◀ Back to contents

Tune-notch wizard

For how the measurement behind this display works - a played logarithmic sweep deconvolved against the reference to recover the notch's response in one pass - see Theory of operation ▸ Frequency response.

A passive notch filter (typically a twin-T) ahead of the input is the classic way to measure distortion below the analyser's own floor: it removes the fundamental before the ADC, so the FFT sees the harmonics without the converter's own distortion products. But the notch only works if its null sits exactly on the test frequency - and on a twin-T the centre and the depth move together as you turn the trimmers. The wizard gives you a live, continuously refreshing picture of the notch so you can watch the null move while you adjust the filter with a screwdriver in hand.

Open it from the main menu: Tools ▸ Tune notch.... While the dialog is open it takes over playback and capture - the generator, oscilloscope, FFT and any frequency-response sweep stop for the duration. Wire the signal path as for the later distortion measurement: output ▸ notch filter ▸ input.

All four values are remembered between sessions. Editing the start / stop fields re-scales the frequency axis immediately.

What you see. The plot is rebuilt roughly ten times a second from a short looping sweep (about a quarter of a second per pass) - each turn of a trimmer shows up within a fraction of a second. Both input channels are measured on every sweep; the L / R buttons at the top-left of the plot choose which channel is drawn, while the other keeps being measured in the background. The vertical axis fits itself to the trace automatically. Three live elements sit on the plot:

1. Notch readout (top-right corner): the frequency of the deepest null found in the current sweep and its depth in dB, for the displayed channel. It is the deepest point of the drawn curve, found by scanning the trace's own pixel columns, so the number always agrees with the dip you can see.
2. Target marker: a dashed vertical line at the target frequency. Its colour is graded relative to the null this sweep actually reached: within 1 dB of the notch floor it is green - the target is in the notch - and it fades to red at 20 dB above the floor, linearly in between. The colour reflects the attenuation at the target frequency, not at wherever the null currently sits - a deep notch parked at the wrong frequency still shows red. Because the scale follows each sweep's own floor, a network that can only reach −70 dB still turns green when it is trimmed onto the target; there is no absolute depth to reach.
3. Status line: the running sweep count and the progress toward the next result.

Tuning workflow. Set the target to your distortion-test frequency and bracket it with the start / stop band. Turn the notch's frequency trimmer until the readout's null frequency lands on the target and the marker turns green, then adjust the balance / depth trimmer for maximum depth; on a twin-T the two interact, so iterate a couple of times. When the null sits on the target at full depth, close the wizard and first measure the notch's own frequency response with a normal sweep, saving it to a .frc file. Load that as a calibration in the FFT pane so the notch's shape is de-embedded from the reading, then run the distortion measurement there. The measurement behind this display is the same played-sweep deconvolution as a normal frequency-response run - see Theory ▸ Frequency response.