Originally Posted by jakeman
The IMD tests are not very meaningful as he lacks the proper equipment and the two-tone test as conducted is likely catching other distortions including THD. Without a Klippel Analyser ( very expensive I know) it would be near impossible to isolate the modulation components using Illka's methodology. It's not that simple since subs with high excursion drivers will usually have greater amounts of non-linearities to contend with, some non-linearities incapable of being measured directly using his procedures. One cannot conclude that subs with low THD have low IMD since despite his attempt its unclear whether THD has been removed from the test results.
I'm keen to learn here. Can you give us some reasons why the Klippel Analyser would be better suited to isolating and measuring modulation components than Ilkka's method?
Without Ilkka being able to respond here himself, it is more than a little churlish to start questioning his methods. So I will restrict myself to a few observations regarding his latest tests and some of the reactions here to them.
Firstly, I am a little concerned that he has pushed a number of these subs into very serious compression. When I test a sub, I generally stop when power compression exceeds 3dB unless I am sure that the sub's limiter can handle it. So far I have had no "casualties" but I am always mindful that many of the subs I test are the property of private individuals. They have the ultimate veto on how far we go. Testing to destruction is not something I would recommend. Hence some of our maximum outputs do not reflect absolute maximum output levels. Measurements of distortion products where they go off the scale are of limited value anyway, I feel.
The IMD tests are interesting, but the problem with measuring IMD in the subwoofer range is that the restricted frequency range available to us means that IMD and HD components do tend to end up overlapping. This seems unavoidable. We need to see more examples to be able to draw any meaningful conclusions as to whether high HD and high IMD do in fact correlate.
As others have noted, the spectral contamination tests do look to be the most interesting innovation he has introduced. But when it comes to any comparison to real-world signals, individual sine wave frequencies separated by a Fibonacci ratio would not correspond with any kind of music that I am familiar with or would ever wish to be. :p
None of these test signals we are using closely mirror any real-world signals that a typical subwoofer would be expected to reproduce. Shaped tone bursts are just another artificial signal. Very short tone-bursts are problematic to record while still maintaining adequate signal-to-noise ratio. Longer sine wave sweeps allow better S/N ratio and accuracy in the results. We used a longer sweep for reverse sine sweep testing at AV:Talk because it was a requirement of the software we are using (WinMLS) to extract the harmonic distortion components from the resulting impulse response while still retaining adequate S/N ratio.
If Ilkka were to plot his harmonic distortion by component and overlay it with the CEA stepped limit thresholds, he would have had to plot separate spectra for each of the 1/3 octave frequencies he chose to plot. That's 7 graphs as opposed to 1. Bandwidth on the hosting site is not limitless.
Since we are on the topic of the CEA-2010 standard, I am slightly puzzled as to why some people seem to think this will advance the state of knowledge very much. As far as I can see, the standard only requires test results to be stated in terms of 2 numbers: maximum SPL in the low bass and ultra-low bass region. There is no requirement to show the underlying graphs or the actual limiting-factor in each case. In the absence of any other published information from manufacturers, this might just be a small step forward. But as a replacement for some of the measurements that reviewers are trying to give you (for free in the case of Ilkka and myself, I might add!) I doubt that it will really give you more insight into the performance of any given sub. :confused:
While the SC plots are interesting, what is very hard to glean from them is what correlation they have to what sounds good. While it is possible to say that the absence of visible "mush" in the output between the original drive frequencies should correlate with "cleaner" reproduction, where there is such mush present, it is impossible to say whether it would sound objectionable or even be perceived at all without doing a lot of corresponding listening tests. :cool: