Sallen-Key input Z

I came up with that twist is the early '70's while trying to make filters at low very frequencies. Zeroed-in on the fact that capacitor dissipation factor can screw very low frequency designs. Padded it to get predictable performance... gained other improvements by accident ;-) ...Jim Thompson

Zverev's work is just a BIG book of tables... not very elucidating. ...Jim Thompson

Agreed, lots of front-end diffamps and such can't have much filtering, one pole if that [1]. I usually assume that a transducer will have a big "DC" signal and a modest amount of noise, from vibration or magnetic fields or Johnson/shot noise, and a bunch of RF. If anything is going to clip, it's the front end, *before* an anti-aliasing filter. What the filter sees is +-full scale signal plus some smaller amount of noise. Numbers like 2:1 or so headroom in the front end and in the filter are reasonable for instrumentation. The filter is the easy part.

But I'd rather design things case-by-case than to constrain myself with arbitrary rules.

quoted text -

Wait until you graduate to a quad 16QFN and 0201 :-)

Then your techs need a sign above the lab benches "No sneezing within 10 feet".

quoted text -

Whenever I need a surfmount resistor or cap from stock, I pull 5. I usually wind up losing or tidly-winking a couple before I get one safely soldered down. Good thing they are so cheap.

Careful, that can backfire. I've had cases where I had to drop GBW when tasked with a redesign. Mainly when strong radio signals got in there. Not via BE junction recification but smack dab in the signal band.

with

Hide quoted text -

The scare is when it snaps out of the tweezers, is nowhere to be found, and there's your coffee mug right next to all the action. Did it land in there?

Or what happened to a friend: Snip ... snap ... tchk ... *PHOOFFFF* ... $2k worth of damage in a piece of lab equipment it had sailed into.

Excuse me, but I am 100% correct here. You can do whatever s*it you want, but dynamic range adjustment is filters 101. It costs you nothing. I would hate for someone to read this thread and think your method is good engineering.

Digital filters were 100% irrelevant to the conversation.

I'm afraid not. There is simply no reason to design a filter with nodes that don't swing at equal amplitudes at all op amps. There is no argument here. I am 100% correct on this.

^^^^^^^^^^^^^^^^

That's called a "dodge" ;-) ...Jim Thompson

-- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at

Nope. I never read Williams book. But anyone simply meditating on the task at hand can see that IF you have enough degrees of freedom, it makes total sense for each op amp to reach the same signal swing. If you don't, you are simply adding more noise to the system for no good reason other than laziness or incompetence.

Now not every topology has enough degrees of freedom to adjust the swing of all nodes.

If you go to this power point presentation, Dr. Temes explains the procedure of dynamic range adjustment. The discussion starts on page 21.

Google hides the direct link unfortunately. I would guess this is an undergraduate class based on what I read.

You're cracked. DC accuracy is such a reason, as is board space, as is component availability. Cheer up, there is life beyond chip design.

Cheers

Phil Hobbs

You said "This is what is called good engineering. It is independent of technology."

"Good engineering" usually means "the way I always do things."

If DC accuracy is so sacrosanct, what about offset? ...Jim Thompson

You can always Spice the filter, plug in slower opamps, and futz values to trim the Q of each stage. You can do that in an active filter, within reason, without getting "lost in space."

Hand tweaking LC filters is usually frustrating. Things tend to diverge.

Modern chop amps are amazing, but I was talking about gain more than offset.

Affordable resistors are somewhat less amazing, and decent instrument accuracy is usually ~ 0.1%. Having to use all 10-ppm/K resistors to build filters with is nuts, especially when it's generally simple to make sure you have enough headroom for a few dB of filter peaking.

Cheers

Phil Hobbs

You snipped, and didn't address, my question:

"OK, how would you do a non-Sallen-Key lowpass filter that has a gain tempco in the single digits of PPM? Mortgage all your property to Vishay?"

That's a ridiculous overgeneralization. Active filters are horribly noisy no matter what you do, so getting all hoity-toity about a couple of dB makes no sense whatsoever. They just aren't suitable for wide dynamic range applications. Since you have to reduce the dynamic range to use them anyway, what's the big deal?

In a built-up circuit I can drop the resistor values and use bigger caps if I like, and get more of an improvement than I would by switching filter topologies. For low order filters of lowish Q, the low parts count and excellent DC gain stability make S-K filters a win.

Sounds sensible, but not always necessary. Good charts though, thanks.

Cheers

Phil Hobbs

I don't know. I like Zverev's plots, and besides, I've actually built filters out of that book, and they worked!

I especially like his equiripple group delay filter.

Cheers

Phil Hobbs