JPC, DD, at it again w/LTD02 stirring the pot...

[SME]

ah, i was not aware of that. do you happen to have a link/reference for this adjustment?   

It's mentioned in a few places.  Here's a technical paper on the subject.

 

JSS provided a very good answer to your second question about why the CEA tones look the way they do.

[mojave]

He's also severely clipping the waveform graph.

How can you tell that? 

 

Edit:  Do you still use a display range of 60%. It doesn't show anything on your screencaps. Your inf setup file that you provided a while back uses a display range of 60% and shredhead also uses 60%. I'm not sure why you would not want to show the full waveform range. I guess if you are mic'ing the signal and the levels are too low it looks like you are using higher levels when using 60%.

[Bossobass Dave]

JPC rants on (over at AVS) about full range, etc. Apparently he was name calling me (the Nazi stuff again) but the mods deleted those posts. Great stuff, as usual. Here's the full range settings JPC wants everyone to use. His frequency scale is LOG, not LINEAR and is from 3-24,000 Hz. This means no divisor on the FFT input size and no decimation of the sample rate in order to get bandwidth to 24k Hz.

 

That leaves about 4.5 bars in the entire 0-10 Hz bandwidth. No wonder SL spectrographs don't indicate low end content details for him.

 

Those setting are useless for MWB as the comparison shows (well, shows everyone but Jay, that is).

 

 

You can see his stereo full range spectrograph. The top end rolls off to zero around 10k Hz, which is correct whether you use SL or RTA. This is how recorded music and movies are mixed and there is rarely much content above around 10k Hz. Thank goodness for that. If there was equal content and energy across the entire 3-24k Hz bandwidth, you ears would bleed and it would sound like butt.

 

Max and Shredhead have detailed all of that very well with "a bunch of those pretty graphs 'n stuff".

 

The Nazi name-calling stuff?...Awesome... I wouldn't feel at home reading your posts at AVS without it. Great settings and good luck with all of that as well.

[Bossobass Dave]

John (LTD),

 

Regarding the content of a door slam, the spectrographs show most of the content <20 Hz because, well... most of the content IS <20 Hz in a door slam.

 

This is an actual door slam full bandwidth and zoomed. 3dB/octave from 4 Hz to 2k Hz is around 26dB drop total. The spectrograph (settings to allow to 4k Hz capture) show the content is down -60dB by 2k Hz with not much content above 30 Hz, relatively speaking.

 

[Bossobass Dave]

How can you tell that? 

 

Edit:  Do you still use a display range of 60%. It doesn't show anything on your screencaps. Your inf setup file that you provided a while back uses a display range of 60% and shredhead also uses 60%. I'm not sure why you would not want to show the full waveform range. I guess if you are mic'ing the signal and the levels are too low it looks like you are using higher levels when using 60%.

 

I didn't notice the 0.5% range setting used in the cap. Sorry, I assumed the expert who post the graph would have the AB % set correctly. So, no, I wouldn't know and neither would the user with that setting.

 

Thanks for catching that one.

 

I normally only mic. I don't use the AB for amplitude calibration, I use the interface to assure no clipping, which always coincides closely enough (to the 60% setting) anyway. I normally cut the AB out of my caps because that's not my focus (the total amplitude in the time domain). I included it for a while after a previous rant by JPC that disparaged the frequency focus setting I use vs his  setting that shows time domain data and virtually zero frequency detail (see post below).

 

I show it occasionally for time domain data and recently included it because I was clipping the interface input with the subs +10dB hot.

[Shredhead]

I made 2 tones with the same duration and amplitude and played them in both SL and REW. I found that I couldn't get REW's RTA to show equal levels regardless of what settings I changed:

 

I then played a transient scene in Thor and it shows the same difference in levels as ~100Hz.

Here you can see that I overlaid REW's RTA in white over the average in red in speclab. I found it interesting that most of it agrees but REW's gets progressively higher in level around 100Hz. I didn't really expect to see any difference. I think that this points to SL being more accurate in levels than any other option that we have at our disposal for analyzing disc content.

[maxmercy]

REW uses a cal file, make sure it isn't engaged, nor any meter compensation, and ensure the input is dBZ.  If all of that doesn't fix the problem, report the problem to JPM over at HTS, and he'll get it sorted.

 

JSS

[SME]

Here you can see that I overlaid REW's RTA in white over the average in red in speclab. I found it interesting that most of it agrees but REW's gets progressively higher in level around 100Hz. I didn't really expect to see any difference. I think that this points to SL being more accurate in levels than any other option that we have at our disposal for analyzing disc content.

What?  Isn't REW's RTA feature just emulating the behavior of analog RTAs?  If so, it has nothing to do with accuracy and everything to do with presenting the same measurements differently, as discussed earlier in this thread.

[maxmercy]

If The 'Spectrum' Function is engaged, it should not function as a typical RTA.

 

JSS

[mojave]

If The 'Spectrum' Function is engaged, it should not function as a typical RTA.

 

JSS

Correct. You can use the Spectrum which is an FFT or the RTA mode when the RTA window is running. You can also switch between the FFT and RTA in Omnimic. Makes it easy to view actual content during playback.

 

Here is 4 seconds of John Wick during the shootout at the club.

 

Here is two full minutes showing full bandwidth.

 

[Bossobass Dave]

Yeah, I'm ignoring the bottom graph. The top graph interests me because there are 3 graphs overlaid and not very well at that.

 

When I copy the SL trace and RTA trace and scale them properly, they look far less exaggerated:

 

 

I can't speak for anyone else, but I'd accept either of those scales if both were accurate in their intended relative differences. Trouble is, the RTA is not even close to the accuracy of the SL graph, which has quadruple the data points and far more accurately tracks peaks (transients). This is obvious to the casual observer at-a-glance. This is not to mention the obvious, that SL goes much, much farther in providing data than a peak hold trace.

 

It would have been a lot easier if you had posted the 3 graphs separately. I'm fairly adept at normalization of disparate graphs. And, it would be better to post examples that don't disappear into the noise floor.

 

The point here, though, is that the ULF, that is <20 Hz content, will not have a disparity of 15dB  from 40 Hz or 100 Hz, or whatever it is you've claimed. You really need to just retract that stuff and face reality.

 

The spectrographs we post do not mean that a 15 Hz tuned ported subwoofer will do much of anything <15 Hz, much less accurately portray the encoded content, in-room or on Nibiru. They are not meant to have anything to do with human hearing, equal loudness curves, logarithmic sine wave sweeps, logarithmic anything or Real Time Analyzers, although they are certainly relative to all of the above.

[BeastAudio]

Where is Nibiru? That next to Gastonia? 

[MKtheater]

The one question is there are many spec lab peak and average graphs that are flat or near flat, where is the 3dB/ octave increase? 

[Bossobass Dave]

The one question is there are many spec lab peak and average graphs that are flat or near flat, where is the 3dB/ octave increase? 

 

Take the HULK scene you posted...

 

[Bossobass Dave]

Where is Nibiru? That next to Gastonia? 

 

THIS^^^ was FUNNY!!

 

...had to be there.

[MKtheater]

Tell that to my room!  I have never needed to fix so many problems with this much bass, and I love it!  Every time I hear that very low weight of the Hulk the spec lab confirms.  Addicting!

[Madaeel]

Tell that to my room!  I have never needed to fix so many problems with this much bass, and I love it!  Every time I hear that very low weight of the Hulk the spec lab confirms.  Addicting!

Yah I thought the same thing. I've never looked over at Speclab, while it was running mind you, and thought no f'n way the low end is that hot. You get used to ULF after awhile and you can tell if, and when, it's hotter than the rest of the frequencies. Of course they're saying it's ONLY on full BW transients which I'll admit would be harder to detect, but even that sweep in Elysium was full range down to 15hz and both instances you could tell it was centered around 18-20hz.

 

OTOH if they're right I'm telling Dave to take this pos Raptor system back so I can get those do all ported subs.

[MKtheater]

Well, take The Incredible Hulk since that is the most tilted graph(it feels it too!).  Again, there are many flat spec lab graphs but if there is a flaw won't all graphs look like Hulk?  It does not make sense to me. I know that 25 dB under reference or under 100 dB at 10Hz won't do much in my room.  So when it is showing purple my screen and floor ripples.  It has to be 125 dB not 100 dB. 

[LTD02]

Tonebursts that have equal lengths of time will have equal intensity on the waterfall.  Since the CEA burst length is inversely proportional to increasing frequency, it appears to have less intensity. Since you are graphing linearly on the x-axis, the higher freq bursts are more 'spread out'.

 

that's the bugger.  we are looking at a combination of level and length in the waterfall intensity.  to the extent that higher frequencies are inversely proportional in duration to low frequencies, higher frequencies will always appear to have less intensity when an equal number of cycles is analyzed.  apparently, there is no simple answer to the question what frequencies are present and at what levels.

[LTD02]

John (LTD),

 

Regarding the content of a door slam, the spectrographs show most of the content <20 Hz because, well... most of the content IS <20 Hz in a door slam.

 

This is an actual door slam full bandwidth and zoomed. 3dB/octave from 4 Hz to 2k Hz is around 26dB drop total. The spectrograph (settings to allow to 4k Hz capture) show the content is down -60dB by 2k Hz with not much content above 30 Hz, relatively speaking.

 

that's the same problem.  the 3db/oct is the slope when using pink noise, log sine sweep, etc.  with a transient, such as a door slam, the slope is even steeper.  i'm not even sure if an fft can measure a full bandwidth transient.

 

as a test, I created three test signals in audacity.  all had the same amplitude.  each was 3 cycles in length.  then I measured their intensity in spectrum lab and measured significantly greater decline than 3db/oct. (appears to be closer to 30db/decade than 10db/decade).

 

 

with constant sine waves, there is no rolloff with frequency (the blue dashed line is flat).

 

so what is happening is the shorter the number of cycles, the steeper the slope of the rolloff (blue dashed line).

 

with a single transient pulse, with equal level of output at all frequencies, the rolloff would be very dramatic, perhaps not all that different from what you presented.

[Bossobass Dave]

If soundtrack low frequency effects were 38 ms in duration, you'd have a point worth looking at, but we would already have done that by now without any assistance.

 

No one has ever successfully measured a transient of less than 40 ms duration by any means that I'm aware of. Thankfully, that isn't what we measure here because it doesn't occur in soundtracks.

 

Have a look:

 

 

These effects are a creation of sound design to give the illusion of explosions from weapons of various sorts, including alien technology. You see no tilt because the effects are created and mixed to sound correct according to one specific artistic license.

 

If you had a subwoofer capable of ULF reproduction at reference levels, you would know that the spectrograph is not exaggerating the 1st decade by 60dB. This is not debatable and is actually more than a bit disturbing that it's still being debated.

 

The famous Open Range gunshot transients vs the much more full range gunshot transients from John Wick:

 

 

The difference in presentation of the 2 scenes is immediately discernible. That would fly in the face of the ELC arguments if the JW gunshots were really down -50dB <20 Hz. It doesn't because they aren't.

 

Notice that none of the transients, in movies or in real life, are comprised of a single frequency. They all naturally have content to DC below the fundamental and to F*X above the fundamental, plus any and all associated harmonics. They are then combined with other sounds and/or sound effects. They are then processed. They are then mixed.

 

If you really want to measure transients, you need to adjust your SL settings to reflect a higher time resolution than frequency resolution. But, that would be a subject for another thread, and you could post RTA graphs of your creation.

[SME]

that's the bugger.  we are looking at a combination of level and length in the waterfall intensity.  to the extent that higher frequencies are inversely proportional in duration to low frequencies, higher frequencies will always appear to have less intensity when an equal number of cycles is analyzed.  apparently, there is no simple answer to the question what frequencies are present and at what levels.

There *is* a simple answer to the question "what frequencies are present and at what levels," but it's not a particularly useful answer.  Just do an FFT and you have your simple answer.

 

The question that gets a complicated answer is "what frequencies are present and at what levels at a particular time?"  This is the question we want answers to because we perceive sound roughly in terms of level of frequency versus time.  Unfortunately, the answer depends entirely on the kind of analysis, and many kinds of analysis are possible.  The spectrogram is, perhaps, the simplest kind of analysis possible, but it is defined in terms of an adjustable parameter that trades time resolution for frequency resolution.  Incidentally, the spectrogram contains all of the information required to reconstruct the original PCM data, no matter what value is used for this parameter.  Nevertheless, the parameter itself has a dramatic effect on the appearance of the spectrogram.

[SME]

that's the same problem.  the 3db/oct is the slope when using pink noise, log sine sweep, etc.  with a transient, such as a door slam, the slope is even steeper.  i'm not even sure if an fft can measure a full bandwidth transient.

 

as a test, I created three test signals in audacity.  all had the same amplitude.  each was 3 cycles in length.  then I measured their intensity in spectrum lab and measured significantly greater decline than 3db/oct. (appears to be closer to 30db/decade than 10db/decade).

 

 

with constant sine waves, there is no rolloff with frequency (the blue dashed line is flat).

 

so what is happening is the shorter the number of cycles, the steeper the slope of the rolloff (blue dashed line).

 

with a single transient pulse, with equal level of output at all frequencies, the rolloff would be very dramatic, perhaps not all that different from what you presented.

This is a very clear example of the frequency vs. time uncertainty principle at work.  As the transient events get shorter, the range of frequencies involved becomes wider.  There's two things at work here.  First, each time you double the frequency, the total duration decreases by half, so the total energy drops by half or 3 dB.  Second, because the duration dropped by half, the energy is forced to spread out across twice as many frequencies.  The result is actually a 6 dB/octave drop in "Speclab level at center frequency".  On the other hand, if you summed up the Speclab magnitudes across frequencies, the trend would only be 3 dB/octave again.

 

A single transient pulse with equal level of output at all frequencies is actually, by definition, an impulse, and the waveform is full-scale at the time of the impulse and zero everywhere else.  In Speclab, this will actually look like a uniformly colored horizontal bar regardless of the time/frequency parameter used to generate it.

[SME]

If soundtrack low frequency effects were 38 ms in duration, you'd have a point worth looking at, but we would already have done that by now without any assistance.

 

No one has ever successfully measured a transient of less than 40 ms duration by any means that I'm aware of. Thankfully, that isn't what we measure here because it doesn't occur in soundtracks.

Soundtracks have transients under 40 ms all over the place, and there's no problem measuring these.

 

On the other hand, transients this short won't appear as such in SpecLab unless you give up a lot of frequency resolution.  Too much frequency resolution will compromise time resolution and make these transients look longer than they are.  In the examples you posted, one can't even see transients under 40 ms in the wave form at the far right.  Why not?  Your image is 680 pixels high and contains 2 specs depicting about 35-40 seconds of content.  Doing the math, a single pixel represents at least 120 ms of time.  How do you expect anyone to see 40 ms transients when the time resolution in the picture isn't even high enough to discern them?

[Bossobass Dave]

How do you expect anyone to see 40 ms transients when the time resolution in the picture isn't even high enough to discern them?

 

I don't. That's a settings matter. Sacrificing frequency resolution for that kind of time domain preference makes for a spectrograph that's useless to the content forum.

 

I expect THEM to show the data to support what they're yammering on about, which they seem averse to do. John has copied my settings to make his point, which is contraindicated.

 

It would also be OK if you just post one scene of low end effect that has <40 ms duration, then I might take your word that it happens "all over the place".

JPC has many a disparaging remark regarding the bandwidth and frequency resolution settings I use. He prefers 24k Hz bandwidth, which requires a minimum of 48k Hz sampling rate with no divisor or decimation. That changes the 0-120 Hz bandwidth we use dramatically: