This notch circuit is intended for distortion measurements, where the fundamental is 1.0 kHz, and the harmonics (2kHz,3kHz, etc) are very faint. In such a case, the fundamental may overload the measuring devices, so a notch circuit is introduced to reduce its amplitude, and allow the harmonics to be measured.
Here is a notch circuit (in attachment) that gives a sharp null for a not-so-great-Q inductor. The values shown will null 1 KHz, to better than 100db, and negligible insertion loss if less than 1 octave below or more than 1 octave above notch.
The circuit is derived from How to calculate LC notch filter - online calculator written by M.Shenberg, from an article in "Radio" magazine, May 1956.
I have added the "tune" part of the circuit to accommodate +- 0.5% changes in the frequency to be notched.
OPTIONAL
For the circuit shown in the attachment, the following are guidelines if you want to change the circuit.
C1 + C2 and L1 should resonate at notch frequency. I chose L1 arbitrarily - it was the better part of a 1/2 lb spool of #28AWG magnet wire (air cored). If you want to notch at different frequencies then follow the standard rules of scaling for LC circuits: where X is ratio of frequency new/old: Lnew=Lold/X Cnew=Cold/X
.R1 is the resistance value that gives a null
.R1=(pi * f * L1 * Q1)/2 Q1 is Q of L1
.In real life, R1 ranges 1k - 5k ohms
.R1=R1a + R1b (notch adj., fine & coarse)
.R2 should be roughly 10x R1
.C3 should be roughly 5% of C1 or C2
.SPICE simulations work reasonably well, except for modelling the effects of nearby metallic objects ,and radiated fields.
If you plan to use the notch circuit to its extreme rejection, you should use 10 turn pots for R1a, R1b, R2.
R1b is used to get the notch in the center of the range for R1a. If you can't get 10 turn pots, then add lower value pot in series for finer adjustments. The null depends on the stability of your 1khz signal and the temperature coefficients of the components.
L1 should be air cored. Any ferromagnetic core will cause distortion in the notched output. Avoid ferromagnetic case and hardware. An aluminum case will drop the values of inductance for L1, so C1 and C2 need to be changed accordingly. I had to add more than 15 nF because of my aluminum case. I didn't see any distortion effects from the aluminum case (down to my measurement limit of -170dbV). Try to keep the coil more than 2 inches (5 cm) away from the case surface.
R1a, R1b, R2 should have low contact noise, or it will be hard to null.
The load impedance should be higher than 50kohms. Devices on the notch input will "see" 2 * R1 in parallel with R2 and the load impedance.
The source impedance should be lower then R1/10. Devices on the notch output will "see" 2*R1 in parallel with R2 at resonance. Away from resonance, output devices will see the impedance feeding the notch circuit. That will give you an idea of how to get best noise performance of your preamp on the notch output.
Non-optimal source and load impedances will affect notch sharpness.
The bad side of this filter is related to the air-cored inductor. It will pick up the power line frequency and its harmonics, plus any switching noise of your instrumentation. The filters noise pickup is position and orientation sensitive. A .5-1 KHz 2nd or 3rd order hi-pass filter is recommended before your analyzer.
Since you are looking for discrete frequencies (2 KHz, 3 KHz, etc) you can use very narrow bandwidth functions on your analyzer to reduce the effects of noise and extraneous interference. If your analyzer is tuned to a harmonic, you can make quick pass-nogo testing of component distortion.
Split-C vs parallel-T
.....................................
Split-C: .better notch rejection
.Narrower band
.lower pass-thru impedance (less thermal noise)
.all passive
.no measurable distortion (air core, & good C's)
.can be tuned (+- 0.5%)
.will pick up stray magnetic fields
Parallel-T: .much more compact
.requires precision or matched components
.active circuit for narrower bandwidth
.no coils, no magnetic sensitivity
.circuit needs modification for tuning and notch depth
.no distortion if no active ckt, & good R's & C's
.High impedance (typically)
Here is a notch circuit (in attachment) that gives a sharp null for a not-so-great-Q inductor. The values shown will null 1 KHz, to better than 100db, and negligible insertion loss if less than 1 octave below or more than 1 octave above notch.
The circuit is derived from How to calculate LC notch filter - online calculator written by M.Shenberg, from an article in "Radio" magazine, May 1956.
I have added the "tune" part of the circuit to accommodate +- 0.5% changes in the frequency to be notched.
OPTIONAL
For the circuit shown in the attachment, the following are guidelines if you want to change the circuit.
C1 + C2 and L1 should resonate at notch frequency. I chose L1 arbitrarily - it was the better part of a 1/2 lb spool of #28AWG magnet wire (air cored). If you want to notch at different frequencies then follow the standard rules of scaling for LC circuits: where X is ratio of frequency new/old: Lnew=Lold/X Cnew=Cold/X
.R1 is the resistance value that gives a null
.R1=(pi * f * L1 * Q1)/2 Q1 is Q of L1
.In real life, R1 ranges 1k - 5k ohms
.R1=R1a + R1b (notch adj., fine & coarse)
.R2 should be roughly 10x R1
.C3 should be roughly 5% of C1 or C2
.SPICE simulations work reasonably well, except for modelling the effects of nearby metallic objects ,and radiated fields.
If you plan to use the notch circuit to its extreme rejection, you should use 10 turn pots for R1a, R1b, R2.
R1b is used to get the notch in the center of the range for R1a. If you can't get 10 turn pots, then add lower value pot in series for finer adjustments. The null depends on the stability of your 1khz signal and the temperature coefficients of the components.
L1 should be air cored. Any ferromagnetic core will cause distortion in the notched output. Avoid ferromagnetic case and hardware. An aluminum case will drop the values of inductance for L1, so C1 and C2 need to be changed accordingly. I had to add more than 15 nF because of my aluminum case. I didn't see any distortion effects from the aluminum case (down to my measurement limit of -170dbV). Try to keep the coil more than 2 inches (5 cm) away from the case surface.
R1a, R1b, R2 should have low contact noise, or it will be hard to null.
The load impedance should be higher than 50kohms. Devices on the notch input will "see" 2 * R1 in parallel with R2 and the load impedance.
The source impedance should be lower then R1/10. Devices on the notch output will "see" 2*R1 in parallel with R2 at resonance. Away from resonance, output devices will see the impedance feeding the notch circuit. That will give you an idea of how to get best noise performance of your preamp on the notch output.
Non-optimal source and load impedances will affect notch sharpness.
The bad side of this filter is related to the air-cored inductor. It will pick up the power line frequency and its harmonics, plus any switching noise of your instrumentation. The filters noise pickup is position and orientation sensitive. A .5-1 KHz 2nd or 3rd order hi-pass filter is recommended before your analyzer.
Since you are looking for discrete frequencies (2 KHz, 3 KHz, etc) you can use very narrow bandwidth functions on your analyzer to reduce the effects of noise and extraneous interference. If your analyzer is tuned to a harmonic, you can make quick pass-nogo testing of component distortion.
Split-C vs parallel-T
.....................................
Split-C: .better notch rejection
.Narrower band
.lower pass-thru impedance (less thermal noise)
.all passive
.no measurable distortion (air core, & good C's)
.can be tuned (+- 0.5%)
.will pick up stray magnetic fields
Parallel-T: .much more compact
.requires precision or matched components
.active circuit for narrower bandwidth
.no coils, no magnetic sensitivity
.circuit needs modification for tuning and notch depth
.no distortion if no active ckt, & good R's & C's
.High impedance (typically)