Analog filtering for very low frequency signals

[Alessandro Convertino]

Hi to all,
I would like to know some informations about bandpass filtering on
frequency range between 0.05-30 Hz. In the past I've used a Sallen-key
active filter, and I would like to find a more performing solution.
What IC can I use to implement it?

Thanks for your attention.

Alessandro

 

[Jim Thompson]

Hi to all,
I would like to know some informations about bandpass filtering on
frequency range between 0.05-30 Hz. In the past I've used a Sallen-key
active filter, and I would like to find a more performing solution.
What IC can I use to implement it?

Thanks for your attention.

Alessandro

VERY LOW FREQUENCY filters are difficult since the dissipation factor
of typical capacitors dominates (and deteriorates) the performance,
particularly in a production environment where repeatability is
important.

For a way to swamp out dissipation effects, look over
"StateVariableFilter(P+1).pdf" on the SED/Schematics page of my
website.

I used this method for telephone DTMF filters in the early '70s, using
thick film hybrid technology for the production.

I even devised a probing and adjusting scheme that turned the circuit
into an oscillator (AGC'd) at the desired pole frequency, so that pole
location and Q could be adjusted independently... and automatically,
using sand blasting ;-)

...Jim Thompson

 

[Ken Smith]

Hi to all,
I would like to know some informations about bandpass filtering on
frequency range between 0.05-30 Hz. In the past I've used a Sallen-key
active filter, and I would like to find a more performing solution.
What IC can I use to implement it?

How good does it have to be and are you making one or a million?

Making a low noise high performance analog filter near 0.05 is next to
imposible. The capacitors you need to use are large and costly. If this
is just a one-off or cost is no object, There are 2 and 3 op-amp circuits
you may want to consider.

If the cost is an object, I'd suggest you try to find a way to not need
the filter. A micro controller with an ADC and DAC could implement a very
sharp filter, but there you have the LSB rattling issue to deal with.

If I was going with the micro method I'd do this:

If needed add some high frequency noise to the signal to ensure that the
noise is at least 2 LSBs RMS. Run the ADC circuit at as high of a
frequency as you can. Construct a multi-stage IIR filter in software.
Pump the output values to the DAC much faster than the Nyquist. Add a
high frequency dithering (PWM to fake more bits) to the data for the DAC
that is large enough to blur out the dif-nonlinearities. Follow the DAC
with a low pass filter to remove the dithering.

 

[Ban]

Hi to all,
I would like to know some informations about bandpass filtering on
frequency range between 0.05-30 Hz. In the past I've used a Sallen-key
active filter, and I would like to find a more performing solution.
What IC can I use to implement it?

Thanks for your attention.

Alessandro

With this broad range you can only cascade a highpass with a lowpass filter.
With two filters of second order with maybe Bessel-characteristic you will
get a very good performance with very small overshoot and constant delay in
the passband.
It doesn't matter which topology you implement the filters, because the
performance will be the same. Sallen-Key is indeed the most suitable
topology, but keep resistors low and (unfortunately) capacitors high. If you
need steeper filter characteristic, you will have to increase the filter
order.
It is very important to choose the right opamps for this low frequency task.
I recommend the AD797 or if it has to be cheap the NE5532. Fet- and CMos
opamps are unsuitable because of their low-frequency unstability. Use caps
of at least 1000uF for the high-pass. I think the noise performance will be
the most important issue, as well as THD. But it all depends on your
application.

 

[Joerg]

Hello Alessandro,

This could be nicely done with a DSP. ADC and DAC for data conversion
are cheap and plentiful, and so are DSP.

A microcontroller solution might work as well if it has to be rock
bottom in cost. Some MSP430 versions contain a 16 bit multiplier.
However, it can be a hassle to pipe the data in and out through the ports.

Ciao, Joerg

 

[Jim Thompson]

With this broad range you can only cascade a highpass with a lowpass filter.

"...only..."?? Not so.

[snip]

Sallen-Key is indeed the most suitable
topology, but keep resistors low and (unfortunately) capacitors high.

[snip]

Sallen-Key sucks... why do you declare it "most suitable topology"?

...Jim Thompson

 

[John Larkin]

VERY LOW FREQUENCY filters are difficult since the dissipation factor
of typical capacitors dominates (and deteriorates) the performance,
particularly in a production environment where repeatability is
important.

How so? I'd suspect a polycarb or something decent would have a very
low df in the sub-Hz range. Numbers like 1 uF and 10 megohms seem
perfectly nice to me, with fet opamps.

John

 

[John Larkin]

Making a low noise high performance analog filter near 0.05 is next to
imposible. The capacitors you need to use are large and costly. If this
is just a one-off or cost is no object, There are 2 and 3 op-amp circuits
you may want to consider.

0.05 Hz is an omega of 0.3, so 1 uF and 3 megs is the ballpark RC. Or
0.33 uF and 10M. Those are perfectly happy numbers to work around a
fet opamp. Sounds easy to me.

John

 

[John Larkin]

the passband.
It doesn't matter which topology you implement the filters, because the
performance will be the same. Sallen-Key is indeed the most suitable
topology, but keep resistors low and (unfortunately) capacitors high.
Why?


Fet- and CMos
opamps are unsuitable because of their low-frequency unstability.

What? Fet opamps don't work at DC?

John

 

[Jim Thompson]

How so? I'd suspect a polycarb or something decent would have a very
low df in the sub-Hz range. Numbers like 1 uF and 10 megohms seem
perfectly nice to me, with fet opamps.

John

John,

What kind of df do you get with polycarb caps?

My experience is primarily with (NPO) chip caps for hybrids, which is
why I utilized the "p+1" method... then you can tolerate horrible df
before it even tweaks you a dB.

...Jim Thompson

 

[John Larkin]

John,

What kind of df do you get with polycarb caps?

My experience is primarily with (NPO) chip caps for hybrids, which is
why I utilized the "p+1" method... then you can tolerate horrible df
before it even tweaks you a dB.

...Jim Thompson

I've used them in sample-hold and precision integrator circuits, and
the DFs are tiny, below 0.1% I'm guessing. Polypropylene is good too,
polysulfone (hard to find nowadays) is better. Mylar sucks.

Larger ceramics are terrible, and even NPOs are mediocre. The biggest
NPO I've ever seen was only 22 nF.

Right, there are filter designs that adjust for lossy caps.

John

 

[Terry Given]

I've used them in sample-hold and precision integrator circuits, and
the DFs are tiny, below 0.1% I'm guessing. Polypropylene is good too,
polysulfone (hard to find nowadays) is better. Mylar sucks.

Larger ceramics are terrible, and even NPOs are mediocre. The biggest
NPO I've ever seen was only 22 nF.

Right, there are filter designs that adjust for lossy caps.

John

You can buy much bigger NPO caps (say 1uF 200V) but they cost $$$$$ and
are made in very low volumes.

Cheers
Terry

 

[Terry Given]

John,

What kind of df do you get with polycarb caps?

My experience is primarily with (NPO) chip caps for hybrids, which is
why I utilized the "p+1" method... then you can tolerate horrible df
before it even tweaks you a dB.

...Jim Thompson

Hi Jim,

whats the "p+1" method?

Cheers
Terry

 

[James Meyer]

With this broad range you can only cascade a highpass with a lowpass filter.

"...only..."?? Not so.

[snip]

Sallen-Key is indeed the most suitable
topology, but keep resistors low and (unfortunately) capacitors high.

[snip]

Sallen-Key sucks... why do you declare it "most suitable topology"?

...Jim Thompson

If all you have in your toolbox is a hammer, every problem begins to
look like a nail.

Jim "The other one."

 

[Tim Wescott]

Hello Alessandro,

This could be nicely done with a DSP. ADC and DAC for data conversion
are cheap and plentiful, and so are DSP.

A microcontroller solution might work as well if it has to be rock
bottom in cost. Some MSP430 versions contain a 16 bit multiplier.
However, it can be a hassle to pipe the data in and out through the ports.

Ciao, Joerg

Sampling at 150Hz will make the anti-aliasing easy, and even an 8051
ought to be able to make a pretty credible filter at that sample rate.
There are some very good 24-bit Sigma-Delta ADCs that have output sample
rates that fast, and input sample rates high enough that you anti-alias
with any R and C you find kicking around on the floor.

You still need an anti-aliasing filter, though. The advantage to doing
it with digital hardware is your frequencies are set by the uP
oscillator frequency, which will be pretty darn stable.

 

[Jim Thompson]

With this broad range you can only cascade a highpass with a lowpass filter.

"...only..."?? Not so.

[snip]

Sallen-Key is indeed the most suitable
topology, but keep resistors low and (unfortunately) capacitors high.

[snip]

Sallen-Key sucks... why do you declare it "most suitable topology"?

...Jim Thompson

If all you have in your toolbox is a hammer, every problem begins to
look like a nail.

Jim "The other one."

ROTFLMAO!

...Jim Thompson

 

[Jim Thompson]

Jim Thompson wrote: [snip]

My experience is primarily with (NPO) chip caps for hybrids, which is
why I utilized the "p+1" method... then you can tolerate horrible df
before it even tweaks you a dB.

...Jim Thompson

Hi Jim,

whats the "p+1" method?

Cheers
Terry

For a way to swamp out dissipation effects, look over
"StateVariableFilter(P+1).pdf" on the SED/Schematics page of my
website.

...Jim Thompson

 

[Joerg]

Hello Alessandro,

In case all the other methods are not to your liking here is another: Switched capacitor filters. They are not so much 'en vogue' anymore these days since everyone uses DSP but I believe LTC and National still offer these chips. In case this is for series production you'd have to make sure the chips stay around for a while though.

That gets you around most of the large capacitor tolerance issues.

Regards, Joerg

 

[john jardine]

With this broad range you can only cascade a highpass with a lowpass filter.
With two filters of second order with maybe Bessel-characteristic you will
get a very good performance with very small overshoot and constant delay in
the passband.
It doesn't matter which topology you implement the filters, because the
performance will be the same. Sallen-Key is indeed the most suitable
topology, but keep resistors low and (unfortunately) capacitors high. If you
need steeper filter characteristic, you will have to increase the filter
order.
It is very important to choose the right opamps for this low frequency task.
I recommend the AD797 or if it has to be cheap the NE5532. Fet- and CMos
opamps are unsuitable because of their low-frequency unstability. Use caps
of at least 1000uF for the high-pass. I think the noise performance will be
the most important issue, as well as THD. But it all depends on your
application.

(My Krohn-Hite 3342 switched, lo-pass hi-pass, lab filter goes down to
0.001Hz. Biggest caps in there seem 10uF polyprops, using discrete semis!)
regards
john

 

[Joerg]

Hi Tim,

Sampling at 150Hz will make the anti-aliasing easy, and even an 8051
ought to be able to make a pretty credible filter at that sample rate.
There are some very good 24-bit Sigma-Delta ADCs that have output
sample rates that fast, and input sample rates high enough that you
anti-alias with any R and C you find kicking around on the floor.

You still need an anti-aliasing filter, though. The advantage to
doing it with digital hardware is your frequencies are set by the uP
oscillator frequency, which will be pretty darn stable.

The only chore would be to pipe the data in and out. With a typical
limitation of two or three 8bit ports where some pins need to be shared
for JTAG programming it almost boils down to serial interfacing, or at
least muxed data tranfer.

BTW, I think your new paper on your web site about block diagrams in
control systems is great.

Regards, Joerg