For starters, if your amp is sensitive to the load then any load that is different from the "actual load" will not give you the results you want. While an actual speaker (in the actual enclosure) will provide a complex load, any other speaker in any other enclosure will provide a different complex load. In essence, the graph of impedance vs. frequency will be different for the two loads-- in some cases significantly different. All of the notches and peaks in the graph will be at different places.
If your amp is sensitive to the load, then placing a significantly different load on it from the actual load will be no more valid than using a simple resistor load.
You should ask yourself if an amp should be sensitive to the load. Most people (including myself) would say that it should not be. But there are some niche applications where some other attribute is more important than audio clarity-- like low power or intentional distortion.
One important reason to not have an amp sensitive to the load is that the load will change over time. Speakers age. Components age. Sometimes speakers get damaged and replaced. If your amp is dependent on the speaker then it will not be guaranteed to work over time.
If your amp is sensitive to the load then the audio quality will also be load dependent. That means that you must test with the actual speaker. Additionally, the enclosure that you put the speaker in will change the loading-- so you must put the speaker in the final enclosure.
In my opinion, there is no point in making a dummy load for this amp. The dummy load will not match the final load, so you don't know if your amp will be stable or even work properly with the final load. You will also not know if it sounds good with the final load. Use the actual final load, or just a simple resistor load.
If you choose to use the final load, but you don't want to be blasting out noise all of the time then put the speaker (plus enclosure) into a sound-isolating chamber. The bigger the chamber the better, but it should be at least 5x the volume of the speaker enclosure. A "cheap" way would be to buy a used refrigerator and line the inside with rigid fiberglass panels. A used fridge can cost as little as US$150, or much less if it is broken.
Assume that you have a conventional permanent-magnet eight-ohm loudspeaker. Your measurements will be roughly close to absolute value of Zspkr. Be aware that your results will change upon mounting (or dis-mounting) your speaker in an enclosure. Here is a simplified equivalent circuit for a loudspeaker at low audio frequencies, where cone motion is uniformly in and out:
simulate this circuit – Schematic created using CircuitLab
R1 is the resistance of the wire on the speaker's voice coil that reacts with the magnetic field produced by the permanent magnet. R1's resistance of about six ohms is common for a speaker considered to be " 8 ohms". You can measure this resistance with an ohmmeter.
Since your speaker operates like a motor, it can generate voltage as well as respond to applied current. The components R2, C1, L1 are not real, but they model the back-emf effect of the voice coil's motion in a magnetic field. They are also affected by the mechanical mounting compliance of the voice coil, and the mass of the moving parts. This is a simplified model, because the power transfer from electrical current to the motion of air is not shown - I'm assuming that you want to see the impedance between the speaker's two terminals.
Your loudspeaker goes through a low-frequency resonance (perhaps below 400 Hz) where C1 and L1's impedance cancel, leaving you with a measurement of R1 + R2. You might measure about 20 ohms at resonance using your method.
Above this frequency, your measurements will yield a lower value than at resonance, which dips close to eight ohms, then rises again at still higher frequencies. This model is not adequate to properly represent speaker impedance for these frequencies, deviating considerably at high frequency.
Resonance modelled by R2, C1, L1 is considerably affected by the air path/load on the speaker cone. Model value for R2 also depend on the permanent magnet's strength.
Your measurement method gives results that are very nearly correct at three frequencies where phase angle between voltage and current is near zero:
- at low frequency near 0 Hz
- at resonant frequency
- above resonance, where impedance dips to a minimum
To measure speaker impedance more accurately, you would measure voltage from one speaker terminal to the other, and you would measure speaker current, paying attention to the phase relationship between these two measurements. A two-channel oscilloscope would be a useful measurement tool.
Best Answer
For an explanation of the flattening of the post-resonance increase in impedance, take a look here - for a slightly less easy-to-read explanation of how to flatten the resonance peak (and the post-resonance ramp), take a look here.
BTW - I found both of these using Uncle Google and I learned something in the process.