Sallen-Key input Z

ISBN: 0138800057 / ISBN-13: 9780138800055

Some of us have to live outside the laboratory in the consumer world... dazzling performance at dirt-cheap pricing ;-) ...Jim Thompson

Issued after I bought. My copy has only the Library of Congress Number. ...Jim Thompson

You can certainly _build_ from Zverev's book, but there's minimal instruction/education there-in. ...Jim Thompson

The future's so bright, I gotta wear shades. ;)

Which 10 ppm/K resistors would you be wanting to use, then? The simplest S-Ks are cheap, small, and inherently have a DC gain of 1. Very few applications need 0.1% accuracy in the corner frequency, but lots do at DC.

Cheers

Phil Hobbs

I've made bucket-loads of active filters using 1% resistors and 2% NPO/COG capacitors.

In the hybrid world I've developed trivial-to-do sand-blaster trim procedures that produce better than 0.1% accuracy.

If you avoid S-K configurations there are no component sensitivity issues to contend with.

Add zeroes and component sensitivities get _really_ low. ...Jim Thompson

Luckily, my customers will apply a DC input to check our accuracy. Very few will care about the precision of a lowpass filter rolloff.

On did, though. They were counting pills on a load-cell platform, and wanted to go fast, and settling had to be ASAP to 0.02% or some such. I'd done a digital filter, programmable from a list of cutoff frequencies, and I cheated and rounded all the coefficients to powers of 2, so I could use shifts instead of multiplies. That added a response tail that messed up his throughput. We modified the filters to replace one of the shifts with a proper multiply, and tuned that.

My actual filter design annoyed a couple of experts. I designed an analog state-variable filter, and then modeled it digitally, in 68K assembly language. This turns out to be a seldom-used digital filter form, and it's nice because all the coefficients are pretty sensible numbers and the computational load is low.

Load cells are awesome.

.SBTTL LOWPASS FILTERS

; THIS IS A 4-POLE, STATE-VARIABLE (INTEGRATOR-BASED) GADGET WITH A ; 'TRANSITIONAL' (EG, HOMEBREW) TRANSFER FUNCTION.

; NODE 1 NODE 2 NODE 3 NODE 4 ; | | | | ;IN-->--(+)--[K1:INT]---(+)--[K2:INT]-->--(+)--[K3:INT]---(+)--[K4:INT]-->--OUT ; | | | | | | ; | | | | | | ; | ^---(-1)--- | | ^---(-1)--- | ; | | | | ; | | | | ; '------

Was your copy chiseled out of stone or inscribed on papyrus?

ISBN was contrived in 1965 and not widely used until after about 1972. So you're correct, a book published in 1966 will probably not have an ISBN number on the copyright page:

My 1967 copy of "Handbook for Filter Synthesis" by Anatol I. Zverev also does not have a printed ISBN number. Thumbing through its dusty pages, my guess is about 100 of the 576 pages are tables. It's not just a book of tables.

Why are you screaming? Got your balls jammed by Caps Lock key?

I always program assembly and Basic in uppercase. We honor tradition here.

It's easier to read on 11x17 fanfold dot-matrix listings, too.

What? You're not impressed ?>:-} ...Jim Thompson

Maybe I found it just a book of tables because active filters have been a hobby of mine forever (*)... thus Zeverev taught me no new tricks.

(*) I even used an ANALOG computer to factor transfer functions. Forgot its name... some company in Oregon... blue case like Tek, but not Tek... cost me (company, Dickson) around $10K in 1970 :-( ...Jim Thompson

Bingo. When you crank stuff out by the millions, close isn't good enough.

Has it occurred to you that some requirements need AC accuracy. Your SK filters are simply s*it for that use.

I know you have a PHD and don't like being pointed out when you are wrong, but give it a rest here.

Not to mention laser trimmed hybrids.

Actually the LT1062 does that. Nello is a good filter designer.

Except it costs you nothing to dynamic range the design. Hence Dr. Temes put it in his lecture. Why do a s***ty job when for no additional components, you can do it correctly?

Further, with low noise op amps, you can make quiet active filters. There are applications where it is best to avoid the A-D and D->A steps.

It has a number of chapters on filter synthesis. Of course, with computers, perhaps the techniques are old fashioned.

I don't recall seeing any other sources on crystal filters. I suppose they exist.

You can predistort for Q enhancement, but you run the risk of tempco effecting the filter.

If you go for a ladder approach, the localized Qs tend to be lower. I'm not sure if localized Q is a generally accepted term, but I learned it from Dan Senderovich. The idea is every node is a filter, even if only one node has the desired frequency response. If you examine the response at intermediate nodes of a cascade of biquadratic sections versus the nodes of a ladder, the ladder "localized" Q is lower. I don't believe there is a strict proof of this, but anecdotal evidence shows this.

The other advantage to a ladder is that the all the components interact in a manner where the sensitivity to an individual component is reduced.

Senderovich is quite the guru at this. I learned how to do bandpass and highpass ladders from him, something textbooks don't cover. His approach is totally signal flow graph. In fact, if you want to ponder dynamic range adjustment, signal flow graph is the way to go.

In ladder design using signal flow graph, the upper nodes represent voltage, and physically are op amp outputs. The lower nodes represent current. Going from the upper nodes to the lower nodes is done with integrators (physically an op amp with a cap as feedback, positive input to ground if single ended.).

Say the node is peaking. To dynamic range adjust a filter as shown on the signal flow graph, you scale down (attenuate) all the paths entering the node, then scale up (amplify) all the paths leaving the node. If you do this, the final output of the signal flow graph has remained unchanged.