The residual visualizer seems to work as intended, but how can I zoom out once I’ve zoomed in (by drawing a boundary box around the part I want to zoom in on)?
Are there hot keys to zoom in/out and for any other functions in the visualizer?
The only way I can ‘reset’ the zoom now is by closing the box and relaunching it.
Hi @Rammis, right click in the display area and pick “set scale to default” or “undo all zoom” to get back to where you were. The first is helpful if you have way overzoomed and lost your way.
I seem to get a residual that is 180deg out of phase. This is compared to another application I’m using. Is it possible that there is a flip in the residual?
Under a little closer examination, you can see that the asymmetric residual seems to align correctly to the signal but both are plotted 180deg out like it’s just a neg vs pos in the graph.
Another request for possible improvement with the residual visualizer - would it be possible to display the residual when both tone generators are active? Or even the multitone generator?
Hi Mark, I think theoretically it could. There is a very deep notch for the single tone operation. It could be extended to other tones. But I think the phase would start to make things very hard to interpret. For example, by looking at the residual you can see if the distortion is cross-over related, and maybe just positive or negative crossover. Or clipping, Or slew. But if the phase is skewed up (due to the filter at another frequency) it’s really hard to understand what you are looking at.
I would generate the residual by subtraction, not digital filtering. Generate the pure tone from the signal peak information, subtract sample-by-sample. It can also be done with FFT, chop out the relevant peak(s), IFFT…
A technique I’ve used in the path to measure HD and calculate the residual avoids taking a spectrum and doesn’t need to a priori know the fundamental:
First use zero-crossing to estimate signal freq + phase. Use these to generate a model waveform and correlate it separately with the first half and last half of the measured waveform. The amplitude and phase information from the 2 correlations can be used to refine frequency, phase and amplitude of the model, then repeat with refined estimated values. This is linear time in the number of samples.
I would generate the residual by subtraction, not digital filtering.
Hi @MarkT, yes, lots of options here for sure. The origin of the feature was to mimic the residual output on the back of an HP8903. But next time it’s overhauled it should be a lot more flexible.