47

For the magnitude of nF or µF capacitors, I hope I can build them on a PCB board. The capacitor is like a two metal layer and something between them.

Is this possible?

Not buying the capacitor, just design the capacitor on the PCB board. Double metal layers on the PCB board.

Peter Mortensen
  • 1,676
  • 3
  • 17
  • 23
alan
  • 787
  • 2
  • 7
  • 15
  • 8
    +1, a question of interest to many experimenters, I'm sure. – Anindo Ghosh Jan 10 '13 at 06:33
  • 1
    Some good worked examples here. But a larger value is possible between specific layers in a multilayer stack. Check with PCB manufacturer, but an 0.15mm layer is realistic for an order of magnitude bigger C. –  Jan 10 '13 at 11:52
  • 1
    To add to @BrianDrummond's comment, even 0.1 mm is readily available, making 1 - 10 nF reasonable values to reach in a multi-layer board, but 1 uF is probably still out of reach. – The Photon Jan 10 '13 at 21:27
  • 2
    [In 1947 you could build capacitors into circuit boards](http://www.youtube.com/watch?v=3QHizDkrFqg). – RedGrittyBrick Jun 30 '13 at 20:33
  • I could use this question for my question: what capacitance could you get from 2 sided un-etched PCB, dimensions 10x15cm? I'm asking about very rough numbers, such as are we talking about pF or nF. – Jakey May 31 '17 at 08:22

8 Answers8

51

You will have a hard time achieving 1 nF by just laying out copper on a standard two-layer FR-4 board. The capacitance is given roughly by the parallel plate equation:

\$ C = \dfrac{ \epsilon A }{ d } \$

In this case

\$ C = \dfrac{(4.7)(8.854 \times 10^{-12}) A }{ (1.6 \times 10^{-3} ) } \$

or

\$ C = A (2.6 \times 10^{-8}\mathrm{F/m^2}) \$

Meaning you'd need .038 m2 or 380 cm2 of copper area to achieve 1 nF. I used 4.7 as a typical dielectric constant (relative permittivity) for FR-4 and 1.6 mm as a typical board thickness.

It is not uncommon to make pF scale capacitors by parallel copper regions, but it's normally done in multilayer boards where the d term can be much smaller. This kind of constructed capacitor can achieve lower ESR and ESL than a discrete capacitor, so it is valuable for bypassing power supplies in very high frequency circuits.

There are also companies that make special materials that can be laminated up in a multilayer PCB to provide a high-dielectric-constant layer, enabling construction of even larger capacitor value by metal patterning. 3M is one. These are often called embedded capacitors or buried capacitors. Contact your PCB fabrication shop to see if they support this type of material.

Ricardo
  • 6,134
  • 19
  • 52
  • 85
The Photon
  • 126,425
  • 3
  • 159
  • 304
  • +1 as well... The computed values are useful! – Anindo Ghosh Jan 10 '13 at 06:28
  • 1
    1nF with 380 cm2, but that is one low inductance capacitor for your plane, that will decouple high frequency well. – Kortuk Jan 10 '13 at 06:31
  • 3
    @Kortuk, sure but when did anybody ever have a spare 60 square inches on a PCB design? – The Photon Jan 10 '13 at 06:33
  • @ThePhoton Heh, you beat me by some minutes (+1) . So let's say our answers are cross checking themselves ;-) – zebonaut Jan 10 '13 at 06:40
  • 8
    keep in mind that FR-4 is hydroscopic and absorbs water moisture. Expect the value to change unless sealed. Another reason not to do this. – placeholder Jan 10 '13 at 06:43
  • @rawbrawb All the more reason to consider a PCB material like Dupont Pyralux: Tight dimensional tolerance, non-hygroscopic, and really thin substrate options. FR4 certainly doesn't cut it here. – Anindo Ghosh Jan 10 '13 at 06:50
  • 2
    Of course, you're not stuck with 1.6mm boards - 1.0mm boards are easy enough to get. – Nick Johnson Jan 10 '13 at 09:57
  • Interesting, so a typical prototype blank of un-etched double sided stock would be on the order of a nanofarad. Will have to try measuring one some time. – Chris Stratton Jan 10 '13 at 15:20
  • @rawbrawb It's correct that FR4 is hydroscopic, but I don't think it will be an issue in this case for two reasons: 1. Even in tropical jungle conditions it takes many many months for moisture to penetrate more than a fraction of a mm into the material; 2. The alternative solution is a ceramic capacitor that's likely to have +/- 15% tolerance over temperature anyway. – The Photon Jan 10 '13 at 17:13
  • @Kortuk, another issue with using such a large capacitor is that for the high frequencies where you want it to work it's no longer a lumped transistor but a distributed element. .038 m^2 means a 20x20 cm square, which is already 1/2 wavelength at 350 MHz. – The Photon Jan 10 '13 at 17:19
  • @ThePhoton Now if we could easily calculate an inductance... – Kortuk Jan 10 '13 at 17:22
  • @ThePhoton I wanna ask, if the PCB capacitor can work for the low frequency ? I mean, the main frequency sine signal can work on the pcb capacitor ? I thought the pcb capacitor can just work for high frequency like hundred KHz. – alan Jan 17 '13 at 16:08
  • @alan, the only reason built-in capacitors aren't normally effective at low frequencies is that they're practically limited to fairly small values as I showed in the answer. For operations at say 1 kHz and below you often want capacitors in the range of 1 - 100 uF or more, and those aren't easy to build in to the board. – The Photon Jan 17 '13 at 16:19
13

It is possible to build capacitors that way, but you can forget µF. It would most likely be in the pF range.

I think the formula for calculating the capacitance of a plate capacitor would be appropriate here. \$C = \dfrac{\varepsilon A}{d}\$

It will be hard to build a large area on a PCB and you can not make the plate separation arbitrarily small as it will we hard for you to build it that way and you also probably will want it to be able to have some voltage across it.

And yes,this means that you get capacitance on the board from the traces, it's usually not a large value but it matters, especially if you have long traces close to each other and you are running a high frequency.

Gunnish
  • 658
  • 1
  • 8
  • 20
  • 1
    +1... Adding in the epsilon value for some suitable flexible PCB material might help the OP. – Anindo Ghosh Jan 10 '13 at 06:26
  • @Gunnish You might want to add to your comment that because this is possible, you shouldn't make too large areas on your PBC when you do _not_ want pFs. –  Jan 10 '13 at 07:09
8

For a capacitor on a PCB, we need to look at the common formula for a parallel-plate capacitor with an area A, a distance of d between the plates, and a relative permittivity \$\varepsilon_r\$.

\$C=\varepsilon_0\varepsilon_r \frac{A}{d}\$

Let's use some common numbers: Our PCB has an area of 100 mm x 100 mm = 0.01 m2, the thickness of the core is 1.5 mm, and FR4 (a.k.a. "PCB-type epoxy") as an \$\varepsilon_r\$ of approx. 4.2. Thus,

\$C=8.85 \cdot 10^{-12} \frac{\mathrm{F}}{\mathrm{m}} \cdot 4.2 \cdot\frac{0.01\space\mathrm{m}^2}{0.0015\space\mathrm{m}}\$

\$C=248\space\mathrm{pF}\$

Even if we used a thinner dielectric (FR4 core), and maybe even a multilayer board for more than two plates, getting towards nF will be big, and we are far from getting into the µF range.

However, you can use some capacitors on the edges of your board, and distribute their voltage across the board using two copper planes acting as a capacitor. The discrete capacitors paralleled with your PCB capacitor can act as one nearly perfect lumped capacitor, giving your fast logic or power design the warm fuzzies.

You won't use a PCB capacitor if you need exact or big values, but you can use it to create a really good power distribution system on your complete design.

zebonaut
  • 18,170
  • 4
  • 60
  • 104
4

A more esoteric form of capacitor uses fringing fields and lays both electrodes out on both layers in a intertwined fractal pattern. There is no closed form solution and it is very manufacturing tolerance sensitive, so practically useless in this case. Boost in capacitance would be in the range of 4X to 5X. Just mentioned for completeness. NOT at all advised.

placeholder
  • 29,982
  • 10
  • 63
  • 110
  • 8
    Do you have a reference for that? It might be useful in the design of capacitive sensors. – drxzcl Jan 10 '13 at 12:14
  • I am not sure I can bite on that. On a down side, you will blow up your inductance, but if you are chopping up a completed ground plane with cuts you are going to reduce your ground plane, not increase it. – Kortuk Jan 15 '13 at 18:42
  • 3
    @drxzcl He's most likely referring to *fractal capacitors*. They are useful in the sense that vertical spacing of plates don't scale as well as lateral spacing, so fractal capacitors take advantage as much as possible of small lateral spacing. See http://www.chic.caltech.edu/Publications/fractal_caps.PDF – apalopohapa Jan 16 '13 at 00:29
  • 2
    That's actually very good for me! I've got a design that relies on treating the sensor as a co-planar capacitor, but I found it challenging to get enough capacitance out of it. This might be just what I need. See also http://zerocharactersleft.blogspot.nl/2011/11/pcb-as-capacitive-soil-moisture-sensor.html – drxzcl Jan 16 '13 at 18:51
3

As an experiment last year I attempted to build a capacitor by wrapping sheets of aluminum foil separated by a sheet of paper around a roll a few times. I think I only got something around 20 nF or so. Very little. Would be hard to get anywhere near that on a pcb since I was using relatively huge sheets of Al.

JSideris
  • 619
  • 1
  • 7
  • 13
2

I have been building double side cap's with "double sided PCB" boards for awhile. I range about 30-150 pf. I always coat the pCB on the surface and edges to help increase the voltage breakdown capability. I would NEVER subject them to more than a few hundred volts, because at RF frequencies, they can get pretty hot!! I use them in trap-coils for antennas, and if properly designed can handle up to about 300w(PEP) with no problems. I doubt the could handle much more than that. I sure wouldn't give them any guarantee to work at those levels. I use them in trapped antennas at my QTH and on radio outings, but we are always at "barefoot" power levels.

cheers Noted the data a little late> apologize if this is not what was expected.

dave
  • 21
  • 1
2

I am often using this method for high reactive power high frequency systems. However, I want to warn, the "normal" PCB material like FR4 glass-fiber textolite acts not as expected. It has tan(fi) around 0.035 what means that in my constructions the 100 pF tank capacitor at 4 kV and 10 Amp of 100 MHz gets "little bit" hot.... In the first seconds 200 C and after minute 400 C.

Some time I tried glue-on the radiators both sides, tried to immerse it into coolant etc. Logically it’s not nice at all. The Infrared photo shown the uniform indeed temperature field by a surface, with no any altered patch around wire sticking, thus sure for the reason stand dielectric heating and not a Foucault effect in copper.

The ultimate solution I found in my case, was Rogers Inc. produced (at Belgium manufacture) teflon-based PCB, which (there are different materials, I give number for the best of) has tan(fi)=0,0003. The difference is worth the money, indeed. And sure this capacitor is muuuuch cheaper as Vishay of kVAR series or Jennings etc.

Secondly: Often the "Tesla coils people" needs stuff like 40 kV caps, and they are working at so low as kHz range frequencies, thus the dielectric heating is not so important for them. Then there is nothing better than floor carpet PVC tiles, the semi-hard type in roulons, about 2...3 mm thick. Put two copper folia between and roll in into the "sausage". This material "as it is" may persist up to 40 kV or at extreme 50, and it has epsilon between 2.7 and 3.3 with dissipation factor between 0.006 and 0.017. Thus, except that copper may "walk" slightly or form an air-pockets, the PVC ought be seen as much better material for capacitors in comparison with glass-fiber-epoxy PCB.

3) I read here about one's trials about paper. It stays written that figures on paper products: cellophane film: e=6.7...7.6 and tan=0.065...0.01, paper fibers 6.5 and 0.005; kraft tissue 1.8 and 0.001-0.0015; rag-cotton tissue 1.7 and 0.0008-0.0065; pressboard 3.2 and 0.008. For the case of impregnated sorts of paper, logically, the impregnating chemical makes a main impact. Thus, the paper is rather lossy material, however even it acts better than PCB.

winny
  • 13,064
  • 6
  • 46
  • 63
Janis Blahins
  • 51
  • 2
  • 1
    4 year old question but an interesting read. I cleared up some language and formatting for you. However, what is "roulons"? – winny May 31 '17 at 08:47
2

Is this possible? YES!

If I take your question verbatim and literally, you can build caps of that magnitude on PCB of a very large size. I don't know the equation of calculating the PCB size but I assume it would be fairly larger than the cost of the capacitor that you want to build on the PCB.

Chetan Bhargava
  • 4,612
  • 5
  • 27
  • 40