Hi Logician,
Interesting comment, and I tend to agree with you.
Can you elaborate more on your "MICRODYNAMICS"?.
Best Regards,
Bohdan
...microdynamics are ever present in live music
but sadly missing from some contemporary DSP'ed replay.
In my view; no matter how smooth the frequency response, how pleasing the timbres or how pleasant the tones etc. Such may be considered separately from microdynamic expression which is quantitatively group delay related and qualitatively temporal in nature.
When corrupted; this may be all too readily identified by the frustrating annoyance arising from no longer hearing clearly and easily that which one knows or expects should be apparent. This experience is usually coupled with some measure of fatiguing dissatisfaction coming from a listeners subliminal perceptual system constantly haranguing them over this loss of acoustic integrity.
So; not such an empty word after all or indeed one any more meaningful for being capitalised but just a shorthand pointer to what's tangibly all too real in the above phenomena.
As it is not a term I have used I will not comment on the above pejorative reference to the corollary except to say that it is more often compromised by non minimum phase effects...
but sadly missing from some contemporary DSP'ed replay.
In my view; no matter how smooth the frequency response, how pleasing the timbres or how pleasant the tones etc. Such may be considered separately from microdynamic expression which is quantitatively group delay related and qualitatively temporal in nature.
When corrupted; this may be all too readily identified by the frustrating annoyance arising from no longer hearing clearly and easily that which one knows or expects should be apparent. This experience is usually coupled with some measure of fatiguing dissatisfaction coming from a listeners subliminal perceptual system constantly haranguing them over this loss of acoustic integrity.
So; not such an empty word after all or indeed one any more meaningful for being capitalised but just a shorthand pointer to what's tangibly all too real in the above phenomena.
As it is not a term I have used I will not comment on the above pejorative reference to the corollary except to say that it is more often compromised by non minimum phase effects...
Last edited:
logician,
Seems we have very similar ideas of what the cause is. Many countering arguments though by those who don't think we can detect phase information. I think if you can't get the impulse response timing right and keep the phase from doing multiple loops all is lost for those small details.
Seems we have very similar ideas of what the cause is. Many countering arguments though by those who don't think we can detect phase information. I think if you can't get the impulse response timing right and keep the phase from doing multiple loops all is lost for those small details.
Visuals
...Pure electric filters to replica speaker system bandpass IRR 35-20kHz, plot 1-5 have system HP as BW2 plot 6-10 have BW4, XO point 2500Hz shown with FIR LR4, IRR LR4 and IRR compromise Harsch XO type.
Link: http://www.diyaudio.com/forums/multi-way/277691-s-harsch-xo.html#post4402917.
Filters set in JRiver DSP engine REW listening in soundcard loop.
Attachments
BYRTT,
I'll have to read through that thread. I haven't played with any of the digital software yet to see how it all works. I do have REW on my computer and have been following the rephase thread. I do understand this from a speaker design standpoint and have done all the things i can to make the best devices I can do. I will gladly give up efficiency to have better linearity and that is part of what I have done. I've also addressed inductance and the next step is to do what can with network design. I expect to have IIR or FIR processing in the networks when this is all done and finished.
I'll have to read through that thread. I haven't played with any of the digital software yet to see how it all works. I do have REW on my computer and have been following the rephase thread. I do understand this from a speaker design standpoint and have done all the things i can to make the best devices I can do. I will gladly give up efficiency to have better linearity and that is part of what I have done. I've also addressed inductance and the next step is to do what can with network design. I expect to have IIR or FIR processing in the networks when this is all done and finished.
Hi Logician,
Your comments about correct implementation vs. improper implementation of FIR filters seem to resonate with some of my experiences with FIR filters. So perhaps comparing the notes on microdynamics may shed some more light on this issue.
I have experimented, built and used FIR filters for several years now. All of the filters I used, are equalizing filters. They use HBT-type of equalization plus phase linearization of each individual driver, therefore you may call them “custom-built” filters, based on FFT convolution. The final outcome is acoustical SPL and phase of each driver are both flat lines. So all time-domain responses (impulse response, step response and square wave playback) are perfect.
Yet, some of those filters sounded better that the others.
In the early days, I was inspired by Floyd Toole’s comments about bass resolution being essential, so my initial FIR filters had IR length of 65536 bins, bass resolution of 48000/65536 = 0.73Hz. Data block size was 1365.3ms. This filter measured perfectly, but did not sound right, it sounded dull, lifeless, and lost dynamics on any type of music, particularly percussion and drums instruments.
Next FIR filter was 32768 long with bass resolution of 1.47Hz and data block size of 682.6ms. Acoustically, I had similar observations to the previous version. Next:
16384 bins with 341.3ms data blocks and 2.93Hz bass resolution, (sounded better)
8192bins with170.66ms data blocks and 5.86Hz bass resolution,
4096bins with 85.3ms data blocks and 11.72Hz bass resolution,
2048bins with 42.66ms data blocks and 23.44Hz bass resolution,
1024bins with 21.34ms data blocks 46.86Hz bass resolution,
512bins with 10.7ms data blocks 93.76Hz bass resolution,
To keep the long story short, it was not until I reduced the data blocks to 170ms and less, that the dynamics was restored – this is by my ears. Another person may have different opinion.
However, at the same time, I lost accuracy of subwoofer and room equalization.
The solution to these issues was partitioned convolution. Current implementation allows me to run filters as long as IR=16384 bins with partition size of 1024bins (or data block size of 21.34ms) and bass resolution of 2.93Hz.
Given that all FIR filters were transient perfect as I described them before, my conclusions from all these was, that reducing data blocks to 20-40ms preserves short-term envelope (or dynamics) of the sound – is this related to your microdynamics?.
Best Regards,
Bohdan
Linkwitz Transform as FIR or IIR filter
Dear all,
I have a PC-driven (FIR filters in JRiver) 4 way active system. Now, I would like to flatten the responses of my subs by using the Linkwitz Tranform. Does anyone of you know if it's possible to save the Linkwitz Transform as a FIR filter in order to be convolved with the corresponding XO filter?
I know that JRiver implements the functionality of Linkwitz Transform in its PEQ, but I need for testing pourposes to have it upstream, by using the ASIO drivers of my board.
Best Regards
Dear all,
I have a PC-driven (FIR filters in JRiver) 4 way active system. Now, I would like to flatten the responses of my subs by using the Linkwitz Tranform. Does anyone of you know if it's possible to save the Linkwitz Transform as a FIR filter in order to be convolved with the corresponding XO filter?
I know that JRiver implements the functionality of Linkwitz Transform in its PEQ, but I need for testing pourposes to have it upstream, by using the ASIO drivers of my board.
Best Regards
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.