Automatic cold sky feature
On the Noise Measurement window, you can set the "cold sky" reading either manually or automatic (automatic since version 1.0.0.20). If selecting the automatic mode, SimpleCalc will search and try to find the proper cold sky reading. This will be from the data currently visible on the Noise Measure window, if the window is zoomed, only the visible data will be used for the search.
To use it, right-click the graph and choose Find Cold Sky value automatically from scan to have SimpleCalc determine the cold-sky reference from the recorded data instead of picking a point by hand.
Two situations, handled differently
The function first checks whether there is a Sun in the data at all, by comparing the quietest and busiest parts of the trace. Empty sky varies by a few tenths of a dB; even a weak, narrow Sun spans several dB, so the two cases separate cleanly.
With a Sun present, it looks for the lowest sustained plateau. A window is slid across the data, each window averaged in linear power, and the lowest one sets the reference; other windows within a small tolerance are treated as the same level and included. Because contamination is one-sided — the Sun, RFI and ground pickup only ever add power — the lowest sustained level is the best estimate of empty sky.
With no Sun present, there is no plateau to search for: the whole record is empty sky, so all of it is averaged. This is the normal way to set the reference before the Sun arrives — point at blank sky, record for a minute, and run the function. Searching for a lowest section here would simply track the low side of the slow baseline wander and read slightly low.
Two approaches it deliberately avoids in both modes:
- Averaging the lowest N samples.That selects the downward noise excursions and always reads low, with the error growing as the trace gets noisier.
- Averaging everything below a threshold.The stretch just after the Sun leaves the main lobe is not empty sky — it still carries far-sidelobe and ground pickup decaying away. Including it biases the result high by several tenths of a dB.
All averaging is done in linear power. Averaging dB values would form a geometric mean and read low.
Zoom is respected
Only the samples currently visible are used, and the Sun-present check is re-evaluated on that range. If part of a recording is spoiled by interference, a slew or a second source, zoom to the section you trust and run the function on that alone. Zooming to a clean, flat stretch of sky gives an exact result.
One case needs a little care: if you zoom to a region that still contains part of the Sun's decaying skirt, that range can look flat enough to be taken for empty sky, and the result will read slightly high. A gentle skirt and a slow baseline wander cannot be told apart reliably from the data, so rather than guess, the function warns when the visible range is not flat and suggests zooming to a flatter section. In practice, either leave the view unzoomed — the plateau search handles the skirt correctly on its own — or zoom to a stretch that looks visibly level.
What it reports
Before anything is applied, a dialog shows the level found, how many samples were averaged, the sample noise, and the difference from the cold-sky value currently set. With a Sun present it also shows which samples the plateau was taken from, the peak level and the implied Y-factor; with no Sun it reports the level spread across the range instead. You then choose whether to accept it.
It warns when:
- the visible range is not flat, and may still contain part of the Sun's skirt;
- the quiet stretch is short — barely enough data to average;
- the level differs by more than 0.1 dB before and after the Sun (or, on a Sun-free recording, between the start and end), indicating gain or atmospheric drift;
- the Y-factor is below 3 dB, suggesting the Sun was not properly in the beam. This is only raised when a Sun was actually found, so recording empty sky on its own is never flagged as a failed measurement.
Accuracy
Tested against simulated scans with a known baseline, the value returned is typically within 0.01–0.05 dB of the true level, including on heavily averaged traces where sample-to-sample noise is only a few thousandths of a dB, and on records with slow baseline undulation. The manual Set Cold Sky value to point option remains available if you prefer to choose the level yourself.
Below are a couple of examples of the "automatic cold sky" feature.
First example finds cold sky automatically from the full dataset currently recorded, the second example shows the search done on a zoomed part of the dataset.
The third example the data are zoomed and the peak from the measurement of 50 ohm termination on LNA is visible. In this case, SimpleCalc warns you about this, but still manages to find a plausible cold sky reference.
