A high pass filter question

[SolarSunrise]

Hello there! I have a really silly and stupid question. Here we go:
I'm making a high pass filter with a corner frequency of around 0.7hz. After a simple calculation, I decided to go with 1MEG resistor and 0.22uF cap. However, I couldn't find any surface mount 0.22uF cap in a small package, and the price was monstrous - $1.4!
So I thought, why not use a 2MEG resistor and a 0.1uF cap? There are plenty of 0.1uF PPS film caps out there... Does this make the filter more vulnerable to noise? Since I am building a high precision circuit (dealing with 1uV here), is this is a dumb and hopeless idea or is it okay? If it is not good at all, can someone give me an alternative? Thanks!

[KJDS]

What are your source and load impedances?

[Bloch]

What are your source and load impedances?

Maybe even a schematic ?

[SolarSunrise]

What are your source and load impedances?

I'm not sure about impedance but the source impedance is coming out of an opamp (unity gain) and the load impedance is an input to a 100 gain op amp. The datasheet says that the opamp's typical input bias current is +/- 200fA (The opamp is LMP7704 quad, which I intentionally chose for its high input impedance). Also, there is a 1 meter of unshielded wire in between the two opamps, which might induce some noise.

[zapta]

... If it is not good at all, can someone give me an alternative? Thanks!

Have two 0.1 caps in parallel?

[Jon86]

What are your source and load impedances?

I'm not sure about impedance but the source impedance is coming out of an opamp (unity gain) and the load impedance is an input to a 100 gain op amp. The datasheet says that the opamp's typical input bias current is +/- 200fA (The opamp is LMP7704 quad, which I intentionally chose for its high input impedance). Also, there is a 1 meter of unshielded wire in between the two opamps, which might induce some noise.

A meter of unshielded cable? with uV precision? You might need to get your cable shielded. I think you're going to have so much noise you're not even going to have a signal that you can make sense of, especially with a 2M resistor.

[SolarSunrise]

What are your source and load impedances?

I'm not sure about impedance but the source impedance is coming out of an opamp (unity gain) and the load impedance is an input to a 100 gain op amp. The datasheet says that the opamp's typical input bias current is +/- 200fA (The opamp is LMP7704 quad, which I intentionally chose for its high input impedance). Also, there is a 1 meter of unshielded wire in between the two opamps, which might induce some noise.

A meter of unshielded cable? with uV precision? You might need to get your cable shielded. I think you're going to have so much noise you're not even going to have a signal that you can make sense of, especially with a 2M resistor.

I'm sorry for the lack of details regarding my project! The reason I need to measure in the micro volt range is because I'm building an EEG amplifier. If my understanding is correct, only passive electrodes (silver electrodes with conductive saline gels) need a shielded cable, like you've said. However for active electrodes, a voltage buffer/follower is directly attached to the electrode itself thus the signal coming off of the electrode has very low impedance. My design utilizes an active electrode, therefore this is the reason why I decided against using shielded cables. This is a good thing considering shielded cables are not money-friendly  .

Also I decided not to use 2M resistor. I would rather sacrifice the frequency bandwidth of my EEG amplifier rather than degrading the whole system's performance. The question still lives on though. Does it still work?

[FrankBuss]

However, I couldn't find any surface mount 0.22uF cap in a small package, and the price was monstrous - $1.4!

Which shop sells it for $1.40? Try "ceramic capacitors" at Digikey, filter for "in stock" and RoHS/Lead free, if you need it:

http://www.digikey.com/product-search/en/capacitors/ceramic-capacitors/131083?stock=1&pbfree=1&rohs=1

Then set "quantity" to 1, filter for 0.22uF and sort for price. 0402 packages are really small, but this is a nice on as 0603:

http://www.digikey.com/product-detail/en/C1608X7R1C224K080AC/445-1318-1-ND/567690

$0.10 each and $0.066 if you buy 10 of it.

[SolarSunrise]

However, I couldn't find any surface mount 0.22uF cap in a small package, and the price was monstrous - $1.4!

Which shop sells it for $1.40? Try "ceramic capacitors" at Digikey, filter for "in stock" and RoHS/Lead free, if you need it:

http://www.digikey.com/product-search/en/capacitors/ceramic-capacitors/131083?stock=1&pbfree=1&rohs=1

Then set "quantity" to 1, filter for 0.22uF and sort for price. 0402 packages are really small, but this is a nice on as 0603:

http://www.digikey.com/product-detail/en/C1608X7R1C224K080AC/445-1318-1-ND/567690

$0.10 each and $0.066 if you buy 10 of it.

Can I use Ceramic caps as signal processing filter? Wouldn't it degrade & induce noise to the signal? If not, I'll definitely check them out! Thanks!

[JackOfVA]

However, I couldn't find any surface mount 0.22uF cap in a small package, and the price was monstrous - $1.4!

Which shop sells it for $1.40? Try "ceramic capacitors" at Digikey, filter for "in stock" and RoHS/Lead free, if you need it:

http://www.digikey.com/product-search/en/capacitors/ceramic-capacitors/131083?stock=1&pbfree=1&rohs=1

Then set "quantity" to 1, filter for 0.22uF and sort for price. 0402 packages are really small, but this is a nice on as 0603:

http://www.digikey.com/product-detail/en/C1608X7R1C224K080AC/445-1318-1-ND/567690

$0.10 each and $0.066 if you buy 10 of it.

That is an X7R dielectric part and it will have a capacitance versus voltage curve that makes it somewhere between undesirable and unusable in a filter application depending on topology. It will also not have good temperature stability, but that might be less important in an indoor environment.

If a ceramic capacitor is to be used in a filter, a C0G/NP0 is the part of choice. Very little to nil change in capacitance versus voltage and excellent temperature stability. A pair of 0.1uF C0G parts in parallel along with a 0.02 would do the job. I know the 0.1uF C0G ceramic is available in 1206 footprint as I use that in some of my products, have not looked at smaller footprint parts, however.

With a corner frequency of 0.7 Hz, it's probably necessary to pay attention to dielectric soakage as well.

[FrankBuss]

Can I use Ceramic caps as signal processing filter? Wouldn't it degrade & induce noise to the signal? If not, I'll definitely check them out! Thanks!

Yes, ceramic capacitors should work for your application as a high pass filter. But why do you need it at all? I would filter any slow changes or DC just in software, much easier.

How does your circuit look like? Amplifying microvolts is not easy. There is an open source project for an EEG amplifier, this is the input stage:

http://openeeg.sourceforge.net/doc/modeeg/modEEGamp-v1.0.png

It uses instrumentation amplifiers ( http://en.wikipedia.org/wiki/Instrumentation_amplifier ) and a low pass filters for the inputs, which is more important, because the cable is a very good antenna. And of course, you should always use batteries to operate your device and optocouplers, if you connect it to a PC.

That is an X7R dielectric part and it will have a capacitance versus voltage curve that makes it somewhere between undesirable and unusable in a filter application depending on topology. It will also not have good temperature stability, but that might be less important in an indoor environment.

The capacitance vs. voltage curve doesn't matter for this application. See for example this X7R description, page 14:

http://www.ece.ucdavis.edu/vcl/asap/asap_v1/docs/X7R_C.pdf

Capacitance will change max. 10% and this is for full voltage rating. My example was rated for 16V, but it is only used for microvolts (there is a simple voltage follower at the side of the active electrodes). I would expect that it is very stable and would be good for a high pass filter.

Temperature is maybe just 5%, for room temperature in winter and summer. But it would be not important anyway for an EEG filter, if your cutoff frequency is 0.7 Hz or 0.6 Hz.

But the filter will introduce some noise. There is a nice calculator for the thermal resistor noise:

http://www.daycounter.com/Calculators/Thermal-Noise-Calculator.phtml

At 20°C temperature and 10 Hz bandwidth, a 1 megohm resistor will create 0.4 uV noise. Might be in the range of the EEG signal.

[Jon86]

What are your source and load impedances?

I'm not sure about impedance but the source impedance is coming out of an opamp (unity gain) and the load impedance is an input to a 100 gain op amp. The datasheet says that the opamp's typical input bias current is +/- 200fA (The opamp is LMP7704 quad, which I intentionally chose for its high input impedance). Also, there is a 1 meter of unshielded wire in between the two opamps, which might induce some noise.

A meter of unshielded cable? with uV precision? You might need to get your cable shielded. I think you're going to have so much noise you're not even going to have a signal that you can make sense of, especially with a 2M resistor.

I'm sorry for the lack of details regarding my project! The reason I need to measure in the micro volt range is because I'm building an EEG amplifier. If my understanding is correct, only passive electrodes (silver electrodes with conductive saline gels) need a shielded cable, like you've said. However for active electrodes, a voltage buffer/follower is directly attached to the electrode itself thus the signal coming off of the electrode has very low impedance. My design utilizes an active electrode, therefore this is the reason why I decided against using shielded cables. This is a good thing considering shielded cables are not money-friendly  .

Also I decided not to use 2M resistor. I would rather sacrifice the frequency bandwidth of my EEG amplifier rather than degrading the whole system's performance. The question still lives on though. Does it still work?

Transmitting low-voltage signals like that across a large distance in a high impedance circuit is probably not going to work that well with unshielded cable. Try it, hook it up to a scope and see if the output signal looks like what you need. Then add in some shielding if necessary.

[JackOfVA]

Can I use Ceramic caps as signal processing filter? Wouldn't it degrade & induce noise to the signal? If not, I'll definitely check them out! Thanks!

Yes, ceramic capacitors should work for your application as a high pass filter. But why do you need it at all? I would filter any slow changes or DC just in software, much easier.

How does your circuit look like? Amplifying microvolts is not easy. There is an open source project for an EEG amplifier, this is the input stage:

http://openeeg.sourceforge.net/doc/modeeg/modEEGamp-v1.0.png

It uses instrumentation amplifiers ( http://en.wikipedia.org/wiki/Instrumentation_amplifier ) and a low pass filters for the inputs, which is more important, because the cable is a very good antenna. And of course, you should always use batteries to operate your device and optocouplers, if you connect it to a PC.

That is an X7R dielectric part and it will have a capacitance versus voltage curve that makes it somewhere between undesirable and unusable in a filter application depending on topology. It will also not have good temperature stability, but that might be less important in an indoor environment.

The capacitance vs. voltage curve doesn't matter for this application. See for example this X7R description, page 14:

http://www.ece.ucdavis.edu/vcl/asap/asap_v1/docs/X7R_C.pdf

Capacitance will change max. 10% and this is for full voltage rating. My example was rated for 16V, but it is only used for microvolts (there is a simple voltage follower at the side of the active electrodes). I would expect that it is very stable and would be good for a high pass filter.

Temperature is maybe just 5%, for room temperature in winter and summer. But it would be not important anyway for an EEG filter, if your cutoff frequency is 0.7 Hz or 0.6 Hz.

But the filter will introduce some noise. There is a nice calculator for the thermal resistor noise:

http://www.daycounter.com/Calculators/Thermal-Noise-Calculator.phtml

At 20°C temperature and 10 Hz bandwidth, a 1 megohm resistor will create 0.4 uV noise. Might be in the range of the EEG signal.

I can't locate a C versus V plot on the data sheet you linked to, but the plot below is one I made for a couple of 1uF 1206 X7R capacitors, one rated at 25V and one at 50V. Data taken with an HP4192A impedance meter and internal bias voltage generator.  The bias cycles so as to yield the classic "butterfly" C versus V response.

In some limited cases, where one does not care about distortion, linearity, accuracy and low Q / high D for the capacitor, an X7R dielectric might be suitable. It might well work with zero bias voltage and a microvolt signal level, but I do not wish to leave the impression that an X7R capacitor makes sense for filters except in extremely low performance applications.

I would draw your attention to a couple of points in the plot. First, note that there is significant capacitance change with applied voltage even about the 0 volt point, although it is somewhat flatter there than other parts of the C versus V curve.

Second, note the hysteresis exhibited in the plot, due at least in part to stored charge. Not a good thing for a filter that will be working at low frequencies, to say the least.

In addition, X7R capacitors are well known to exhibit piezoelectric effect where applied voltage causes a change in mechanical dimensions and, of more importance in a filter application, where board flexing will cause an induced voltage in the capacitor. And, the board flexing may be something as trivial as nearby fan coupling or even air pressure changes from sound waves at some distance.  This will also be a problem at the microvolt level.

[SolarSunrise]

Thanks for everyone who has given me some valuable info! So everything boils down to this: Does a high pass filter with X7R 0.2uF cap and 1MEG produce more noise than 2MEG with PPS film 0.1uF cap? Both will have the same 0.7hz cut off freq, so I was wondering which one to choose...

Oh, by the way, happy 2014!

[FrankBuss]

I can't locate a C versus V plot on the data sheet you linked to

but the plot below is one I made for a couple of 1uF 1206 X7R capacitors, one rated at 25V and one at 50V. Data taken with an HP4192A impedance meter and internal bias voltage generator.  The bias cycles so as to yield the classic "butterfly" C versus V response.

Thanks, that's interesting. Looks like they are a bit too optimistic (10% in the diagram vs 100% what you measured). Maybe they tested capacitors with lower max rated voltages.

In addition, X7R capacitors are well known to exhibit piezoelectric effect where applied voltage causes a change in mechanical dimensions and, of more importance in a filter application, where board flexing will cause an induced voltage in the capacitor. And, the board flexing may be something as trivial as nearby fan coupling or even air pressure changes from sound waves at some distance.  This will also be a problem at the microvolt level.

Right, this could be a problem. I found another good datasheet with some background:

http://cds.linear.com/docs/en/datasheet/1964fb.pdf

On page 13, figure 4, it shows an oscillogram of the noise when light tapping a ceramic capacitor with a pencil, and it has an amplitude of more than 2 mV.

I still think it should work with a X7R type capacitor for this application, if you don't tap it with a pencil but would be interesting to compare different types of capacitors. So maybe SolarSunrise should buy one of those expensive C0G/NP0 capacitors, and the cheap X7R ones, and maybe another old style through-hole film capacitor (for hobby usage it can be soldered to SMD pads as well), and compare it.

[Hideki]

What makes you think ceramic caps are more noisy than PPS film caps? Thermal capacitor noise goes DOWN with increased capacitance, and resistor noise goes UP with increased resistance, so if that noise is all you care about, then a high capacitance combined with a low resistance is what you want. Ideally much less than 1M.

If possible, I would try to get rid of the filter altogether. If the exact cutoff frequency isn't too important, use smd electrolytic caps. Sure, they are a bit nonlinear as well, but have no piezo effect.

[SolarSunrise]

Thank you everybody for your suggestions! Instead of going with 2 MEG Resistor with 0.1uF PPS cap, I decided to go with 1 MEG resistor with two 0.1uF C0G caps in parallel (total 0.2uF).