Wrong data for frequency response in 35670A

[Rmano]

First of all: thanks for the great code, it has been quite useful so far (the problem below is easily corrected after-the-fact...

In

sdfascii/sdfascii.py

Lines 983 to 987 in 255b2b2

	# FIXME: I'm cheating if this is a 35670A measurement and
	# converting from the Vpk^2 (native units) to Vrms. Convert from
	# Vpk^2 to Vpk and then to Vrms
	if sdf_hdr['file_hdr']['application'] == 'HP 35670A':
	sdf_data = np.sqrt(sdf_data) / np.sqrt(2)

the library converts the units of the file unconditionally if the instrument is a 35670A signal analyzer; I am unsure why it does this.

1. V²/Hz is a perfectly valid measurement unit (and by the way, the better one if you need to integrate non-uniform noise on a band, giving that the square root of the integral is not the same as the integral of the square root 😉)
2. in the case that the file is a transfer function from a sweep sine measurement, the result is simply completely wrong.

I suggest not doing the scaling at all (so that the data is coherent with the one exported in ASCII mode); as a second option, do the scaling only if y_is_complex is False...

If you want, I can upload files of different measurement types in ASCII and SDF for checking.

Thank you again for the software.

[matthewrankin]

I was needing Vrms for my application, which is why I cheated for the 35670A.

Could you please upload a few different measurement types?

[Rmano]

Thanks for looking into it!

For now, I have only frequency response (swept-sine) measurements in SDF format. After Easter break, I can try to do a power spectrum and send it.

FreqResp.zip

The data above is the instrument .DAT files and the .csv after the correction; I graphed them with gnuplot

set datafile separator ',';
set logscale x;
set grid;
set ylabel 'Magnitude (dB)';
set xlabel 'Frequency (Hz)';
plot 'TRAC9.csv' using 1:(20*log10($2)) t "1→1" with lines;
replot 'TRAC10.csv' using 1:(20*log10($2)) t "2→1" with lines;
replot 'TRAC11.csv' using 1:(20*log10($2)) t "3→1" with lines;
replot 'TRAC12.csv' using 1:(20*log10($2)) t "4→1" with lines;

Obtaining the following graph which is correct:

I used this program to do the translation:

#!/usr/bin/env python3

import sys
from pathlib import Path
import numpy as np
import sdfascii


def sdf_to_csv(sdf_path, csv_path):
    print("Reading SDF...")
    hdr, data = sdfascii.read_sdf_file(sdf_path)
    print("SDF read done")

    # print("Debug SDF header", hdr)
    # Undo incorrect HP35670A conversion in sdfascii for FRF data
    if (
        hdr["file_hdr"]["application"] == "HP 35670A"
        and hdr["data_hdr"][0]["data_type"] == "Frequency response"
    ):
        data = 2.0 * data**2
        print("Forcibly correct data due to sdflib bug...")

    data = np.asarray(data)
    print(f"Data length: {len(data)}")

    dh = hdr["data_hdr"][0]

    n_points = int(dh["num_points"])
    f0 = float(dh["abscissa_first_x"])
    dx = float(dh["abscissa_delta_x"])
    x_type = dh["x_resolution_type"]

    print(f"n_points={n_points}, f0={f0}, dx={dx}, x_type={x_type}")

    if len(data) != n_points:
        raise ValueError(f"Header says {n_points} points, but data has {len(data)}")

    print("Building frequency axis...")
    if x_type == "Logarithmic":
        freq = f0 * (dx ** np.arange(n_points))
    else:
        freq = f0 + dx * np.arange(n_points)

    print("Computing magnitude and phase...")
    magnitude = np.abs(data)
    phase_deg = np.angle(data, deg=True)

    print("Stacking output...")
    out = np.column_stack((freq, magnitude, phase_deg))

    print(f"Writing CSV to {csv_path} ...")
    np.savetxt(
        csv_path,
        out,
        delimiter=",",
        header="f_Hz,module,phase_deg",
        comments=""
    )
    print("Done")


def main():
    if len(sys.argv) != 2:
        print(f"Usage: {sys.argv[0]} <file.DAT>")
        sys.exit(1)

    sdf_path = Path(sys.argv[1])

    if not sdf_path.exists():
        print(f"Error: file '{sdf_path}' not found")
        sys.exit(1)

    if sdf_path.suffix.lower() != ".dat":
        print("Warning: input file does not have .DAT extension")

    csv_path = sdf_path.with_suffix(".csv")

    sdf_to_csv(str(sdf_path), str(csv_path))


if __name__ == "__main__":
    main()