I've fully re-read this entire thread today and have a better understanding. For the benefit of others and myself, given the length of this thread, I am trying to summarize the concepts and procedures discussed in the preceding 7 pages, both through my (layman's) explanation below and by means of my questions and hoped for answers that follow:
-- I see that to use multi-driver/multi-ch filters in MC (meaning needing more than one filter) one accesses the desired filters by creating a config file following the specs for the Convolver config file format:
http://convolver.sourceforge.net/config.htmlhttp://convolver.sourceforge.net/configegs.htmlIn that text config file one puts the path to all the desired correction filter files to be used in MC's convolution DSP. In addition, one can add additional delays and other criteria. This answers my prior question above as to how to use individual driver/speaker corrections given that MC only accepts a single file.
I also understand that in addition to (or in place of) the above one can use the Room Correction and/or the PEQ sections of MC DSP studio for similar purposes, e.g. delays, channel mapping, x/o's, etc, etc
After following all of the links and/or references in this thread and reading everything on all referenced sites (WOW!), I understand that in general, X/O and DRC corrections are best handled via convolution filters, rather than the more simple PEQ and/or MC Room Correction, because the former will better handle time domain, specifically delays and phase shifts, in addition to frequency adjustments handled by the latter. So even when speaking of a seemingly simple single-speaker X/O (vs an even more complex room correction), being able to include delays and phase shifts (and other components) moves one out of minimum-phase adjustments and into linear-phase adjustments yielding textbook results, potentially perfect in both frequency and time. This is the fundamental difference between electronic (passive) X/O and computer (active) X/O. Thus, if using the PEQ and/or Room Correction components of MC DSP studio and NOT the Convolution component, one is, at best, matching passive X/O and never achieving active X/O potential. On the other hand, two important consequences of adding delay and phase (linear-phase adjustments) are enhanced: 1) correction artifacts, i.e. "pre-ringing" (hearing parts of the correction sooner in time than the direct signal) and, 2) even greater single-listening-position sensitivity (all other locations will likely be even worse). I'm sure I've at least partially misstated this, but hopefully not fundamentally so.
Understanding the above (and welcoming any corrections others will offer), I have the following questions:
1) Do we manually create the config text file, or will programs generate this for us (likely in need of editing). E.G. I understand that DRCDesigner may create the config file for DRC generated correction filters.
2) Recognizing that REW may not be as robust as other products like Acourate and Audiolense, if desired can our config file point to REW correction files, or are they somehow not compatible with the config file usage?
3) If one desired, can a config file point to filters made by more than one program (e.g. if one chose to do per speaker corrections in REW and then combined post-adjusted room correction in another program, assuming all filters are .wav files)
4) Recognizing the limitations described above, if one wanted to use MC PEQ and/or Room Correction in addition to convolution, what should be the order of each DSP component, i.e. should PEQ come before Convolution? Why?
5)
@ TheLion describes he uses Accurate for filters, but then uses Audiolense to calculate delays, which he then manually adds to the config file, because Accurate does not calculate delays (as well or at all?). Why not do it all in Audiolense?
6)
@ TheLion describes he uses hardware DSP on his subs and SW DSP on everything else. Why not do it all in SW? To do otherwise implies a HW solution is superior to 64bit convolution in its present state-of-the-art. Is that your feeling? Then why not do it all in HW?
7)
@ TheLion describes he Audiolense does not correctly calculate the delay for his subs. How do you know what the correct delay SHOULD BE? Is it only based on relative distance, or other factors? Why do you think the SW calculation is wrong?
8 ) Being sensitive that
Uli has contributed greatly to this thread and thus MC's execution of convolution, what are the pluses/minuses of Acourate vs Audiolense vs DRC, REW, etc
9) As
Uli pointed out, because
@ TheLion is summing both the frequencies below 80hz from all 7 separate speaker channels and the LFE channel below 160 hz, all to his subs , and because each of those 7 speakers/channels are at different distances and thus have different delays (and are different than the LFE), the delay (and likely the phase) from the summed signals to the subs must be all messed up (7 different delays/phase x 2 signal components coming to 2 subs, it must be completely smeared), because at present MC does not support additional parameters for summed delays. Given your precision, why isn't this a problem for you?
10) In concept, with a multi-driver/multi channel setup such as
@ hulkssor or
@ TheLion, what is the workflow to create X/O's, balance each speaker and apply room correction?
I prefer to do XO and DRC incrementally. By this I mean I would like to proceed asfollows:
--A) Set speaker locations while running RTA w Pink PNoise, triangle formulas, laser measurements and other conventional methods;
--B) Take near-field frequency response measurements ofeach driver with band limited frequency sweeps to determine sweet range of eachdriver and successive driver to set X/O points and slopes;
--C) Measuring mid-field and in exact path betweenspeaker and listening position and at exact listening height, run DRC SW tocreate linear-phase filter to set delay and phase adjustments for the driverson individual speaker basis, based on that speaker's relative acoustic centerfor its individual drivers;
--D) Load these per speaker correction filters into DRC software or JRiver MC vialoopback feature
http://yabb.jriver.com/interact/index.php?topic=70242.0and then use DRC SW for combined response (single sweep to allchannels/speakers) adjustment room correction based on desired "housecurve".
--E) Create required Convolver config text filepointing to the above generated filters and load into JRiver MC convolution
--F) Load these combined correction filters into DRC software or JRiver MC via loopbackfeature with convolution applied, re-measure from listening positionand adjust existing filters, as desired/needed.
Does anyone know in DRC SW one can I load incremental and/orfinal adjustment filters and perform additional measurements/corrections "ontop of" existing filters?
Sorry for the long post, but as stated above Im hoping to bring together much of what has been discussed and to reconciled some inconsistencies from the preceding 7 pages of this thread.
I look forward to your wisdom, answers and corrections.
Thanks!! and thanks Matt & JR for this feature!!
Author
Topic: Native JRiver 64bit fp convolution engine for Room Correction (FIR filters)? (Read 340713 times)