Mid scoop/ Twin T Filter

[tocs100]

Okay, I'm trying to figure out the values for a twin T notch at 1kHz. I know this is the right filter'cause I want a Q of .3 (about 3 octaves, 375Hz to 3kHz).

The schem says Fc = 1/    2 x pi x R x C
R = 100k  C = 0.0027 = Fc 589.7 Hz

But when I multiply the given values, I get 1.6956!
What am I doing wrong?

[slacker]

I think you're using the wrong value for the 100k resistor. In that formula R is in Meg ohms so 100k is 0.1 (100,000/1,000,000)

so you should be doing 1/(2 x pi x 0.1 x 0.0027) which equals  589.46

[B Tremblay]

Yup, it's the units that are throwing it off.

Whenever I use that formula, I keep my resistance in Ohms and my capacitance in Farads.

589 = 1 / (2* Pi*100000*.000000027)

[tocs100]

Thanks!

So for a notch around 1kHz, I've got 2 x pi x .1 x 0.0017 =1.06.
1/106 = 9.43 = 943Hz. Correct?

Any suggestions on which components I should slightly mismatch so I don't get an infinite notch? -15dB to -20dB is my target.

Should I keep the caps at 0.0017, the resistors at 100k, and use 25k instead of 50k for 0.5R?

[John Lyons]

Here's a way to simulate it easily.

T-notch filter:
Go to Duncan's amp page and download the "tone stack calculator"
Click on big muff tone stack and plug in these values
Make the resistor to ground at the top 20M or so to take it out of the circuit effectively.
Make the bridging resistors 330K or so. One will be the pot, slide that all the way to the right.
Make the bridging cap 1nf and then the cap to ground 10nf
Just tweak from there until you get something that looks good.

John

[earthtonesaudio]

I was looking for the same type of info, and stumbled upon this:

http://books.google.com/books?id=bkOMDgwFA28C&pg=PA280&lpg=PA280&dq=bootstrapped+twin+T&source=web&ots=F28fUFe-Xw&sig=Et73g8QPfDvE7AoMn-m-55zol7E#PPA284,M1

A few pages down (page 281) is a different type of notch filter called a "bridged differentiator."  It's simpler than the twin-T, fewer components, and it worked on the first try.  I'm inverting it for use as a bandpass/wah effect, and the first thing I noticed about it is that the R_adjustable sweeps frequency from high-low-high as shown.  For low-high, just flip the pot's wiper and one of the terminals.  I think the value of the caps have the biggest impact on the sound.  I'm about to go tweak it a little more, so I'll report back if I find anything good.

[John Lyons]

How do you invert that bridge differentiator for band pass?
Have you seen the colorsound inductorless wah?
It's pretty simple and does a nice quacky wah.

Thanks for that link BTW!

John

[earthtonesaudio]

How do you invert that bridge differentiator for band pass?
Have you seen the colorsound inductorless wah?
It's pretty simple and does a nice quacky wah.

Thanks for that link BTW!

John

Inverting notch filters for band pass is forehead-slapping simple: you take that circuit fragment, a twin-T or bridged differentiator or whatever, and stick one end on the input of an inverting opamp, transistor, CMOS inverter, etc. and the other end on the output.

R.G. says:
"The Twin T passes all frequencies outside its notch, but attenuates the frequencies near its notch. Since this is negative feedback, the frequencies we pass through most easily reduce the gain most, and the frequencies that are not passed through the feedback network are not reduced by feedback, so they have a great deal of gain. The negative feedback connection therefore inverts a frequency notch to a frequency hump, which is exactly what we need for a wah."
Source: http://www.geofex.com/Article_Folders/wahpedl/wahped.htm (under "twin-T" section)

But I believe there is also something going on with phase inversion, but that's probably deeper than you need to go.  I have seen pictures of the colorsound wah, but no schematic, and no sound samples.  I'd love to see/hear more, from what people have wrote it seems cool.

-Alex

[tocs100]

Here's a way to simulate it easily.
T-notch filter:
Go to Duncan's amp page and download the "tone stack calculator"
Click on big muff tone stack and plug in these values
Make the resistor to ground at the top 20M or so to take it out of the circuit effectively.
Make the bridging resistors 330K or so. One will be the pot, slide that all the way to the right.
Make the bridging cap 1nf and then the cap to ground 10nf
Just tweak from there until you get something that looks good.

Thanks, but my computer has no harddrive to download software. (Is that what the Big Muff tonestack is, a Twin T notch with .3 bandwidth?)

Someone please clear-up my two concerns in post #4.

[George Giblet]

> So for a notch around 1kHz, I've got 2 x pi x .1 x 0.0017 =1.06.
1/106 = 9.43 = 943Hz. Correct?

I prefer the to keep to ohms and Farads but your calculations look OK.

> Any suggestions on which components I should slightly mismatch so I don't get an infinite notch? -15dB to -20dB is my target.
> Should I keep the caps at 0.0017, the resistors at 100k, and use 25k instead of 50k for 0.5R?

Firstly you wanted a Q=0.3, IIRC the passive form of the twin-T has a Q of 0.25 which is quite a wide bandwidth.

To ensure the Q is kept high it is better to detune the twin-T by making the 50k a 100k, that will you a -20dB notch.  Unfortunately by changing  the resistor you have now changed the frequency - in this case it lowers the frequency.  The way to compensate is to multiply the *all* the caps by a scaling factor (which makes the caps smaller).  The problem is choosing the scaling factor.  As a first guess doubling a single resistor requires you to scale all the capacitors by 1/2^(1/3) = 0.794.  The twin-T is a complex circuit to analyse and it turns out that a better scaling factor is to scale all the capacitors by 1/2^(1/4) = 0.841.

You might have a go at a simpler Bridged-T notch but I suspect you won't get a -20dB notch with such a low Q.


[don't forget one of the caps is twice a big as the others]

[tocs100]

> So for a notch around 1kHz, I've got 2 x pi x .1 x 0.0017 =1.06.
1/106 = 9.43 = 943Hz. Correct?

I prefer the to keep to ohms and Farads but your calculations look OK.

> Any suggestions on which components I should slightly mismatch so I don't get an infinite notch? -15dB to -20dB is my target.
> Should I keep the caps at 0.0017, the resistors at 100k, and use 25k instead of 50k for 0.5R?

Firstly you wanted a Q=0.3, IIRC the passive form of the twin-T has a Q of 0.25 which is quite a wide bandwidth.

To ensure the Q is kept high it is better to detune the twin-T by making the 50k a 100k, that will you a -20dB notch.  Unfortunately by changing  the resistor you have now changed the frequency - in this case it lowers the frequency.  The way to compensate is to multiply the *all* the caps by a scaling factor (which makes the caps smaller).  The problem is choosing the scaling factor.  As a first guess doubling a single resistor requires you to scale all the capacitors by 1/2^(1/3) = 0.794.  The twin-T is a complex circuit to analyse and it turns out that a better scaling factor is to scale all the capacitors by 1/2^(1/4) = 0.841.

You might have a go at a simpler Bridged-T notch but I suspect you won't get a -20dB notch with such a low Q.


[don't forget one of the caps is twice a big as the others]

Awesome George, thanks a ton!

[earthtonesaudio]

Also, if you're looking for higher Q (more narrow notch), you could try bootstrapping the filter (see geofex article: http://www.geofex.com/Article_Folders/EQs/paramet.htm ).  But as it has been noted, if you change one thing in a twin-T, bridged-T, or bridged differentiator, you change everything.  By altering one component to change frequency, you end up changing gain and Q, etc.  Bootstrapping allows for variable Q which can be useful, especially in conjunction with variable frequency.  But this adds extra complexity (requires another gain stage) so you might decide you'd be better off using a state variable filter instead of a twin-T.

I have a book that explained the math in a really simple intuitive way:

1.) Basic equation:
Freq = 1/(2piR*C)
Variables:
R1 = R2
R3 = 0.5 * R1
C1 = C2
C3 = 2 * C1

2.) Pick your center frequency (how about 1000Hz), and then pick an arbitrary value for C1 (whatever you have on hand), for instance 0.1µF

3.) Rearrange the basic equation to solve for R1:
R1 = 1/(2pi * 1000 * 0.0000001)

R1 = 1591.5Ω, but 1.5k is close enough. 

4.) Now you just figure out the other values you need based on the variables listed in #1 above.


I know your question has already been pretty much answered, but my thought is, a little more info is good and too much is just right 

[tocs100]

...the passive form of the twin-T has a Q of 0.25 which is quite a wide bandwidth.

To ensure the Q is kept high it is better to detune the twin-T by making the 50k a 100k,

that will you a -20dB notch [which you requested] 

Unfortunately by changing  the resistor you have now changed the frequency - in this case it lowers the frequency.  The way to compensate is to....

I think I need a -30dB notch (Q=0.25). How do I achieve this? 150k resistor instead of 50k?
Thanks!

[WGTP]

I think these mid-range filters really are helpful in gettting a good tone.  I have found using the BMP filter with a trim pot on the breadboard is nice because I can use the Duncan TSC and get a notch where and as deep as I want, and then have the trim pot for a little tweaking, balancing the highs and lows.