A Universal Sallen-Key Filter Board

Where we discuss new analog design ideas for Pro Audio and modern spins on vintage ones.
Gold
Posts: 677
Joined: Sat Dec 01, 2007 5:20 pm

Re: A Universal Sallen-Key Filter Board

Post by Gold »

I've always considered remounting and using my Neumann HT75 HPF/LPF a stop gap measure. The switches are old and can't be replaced. I have a replacement in mind that I've discussed with Wayne off list. I have it all drawn out. Now all I need is a giant stack of these things. It will be pretty slick if I do say so myself. I also have another filter set idea that will require giant pile as well.
User avatar
mediatechnology
Posts: 5443
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: A Universal Sallen-Key Filter Board

Post by mediatechnology »

Anyone looked into the out of band anomalies WRT Sallen Key filters?
Yes. With careful selection of components it can be minimized or be insignificant.
Pease wrote about it here: http://electronicdesign.com/analog/what ... uff-anyhow
Bonnie Baker also wrote a good article: http://www.ti.com/lit/an/slyt306/slyt306.pdf

In the low pass topology the rising closed loop output impedance vs. frequency of the op amp contributes to the filter response.
If the feedback elements present a low impedance relative to the op amp's closed loop output impedance an out-of-band HF "hump" can form in the stop band.
If the resistor values are made large enough and the capacitors small it's usually not an issue.
A passive input pole can also be added to increase stop band attenuation.

Anti-aliasing and reconstruction filters are an area where an MFB shines.

I'm about to build a 27 kHz Sallen-Key and I'll see with sane capacitor values and a fast op amp if it really matters.
Something tells me that if there's line-level input at 100 kHz and it's forming an out-of-band hump I've got bigger problems to solve.
With typical spectral distribution of audio falling vs. frequency it shouldn't be a problem.

The unity gain Sallen-Key has so much going for it in terms of simplicity I think it's the best option for a board like this one.
(Which I hope I can get around to doing.)


A great calculator for Sallen-Key filters is here: http://sim.okawa-denshi.jp/en/OPseikiLowkeisan.htm
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: A Universal Sallen-Key Filter Board

Post by JR. »

raf wrote:Anyone looked into the out of band anomalies WRT Sallen Key filters?
?? I've used Sallen and Key topology among others many times over the decades. What out of band anomalies are you speaking of?

The closest to an issue that I've experienced was in the anti-image (smoothing) filters immediately following a BBD analog shirt register where high frequency clock spikes could cause problems for medium bandwidth (slow) op amps available back in the 70s. I suspect a similar issue could occur from I to V conversion from early DACs. My remedy for the BBD designs was to incorporate a real pole at the ASR output node, a part of a multi-pole cascaded filter. (I also placed under damped filter stage in front of the BBD anti-alias section, and over damped stages post the BBD for some free pre-/de-emphasis.)

I spent far more bench time and killed brain cells chasing out quirks of state variable filters. A poorly understood aspect of SVF is the gain structure of the different stages. Some schemes for adjusting Q result in higher gain for the HP or LP section than the bandpass, making it possible for out of band clipping that isn't in the direct path, but can cause distortion in the passband.

I am not aware of specific issues with Sallen and Key topology and didn't find anything obvious from a quick google search, while I did not search rigorously. I expect op amp drive capability and output impedance could matter for some component selections, op amp lag could affect extreme tuning.

JR

[edit- Wayne did a better job than I did, and stop band deviation at 20-30x tuning is mostly innocuous except perhaps in the post BBD application I already mentioned. My fix for that was a real pole before the active filter stages and that really worked. [/edit]
Cancel the "cancel culture", do not support mob hatred.
Gold
Posts: 677
Joined: Sat Dec 01, 2007 5:20 pm

Re: A Universal Sallen-Key Filter Board

Post by Gold »

Lancaster's Active Filter Cookbook is always out and open. I look at it all the time. His answer is basically "it doesn't hurt to run the numbers but pick a good op amp and don't worry about it."
User avatar
mediatechnology
Posts: 5443
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: A Universal Sallen-Key Filter Board

Post by mediatechnology »

r
The closest to an issue that I've experienced was in the anti-image (smoothing) filters immediately following a BBD analog shirt register where high frequency clock spikes could cause problems for medium bandwidth (slow) op amps available back in the 70s. I suspect a similar issue could occur from I to V conversion from early DACs.
Those are about the only two places where i think it could be an issue in audio.
But for general audio filtering applications, no.
The only reason I brought it up is that someone else would.

This is what it looks like with slow and fast op amps and low-ish component values.

Image
Stop-band limitations of the Sallen-Key low-pass filter, Bonnie C. Baker, TI App Note SLYT306 http://www.ti.com/lit/an/slyt306/slyt306.pdf
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: A Universal Sallen-Key Filter Board

Post by JR. »

I designed many cascaded multi-pole filers back in my active delay line design days. What generally matters for cascaded filter designs is response say an octave below and octave above tuning. I've mixed different filter topologies, which is different than filter alignments, in the same filter design. I even did one mother of all filter alignments where i combined a 12 dB pre-/de-emphasis network into a chebyschev alignment where the real pole of the chebyschev mated with the de-emphasis section... Good for S/N through a low dynamic range BBD path and the real pole down at a fraction of the overall tuning was great for squashing clock artifacts after the BBD.

Back in the '70s we (I?) didn't have fancy filter simulation software so i wrote my own filter software** that allowed me to cascade multiple active filter sections with real 5% component values plugged in. It printed out a crude frequency response plot using the tab function of my dot matrix printer (while printing actual response). The math was a real PIA for a 7 or 9 stage active filter by hand even with a programmable calculator, but the computer could crunch it and print it out almost as fast as I could enter new values. Much better than the build a breadboard and use trial and error to tweak it in. 8-)

JR

** The program was written in Basic. The code on an 8" floppy and the hardware to run it on (H-11) are long gone.
Cancel the "cancel culture", do not support mob hatred.
Gold
Posts: 677
Joined: Sat Dec 01, 2007 5:20 pm

Re: A Universal Sallen-Key Filter Board

Post by Gold »

mediatechnology wrote: If the idea makes sense it could be extended to MFB filter types which for LP responses may be better in some applications.
This slipped by me the first time. Multiple Feedback filter topology is a new one to me. It's not covered in the Lancaster book, which is (was?) my bible. From the little I've read it looks like Sallen-Key is better for unity gain HPF/LPF. That's pretty much what I want.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: A Universal Sallen-Key Filter Board

Post by JR. »

Gold wrote:
mediatechnology wrote: If the idea makes sense it could be extended to MFB filter types which for LP responses may be better in some applications.
This slipped by me the first time. Multiple Feedback filter topology is a new one to me. It's not covered in the Lancaster book, which is (was?) my bible. From the little I've read it looks like Sallen-Key is better for unity gain HPF/LPF. That's pretty much what I want.
You should use what you want...
=====
I learned filter design using the Burr Brown "Operational Amplifiers" C 1971 they called Sallen and Key topology VCVS back then (perhaps sallen or key was still alive?).

I think MFB may also have been called infinite feedback...

The different topologies have different pros and cons depending on application and circuit criteria... If one was always better we wouldn't still use different types.

The sallen and key is similar to the old school filters that were built around a single transistor emitter follower or perhaps a darlington in place of an op amp, that used to be expensive. IIRC the active EQ used for BOSE 901 speakers used a darlington follower.

JR
Cancel the "cancel culture", do not support mob hatred.
ricardo
Posts: 262
Joined: Sun Apr 10, 2011 7:24 am

Re: A Universal Sallen-Key Filter Board

Post by ricardo »

I've done loadsa S&K filters in commercial designs and have never had problems with stop band stuff ... but I don't do DAC reconstruction filters or anti-aliasing active filters.

My main grouse with S&K filters is they are more sensitive to component tolerances than practically any other filter topology. You see this especially in higher order filters with cascaded HiQ S&Ks. But they also need less 1% caps than other filters :mrgreen:

Another problem is that when overloaded, with certain OPAs eg TL07x, they exhibit latch up and polarity reversal. This is easily cured by using eg 5532 but MFB filters are much nicer in this respect.

OTH, I believe they have the best noise performance.[*] There's an EDN article by someone, with a double barrelled name that I've forgotten, on the pros & cons of Digital Filters and he uses S&Ks as the quietest of the analogue filters.

I second JR's caveats about stage gains in SVFs.

[*] I might be wrong on this. The double integrators of SVFs are very nice. But its a long time since I was obsessed with analogue filters. :D
User avatar
mediatechnology
Posts: 5443
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: A Universal Sallen-Key Filter Board

Post by mediatechnology »

Thanks ricardo and everyone.

The stop band hump is only in the Low Pass implementation too.
The HP doesn't have it.

One of the big advantages of the unity gain Sallen-Key in MS work is that in the pass-band they really are, for all practical purposes, unity gain.
In terms of gain accuracy they win.
One less resistor too.

I did a DAC filter comparison between Sallen-Key and MFB and IIRC the Sallen-Key was indeed quieter.
Post Reply