identify this schematic

[bergmann disney]

hello,
this is a link to a part of a schematic i am trying to understand :
http://twin-x.com/groupdiy/displayimage.php?pos=-1009

by where it's placed in the schematic, i can nearly 100% be shure this is a lowpass.
but does anyone know somthing about this components setting (am i wrong, is there a formula to determinate the order and cutting freq of this filter if i am not, or anything else...) ??

 

[burdij]

Don't think its just a lowpass filter. There appears to be a positive feedback path and one stage of passive lowpass. An "allpass" (phase shift) filter?

 

[PRR]

With 2 R 2 C, it is a standard Unity-Gain Sallen-Key low-pass. Look it up. Or get the Active Filter Cookbook.

With that 3rd RC tacked in front, I never saw that. With the C tied to ground, it would be a common way to get a 3rd-order response with just one amplifier. To get anything approximately optimum needs a large impedance gain (low source Z) and lots of figuring or tinkering.

 

[JohnRoberts]

I'll echo what PRR said.. Also, when cascading an extra real pole in front of an active two pole the impedance of the passive pole is typically dropped an order of magnitude to reduce interaction with the following components.

I've also seen darlingtons used instead of single common collector stage for higher input impedance.

It's clearly a low pass filter but I doubt you'll find that specific topology in a standard filter text because of the odd connection of the bottom of the first C1.

JR

 

[bcarso]

I believe the values shown are way off for approximating typical lowpass filter shapes. Also, if I recall correctly the second feedback connection to the cap closest to the input actually does not realize the third-order lowpass regardless of values---I used this once thinking it might work better and never could get what I wanted.

Putting an R-C in front of an existing Sallen-Key unity-gain lowpass as PRR mentions is standard, and there is a solution for all three R equal. Williams (in Electronic Filter Design HB) comments that the determination of the C values is best done with a computer :roll:

It would be a good exercise for someone to write down the transfer function of the filter with double feedback to see if it does get you there (say, allow a maximally-flat filter shape) for the right combination of values. As I say I used it once and got close, but it was just for a demo system and before I had access to the tabulated values for the more capable circuit. I also enjoyed bedeviling another engineer, who came in one morning and said "Sir, for the record: I could not find that circuit in any of my filter books". He implied that it should not be used on that account. :razz:

In addition, a emitter follower is hard-pressed to do an accurate job as a unity-gain voltage amplifier in these cases, even for just a two-pole filter. To expect decent performance for a three-pole is pushing it, although as John suggests just putting and R-C in front of an existing two-pole and making the impedance a lot lower (thus loading down the source) will be o.k.

 

[SonsOfThunder]

Typically when you see this sort of thing implemented it will exhibit a 6dB per octave rolloff in the first half octave up from the cutoff freq and then transition into 18dB per octave from there on. So it gives a more gentle rolloff rather than a sharp transition. Which could be desirable (or at least acceptable) in some cases.

Some "major" manufacturers in the car audio world have done this sort of thing in their products and passed it off as 18dB per octave...it is sorta kinda...but it doesn't sound like it when you use your ears. To pull it off right requires a second opamp...but that costs money!

HTH!

 

[PRR]

> I believe the values shown are way off for approximating typical lowpass filter shapes.

What "values shown"?

Assuming equal values: SOT nailed it. It is a gentle slope at first. By an octave out, it gets near 12dB/oct. It does finally approach 18dB/oct.... around a decade up and 40dB down.

For most musical purposes, it ain't a whole lot better than two cascaded (un-Active) R-C nets.

I tried a few un-equal values and did not find an easy path to a sharp corner. It may exist. But like your bedeviled colleague, if it is not in the books (preferably very simple books), it's probably not worth my effort to explore.