If Rs is small compared to the reactance of the leakage inductance, then a peak in the response can appear. The actual situation in a transformer is more complex than shown here, since the leakage inductance and stray capacitance is distributed throughout the windings. Depending on the winding structure, there may be local resonant circuits that can give a complex frequency response with many peaks and valleys.

    It is important that the output voltage level at the oscillator output be held constant over all frequencies tested. Wein-bridge oscillators without automatic level control tend to vary in their output. Even oscillators with output leveling, such as in the Sound Technology 1700B, need to have their levels checked, since their output impedance (600 ohms for the 1700B) is often a significant fraction of Rs. By keeping the level constant, the internal impedance of the generator can be ignored for this kind of frequency response test.
    To accurately test the low frequency behavior of a transformer, full signal strength (hundreds of volts on the primary) is required and, in the case of conventional air-gapped single-ended transformers, DC current must be applied. A rough test at low signal levels can be done with the test set-up of fig. 1. High frequency measurements are much less affected by signal level and DC current (especially since the leakage flux, which causes leakage inductance, is mostly outside of the iron core), and so these tests are pretty accurate with the fig. 1 set-up. However, these were the measurements that were this time very inconsistent.
    Before jumping into the reasons for the measurement differences, let's see what an output transformer looks like at high frequencies. Figure 2 shows the major parasitic effects: leakage inductance and stray capacitance. The leakage inductance in combination with Rs forms a low-pass filter when it sees capacitance to ground. Transformer designers reduce the leakage inductance by interleaving the primary and secondary, but this tends to increase the primary-secondary capacitance.

    Let's also look at the output circuits of audio oscillators and the input circuits of audio level meters. The equipment examined are the H-P 200CD and 400L, the Sound Technology 1700B, and the Audio Precision System 2. Their circuits are representative of most audio test equipment.

    The output section of the Sound Technology oscillator is shown in figure 3. It has the option of "floating" the internal analog ground from the chassis (and thus power line safety) ground, but the whole internal instrument uses the analog ground, so the signal isolation is questionable. Most inexpensive oscillators use this circuit, but without the ground isolation.