Originally Posted by bossobass
Originally Posted by Mark Seaton
While most drivers with sufficient excursion capability can be made to behave similarly with enough EQ, especially when using 16 of them, as soon as you push them a little there are differences related to what the behavior is before EQ is applied.
You can't increase Bl and hold all else the same.
How exactly do you push a 16 driver system in you home theater? I contend that it's they which will push you.
Looking back I see had missed this question. My last statement here was misunderstood and not clearly stated in that I was agreeing that when using many drivers (I said 16 to make the point), most drivers are very consistent as they rarely get past a fraction of Xmax. In the latter part I should have said "as soon as you are using fewer and need to push them a little (lets say use to 1/2 to full Xmax), differences related to the native response and Fb."
I get the sense that some of the angst was coming from thinking I was suggesting that this is an issue no matter how many drivers are used, which is not what was intended. My point was that most drivers are far from ideal and when using somewhere between 1/4-3/4 of Xmax or some level of current through the coil, behavior changes and can be audible enough to create measurable and subjective differences. An extreme example of this is comparing a compact sealed box vs. an IB when both are EQ'd to the same frequency response.
Originally Posted by bossobass
Originally Posted by Mark Seaton
I say that to prompt others to take a deeper look on their own and come to their own conclusions, not proclaim some ideal solution or conclusion for them. Statements of absolutes are not my style as there are always multiple factors to consider, and most real phenomena onset gradually rather than hitting a specific point when things change dramatically. The argument of where on such curves is "good enough" or inaudible is fuzzy and open for debate and plenty of varying opinions.
Whenever response changes with level the fingerprint of the original driver is observed. While it would require a screen capture to directly post here, 2 good examples of this are found at Data-Bass by selecting the "Extended Charts" tab for the measurements of the Stereo Integrity HT18D2
and the Fi SSD 18D2
. Looking at the lower "Long Term Compression (Magnitude)" graphs the Stereo Integrity is rather consistent at lower levels where the driver starts it's more obvious change in response shape around and past the nominal "110dB" curve. By comparison, the same graph of the SSD 18D2 shows more variation at low levels and start where the driver starts it's more obvious change in response shape past the nominal "105dB" and "110dB" curves.
While the SSD looks like it compresses twice as much at first glance, a check of the levels shows it was driven harder into overload. Separate from any argument of preference is the fact that the *shape* of the dynamic change is different, and they will sound different. The detail not easy to quantify from these graphs is what happens when high excursion from low frequency content is coincident with content above ~40Hz, which will make some of these problems onset much sooner, and introduce new variations which the single swept sine doesn't test.
Although I get the point when citing single driver systems with no signal shaping and being fed sine sweeps outdoors and I agree that these do not properly examine the DUT with actual complex program, these are snapshots and thus misleading.
The point is that signal shaping can only be applied to one of the curves seen in those outdoor measurements. The relative changes between the curves remain.
In-room is what matters and designing the system for the room is the task. Sorry, but your statements on the subject do indeed give the impression that reversion to anechoic response is a fact of life "when pushed".
I never said the response reverts to the anechoic response, just that the changes are related to driver non-linearities and Fb of the system.
notnyt's sweeps and my own show no reversion to anechoic response to desired playback under sine sweeps.
And with extremely difficult actual program played back at the desired level, again, no reversion to anechoic response:
And these are way too coarse a display to see what is going on from curve to curve. A direct subtraction of the data between the curves is required to see finer details of what is or is no changing between curves. The spectrograms can be very interesting, but would tell dramatically more if they were instead difference measurements showing the deviation rather than the total result where we can only see distortion or omission in areas of limited content with a fairly coarse scale.
Thinking the response effectively reverts back to the anechoic response suggests a misunderstanding of what the anechoic measurements are really showing.
At some point in raising the level 5dB more enough times, any amp will clip. Of course with a constant Voltage input vs. frequency it clips first where the EQ's transfer function has the highest gain. The misunderstanding comes from seeing that we can keep driving the upper range in a sweep that only tests a single moving sine wave through the frequency range. In real use, it doesn't matter if 80Hz causes the clipping or 10Hz, the amp has no more to give at that point, so there is no room for the upper range to keep going further. Adding a limiter only changes how the amp sounds when it approaches clipping. Assuming it's a limiter of max Voltage and not a more complex function that is frequency dependent, at some point the level stops getting louder as the limiter engages or the amp clips. Amp clipping is a form of limiting; it just often sounds worse if you get there. Any signal input that hits the limits caps anything else coming through at the same time. There most certainly are devices with more complex limiting, but it is hard if not impossible to decipher which is being used when only looking at a set of sine sweeps.
If response changes with level, the hardware is not up to the task at hand. Continually focussing on what to expect with that failure to execute seems pointless to me. It bears mentioning as 'what not to do', but the discussion should definitely be more of 'what to do'.
"What to do" when gross excess is not practical comes back to understanding the problem and what causes it. From that it is possible to look at outdoor measurements to evaluate the more ideal limits and how much things change as you get past those ideal limits. The 2 drivers I grabbed for reference both have similar Xmax but show notably different changes in their frequency response. Assuming much of their Xmax is put to use (so as to not be wasteful of resources) they will have somewhat different sonic signatures. For those not using many drivers at a fraction of their maximum capability, these differences can matter, and in the complex, modulated situation of real program material, the differences hinted at in the incremental sine sweeps are magnified.