Hi GTNubi,
zum Verständnis hier ein Kommentar des Entwicklers zur Funktion des AM und die Einbindung über die Phaseneinstellung.
Das Problem mit der Beschreibung von solchen akustischen Geräten ist: Entweder ist die Erklärung höchst akademisch (bzw. Betriebsgeheimnis des Herstellers), oder naiv platt. Ich habe noch keinen Mittelweg gefunden.
Dear all,
I have also added the subwoofer-main speakers integration advice to the FAQ's, since it is also one of the most common source of misunderstandigs. In a nutshell, it doesn't matter what EQ is used for the subwoofer, as the only parameters that need to be attended for the integration are phase, level and cross-over. These all are being controlled by any AVR.
Generally, no subwoofer EQ can match the phase optimally to the mains, because a subwoofer is always "late" because of it's moving mass and reflex port acoustic reactance. While this can be compensated in frequency domain with adjustable phase on the subwoofer or subwoofer EQ, it will not time-align the subwoofer to the main speakers optimally. To time-align the subwoofer to the main speakers seamlessly, main speakers must be delayed to match the slower output of the subwoofer, and this is practically what happens in every AV-Receiver. Except the delay is actually called "distance" in the AVR. Adjusting the distance is exactly the same as adjusting total group delay, which is exactly the sam eas adjusting the derivative of the phase function. The distance is not hence always (=never with any extra A/D -DA loop) the physical distance of the subwoofer to the mic, but more of a distance equal to the acoustic delay it introduces (and must be compensated by delaying the mains). Of course, in addition the level and cross-over needs also be matched for the best result, which is a piece of cake for any modern AV-Receiver.
Also, I get a lot PM's about FIR vs. IIR on other forums, so while it is definitely off-topic to put in tohe 8033 FAQ, it is perhaps something that needs to be uncovered. I will answer every PM I ever get (eventually), but a pre-emptive strike is better than answering the same answer to the same question on daily basis, so here we go (apologies for the marathon post):
IIR (Infinite Impulse Response) means that the filter has poles in Z-domain, and feedback coefficients in it's computaional structure. It can be seen as the denominator of the transfer function polynomial. FIR (Finite impulse response) on the other hand has zeros in Z-domain, and only feedforward coefficients in the computational structure. Usually the IIR filters also have FIR part, and if they don't they are called all-pole structures. So where do you use them to take the most advantage?
In low-frequency correction, the group delay of the system must not be excessive, but the needed corrections require quite steep filters. This leads to one conclusion that using the most basic straight forward linear-phase FIR's are pretty much out of the question, because while they have some nice properties, they would inflict group delay that cannot be tolerated in real-time system. You can use linear phase structure for example in digital cross-overs at higher frequency (typically midbass/tweeter), and this is what I preferred to use in the XO of the DSPeaker Servo 299. It should be noted, that LP-FIR does not remove the amount of 'echoing' or 'ringing' in non-amplitude-complement (for example power complementary XO) systems. It does have less post-ringing, but it is in a way shifted to the pre-ringing. In exactly amplitude complement systems this would of course be not problem, because the two pre- and post-ringings would cancel out each other, at least on the acoustical axis.
Excluding the LP-FIR doesn not exclude the using of FIR's in general on low frequency, since any FIR can be minimum-phase (have all it's zeros inside the unit circle) or almost-minimal (having most of the zeros inside unit circle), which will give minimal group delay (for example the anti-bafflestep EQ in DSPeaker Servo 299). MP-FIR's have nice properties like easy interpolation between both the frequency and time-domain (even linear interpolation between the coefficients will do in most cases). This is why it is popular for real-time interpolating applications, like dynamic HRTF-auralizations (like DSPeaker HeaDSPeaker 5.1). In addition, multiple MP-FIR channels can be time-aligned easily (also as a part of DSPeaker HeaDSPeaker 5.1 HTAR-algorithm)
Also, all FIR's in general can easily be computed in Multirate-structure with either sample-rate conversion with subsampling or more clever cyclic structure (again, like in DSPeaker Servo 299). The first approach introduces aliasing, which can be removed with appropriate anti-alias filter. The later introduces less harmful spectral imaging, which can be removed completely by Anti-Image filter (natural A-I filter in DSP-based loudspeakers is the cross-over).
However, at the point where the all-zero structure (strictly feedforward) will start to attend low-frequency modal issues without unbearable lag, they are actually imitating (simulating the impulse response)of more complicated IIR structures. The reason for this is, that a Z-domain representation of the room has both poles and zeros, and a counter model must thus have both of them in order to affect decay character properly. Adding a pole to even a long FIR will turn it to IIR, even if FIR order (nominator poly) is thousands and the IIR order (denominator) less than 3. What is usually forgotten when compared the amount of filters is, the order of the filter. Ten 5th order cascaded FIR's are exactly same (you can even pre-convolve them) as one 49th. order FIR. Same math goes with IIR, but the comparation of FIR order to IIR order is more of a 'waste of time'. Even 6th order IIR is superior to 21st order single-rate FIR in terms of frequency behaviour (steepness, stop-band rejection etc.), while the 21st order FIR can give more independent phase characteristics. So it's pretty much oranges and apples comparison.
While FIR's are a lot easier and more efficient to code with VSDSP-assembler (or any Multiply-Accumulate optimized DSP) compared to IIR's, they require more memory. This can be addressed by choosing a better (lower) sampling rate for sub-only systems. On the other hand, while linear interpolation between MP-FIR coefficients is appealing solution for multipoint-compensation, it is not the optimal one in terms of how many measurements must be done, and how their relative position to each other must be known by the algorithm.
Approaching the problem on Z-domain will lead to more robust method that can be viewed as spline-interpolation in Z-domain, which is essentially what 8033AM is using in it's Gradient-method calibration. This measurement will reveal the characteristics of the larger area with only one additional, well-placed measurement (well-placing meaning a place the algo can relate to the other point = toward all surfaces from the first point in Anti-Mode).
However, no equalizer will correct the listening area ideally to everywhere, because the degrees of freedoms simply don't allow for such. If at point A you have 20dB modal resonance, and at point B there is not a sight of it, there is no one solution to correct the 20dB away from point A without putting a hole in to the point B. The is only what can be described as "Best effort", which is to correct the point A to some extent, while minimally ruining the point B. This leads to a concept of "psychoacoustic cost function", which is common in every larger area equalizer, including 8033AM.
So to sum it up, talking about IIR and FIR for each DSP-application is pretty much evaluating the tips of the ice-berg to speculate the underwater structure. And yes, 8033 has both FIR's and IIR's, because the real-life room has both poles and zeros.
. Ich frage mich immer noch, wie das Dröhnen durch das Gerät vermindert wird, bzw. das Überlagern einer bestimmten Frequenz. Nur durch das Absenken der Lautstärke? Das kann ja noch nicht alles sein, oder? Weil Dröhen würde es ja auch bei geringerer Lautstärke. Ich würde es einfach gerne besser verstehen.
