Maker Pro
Maker Pro

Single Supply Salen Key

P

Phil Hobbs

Jan 1, 1970
0
This is for an anti-aliasing filter. I found a good ref here
http://www.eng.yale.edu/ee-labs/morse/compo/sloa058.pdf page 16 which
uses split supply with two resistors as per usual.

How does that compare with say using one of those ICs that generates
a negative supply and using that instead? (so that the op-amp has
+=0=- supply lines from a single positive supply)
That circuit on P. 16 of the app note isn't just a lowpass filter--it's
AC-coupled at the input. That's why it needs its own bias resistors.

A plain-vanilla Sallen-Key lowpass is a DC-coupled noninverting
amplifier with unity DC gain, making it a reasonable choice for
antialiasing. Its output basically just follows its input, so you
shouldn't need a separate source of bias.

You do need to pay attention to what happens to the op amp's output
impedance at high frequency, though. "Rail-to-rail" outputs are
intrinsically high impedance, and rely on feedback to get low
closed-loop output impedance. (LDO regulators have the same issue, for
the same reason.) That means that the op amp's output impedance rises
more or less linearly with frequency, i.e. it looks like a lossy
inductor. At some frequency that will mess up your filter
characteristic, so you need to calculate it fairly carefully.

Some RRIO amps (e.g. the AD8605) use fast internal feedback around the
output stage, which makes them quite a bit stiffer. Those ones make
much better ADC drivers as well as better filters, compared with
normal RRIO amps.

If you use an amp with totem-pole outputs, you'll need that charge pump
to get the output to swing to ground, but with an RRIO with stiff
outputs, you should be able to get away without it.

One very useful tip is to make sure that you have a small positive
offset, i.e. that zero signal gives you an on-scale reading from the ADC.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510 USA
+1 845 480 2058

hobbs at electrooptical dot net
http://electrooptical.net
 
H

Harry Dellamano

Jan 1, 1970
0
Jim Thompson said:
That circuit on P. 16 of the app note isn't just a lowpass filter--it's
AC-coupled at the input. That's why it needs its own bias resistors.

[snip]

Look at p13, non-inverting. Amazing how bad apnotes, datasheets _and_
Spice models have become :-(

...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| San Tan Valley, AZ 85142 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |

I love to cook with wine. Sometimes I even put it in the food.

So, fig. 3 was ok?
Hey Jim, don't take any wooden nickels.

Harry
 
That circuit on P. 16 of the app note isn't just a lowpass filter--it's

AC-coupled at the input. That's why it needs its own bias resistors.



A plain-vanilla Sallen-Key lowpass is a DC-coupled noninverting

amplifier with unity DC gain, making it a reasonable choice for

antialiasing. Its output basically just follows its input, so you

shouldn't need a separate source of bias.



You do need to pay attention to what happens to the op amp's output

impedance at high frequency, though. "Rail-to-rail" outputs are

intrinsically high impedance, and rely on feedback to get low

closed-loop output impedance. (LDO regulators have the same issue, for

the same reason.) That means that the op amp's output impedance rises

more or less linearly with frequency, i.e. it looks like a lossy

inductor. At some frequency that will mess up your filter

characteristic, so you need to calculate it fairly carefully.



Some RRIO amps (e.g. the AD8605) use fast internal feedback around the

output stage, which makes them quite a bit stiffer. Those ones make

much better ADC drivers as well as better filters, compared with

normal RRIO amps.



If you use an amp with totem-pole outputs, you'll need that charge pump

to get the output to swing to ground, but with an RRIO with stiff

outputs, you should be able to get away without it.



One very useful tip is to make sure that you have a small positive

offset, i.e. that zero signal gives you an on-scale reading from the ADC.



Cheers



Phil Hobbs



--

Dr Philip C D Hobbs

Principal Consultant

ElectroOptical Innovations LLC

Optics, Electro-optics, Photonics, Analog Electronics



160 North State Road #203

Briarcliff Manor NY 10510 USA

+1 845 480 2058



hobbs at electrooptical dot net

http://electrooptical.net

Not sure what you are saying here. I take a plain Salen Key low pass - but that needs +0- dual supplies. How do I run it from one supply? Just split the 0v with two resistors? Then I get a bias of V/2 at the output hence the ac coupling at the output to get rid of the dc bias - and at the input as you pointed out.
 
"Jim Thompson" <[email protected]> wrote in
message news:[email protected]...
On Sun, 11 Aug 2013 05:20:02 -0400, Phil Hobbs

On 8/10/2013 11:53 PM, [email protected] wrote:
This is for an anti-aliasing filter. I found a good ref here
http://www.eng.yale.edu/ee-labs/morse/compo/sloa058.pdf page 16 which
uses split supply with two resistors as per usual.

How does that compare with say using one of those ICs that generates
a negative supply and using that instead? (so that the op-amp has
+=0=- supply lines from a single positive supply)

That circuit on P. 16 of the app note isn't just a lowpass filter--it's
AC-coupled at the input. That's why it needs its own bias resistors.


[snip]

Look at p13, non-inverting. Amazing how bad apnotes, datasheets _and_
Spice models have become :-(

...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| San Tan Valley, AZ 85142 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |

I love to cook with wine. Sometimes I even put it in the food.
So, fig. 3 was ok?
Hey Jim, don't take any wooden nickels.



They just used R3 on the wrong opamp.





--



John Larkin Highland Technology Inc

www.highlandtechnology.com jlarkin at highlandtechnology dot com



Precision electronic instrumentation

Picosecond-resolution Digital Delay and Pulse generators

Custom timing and laser controllers

Photonics and fiberoptic TTL data links

VME analog, thermocouple, LVDT, synchro, tachometer

Multichannel arbitrary waveform generators

Ok what about this one in Ficure 12

http://ww1.microchip.com/downloads/en/AppNotes/00682c.pdf
 
P

Phil Hobbs

Jan 1, 1970
0
Not sure what you are saying here. I take a plain Salen Key low pass
- but that needs +0- dual supplies. How do I run it from one supply?
Just split the 0v with two resistors? Then I get a bias of V/2 at
the output hence the ac coupling at the output to get rid of the dc
bias - and at the input as you pointed out.

From the op amp's POV, there's nothing special about single-supply vs
bipolar supplies--all it cares about is that the inputs and outputs stay
in range. (Op amps don't have a pin called 'ground'.) A "single
supply" op amp is one whose input common mode range includes the
negative supply rail, and whose output can swing nearly to the negative
rail as well. That makes it convenient to run from a single positive
supply, but no magic is involved.

Since you're building a filter, it has to get its input from some
earlier stage. I assume this input signal is also unipolar because you
didn't ask about level shifting (e.g. going from a +-5V signal range to
a 0-2.5V digitizer).

The basic Sallen-Key lowpass is a voltage follower at low frequency, so
if its input signal range is centered around 2.5V, say, so will its
output be.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510 USA
+1 845 480 2058

hobbs at electrooptical dot net
http://electrooptical.net
 
M

miso

Jan 1, 1970
0
You are over thinking this. For the Salen and Key g=1, the op amp is
just a follower (buffer). In fact, the wiki does a better job than that
app note.

One cap goes to ground, but any AC ground will work since it is just a
capacitor. Obviously you would use the quietest ground you have.

For any other Salen and Key, you would need the psuedo ground, but not
for g=1.

While Jim is right that the Salen and Key is a shitty topology due to
component sensitivity, it may be suitable for anti-aliasing oversampled
ADCs. That is, the ADC has its own anti-aliasing up to a point, thanks
to the digital filter, but it is still subject to aliasing artifacts at
the Nyquiest frequency of the sigma-delta modulator in the ADC. But
thanks to oversampling, you don't need much of a filter (2 poles is
fine) AND the corner frequency will be in the stop band of the digital
filter. Basically the anti-aliasing filter corner will be far enough
away from the in-band signal that you don't care too much about the
crappy accuracy.

Here is the procedure:
1) Determine the sigma-delta modulator sampling frequency.
2) Divide that frequency by two to get the Nyquist frequency.
3) Pick an anti-aliasing type: Butterworth, Gausian, Bessel, etc. [Note
the 2nd order Bessel doesn't ring if that is your criteria.]
4) Using a table or hacking around with spice, insure that at the
Nyquist frequency, your filter is down (attenuating) sufficient for the
noise floor of the ADC. If you use some sexy 24 bit ADC, that would be
around 110db.
5) Since the Salen Key component sensitivity is horrible, see if you can
live with the error. Alter the components for a worst case. If you are
doing multiple channels, look at phase difference.

If this Salen Key filter is for anything other than an oversampled ADC,
you are just kidding yourself.
 
From the op amp's POV, there's nothing special about single-supply vs

bipolar supplies--all it cares about is that the inputs and outputs stay

in range. (Op amps don't have a pin called 'ground'.) A "single

supply" op amp is one whose input common mode range includes the

negative supply rail, and whose output can swing nearly to the negative

rail as well. That makes it convenient to run from a single positive

supply, but no magic is involved.



Since you're building a filter, it has to get its input from some

earlier stage. I assume this input signal is also unipolar because you

didn't ask about level shifting (e.g. going from a +-5V signal range to

a 0-2.5V digitizer).



The basic Sallen-Key lowpass is a voltage follower at low frequency, so

if its input signal range is centered around 2.5V, say, so will its

output be.



Cheers



Phil Hobbs



--

Dr Philip C D Hobbs

Principal Consultant

ElectroOptical Innovations LLC

Optics, Electro-optics, Photonics, Analog Electronics



160 North State Road #203

Briarcliff Manor NY 10510 USA

+1 845 480 2058



hobbs at electrooptical dot net

http://electrooptical.net

My ADC runs + o - 10v ie bipolar input and runs from a 5v supply. The source
is from a stereo DAT player which I assume has been ac coupled at the output to remove dc.
 
You are over thinking this. For the Salen and Key g=1, the op amp is

just a follower (buffer). In fact, the wiki does a better job than that

app note.





One cap goes to ground, but any AC ground will work since it is just a

capacitor. Obviously you would use the quietest ground you have.



For any other Salen and Key, you would need the psuedo ground, but not

for g=1.



While Jim is right that the Salen and Key is a shitty topology due to

component sensitivity, it may be suitable for anti-aliasing oversampled

ADCs. That is, the ADC has its own anti-aliasing up to a point, thanks

to the digital filter, but it is still subject to aliasing artifacts at

the Nyquiest frequency of the sigma-delta modulator in the ADC. But

thanks to oversampling, you don't need much of a filter (2 poles is

fine) AND the corner frequency will be in the stop band of the digital

filter. Basically the anti-aliasing filter corner will be far enough

away from the in-band signal that you don't care too much about the

crappy accuracy.



Here is the procedure:

1) Determine the sigma-delta modulator sampling frequency.

2) Divide that frequency by two to get the Nyquist frequency.

3) Pick an anti-aliasing type: Butterworth, Gausian, Bessel, etc. [Note

the 2nd order Bessel doesn't ring if that is your criteria.]

4) Using a table or hacking around with spice, insure that at the

Nyquist frequency, your filter is down (attenuating) sufficient for the

noise floor of the ADC. If you use some sexy 24 bit ADC, that would be

around 110db.

5) Since the Salen Key component sensitivity is horrible, see if you can

live with the error. Alter the components for a worst case. If you are

doing multiple channels, look at phase difference.



If this Salen Key filter is for anything other than an oversampled ADC,

you are just kidding yourself.

Don't have sigma-delta adc on this board. I did on another one and it was easy to deal with. Won't be a single op-amp, I will need at least a 4th order or higher and probably to sacrifice a lot of bandwidth - I know its not what I would have hope for. Before sigma-delta we managed ok for a good 30 odd years. It is annoying that designers aren't putting in sigma-deltas as standard.
 
M

miso

Jan 1, 1970
0
Don't have sigma-delta adc on this board. I did on another one and it was easy to deal with. Won't be a single op-amp,
I will need at least a 4th order or higher
and probably to sacrifice a lot of bandwidth - I know its not what I would have hope for.
Before sigma-delta we managed ok for a good 30 odd years.

Those sigma-delta bastards made me find a new specialty. It killed the
SCF business. The technology is good, especially for control systems
since there are no missing codes.

It is annoying that designers aren't putting in sigma-deltas as standard.
The only place a shitty filter like a Salen Key is useful is before an
oversampled data converter. In fact, if gain error is a big deal, I can
make an argument that a g=1 SK is the right filter.

But it appears now you are not using an oversampled converter. The
filter error is now part of the error budget. Expect lots of variance
due to using the wrong filter topology.

Have you priced 1% caps?

I'm outta here.
 
Top