Wideband Notch Filter and Circuit Impedances

[Joncaster]

Hey all,

Thinking of cutting out a bit of mid range, about 6dB, wide Q, from a tube amp circuit.
Non adjustable, lowest loss possible (my gain staging is fine the way it is).

So playing with the Duncan TSC, the Big Muff and Framus cuts are all a touch lossy and adjustable.
If I don't have to drop 4-6db insertion loss, rather 1-2dB, that would be ideal.

The simplest implementation seems to a bandstop RC Hi and Lo pass in parallel (two 1st orders).
Almost the same as the Framus, but without the resistor leg to ground off the second cap.
Like this guy:



My question is about the Source and Load impedance of the circuit:

Can I substitute the R1 and R2 values for the source and load Z respectively?

The low pass would be the Zout of the preceding stage, flowing into its cap to ground.
The Hi pass sees its cap and the input impedance to ground of the following stage.

Like this:

             (cap)
Zout------| |------Zin
          \______/
                |
                = (cap)
                |
              gnd

Or do I need that R1 in parallel to the C2 cap to force the signal through C2 and not just bypass it?
Do I need the R2 to provide a different path to ground?

Let's say I have local negative feedback around the second stage, and no grid resistor...
Will it cause issues to use the feedback resistor (which is also the grid ground reference) to dump the Low frequencies?

Assuming it's all working fine like that:
The trick would be to have a very low Zout and a medium Zin.
Too high a Zin, and the high pass cap needs to be far too small.
Too high a Zout and it forms a voltage divider with the Zin.
That is, in fact, what I have between two stages currently...so that's what got me thinking.

Obviously Ill try it and see, but want to know if i'm missing anything.

Last time I played around with splitting Hi/Lo, its tricky to know exactly what frequencies you are hearing vs what you have calculated.

[antonis]

A brute approximation could be to consider a filter of cut-off frequency 1/2π*A*B, where A = (R1*C2)/(R1+C2) and B = (R2*C1)/(R2+C1)..

But this would stand only in case of R1 << R2 and C1 >> C2..
(negligible loading for each individual first order filter..)

[Joncaster]

A brute approximation could be to consider a filter of cut-off frequency 1/2π*A*B, where A = (R1*C2)/(R1+C2) and B = (R2*C1)/(R2+C1)..

But this would stand only in case of R1 << R2 and C1 >> C2..
(negligible loading for each individual first order filter..)

So I can calculate the R1 and R2 using the Zin/Zout?

My calculations work out to:

Zout = 7k (calculates to 6.92k actually)
Zin = 192k
(12AU7, The second triode has a tiny bit of LNFB that lowers the Zin)

For a 6dB drop at 630hz: (where the two frequencies meet at -9dB):

LPF Fc= 315hz
C1 = 72nf
75n closest value

HPF Fc = 1260hz
C2 = 660pf
680pf closest value

Bandwidth is 945Hz
Q is 0.66
Vout is 0.96 loss...so not even half a dB.

Does that sound right?

[antonis]

So I can calculate the R1 and R2 using the Zin/Zout?

Νο.. 

Both for R1 & R2 are resistors (impedance needs at least one frequency variable parameter..) and for Zin/Zout should be set in series/parallel with the existing items values..

[Joncaster]

So I can calculate the R1 and R2 using the Zin/Zout?

Νο.. 

Both for R1 & R2 are resistors (impedance needs at least one frequency variable parameter..) and for Zin/Zout should be set in series/parallel with the existing items values..

OK, so I make R1 say, 1k and R2 1M, adjust for cap values.
Then I'm calculating R1 as 8k (Zout->R1), R2 as 161k (R2||Zin)...

But what about when you use something like a simple treble cut, or a Vox style PI treble cut?
It comes straight off the coupling cap, so no series resistance (apart from the pot which controls the shelving) to set the Fc apart from the previous stages plate Z. When the pot is turned all the way to zero resistance, the filter is a simple 6dB/oct 1st order at the Fc of the plate and cap, right?

I guess my question is, can you take the resistors out, and the circuit would still work?

[antonis]

"Simple" treble cuts are considered so 'cause we ignore in/out impedances..
(a simple "isolated" LPF..)

When pot is set ot zero ohms, previous stage output impedance comes to the scene..
(or else, there shouldn't exist RC filter - just a cap shunting to GND..)

[Joncaster]

So I've tried all kinds of different ways to implement this mid cut...really can't get it working properly. Not sure what I'm doing wrong. Can get them working seperately, just not together.

Anyway, I've kind of given up on this idea, as I'm liking the response so far, not hearing anything I want to pull out.

[antonis]

Could you plz post a schematic of your particular design..??

[Joncaster]

Yes sorry, I know it's bad form to talk a particular circuit in another circuit without a schem. I don't actually have an updated one, so will have to redo it pretty much.

By the time I started wiring this amp (2 years after I started the design) it was etched into my brain, so I just make notes of the changes, ideas.

Looking at the whole thing now, it's not a mind blowingly complex, or particularly innovative circuit. But it is my first amp, so a long learning process.
I should really have an updated schem...

[PRR]

It may look like some form of Teed Bridge, but it is simpler than that. And has the particular feature/flaw of impedance going to ZERO at infinite frequency (which annoys many opamps).

[Joncaster]

I re-looked at the normal Bridged T (the Framus mid in TSC), and that works just fine for low loss if you adjust the values further than I had tried at first.
Keeping the 2:1 ratio of resistors, and finding the sweet spot where the total series/parallel values work with the Zout/Zin:
You get minimum loss for that circuit insertion.

Now I have to rejigger the triode stage after the bandstop.

Currently it has Local NFB wrapped around it (for a particular touch sensitivity I get with it), which lowers the Zin.
I can get the insertion loss to 2dB there:
This sounds great, and refines the response nicely. I would like a bit more gain though.



If I take out the LNFB, i can get 1dB loss (and a touch more gain overall). Testing needed:



Haven't had a chance to work on the schematic, apologies.

[Joncaster]

OK, current schematic done!

EDIT: updated schem:
EDIT 2: incorrect schem.

Tone Cut before Mid network might change, and the Bass Rotary is still to be decided on.
Well, pretty much everything is still to be decided on...

[PRR]

> current schematic

Rectifier drawn wrong.

Do you know how much a 20W-25W SE tube amp weighs and costs?

[Joncaster]

> current schematic

Rectifier drawn wrong.

Do you know how much a 20W-25W SE tube amp weighs and costs?

You are completely right, missed that edit. I had changed the rectifier, must have saved under a different version.
Thanks!

[Joncaster]

Edited