Hello again
@DrCWO and ASR friends,
It is always great pleasure and very much informative for me (and for our friends visiting this thread, I believe) knowing other people's digital DSP audio setup together with the theoretical background.
I have slightly different approach comparing to yours; I use DSP (XO/EQ/delay) software EKIO as
"system wide" DSP center within my PC which can receive digital audio signal through ASIO routing from any other audio-output software including Roon, JRiver MC, Adobe Audition, Audacity, Web browsers,
etc.
As I wrote
here, EKIO uses IIR filters. The processing is done using a cascade of second order transposed direct form II biquad sections. Every calculation is done using 64 bit floating point numbers.
Furthermore, I use rather "mild" second-order (-12 dB/Oct slope) LR (Linkwitz-Riley) filters in my XO configuration except for the sharp high-cut (low-pass) -48 dB/Oct UHF noise cut-off filter at 25 kHz. The reason for my use of -12 dB/Oct mild filters is that I simulate Yamaha's original passive LCR network in my active configuration; I have been using Yamaha NS-1000 cabinet and its three SP drivers throughout this project. Of course, I have tested higher order filters in XO configuration, and concluded that -12 dB/Oct filters should be the best choices at least for my NS-1000 cabinet and drivers (please refer to
my early post here). In case if I would use different main SP (and SP drivers) other than NS-1000, I will/should start a new project thread, I believe.
Let me touch on the "nature" of my "naive/primitive but steady" approaches in measurements and fine tuning throughout this project.
My fundamental policy would be
"Always use my own well understandable validated reproducible measurement methods in each of my go-up steps (do not over trust existing black-box type measurement tools/software unless I would fully validate and understand it)."
Just as
one typical example, my policy in Fq response measurements; nowadays, I always do
not like
psychoacoustic smoothing* using very short/rapid sine sweep data over all the SP drivers; it gives too much smoothing with much statistical deviations/fluctuations based on
poor (little amount of) air sound data, and hence lack of details of Fq responses of the drivers as well as of the room acoustics.
*In REW user manual: "Psychoacoustic smoothing uses 1/3 octave below 100Hz, 1/6 octave above 1 kHz and varies from 1/3 octave to 1/6 octave between 100 Hz and 1 kHz. It also applies more weighting to peaks by using a cubic mean (cube root of the average of the cubed values) to produce a plot that more closely corresponds to the perceived frequency response."
Any of Fq response measurements includes FFT analysis of the recorded sound data, and FFT Fq analysis is a "
statistical procedure/calculation" on given raw data; the "statistical" accuracy/precision/resolution are greatly dependent on the "richness" of the raw sound data; the higher richness in raw data the better accuracy/precision/resolution/reproducibility of FFT Fq response results.
I fully evaluated and
validated, therefore, my rather naive/primitive
"cumulative white noise averaging method" (please refer to post
#392 and
#393) for my Fq response measurements; the pros and merits were summarized
here in my post #404 where I wrote as follows.
1. the method is universally applicable in the stages of digital out of crossover software (EKIO), DAC's analog out, amplifier SP out, and of course in the actual room SP sound,
2. the method is accurate, sensitive and reproducible, having little or no statistical fluctuation, because of the FFT averaging analysis on the "accumulated rich data" of the recorded sound,
3. the recorded "white noise tracks" can be re-analyzed any way, anytime, afterwards,
4. flexible mix-paste (sound mixing) can be done to virtually simulate any combination of the channels, especially in amplifiers' SP out signals before going into SP drivers,
5. if needed, the environmental "continuous room back ground noise" can be reduced/removed by the Adobe Audition's "noise capture - noise reduction" function,
6. if needed, suitable gain/level adjustment can be applied for "level matched comparison" of Fq response shapes between the different series of the recorded data,
7. flexible and suitable FFT size (as smoothing intensity) can be selected depending on the frequency zone of interest.
Furthermore, I believe the
"cumulative white noise averaging" would be more practically suitable than "pure sine sweep" since our SP drivers always receive mixture of various Fq sound signals in our actual music listening situation, not single Fq sine wave.
I also applied such my
"fundamental policy" in the measurements and fine tuning of "time alignment" all over the SP drivers, as I summarized
here.
And, the same for my
DIY 12-VU-Meter Array, and for
revival of analog LP Player in my DSP multichannel multi-amplifier system.
My
"fundamental policy" throughout this thread also includes the followings;
- First, I (we) need to establish my reference sound system to which I can roll back any time during my exploration (in my case it has been passive single amp stereo system with Accuphase E-460)
- Go up (climb up) step-by-step carefully, not in a hurry
- Do not change multiple parameters/configurations at once; it may cause inter-cancellation of pros and cons to give confusion
- In each step, I should trust my ears and brain in subjective listening test using my own common/consistent preferable "audio sampler music playlist" consists of excellent-recording-quality tracks selected from various genres/categories (like shared here and here)
In any way, I would highly appreciate all of your continuing kind attention and further participation(s) in this rather long-lasting still-on-going project thread.
audiojudgement.com
www.stereophile.com
