Maker Pro
Maker Pro

LC notch filter not working!

J

Joel Kolstad

Jan 1, 1970
0
Phil Hobbs said:
Nope, you've got the square root in the wrong place: a 100:1 transformer
changes the impedance by 10,000:1, not 10:1.

Oops, thanks for catching that Phil! Sheesh...
 
A

AF6AY

Jan 1, 1970
0
From: Tom Bruhns said:
This particular part is not true.

Yes, you are correct...caught me with a low level of caffeine on
Thursday. :)
A parallel-resonant trap placed in
series is not detuned by load capacitance at the output. It's still
parallel-resonant at the same frequency. A simulation shows this
easily.

Yes on the parallel L-C for the trap frequency. But, under low source
impedance and high load impedance, with the approximate L and C given,
there is a voltage increase at a frequency below the trap frequency.
[there
are four combinations of 3 components for L and C circuits, each with
a
peak versus dip impedance response, me has to keep reviewing those to
avoid confusion] To explain, my (later) analysis model was as
follows:

One-Ohm impedance current source. Parallel L-C in series with load,
L1 = 10 uHy with Q of 150, C1 = 14 pFd. Load is 1 MOhm in parallel
with
C2, C2 varying 10, 20, 30 pFd. Capacitors were assumed essentially
lossless since their typical Q at these frequencies can be 1000 or
more.
Minimum voltage response was at a nearly constant frequency regardless
of C2 value. Maximum voltage response frequency varied considerably.
Using 1.0 V RMS reference for 0 db, the response v. C2 value was:

C2 = 30 pFd, Vout peak +22 db at 7.8 MHz, Vout minimum -35 db.
C2 = 20 pFd, Vout peak +26 db at 8.25 MHz, Vout minimum -32 db
C2 = 10 pFd, Vout peak +21 db at 10.4 MHz, Vout minimum -26 db

I could have done the above with L1 Q of 50 but that would simply
decrease the lower frequency peak voltage, show a lesser voltage
minimum at the upper trap frequency, the rest about the same.

* At this point someone will get hot about "ya can't have voltage
* gain with no amplifier!" or equivalent. :) Yes, one can since
* a voltage increase only means a current decrease at one
* frequency...the only power loss is in the Qs of the components.
The same is true of a series resonant shunt trap.

Yes, but only for the series resonance frequency. There's a variation
in the overall voltage response depending on the load resistance and
its parallel load (and probe) capacity. For sure, a series-resonant
circuit across the source is going to affect the gain of the driving
source from its frequency variation of impedance.

This is one of those seemingly-inocuous circuit applications which can
get very tricky to apply with any repeatability. Especially so when
the
source and load were unspecified. It's safe to say that EVERYTHING
interacts over frequency and one cannot just assume anything. That
includes scope probes which far too many apply thinking just of their
10 Meg input resistance and forgetting they all have capacity to
ground
in parallel. :-(

Thanks for reminding me to go back to earlier basics, Tom. A number
of years ago I worked the math on impedance of the four basic 3-
component combinations and wrote it up for a work application (that
would have been a high production failure situation if used as-is) and
thought memory "would always be there." Actually it was but my mind
gets cluttered with other stuff on a disorganized basis. :)

BTW, I used my own LINEA (DOS-only) analysis program and LTSpice
(free Windows compatible full package from Linear Technology) to run
this simple circuit model. Results agreed.

73, Len AF6AY
 
A

AF6AY

Jan 1, 1970
0
Seconds?  You should check out the old Collins 490T antenna tuner.  As
I recall, the spec was something like 5 seconds max to tune any new
load within the specified range, any new frequency within range, but
you knew that if it didn't tune in a second and a half, something was
most likely broken.  Motors don't have to be slow.  The inductor went
from one end to the other in I suppose under half a second.  Seems
like it was 8 or 10 turns.

Other than a pre-tuned Collins commercial transmitter at an Army
station in the early 1950s, the first time I recall seeing an
automatic
antenna tuner was in the T-195 transmitter built by Collins for a USMC
contract (forget the AN/ number, its companion receiver was the
R-392, the 28 V counterpart to the R-390 and R-391). On a quickie
demo in 1955, the officer doing the demo disconnected one of the
Jeep's whip antenna sections. The T-195 retuned its antenna is
a few seconds, indicated by a little lamp on the front panel. Most
amazing to me at the time, used to the huge built-to-last-forever
HF monsters that were always most fussy on manual tuning. :)

Much later I got a PDF of that T-195 TM and believe that this set
might have been the first military radio to incorporate the Bruene
voltage-current detector necessary for the automatic antenna
tuning servos. Any delays in operation might have been just from
the detector-sensor output time-constants in addition to motor
speeds. The "Bruene Bridge" as it is sometimes called, is
the basic form for nearly every other automatic antenna tuner built
since then.
 But there's a long ways between these embryonic ideas
and a working design, and I leave that to you.  ;-)

Yes, the all-important "gestation stage." Ask any mother. :)

73, Len AF6AY
 
T

Tom Bruhns

Jan 1, 1970
0
HI Tom,




Yeah, I suppose that if you start back-biasing your diodes at 100V, suddenly a
volt starts to look like a small signal again.

Someone else mentioned that a straightforward means to drop the voltage swings
is by dropping the "system" impedance. A 1:100 transformer takes 10V to a
mere 100mV, although now the 50 ohm system is 5 ohms so Q has to be 10 times
better to obtain the same notch depth. Still, probably worth pursuing.

Speaking of Collins radios, here's a rather sad site:http://cgi.ebay.com/Collins-490T-1-Radio_W0QQitemZ120112545170QQihZ00...

---Joel

I've trimmed off a couple of the groups since someone seems to have
gotten his knickers all twisted up over the posting in, wow, four
pissibly relevant newsgroups. Wish he'd take his venom out on the
idiots that cross-posted from the wierd alt.* groups a couple weeks
ago.

Anyway, Joel, WHY would you think that the Q needs to be any
different?? You'd scale the impedance, and as Phil noted you got the
impedance ratio vs turns ratio backwards, but you'd want the same Q at
that frequency. It might be practical to scale by a 3:1 turns ratio
or possibly even 4:1 at these frequencies, but I'd be wary of going
beyond that.

Cheers,
Tom
 
T

Tom Bruhns

Jan 1, 1970
0
Yes, you are correct...caught me with a low level of caffeine on
Thursday. :)


Yes on the parallel L-C for the trap frequency. But, under low source
impedance and high load impedance, with the approximate L and C given,
there is a voltage increase at a frequency below the trap frequency.
....

Trimmed off a couple of the groups and much of the message, though all
was noted. Thanks for the additional info; I trust the OP will finde
it useful, if he's still around. (Pet peeve: posters who don't
bother to get back to say "Hey, that helped," or "Huh?" or give some
indication they are still lurking.)

Yes, to be sure the response depends on the load. In fact, even at
the notch frequency, if you start with a high load and add
capacitance, you significantly affect the depth of the notch with the
added capacitance.

It can be quite useful to add another capacitor (or inductor) to a
series or shunt trap, to get the response at a frequency you
specifically want to pass to be high. You can do the same thing with
transmission line stubs, which becomes practical at higher
frequencies. For example, you can put a shorted stub across a line,
where the stub length is 1/2 wave on the frequency you want to
"kill." But then the response at nearby frequencies will also be
attenuated. You can then view that first stub as a reactance at the
frequency you want to pass, and add another stub of the same reactance
magnitude but opposite polarity. You'll find, of course, that the two
stubs total a wavelength, assuming both are shorted at the ends away
from the point the join the through line. With low loss line, this
can be a very effective way to get rid of a large signal in a fixed-
frequency receiver system. The capacitor-or-inductor-added-to-the-
trap is a lumped equivalent of this idea.

I suppose in an absolutely accurate analysis, the impedance versus
frequency charaterisitic of a load that includes capacitance may be
such that the frequency of the maximum attenuation of a finite-Q notch
is shifted ever so slightly, but for sure it won't be shifted enough
to notice; the proximity of the metal in the probe to the coil is
likely to affect the resonant frequency more.

Cheers,
Tom
 
J

Joel Kolstad

Jan 1, 1970
0
Hi Tom,

Tom Bruhns said:
Anyway, Joel, WHY would you think that the Q needs to be any
different??

I was probably unclear in that I meant Q of the inductor (and am assuming Q
of the capacitor is high enough to be ignore); *not* Q of the system.

Say you're in a 50 ohm system. If, at resonance, your series shunt L-C
exhibits a resistance of 0.5 ohms, that's about a 40dB notch. However, in a
5 ohm system, it's only a 10dB notch. You need to get the resistance down
to 0.05 -- implying a Q ten times large than what you started with -- to
maintain the same notch depth.

Do you buy this? :)

Thanks,
---Joel
 
A

AF6AY

Jan 1, 1970
0
...

Trimmed off a couple of the groups and much of the message, though all
was noted.  Thanks for the additional info; I trust the OP will finde
it useful, if he's still around.  (Pet peeve:  posters who don't
bother to get back to say "Hey, that helped," or "Huh?" or give some
indication they are still lurking.)

Yes, to be sure the response depends on the load.  In fact, even at
the notch frequency, if you start with a high load and add
capacitance, you significantly affect the depth of the notch with the
added capacitance.

It can be quite useful to add another capacitor (or inductor) to a
series or shunt trap, to get the response at a frequency you
specifically want to pass to be high.  You can do the same thing with
transmission line stubs, which becomes practical at higher
frequencies.  For example, you can put a shorted stub across a line,
where the stub length is 1/2 wave on the frequency you want to
"kill."  But then the response at nearby frequencies will also be
attenuated.  You can then view that first stub as a reactance at the
frequency you want to pass, and add another stub of the same reactance
magnitude but opposite polarity.  You'll find, of course, that the two
stubs total a wavelength, assuming both are shorted at the ends away
from the point the join the through line.  With low loss line, this
can be a very effective way to get rid of a large signal in a fixed-
frequency receiver system.  The capacitor-or-inductor-added-to-the-
trap is a lumped equivalent of this idea.

I suppose in an absolutely accurate analysis, the impedance versus
frequency charaterisitic of a load that includes capacitance may be
such that the frequency of the maximum attenuation of a finite-Q notch
is shifted ever so slightly, but for sure it won't be shifted enough
to notice; the proximity of the metal in the probe to the coil is
likely to affect the resonant frequency more.

Agreed. However, the coax cable stub idea might be a tad
impractical considering that 13.56 MHz is lower in wavelength
than the 20m band. Stubs could get as long as around 15 feet
at that frequency. :)

I once had to "rotate" the impedance of some SAW filters (8 of
them) for the 60 to 70 MHz region and couldn't get any more
space for the matching other than some slots in a machined-out
chassis. I couldn't have done it without skinny lil 1/8" OD coax
held in place by some RTV. [roughly 3/8 of a rotation on the
Smith Chart] I hate to think about doing that at 13 MHz.

73, Len AF6AY
 
J

Joel Kolstad

Jan 1, 1970
0
Joel Kolstad said:
Say you're in a 50 ohm system. If, at resonance, your series shunt L-C
exhibits a resistance of 0.5 ohms, that's about a 40dB notch. However, in
a 5 ohm system, it's only a 10dB notch.

This should, of course, be 20dB. I know I was thinking 20dB, but clearly I
typed 10dB. Oops.
 
Hi Tom,
What input impedance is your scope? In a very slightly more accurate
theory, and a much more useful one, the impedance does NOT become
infinite, but rather becomes Q times the reactance at resonance. The
reactance in your case is about 850 ohms. The Q I have little idea
about: it could be 10 (pretty easily), it could be 1000 (with quite a
bit of difficulty). You may do much better if you put a lower load
resistance on the output of the filter -- in the RF world, 50 ohms
would be usual, but at least something much lower than a 1 megohm
scope input (as I suspect you're using).

The input impedance of the scope is 1 MegOhms for the passive probes
I use, but i can change the coupling to 50 Ohms as well.

Do you know where the Q in an LC notch filter comes from ? Is this
the Q of the inductor defined as (2*pi*f * L) / R ?

What kind of inductor and capacitor is best suited for a notch filter
at 13.56 MHz (RF frequencies) ?
I use a ceramic trimmer right now, but I am not sure what kind of
inductor
is best suited for RF circuits. It seems that the value of an inductor
(even the L) is
very frequency-dependent.

When you say a low load is much better, do you mean for a parallel
LC notch filter, or for a series LC notch filter ?

Thanks for the great help!
 
F

Floyd L. Davidson

Jan 1, 1970
0
Hi Tom,


The input impedance of the scope is 1 MegOhms for the passive probes
I use, but i can change the coupling to 50 Ohms as well.

What are the circuit impedances?

Consider that your circuit more or less looks something like this,

Rtrap
input signal >-----+----/\/\/\/\/----+-----> output
| |
/ /
\ \
Rin / / Rout
\ \
/ /
\ \
| |
| |
----- -----
--- ---
- -

Hmmm, looks just like a plain old RC attenuation pad! Except the
Rtrap is actually a parallel tuned circuit that is a high impedance
at one frequency and lower impedances at other frequencies. So lets
pick any handy set of values for Rtrap, as an example. Maybe your
LC circuit is more, or maybe less... the effect is what you want to
understand. Lets assume the value for Rtrap approaches 100 Ohms for
non-resonant frequencies, and say 10,000 Ohms at the resonate frequency.

So, if Rin happens to be high, say 100,000 Ohms or more we can just
ignore it. (Which is practical, as all it does is provide a constant
load for your source, and we'll assume it is sturdy and can handle
anything from 0 to 1000 megs!)

That means you have two circuits, one at the resonate frequency and
one at all others, which both look like this,
|
/
\ 100 Ohms, or
/
\ 10,000 Ohms
/
|
+------> out
|
/
\
/ Rout
\
/
|
-----
---
-


It's just a plain old resistance divider. If Rout is 100 Ohms the
output will be 1/2 the input at non-resonate frequencies (insertion
loss), and at the resonate frequency it will be 1/100th of the input.

Obviously if the Rout value is 100,000 Ohms your divider is going to
have virtually no effect at all! And if it is 10 Ohms the effect will
be even greater than it was at 100 Ohms.
 
T

Tony Williams

Jan 1, 1970
0
Floyd L. Davidson said:
Consider that your circuit more or less looks something like this,

Rtrap
input signal >-----+----/\/\/\/\/----+-----> output
| |
/ /
\ \
Rin / / Rout
\ \
/ /
\ \
| |
| |
----- -----
--- ---
- -

Hmmm, looks just like a plain old RC attenuation pad! Except the
Rtrap is actually a parallel tuned circuit that is a high impedance
at one frequency and lower impedances at other frequencies.

Can that circuit ever produce any depth of notch? If Rtrap is
a parallel tuned circuit then it has in parallel with it an
effective resistance of Rin+Rout, and to get any reasonable
selectivity Rin+Rout must be high compared to L/C.R, the
dynamic impedance of the tuned circuit at resonance.

If that is so, are there actually any values for Rin and Rout
that could produce a reasonable selectivity?
 
A

Angelo Campanella

Jan 1, 1970
0
However, I can turn the trimmer (in the range from 10 to 20pF) as much
as I want and I don't see ANY effect at all on my scope.
Any hints ? What am I missing ?

The first thing you must do is determine the self-resonance of the coil.
That can be done with a "grid dipper". If that frequency is below the
frequency you want to filter, it won't work. The next ting to do is
determine the resonsnce frequency this time installed in the circuit
witout a trimmer. Again, if that is below the frequency to be filtered,
it won't work. Only when that resonance is above, 13 MHz in this case,
can a trimmer be applied to tune it.
I have also looked at active notch filters, but this seems to be
rather difficult at these high frequencies (see http://focus.ti.com/lit/an/slyt235/slyt235.pdf).

The stray capacitance, possibly multiplied by the chip gain, may rule
out operation at 13 MHz.

Angelo Campanella
 
Top