- I always wondered about the Sigtechs time behaviour.
The waterfall describes the changes of frequency responses in the time domain. So a reversed waterfall requires a correction that changes over time.
Thinking of the available DSP at the time of the sigtech it was highly questionable that there was anything else but a constant filter.
I found this picture http://www.stereophile.com/images/archivesart/1092AWSIfig5.jpg
in the stereophile article about the Sigtech http://www.stereophile.com/content/i-have-heard-futuredsp-room-acoustics-correction-page-2
and this mentioned picture shows the filter with some change in bass in the background which appears incredible to me. Maybe I have not understood the DSP filter tap programming and someone can explain this to me. Maybe it is filter preringing?
In my understanding the DSP allows a number of filter taps and these can be programmed with a specific level factor which leads to a constant (time domain) frequency response without any dynamic behavior, no change within millseconds. This is clearly visible in the aforementioned picture (if we ignore the early bass).
Comparing the uncorrected response of the DALI Grand (http://www.regonaudio.com/Dali%20fruncor.jpg) with the corrected response shown here
I see a quite linear response from 52Hz to 16kHz at the identic level.
This indicates a linear target curve chosen.
The uncorrected speaker has the usual room effects in bass and a tilted response estimates -2dB@10k (0dB@150Hz)
The description of the sound points out the sound tends towards ideal neutrality then.
Ideal ? The linear bass together with non-tilted target curve lead to a "tinny" sound with is far from pleasant listening and it is far from realistic measurement in an echoic environment too.
The early Room correction with a single DSP.
I have no comparison with the Sigtech but for a response free from allison dips that cannot be corrected, flush-mounted speakers are mandatory or (corner-)woofers that fire against the wall as part of a 2.2 System.
The widely used TacT 2.2X and 2.0S allow Ulis GoodVibration Software to do time windowed (Psychoacoustic) calculation of measurement and "snip" limit the range of correction and other ways of manipulating the target curve.
Today the praises of Sigtech appear inadequate to me because in the meantime more advanced roomcorrection systems have established that have better bass resolution, better target curve design facilities
and a graphic user interface that allows the amateur to adjust the system to his requirements.
Considering the archaic DSP of the Sigtech, the time aspect may appear in the measurement data processing but is not part of performing the correction.
So the only effect on the waterfall is a lowering of the level at specific frequency which leads to earlier passing the hearing threshold. There is no change in RT30. There is no cancellation of early reflection. This is imo common to all DSP based systems.
It is not much that is changed - but I like it...
--- In email@example.com, "Robert" <regtas43@...> wrote:
> Roughly speaking, what the Sigtech algorithm
> did was (mostly minimum phase) optimize
> the time energy buildup in the lower frequencies.
> It was like optimizing waterfall in reverse
> so to speak. Of course perfect impulse
> response would imply this. But impulse
> is never perfect over an area so they picked
> a version of this optimazation that worked
> out the optimal thing to do in listening terms.
> They had determined that the time
> of arrival of bass energy was important.
> (Cf my experience with the Gradient Helsinki--
> that even when EQed steady state flat it lacked
> bass immediacy in some sense).
> This was carried out with smaller time windows
> as frequency went up.
> The result was sensibly flat--but not ruler flat
> as you can see from the graphs shown on www.regonaudio.com
> But in listening terms it worked to my ears very well
> indeed. It sounded neutral and precise without sounding
> over-processed and without spatial instability.
> And you could phase it out(as opposed to just turning the
> correction off at a turnover frequency) as the frequency
> went up so that there was not an abrupt transition from
> full correction to no correction. If one
> were willing to work with this, one could get really excellent
> Incidentally, at the end they had a 96/24 version, if people
> were worried about CD digital standard.
> Also the resolution in the frequency domain--
> as well as the percentage correction-could be
> adjusted(as a function of frequency, as above).
> And of course the target curve was
> user choosable. The program had a lot of flixibility
> But it was also fairly complex to use compared to
> today's point and click programs.