of less suppression of spikes resulting from arcing on the commutator-styled switches.
Vinyasi (talk) 15:00, 4 January 2018 (UTC)
Those chaotic looking currents are symptoms of possible convergence failure. By the way, this is WP:FRINGE. Constant314 (talk) 20:40, 4 January 2018 (UTC)
Overunity is always stressful to the circuit, to the inventor, to our penchant for comfort and security. Needless to say it's also stressful to a simulator trying to make sense of it. We're supposed to take a neutral point of view, here. So, I refuse to interpret 'Fringe' as being even remotely prejudicial. I'd much rather tone it down a bit and call this out-of-the-ordinary which destroys the foundation of verification, here. So be it. Without challenging both mine, and your, comfort zone for mingling where I don't belong - here, I would not have learned anything. So, I thank all of us for our differences. The one thing we all seem to share in common, is that we care enough to get involved. Vinyasi (talk) 04:39, 5 January 2018 (UTC)
Fourth revision indicating that all of the so-called magical gain is provided by the permanent magnet rotor. No magic, here.
I suspect the chaos is the result of trying to squeeze all of the gain from out of the primary voltage source, V1. That created the disruption in the simulated voltage source of the permanent magnet rotor, V3. By tinkering with it some more, I decided not to be so precise in attempting to replicate Newman's results. Instead, I decided upon merely replicating the overall concept and adhere to your comment. So, by diminishing the magnetic field of the rotor, and increasing its voltage, and totally eliminating any spikes by increasing the capacitor, I discovered that the mysterious gain claimed by Newman is not so mysterious after all. It all comes from the massive pressure provided by the permanent magnet. This smooths out all of the traces eliminating chaos from the circuit simulation, reverses the current at the primary voltage source (battery pack) - V1, reduces the voltage provided by V1 to zero, and still manages to provide the appearance of gain on the coil, L1, transferred from the magnet. So, the coil is merely acting as a pickup receiving power from the rotating magnet. Thanks for the criticism. This clears up a lot of the mystery surrounding his device. Vinyasi (talk) 20:56, 5 January 2018 (UTC)
Gah4 I'll tell you what I'm doing. Joseph Newman used a very long wire for the primary coil on his motor. That produced a very high capacitance since the depth of the enamel on his magnetic wire was divided into the huge surface area of a 50-100 mile long coil of wire. But the capacitor in my two circuits are in parallel, not inline, with the inductor, and grounded - just like his motor was not grounded directly, but indirectly, to its immediate surroundings through the capacitance of the insulation on his motor's primary coil creating a huge electric field surrounding it. Since it's the capacitor which is grounded, not the inductor itself, this has the opposite effect of a capacitor placed inline with the inductor and thus throws its voltage back onto the inductor. This is especially magnified since the low-level capacitor of 0.01 pico Farads in the 'Back EMF with bypass filter' circuit on the left-hand side of the image reaches saturation very prematurely. This implies what amplified the inductor's back EMF on Newman's primary motor coil (kind of like using 'reverse psychology', or 'reverse logic in a debate', to accomplish a task). The 'Suppressed back EMF with a bypass filter' circuit on the right-hand side of the image is suggestive of conventional motors since the larger valued capacitor of 10 Farads - again, located not inline with the inductor - is effectively equivalent to the use of shorter motor coil lengths of diminished capacitance on conventional motor coils and that smooths out their response. Conventional EV motors satisfy conventional consumer tastes of fast reaction times in providing acceleration not requiring a lengthy period of 'warmup'. Newman's motor required walking away for several minutes (after getting it started with a hand-crank effect similar to early cars, such as Ford's Model T) to wait for his motor's RPMs to increase and stabilize at its normal idling speed operating without a load. Since the Patent Office denied him a patent, we'll never know what we're missing. But I read his book and understand the basic idea. I agree, this sort of thing is not at all practical. But as a teaching tool, his motor rocks. That's what I intended to do with this circuit simulation: emulate his motor, but in a solid state device. I hope I didn't make this more confusing since, in my circuit, the capacitance is in parallel with the inductor creating a capacitant effect inverse to an inline capacitor. But a normal motor coil would engender an effective capacitance equivalent to its wire length – not inverse to it as in my two circuits.
In the real world, the values of capacitance and inductance of a length of wire would rise or fall together. But in my simulation, they're kept as two distinctly separate values since a simulator can't depict a virtual ground of an inductor, namely: a direct ground of the insulation on an inductor – in contradistinction to a direct grounding of its copper coil (how the National Bureau of Standards errored in their test of the Newman motor, see below). Thus, keeping them separate means I have to do the opposite of what the real world would evoke: I have to lower the capacitor's value in my simulation if I want to emulate a long coil of wire since the capacitor is not inline, but in parallel, to the inductor nearby.
This is why I felt compelled to create a new set of electronic symbols to imply this new way of appreciating coil lengths – both with and without a magnetic core. And also in small sizes with and without a magnetic core.
But this is all OR to admit it now. That's why I distinctly chose not to admit this until now hoping that if I focused on the principle, not the legacy of an inventor, I might have a better chance of success. But obfuscation of motives didn't work. Oh, well. Vinyasi (talk) 04:36, 2 January 2018 (UTC)
Concerning the Patent Office, the National Bureau of Standards admits on their website that: The National Bureau of Standards provided the resistive load which was connected in parallel with the coil. This is an example of current wanting to take the path of least resistance through a parallel load, bypassing the coil (of greater resistance than their test load) to a significant degree by cleverly self-shorting Newman's primary coil and prevent accumulation of HV. Had they performed their test as Newman advises in his book by wrapping a shorter length of secondary coil around the larger primary coil and placing the test load only inline with this secondary and isolated from the primary, then it's entirely possible that the NBS would have produced more accurate results and replicate Newman's. Then, the history of Newman and his motor would have read a bit differently then what transpired. Vinyasi (talk) 04:34, 2 January 2018 (UTC)
Constant314 basics.gif Looks the same to me. {from: The Basics - Bypass Capacitors} The only distinction is that Fig.1 succeeds at suppressing back EMF while Fig.2 does not. Vinyasi (talk) 06:08, 1 January 2018 (UTC)
Enhancement vs Suppression of Back EMF.jpg
OK. So it's synthesis on my part. The older, out-of-date NOR said – To the extent that part of an article relies on a primary source, it should: only make descriptive claims about the information found in the primary source, the accuracy and applicability of which is easily verifiable by any reasonable, educated person without specialist knowledge, and make no analytic, synthetic, interpretive, explanatory, or evaluative claims about the information found in the primary source. – I was hoping that anyone can see from this schematic that it might be better not to fight Mother Nature, but accept and enhance back EMF while also trying to discover ways of putting it to good use. Vinyasi (talk) 20:56, 1 January 2018 (UTC)
Also, your circuit is bizarre. The voltage source starts out at 48V, decreases rapidly to 0 for about 1 ms and then goes rapidly back to 48V. Nothing bizarre with that. But then consider what the capacitor sees. It sees the voltage source, through the 0.1 milliohm switch, go from 48V down to 0.5V and then become a high impedance for a millisecond and then change back to a voltage source at 1.5V which then rapidly returns to 48V. There is no circuit, known to me, that would function in such a way. So, not only would a non-expert not understand the circuit, an expert that did understand the circuit would not see a real world example of decoupling. Constant314 (talk) 01:20, 2 January 2018 (UTC)
───────────────────────── A talk page is not a forum for general discussion. The above is not about discussion of the improvement of this article on decoupling capacitors. I propose to apply the "collapse" templates to the above off-topic discussion. --Wtshymanski (talk) 22:09, 5 January 2018 (UTC)
Fine with me. Constant314 (talk) 22:47, 5 January 2018 (UTC)
I tend to give more leeway to talk page discussions. For one, the discussion might decide that something isn't needed or appropriate for the article, but you won't know that without discussion. Or maybe it belongs in another article, which again you learn through discussion. More specifically, an article might be improved by leaving something out, which still requires discussion on why it is best not in the article, which means understanding it enough to know that. Gah4 (talk) 00:17, 6 January 2018 (UTC)
I agree in general. As long as the discussion was about a circuit simulation that might be useful for demonstrating decoupling capacitor, it was on topic. But it has taken off on a tangent regarding over-unity machines and the idiosyncrasies of SPICE; it is clearly off-topic. Anyone who wants to discuss it further can do so via user talk pages. Constant314 (talk) 01:24, 6 January 2018 (UTC)
Or it could go to the talk page for Newman's_energy_machine. Gah4 (talk) 06:05, 6 January 2018 (UTC)
I agree with Gah4. I'm new here and don't know how to move this discussion to a more appropriate location. Vinyasi (talk) 08:00, 6 January 2018 (UTC)
I just did a copy and paste to Vinyasi's user page. It would still be WP:OR / WP:SYN on the Newman's_energy_machine page, so it is unlikely that it would be welcomed there. Constant314 (talk) 13:38, 6 January 2018 (UTC)
Vinyasi, kindly remove your inline citations that are causing the reference listings at the bottom of the page. These will always go to the bottom of the page rather than the bottom of the section. It is not a problem now, but it will be very annoying when someone starts a new discussion. Constant314 (talk) 13:59, 6 January 2018 (UTC)
Thanks for the tip. Vinyasi (talk) 14:12, 6 January 2018 (UTC)
Thanks for that. Now, I will collapse this discussion. If anyone objects, I will un-collapse it will still be here and you can still edit it. However, it will show up as a simple box. You will have to click the "show" link to see the contents. Editing will be unchanged. Constant314 (talk) 15:28, 6 January 2018 (UTC)
And now we should consider marking the discussion as closed with a note that it can be continued on talk's talk page. Constant314 (talk) 15:36, 6 January 2018 (UTC)

German link[edit]

Link to german wiki is incorrect. It should be to de:Blockkondensator, rather than to smoothing capacitor. Don't know how to edit it myself.--Prandr (talk) 16:32, 7 April 2018 (UTC)

Where does that occur in the article? Constant314 (talk) 17:00, 7 April 2018 (UTC)
I mean the link to the german version of the article.--Prandr (talk) 10:52, 8 April 2018 (UTC)
Is it part of this article? If so, where? Constant314 (talk) 11:48, 8 April 2018 (UTC)
I mean the interwiki on the left panel.--Prandr (talk) 13:06, 8 April 2018 (UTC)

decoupling, filtering, smoothing, etc.[edit]

In the discussion I am used to, the capacitors in the power supply are filtering capacitors, to filter out the power line frequency, or for switching power supplies, the switching frequency. These are relatively low (50Hz to 20kHz) and mostly use electrolytic capacitors. Decoupling capacitors, and the ground plane of PC boards, are for the higher frequency signals generated by fast switching digital signals. (That is, high frequency Fourier components.) These have to be close to the fast switching signals, as PC trace and lead inductance are significant at these frequencies. Larger capacitors, especially electrolytics, have enough inductance that they don't do much at all at higher frequencies. Gah4 (talk) 07:26, 13 April 2018 (UTC)