Capacitor Materials

[Overclocked]

Whats the Difference between Capacitor Materials? Do each have a advantage or Disadvantage? For example, whats the difference between these 3 capacitors?
**broken link removed**
**broken link removed**
**broken link removed**

For My application, the IC requires certain Capacitors, such as Mica or Ceramic, Polystyrene, Mylar, and polypropylene. If Newark doesn't have one type of cap, can I substitute it with another? Will Performance degrade because of it? Im asking because I noticed they dont have Mylar Caps, nor do they have Polystyrene.

To help you further Help me; The OSC pin requires either Mica or Ceramic with a tolerance of 10%. The Ref pins require Polystyrene or Mylar with a tolerance of 20%, the INT pins and Auto Zero pins both need a polypropylene or Polystyrene or polycarbonate as 2nd and 3rd Choices. The Input pin doesn't say what type of Cap I need.

So thats why Im asking, If Possible, Can I use the "best" type of capacitor instead of all those? It would Make my order a lot more simplier, and probably save me some money. I know for a fact that I can use the Best lowest Tolerance Cap I can find, since in that aspect they are like Resistors. Its Just the Materials they are made of I am confused about.

For my application, Im using ICL7106/07 A/D Converter.

 

[dknguyen]

There is no best type. They all have different characteristics and non-idealities that can REALLY come into play in real life such as piezo electric distortions, temperature drift, capacitance density, equivelant series resistance and equivelant series inductance, and frequency response. These things affect what the capacitor can do.

You might be able to replace one type with another, but it depends what you need in the capacitor and why it is there. Capacitance density aside, for example, ceramic is good at high frequencies but should not be used in areas where a signal passes through it because of piezo electric effects which will distort the signal, rather it should be used for filtering.

Or another example is electrolytic is cheap and has very high densities, but is affected by temperature and has a high inductance and cannot be used for high frequencies, so if you were using a huge tantalum (or pretty much any other type) for very low frequency bypassing, you could replace it with an electrolytic since the electrolytic posesses the same characteristics required for the job. But if a tantalum was being used for medium frequency bypassing, filtering, or some other analog function you couldn't replace it with an electrolytic instead because these applications are making use of the fairly good high frequency characteristics of the tantalum...while the electrolytic has pretty crappy frequency characteristics.

https://www.electro-tech-online.com/custompdfs/2006/07/dielectr.pdf

I would recommend you Google and compare the particular characteristics of the capacitor type you need and the type you want to replace it with.

Ceramic disc (traditional) and multi-layer are the same in most aspects except multi-layer uses multiple dieelectric layers and can get much higher capacitance densities into one package. ESR and ESI are generally pretty dependent on the packaging, and since multi-layer can cram more into one package, they tend to have a lower ESR and ESI for their capacitance than discs.

I forget what it is for Mica...Google!

 

[Overclocked]

Well I'll be Dealing with DC as an Input Voltage, and I think somewhere in the Data sheet I read an OSC freq. of 40Khz. Im guessing the other caps might deal with the same signals (either DC or some freq not higher than 40Khz)

There's a note at the bottom, The Integrating Cap must have good dielectric absorption characteristics for the A/D Converter to have optimum linearity.

 

[dknguyen]

Your signal caps might be 40khz, but filtering caps might be working at higher frequencies to filter out noise.

If I had no idea and it was an average circuit, I'd use ceramics for bypassing and tantalums for any filtering or wherever the signal passes through, and maybe an electrolytic or two for large capacitance bypassing. But of course, there is a point where this breaks down (like really high frequencies where tantalums can't handle it). I think MIca and polyseter and such handle high frequencies better than tantalum, but tantalum has higher capacitance density. So yeah...it can dependd on how much capacitance you need vs the frequency you are working at.

The right tantalum dielectric (there are several) is pretty temperature stable (and a couple are ridiculously temperature stable). But then again, so are Micas and stuff. Search up on tantalum dieelectrics to find out more.

or you can look at the technical and application notes on www.AVX.com

It can get pretty complicated (it does), but I think electrolytics start breakin down at 40kHz so I wouldn't use at all them in this case unless you needed one really big cap somewhere to low-frequency bypass the entire board.

 

[hjames]

I wish I hadn't given away my "Art of Electronics" book - there was a table with capacitor materials in it and stuff like loss tangents, tempcoef's etc.

From what I remember, all the film types - polypropylene, polycarbonate are the closest to the ideal capactitors - they're also big/low capacitances, and have some manufacturing drawbacks. For capacitors that are part of an ADC integrator (anything connected to the INT line) you *need* to have a capacitor that is as close to the ideal as possible. As for the other stuff
[edit - apparently polyester(same as mylar) aren't so hot...]


Keywords: "Dielectric absorbtion" - charge goes into the capacitor and ends up modifying the dielectric - if you short an electrolytic cap, and measure the voltage immediately afterwards, you might see it *rising*. This is the dielectric "reverting" to it's original state and putting charge back onto the plates.

"Loss tangent/loss factor" - AC voltage goes in, some energy ends up being dissipated in the capacitor itself.

When dealing with high frequency (oscillation as well as bypassing high frequency circuits), the other issue that comes up is inductance - some film and aluminum electrolytic capacitors are essentially long spools of metal which act as inductors - which won't play well with high frequencies.

There's also the accuracy issues as well as temperature coeffecient issues (sometimes circuits are designed so that certain tempco's balance out and end up stabalizing the circuit over temperature).
Here's a chart with some info
**broken link removed**

And here's a pretty good summary of things - although I wish it were in chart form...
http://www.musicsynthesizer.com/txt/caps2.txt

My take: use polyester for everything connected to the chip except for the osc and bypass lines - use some cheapish (X7R or better grade) ceramic parts for them.

Wikipedia also looks pretty useful: **broken link removed**)

 

[Overclocked]

Some more Info I picked up; ha for weeks I studied the data sheet and didnt think twice about reading these sections. I was off by 8Khz, The OSC req is 48Khz.

Integrating Capacitor
The integrating capacitor should be selected to give the
maximum voltage swing that ensures tolerance buildup will
not saturate the integrator swing (approximately. 0.3V from
either supply). In the ICL7106 or the ICL7107, when the
analog COMMON is used as a reference, a nominal +2V fullscale
integrator swing is fine. For the ICL7107 with +5V
supplies and analog COMMON tied to supply ground, a
±3.5V to +4V swing is nominal. For three readings/second
(48kHz clock) nominal values for ClNT are 0.22μF and
0.10μF, respectively. Of course, if different oscillator
frequencies are used, these values should be changed in
inverse proportion to maintain the same output swing.
An additional requirement of the integrating capacitor is that
it must have a low dielectric absorption to prevent roll-over
errors. While other types of capacitors are adequate for this
application, polypropylene capacitors give undetectable
errors at reasonable cost.

Auto-Zero Capacitor
The size of the auto-zero capacitor has some influence on
the noise of the system. For 200mV full scale where noise is
very important, a 0.47μF capacitor is recommended. On the
2V scale, a 0.047μF capacitor increases the speed of
recovery from overload and is adequate for noise on this
scale.

Reference Capacitor
A 0.1μF capacitor gives good results in most applications.
However, where a large common mode voltage exists (i.e.,
the REF LO pin is not at analog COMMON) and a 200mV
scale is used, a larger value is required to prevent roll-over
error. Generally 1μF will hold the roll-over error to 0.5 count
in this instance

This circuit will also help:

Following the circuit, It Seems that The Int Cap, and AutoZero Caps Come from the Analog signal input.

 

[Overclocked]

I found a chart if anyone is intrested
**broken link removed**

 

[Overclocked]

I have one more question, What value cap would I use to eliminate noise from a PSU? The circuit will be powered from a computer PSU, Im worried there will be too much noise. Can I also use a 5V Regulator to also get rid of noise?

 

[dknguyen]

A 5V regulator might supress some low frequency noise, but not high level noise. The capacitor(s) needed depends on the noise present which is dependent on what the circuit is exactly. It's really complicated if you really get into it. In theory, the larger the cap the better, but in reality the series inductance of larger caps is too big and makes them ineffective for higher frequencies. The best way is to get many MANY small capacitors of the SAME value and use all of them to bypass, that way you get large capacitance and low inductance...but it's overkill unless it's a high frequency circuit. Using multiple values of capacitors will spread the bypass frequency range BUT at the same time it makes the bypassing less effective over that frequency range (due to anti-resonance, Google this with bypass capacitors to find out more). Many times this reduction in effectiveness is enough to make large portions of the frequency range ineffective for noise bypassing.

Try 0.1uF...I guess. Most people seem to. if it's a higher operating frequency, then maybe 0.01uF? But smaller caps have to be closer to the thing they are bypassing than larger caps otherwise they don't do anything.