Home Theater Forum and Systems banner

1 - 7 of 7 Posts

·
Banned
Joined
·
4,838 Posts
Discussion Starter #1
Pre-Ringing can occur in systems that involve digital signal processing (DSP) with some sort of time shifting or correction taking place. Linear Phase systems, FIR filtering, Room Correction, and similar processes can cause pre-ringing. Notice the word can - there are numerous conditions that can make it better or worse or not matter at all. More on that later.

Can You Hear It? I used to doubt it. But now I believe it is audible. That is a change in my experience and belief, not in the nature of pre-ringingi:ponder: ! What do YOU think?

What Does It Sound Like? As is often the case in audio, with more subtle phenomena, knowing what to listen for can make a big difference. Is there anyone who claims they can hear it who would like to tell us what they heard?
 

·
Premium Member
Joined
·
1,562 Posts
Do you have a measurement file you would like to discuss?

My hearing is sub par so my comments regarding sound need to processed with that in mind. I do have lots of measurements to go with my listening experience however and these lead me to suspect that this issue is often misunderstood and overstated. Of course pre-ringing can be a problem if the DSP is done wrong, but I often see concern based on an IR chart and feel that can be very misleading. Can there be a pre-ringing issue if the Phase/GD are positive and the ETC is pretty clean prior to the peak? Is there a problem with the example below?
IR.png

GD.png

ETC.png

Step.png

Phase.PNG

View attachment Pre-ringing.mdat
 

·
Banned
Joined
·
4,838 Posts
Discussion Starter #3
Here is my experience, and I do NOT have much to support this yet in terms of measurements, I am working on that.

What threw me was the misconception that you would hear something before the main sound, like a faint "pre-drum-hit" before the main drum hit, for instance. That is not at all what it sounds like.

Pre-ringing occurs at a fixed frequency (pretty much, depending on the exact cause) and gets excited by the presence of frequencies in the program material at or near that frequency. So you end up with little bits of extra energy at high frequencies which are associated with percussives and with sibilants (vocals), not harmonically related to the program material and not proportionate to general program material amplitude. It ends up sounding like a little bit of HF "hash" added in with the HF passages that excite it.

Image clarity suffers, the "SSSSS" at the end of a word in the vocals spreads out where the rest of that word was nice and pinpoint sharp. It sounds a little rougher, like distortion, although it does not measure like THD, and maybe a little brighter because of that extra energy, although FR measures flat.
 

·
Banned
Joined
·
4,838 Posts
Discussion Starter #4
Based on my understanding, which is still in the early growth stages, I don't see anything in your measurements that looks like a problem. Step response seems like a good place to look for it. Impulse response will probably show it too, being the slope (derivative) of the step response.
 

·
Premium Member
Joined
·
1,562 Posts
I did a Foobar ABX (double blind) test of 2 files some time ago when I first got interested in removing the phase rotation in my system. I didn't find any difference between the 2 files using headphones. One file was 'as recorded' and the other was convolved to include ~720° of additional phase rotation. The idea being that my headphones are relatively linear phase. They do not have a great amount of phase rotation; maybe about 45° at 20Hz and about 30° at 20k? The same file was convolved with an FIR filter to add about 720° to that. I have since considered that this is not the same situation as found in a room with all the various reflections that may also be an influencing factor? I have not tried an ABX test in my room. Regardless of my questionable results, there are many who do report hearing a difference. It is just hard to know how objective their findings are. I tend to be skeptical unless it is pretty well controlled listening test and I can see something in the measurements that suggests there could be a problem.

Excessive leading ripple in the REW IR is often shown because the anti-aliasing filter design chosen has a significant impact on the pre-peak ripple. REW does corrects the freq/phase based charts for the soundcard calibration, but it does not correct the time based charts. The IR, step charts therefore show the uncorrected plot and thus look worse than they actually are. The IR chart I posted above does not appear to have any significant pre-ringing however that measurement was taken using a 96kHz sample frequency. If a 44.1 or 48 kHz rate was used there would have been significantly more apparent IR pre-ringing. The impact to the music would be trivial however and the IR is not the best place to look.

All that said, I do have many music files with noticeable sibilance and/or harshness and my sensitivity to it tends to change day to day. This is easily controlled for me with my choice of house curve and has sometimes resulted in my using more rolloff than most others suggest. I currently have been using a pretty typical curve dropping about 5dB from 300Hz - 12kHz based on ~1m measurements. This results in too much harshness on some program material. In balance though, it is just right for a large percentage of my program material.

That's my listening experience. Not much help. I do have several other thoughts about measurements in regard to this subject. So if there is interest on that side of this, I will be able to comment further. My understanding there is also not very solid. It is just and area of interest that I am looking to better understand.
 

·
Registered
Joined
·
2,454 Posts
Wayne wrote:

Pre-ringing occurs at a fixed frequency (pretty much, depending on the exact cause) and gets excited by the presence of frequencies in the program material at or near that frequency. So you end up with little bits of extra energy at high frequencies which are associated with percussives and with sibilants (vocals), not harmonically related to the program material and not proportionate to general program material amplitude. It ends up sounding like a little bit of HF "hash" added in with the HF passages that excite it.
Are you saying that certain parts the speaker are sensitive to a particular frequency that excites another resonant frequency. This seems to make sense. Is it like the cabinet resonance we get in subwoofers? Those are clearly audible & identifiable. Maybe more so than higher frequencies, but along the same principle. Is this why certain horns benefit from damping material? Is it possible that could explain hearing fatigue? Or are you saying something entirely different?
 

·
Banned
Joined
·
4,838 Posts
Discussion Starter #7
I did a Foobar ABX (double blind) test of 2 files some time ago when I first got interested in removing the phase rotation in my system. I didn't find any difference between the 2 files using headphones. One file was 'as recorded' and the other was convolved to include ~720° of additional phase rotation. The idea being that my headphones are relatively linear phase. They do not have a great amount of phase rotation; maybe about 45° at 20Hz and about 30° at 20k? The same file was convolved with an FIR filter to add about 720° to that. I have since considered that this is not the same situation as found in a room with all the various reflections that may also be an influencing factor? I have not tried an ABX test in my room. Regardless of my questionable results, there are many who do report hearing a difference. It is just hard to know how objective their findings are. I tend to be skeptical unless it is pretty well controlled listening test and I can see something in the measurements that suggests there could be a problem.
John, I have not spent a fraction of the time listening for system phase rotation differences (a quick test, once, ½ hour) that you have. I fully expect that I would near nothing as well, with that being the only variable.

Room reflections are another matter, with their influence on soundstage & imaging (SS&I). Here I have become somewhat of a tyrant, and there are those (Dennis???) who believe I have gone ‘round the bend about it. A topic for another thread.

]Excessive leading ripple in the REW IR is often shown because the anti-aliasing filter design chosen has a significant impact on the pre-peak ripple. REW does corrects the freq/phase based charts for the soundcard calibration, but it does not correct the time based charts. The IR, step charts therefore show the uncorrected plot and thus look worse than they actually are. The IR chart I posted above does not appear to have any significant pre-ringing however that measurement was taken using a 96kHz sample frequency. If a 44.1 or 48 kHz rate was used there would have been significantly more apparent IR pre-ringing. The impact to the music would be trivial however and the IR is not the best place to look.
I have no suggestions for how to measure it accurately, other than using a relative approach with a given measurement system, and then try to listen for the bigger differences. If DAC/filter A measures about the same as DAC/filter B, then they probably sound about the same. If DAC/filter C measures a lot better than DAC/filter A, that difference might be audible.

All that said, I do have many music files with noticeable sibilance and/or harshness and my sensitivity to it tends to change day to day. This is easily controlled for me with my choice of house curve and has sometimes resulted in my using more rolloff than most others suggest. I currently have been using a pretty typical curve dropping about 5dB from 300Hz - 12kHz based on ~1m measurements. This results in too much harshness on some program material. In balance though, it is just right for a large percentage of my program material.
That amount of rolloff might very well be masking any negative pre-ringing effects in normal listening. It is really difficult to say, though.

For me, once I had a tool that pointed it out specifically, with quick A-B capability, then I could hear it. Now I know what it sounds like. With that simple filter type change in an A-B situation - I have listened with headphones and with speakers - I can hear it easily on some program material, in that listening situation. I could NOT walk into a room and say what DAC filter is in use.

What we are calling a standard symmetrical anti-aliasing filter does NOT sound “bad.” It sounds great. With ringing removed, it sounds “slightly greater.”

That's my listening experience. Not much help. I do have several other thoughts about measurements in regard to this subject. So if there is interest on that side of this, I will be able to comment further. My understanding there is also not very solid. It is just and area of interest that I am looking to better understand.
I would like to see a good way (also cheap) to measure it accurately. Higher sample rate, as you suggested, should help (192 kHz?). Part of the problem is that the measurement process adds more ringing.

It would also be cool to come up with some kind of threshold of audibility.

Are you saying that certain parts the speaker are sensitive to a particular frequency that excites another resonant frequency. This seems to make sense. Is it like the cabinet resonance we get in subwoofers? Those are clearly audible & identifiable. Maybe more so than higher frequencies, but along the same principle. Is this why certain horns benefit from damping material? Is it possible that could explain hearing fatigue? Or are you saying something entirely different?
The pre-ringing I am referring to is entirely an electronic phenomenon in the anti-aliasing filter of the DAC in question. It happens to some degree in any DAC. The amount of it can be controlled with the anti-aliasing filter type in use. There are numerous design types possible, and some can make it almost go away entirely.
 
1 - 7 of 7 Posts
Top