Originally Posted by Pete_Hsu
As promised, attached are some frequency response sweeps that I performed in our demo room using a 15" non-servo vented subwoofer. I did sweeps at 80dB, 90dB, 100dB, and 110dB (the point at which the whole room was shaking like crazy, and there was up to 2dB output compression at some frequencies) in succession. Then I immediately re-did the sweep at 80dB, 90dB, 100dB in succession. As you can see, the frequency response is essentially identical between the first run and second run. All things considered, the thermal memory effect with a non-servo subwoofer seems to depend on individual design details, and at least in this sample case seems to have very minimal effect on the measured frequency response. Certainly a very interesting and thought-provoking debate though.
Thanks for doing the test. But the memory effect is not static. It is not between sweeps. Memory effect happens in real time, on the fly. I asked Josh to do a test like that is to make use of existing testing tools. But the best way to observe thermal memory is to device a special waveform that emulates the dynamic changing nature of music. I come up with a simple waveform with two signal strengths at 80hz to represent alternating strong and weak signal in music. The correct waveform is supposed to look like this with two distinctive signal levels.
The same waveform is repeated for 8 times consecutively without interrupt. I call each an 1/8th waveform.
Now, the nonservo first 1/8th waveform looks like this
Notice the large signal strength continue to go down due to compression and then it hands off to the small signal in the 2nd 1/8th waveform.
Two interesting things happen here. First, the small signal will go through an expansion (opposite of compression as the beginning temperature is higher and then it gradually cool down). And then after that another period of large signal comes in. Note there the initial strength of the large signal will depend on how long it has been in the small signal (or cool down) period. That creates a highly complex memory effect that is very undesirable. So in real world, the temperature goes through this heat up and then cool down and corresponds to this compression and then expansion process. Here I use small signal as an alternative signal. If I had use a signal that consumes very little current and yet cause large cone excursion (such as the signal located at the impedance peak point), you can imagine that yet creates a different compression/expansion pattern because the temperature recovers very fast.
To conclude here. I add 3rd 1/8th waveform.
and last 1/8th waveform
Note the scale now has dropped 10%.
Here I want to demonstrate is that compression and thermal memory effect is a complex behavior.
One the other hand, the first and last 1/8th waveform for servo