Originally Posted by bgavin
I don't have a problem with the noise floor. I have a problem with the peak shifting +10dB. My sources are in the 100 to 136 dB range, so the noise floor is inconsequential to the intensity of the sine wave.
As I understand a swept sine, each point is driven at full voltage. This is an automated method of taking discreet measurements at single frequencies. When I sample at point at 1/24 it correlates with the external meter. The same point at 1/3 is higher by 10dB than the meter.
Your example above shows the 500 Hz peak does not change amplitude with resolution. This is exactly what I wanted from Level 4, but did not get. I notice you are using dBu, and I was using SPL.
Let's go over it again. These measurements are not trivial and take a while to understand. This time I'll use the SPL mode to demonstrate that a measured peak does not change in level as you go from 1 to 1/3 to 1/24 octave resolution. The measurement depends strongly on the nature of the signal present in each frequency bin. A bin with a single strong frequency will remain constant in level but a bin with noise (many frequencies) will fall with increasing resolution as the filter is narrowed. Note that a bin with a mix of a small peak (say a distortion component) and noise will reveal the peak better as the bins narrow and the noise is reduced to reveal the peak. Here is a measurement of the speaker in my notebook PC along with lots of ambient noise to add character to the measurement.
In the 1 octave resolution shot above we see the peak at 300 Hz only generally defined with a hint of a second peak at 60 Hz. The peak reads close to the broadband SPL level of 80 dB. The noise floor appears around 50 dB SPL when seen through 1 octave wide filters.
The 1/6 octave resolution shot above reveals the presence of a 2nd harmonic distortion component at 600 Hz and perhaps a 4th harmonic at 1200 Hz. A broad peak is seen near 60 Hz. The 300 Hz peak is seen unchanged at 80 dB SPL but the 60 Hz peak has fallen in apparent level from about 57 dB in the 1 octave measurement to about 53 dB in the 1/6 octave measurement. The noise floor appears around 44 dB SPL when seen through 1/6 octave filters.
Finally, in the 1/24 octave resolution shot above we see the 2nd and 4th harmonics sharply revealed. We also see that the 60 Hz peak is actually two separate peaks at 50 and 60 Hz. Note that the 300 Hz peak remains UNCHANGED at 80 dB SPL while the 2nd harmonic level also remains constant as the 4th drops slightly as noise is eliminated from its bin. The noise floor appears around 38 dB SPL when seen through 1/24 octave filters.
Most importantly, note that the peak amplitudes remains constant as the resolution is changed. The noise floor drops at higher resolution to reveal the smaller peaks. It is easier to see what is happening here than it is with a pink noise response so in a subsequent post I will show what happens with a pink noise response as resolution is changed. I'l also show how to shift a pink noise measurement to calibrate it up to a fixed pitch SPL measurement.
A Personal Note to TrueRTA Users:
Meanwhile, I am busily working on a maintenance release of TrueRTA to fix a verifiable bug affecting a major industrial user. When reliable users report bugs that I can reproduce I take them very seriously even if I can't fix them immediately. I'm not sure the minor "import bug" (discussed above) will be addressed in this release but I will be investigating it further.
You may have noticed that as a small software publisher I do all my own customer support. This takes a lot of time and sometimes my responses are delayed more than you (and I) would like. I really enjoy making powerful audio software available at a price the represenst a high value to users. But I am sometimes tempted to charge a separate fee for support! I would much rather spend my time expanding the software. That is why I am highly appreciative of the AVS Forum hosting this thread, and for you TrueRTA users supporting each other here as you do. This allows me to spend more time developing new software. I actually have in the works a powerful new impulse analyzer that will allow analysis of signal arrival times and reflections. But this sort of project requires focused concentration with lots of quiet time to work out all the details.
I want to offer a warm Thank You to all TrueRTA users for helping to make this product sucessful and supporting my small software business.