I am designing low and high pass op amp (+/- 15V) based active filters near 1 Hz that require capacitor values between 1 and 10 uF. Are there any reasons to not use surface mount multilayer ceramic chip capacitors in such a design ? Is capacitance change with voltage an issue ? The filters will be used at room temperature with little temperature variation.
Multilayer ceramic chip capacitors are available in values to 10 uF or higher with +/- 10% tolerances and lead spacings of a few millimeters for $1. Is it better to use metalized polypropylene or metalized polyester even though they are more expensive and have much larger lead spacing (up to 1 inch) ? Examples inlcude Panasonic ECQ-E(B) or ECQ-E(F) series and Epcos MKT series, available at 5% or even better tolerance.
MLCC device capacitance varies quite widely with applied voltage. A X5R will typically be at 50% nominal capacitance at 90% rated voltage. I somewhere have a graph of such things.
My rule for MLCC devices [1] is never to run them above 50% rated voltage and not use anything worse than X5R except where I can live with +20/-80% variations across temperature and -80% across the applied voltage range (typical Y5V).
Capacitance change with voltage will result in distortion of your signal.
There are different types of ceramic capacitor - basically, you can have low capacitance values and high quality, or large values and low quality.
NP0 and C0G are the most stable and will cause the least distortion to your signal (compared to other ceramic types, I'm not sure how they compare to polyester etc.). Unfortunately, these are not usually available in large values, and if they are, they will be physically larger than the other types of ceramic cap of the same value. You could put a few in parallel , but that will obviously increase size and cost.
X7R is not as stable, but you can get higher capacitance than C0G for a fixed size. X5R and X8R are very similar, but the max operating temp is lower for X5R and higher for X8R. If you are not too worried about distortion, these could be ok. I expect that the 10uF +/- 10% caps you mention are X5R or X7R.
Y5V is not very stable but does offer large values in small sizes. I would NOT use these in a filter (or anything else).
It is the usual compromise between size, cost and performance.
--
-----------------------------------------------------------------------
To reply to me directly:
Replace privacy.net with: totalise DOT co DOT uk and replace me with
gareth.harris
If the tolerance is good enough, it will probably work, The very high K types (Y5V, Z5U, F) will change capacitance a lot with both temperature and voltage, but show this as a wide tolerance. The X5R and X7R are pretty good, especially if you can find units that handle a bit more voltage than your circuit will apply.
Read the data sheet and look for graphs of capacitance versus both voltage and temperature.
Take a look at the Panasonic ECJ-3YB1E106K, 10 uF, 1206 size, X5R, 25 V 10%, $0.55 each from Digikey.
Can you use higher resistor values, so you can use lower capacitor values? If so, C0G ceramics may become possible. You can get them in excess of 0.1uF these days. They tend to be pricey.
Beware that capacitors with dielectric absorption will add to the non-ideal behaviour at low frequencies. AFAIK, C0G are low dielectric absorption, as are polypropylene. I came across another dielectric that I'm not remembering right now that's also good--not one of the "usual suspects." Mylar/polyester are not particularly good. The non-ideal behaviour in this case is that the capacitors will look smaller in value to high frequencies than to low; it's still generally a linear effect, at least, unlike the voltage-dependence of the high-K ceramics.
In the distant past, I inherited a partially-done (barely begun) design requiring some low-frequency filtering, four channels of 8 poles each on a business-card size board, and the filters had to have selectable cutoff frequencies. I soon discovered that the required parts would NOT fit, and the performance using those parts would be disappointing at best. Instead, I ended up digitizing the input and passing it through a small DSP, and back out to ADCs. That was before the days of codecs; now the parts are so good and small, it seems it would be a very reasonable way to do it if you want to keep things small. Power might be an issue for you, though; narrow bandwidth op amps are pretty low power compared with DSPs.
I am using Texas Instrument's Filter Pro software and designing a 4 pole high pass MFB filter. With R1 seed at 100 k, it generates capacitance values of 15 uF. and feedback R1 values of 174 k and 412 k. WIth R1 seed as 1 M, the generated capacitance values drop to 1.5 uF and the R1 increases to 1.74 M and 4.12 M. I am using a quad OPA
4277 op amp. Any sugestions on op am parameters to watch out for when dealing with such high feedback resistor values ?
How high a frequency is needed before capacitors with dielectric absorption begin to show lower capacitance ? i.e. If I am working with 1 Hz filter frequencies, are the offending frequencies always so high that they will always be beyond the filter's cutoff frequency even if the capacitance value changes and effects the actual cutoff frequency ?
Except that it isn't true. An X5R capacitor speced for 10% tolerance will have within 10% of its rated capacitance over its voltage and temperature range. Only Y5V or Z5U and other very high K types will show such extreme capacitance variations with applied voltage. And if you want to keep the voltage effects low with the X5R or X7R types, just get capacitors that have at least twice the voltage rating needed. i.e. 25 volt capacitor for 10 volt operation.
A capacitor is tolerance rated separately for capacitance across temperature and across voltage.
I have some really nice Murata, Kemet and Panasonic datasheets and they show clearly the variation of capacitance of the various types across rated voltage.
C0G (as one might expect) does best, with no more than 10% decrease (typically) of capacitance across the working voltage range. Y5V / Z5U do pretty horrendously, with a typical 70% - 90% decrease in capacitance at rated voltage. Note that this is independent of the temperature spec.
I noted that I overspec by a factor of 2 on DCWV for this very reason.
Also you should know that ceramic caps. are piezoelectric devices. They will generate noise when flexed. If your application will be subjected to vibration, noise may be induced into your circuit by flexing of the board on which the ceramics are mounted. If they are lead mounted, not to worry.
I worked on a program that had this problem a number of years ago. The noise from the caps. was sufficient to cause noise in a modulation circuit. The application was shipboard....on a carrier.
(Sorry for the delayed reply...I missed seeing your followup before.)
When you go to higher resistances, voltage noise gets worse. If that's an issue, you'll need to put some amplification ahead of the filter stage, to get the signal far enough above the noise. Op amps with extremely low current noise help: those will generally be JFET-input op amps. There's also the advantage that input bias current is low, so that higher resistance values won't cause too much additional DC offset voltage. Look out, too, for 1/f noise corner frequencies that are too high. If offsets and very low frequency noise are issues for you, you could consider chopper-stabilized CMOS amplifiers, but they get expensive I suppose. In laying out the circuit, work to keep circuit areas small so that noise has less chance to be picked up. Some sort of electrostatic shielding may be in order. You mentioned it's a HPF, so of course, you need to maintain bandwidth out to whatever your application dictates.
Dielectric absorption can be modelled as an ideal capacitor in parallel with several copies of a series R-C, where each C is small compared with the ideal cap, and the R-C time constants of each vary over a range generally from milliseconds to many seconds. So the effect is seen as a gradually changing capacitance versus frequency, if you look at it in the frequency domain.
I'm not sure what topology your filter software is using, but you may find some advantage to using a state variable topology. The integrators can be high impedance, and their low-pass characteristic will keep noise out of the feedback path. The feedback path itself can be kept low impedance, to keep wideband noise down and to make the circuit less sensitive to electrostatic coupling from outside. The higher parts count may be worth it! It will take four integrators to get your four poles: four capacitors, four op amps. Add to that the amps to sum the feedback with the stage's input signal, for two stages.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.