Amazing video! Thanks!
I am curious how this applies to high end dsp / digital room corrections. As it seems like Dirac and other high end DRC really stress that they use FIR and IIR filters to help with the phase of each speaker. I get how having symmetrical phase in the L and R speaker is important (just wire one of your speakers out of phase to see the effect). But my understanding is that as long as both speakers have the same phase then its ok and not worth chasing (which this video also agrees with). So is there any utility to high end room correction and their focus on phase? Or what am I missing about DRC when they talk about phase correction?
Much like
@KSTR has said already. However, I would like to add one major difference when it comes to DSP/DRC and the use of FIR filtering especially at low frequencies in rooms. Unlike "most" DSP/DRC software for rooms, Acourate, Audiolense and Focus Fidelity (plus Denis open source DRC) provide excess phase correction at low frequencies. Which I contend is audible in the sense of the bass response not being clear sounding, aside from regular room modes.
While room modes are mostly minimum phase, there are low frequency room reflections that are not minimum phase. A good explanation is John Mulcahy's doc on
minimum phase and section on, "a common cause of non-minimum phase behavior in rooms." Here is a practical example of that in my room using a 3 way stereo triamped system looking at the step response:
It is a little hard to read as we are seeing the responses of the individual drivers. First the treble arrives, then the bass/midrange driver and finally the subs, inverted phase to boot. See the part marked maximum phase? This is where low frequencies from sub have added together in the room to produce a huge peak that is higher in amplitude than the direct sound. How do we hear that? It is heard as unclear bass response as we are getting the cue from the direct sound and then a larger cue many milliseconds after the direct sound has arrived. Aside from large specialized bass traps (i.e. Helmholtz resonators) the "only" way to fix this is through the use of using excess phase correction which is "only" available through FIR filtering. IIR filters do not have the capability of excess phase correction.
With the large number of FIR taps required (65,536 or even 131,072 taps) for excess phase correction at low frequencies results in latency, typically around 3/4 of a second. This is one of the reasons why you see other DSP/DRC software use IIR filters at low frequencies so they don't have to deal with the latency issue, especially if the application of DRC is for movies where lip synch is the issue. However, players like JRiver can account for that and delay the video by the calculated FIR delay. The other reason that you don't see FIR filtering at low frequencies in h/w devices is that the DSP chips aboard have a real limitation in the number of FIR filter taps available. Typically 1024 or 2048 taps per channel which translates from a FIR filtering perspective to having 2 bands or eq below 100 Hz. Not very effective, especially where you need it the most.
Aside from the time alignment of drivers from my example above and using linear phase digital XO's so they sum properly in both the frequency and time domain, we see a cleaned up (textbook) step response. No more maximum phase peak:
The result is not only smooth bass, but bass that is crystal clear. That is the big difference between IIR DSP/DRC versus SOTA software like I mentioned above. Not only is the proof in the listening, but also consider this measurement at the listening position of the same system:
This is using REW with the default 500ms window letting all of the reflections in and no smoothing. I am using a FIR filter with 800ms of excess phase correction at 10 Hz which becomes progressively less as we move up in frequency, past Schroeder and into the diffusion zone and then not correcting any room reflections past that. See how the bass phase response is flat up to about 4 x my Schroeder frequency? This is because there are no low frequency reflections messing up the clarity of the bass at the listening position, aside from the smooth magnitude response.
Based on my testing and listening of DSP/DRC products over the past 10 years, there are very few software products that actually do this right and can produce the result you see above. While FIR filtering may be too processor intensive for AVR's and PrePro's (aside from the DSP chip limitation of how large a FIR filter can be hosted), a low power PC like an i3 2 GHz processor works just fine.
Pro tip: if you want the very best bass response from your system, use DSP/DRC software that is capable of providing excess phase correction at low frequencies.
