Jump to content

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


Bossobass Dave

Recommended Posts

please stop telling people what they are agreeing with or not.  each person is capable of doing that for themselves.  :-)

 

i congratulated dd on bringing the topic to light, not for anything having to do with a pb13u subwoofer.

 

when folks measure the frequency response of their systems, a signal with log energy per frequency (equal energy per octave) is what is used.  even you do this because presenting a frequency response with equal energy per frequency would create a 3db per octave rise in the response and would have little interpretation. 

 

the point is simply that the same method should be used to present the content as was used to equalize the system.  otherwise, the lower frequencies will appear to rise at 3db/oct.

 

not much more to it than that.

 

Let's recap:

 

DD said:

 

 

That is a Spectrum Lab graph which has a 3 dB/octave rising slope as you go down in frequency vs REW, Omnimic or any other measurement software. This is why some of the current movie graphs at data-bass using actual disc data show a rising tail down below 10 Hz. In the 3-4 octaves below 10 Hz there is an increase of 9-12 dB.

 

You said:

 

 

very nice dd.

 

 

this evening i was creating and analyzing various audio test files and ended up down in the weeds of pink noise and fft. after some experimentation and much searching around, i (re)discovered this thread.

 

 

but what to do now? the "library" of spectrumlabs is enormous and it would seem that they all suffer from this problem.

 

JPC said:

 

 

Just as a note, an RTA analyzer displays the underlying FFT information in a form that is easy to understand when viewed by eye. A flat frequency response displayed on an octave based RTA analyzer is a valid representation of a flat frequency response based on the way that you hear (octave based).

 

Repeat your SL charts using pink noise and white noise. All the data is easier to see if you expand the frequency range to full bandwidth (0 - 24,000 Hz). Throw in a log based frequency scale chart when needed. Note that the log frequency scale chart in SL does not turn the FFT display into an RTA display.

 

You said:

 

 

^^ thanks. i understand all that now.

 

 

what would be useful is a 3db/oct filter that could be applied prior to viewing content in spectrumlab, thus preserving all of its usefulness while providing a more accurate representation of content.

 

I'm not the least bit confused by what the thread is about. Maybe you are?

 

My posts in the thread simply stated that SL accurately creates a spectrograph of the content in a soundtrack. Further. pink noise and a log sine sweep are irrelevant to the subject. This was only to dispel the gist of the thread which is that we are viewing the graphs improperly and that, because of that fact...

 

DD said:

 

 

Because Spectrum Labs has a different slope than other measurements you can't directly compare the charts to room measurements or subwoofer measurements such as those on data-bass.com. In order to directly compare, the Spectrum Lab chart has to be rotated. Here is an example of The Incredible Hulk as posted on the forum at data-bass.com:

 

This has several implications

When rotated to match room/subwoofer response graphs, the Spectrum Labs graphs show that

    1    bass output peaks somewhere in the 40-50 Hz region and flattens or rolls off as you go down in frequency

    2    subwoofers with limiters can still accurately produce most of the recorded bass without have a "variable frequency response" since the tilted Spectrum Lab graphs match        very well with a lot of the subwoofer measurements at data-bass.com - even at higher output levels.

    3    ported subwoofers are producing more of the original bass than thought which corresponds with most listener impressions at subwoofer GTG's

    4    studios are viewing a different representation of the audio than what Spectrum Labs presents

 

Followed by:

 

 

Here is the SVS PB13 Ultra 20Hz tune maximum output from Data-bass.com. I show both the actual measurement (green) and adjusted for room gain per the Bossobass graph (red).

 

This illustrates that you can play War of the Worlds at 118 dB from 12.5 Hz and up and track completely with the maximum output of the signal from 14Hz and up.

 

So, did you miss all of that? Apparently.

 

Again, the reason I replied to the thread at all is to counter the thought that "we" are viewing the data presented here incorrectly, that the rising traces on the peak hold graphs and the corresponding increase in intensity of the colors in the spectrographs is some sort of FFT anomaly and not the result of the mix being bumped at the low end to create a Movie With Bass.

 

The ported sub drivel is less important to me. No amount of data will ever counter that sort of subjective interpretation of distortion commented on by a small group listening to a tiny slice of source on a sunny afternoon. But, you agreed with the OP and weren't specific about which part you agreed with.

Link to comment
Share on other sites

  • Replies 117
  • Created
  • Last Reply

Hi, I'm DesertDome at AVS. There seems to be a lot of misinformation about why I started the thread at AVS, what gear I have, and things that are the complete opposite of what I said.

 

Why did I start the thread?

1.  In 2011 I noticed that SpecLab and other FFT analyzers were not consistent in how they represented the data. Some had a slope or tilt. I posted this at JRiver's forum on June 7, 2011, "What started me down this path is I'm trying to figure out why the movie frequency plots posted on AVS don't seem to match what I see in MC with spectrum analyzers.

2.  I found out that some FFT analyzers have a 3 dB slope "baked" into them and other FFT's have the ability to adjust the slope. I also found discussion on gearsluz, cockos, krvaudio, and other forums regarding which slope to use when mixing the audio.

3.  I corresponded with an audio mastering engineer and found that he mixes using the pink noise Accumulation Mode in SpectraFoo

 

 When the Accumulation Mode is set to Pink Noise is Flat, the Spectragraph shows the sum of the power of all of the spectral lines in the FFT that fit into each point in the displayed trace. This has the effect of displaying pink noise (or noise that has an equal amount of power in each octave) as a flat line. This mode is traditionally used when the frequency scale of the graph is logarithmic and corresponds to the type of display available on a traditional RTA.

 

4.  I looked at the SpectraFoo manual and found graphs showing how the two methods of accumalation represented the data differently.

5.  I read and tested more over the next few years and finally posted at AVS.

6.  I didn't start the thread because I don't believe in low bass, can't reproduce bass or any other such nonsensical reason.

 

What gear do I have?

1.  It was said "JPC and DD have systems that eliminate 2/3 or more of the ULF".

2.  I am probably the only one on AVS and data-bass that has no signal rolloff. I am flat to DC.

3.  For audio devices I have both a Lynx Aurora 16 Thunderbolt and a MOTU 1248. Both are DC coupled and flat to DC. Both can be used to send control voltages to analog synthesizers. MOTU has instructions on their website for Testing analog outputs for control voltage compatibility.

4.  The mic preamp on my MOTU 1248 is -2.5 dB at 3 Hz.

5  My subwoofer amplifiers are Digital Amp Company Maraschino Cherry's which are flat to DC and stable to a 1 ohm load. The amp designer said, "The Maraschino is DC coupled end-to-end with isolated lock to the preamp (or D/A) GND ---- it can actually be used as a very quiet DC power supply, and since it's 'true-differential', it can output positive or negative voltage, too."

6.  I have Picotech and Tektronix oscilloscopes.

7.  I have a iSEMcon EMM 7101 ICP (IEPE) microphone that is DC coupled and, like all ICP devices, is accurate down to 1 Hz.

8.  I have an infinite baffle subwoofer system using eight Acoustic Elegance IB15 drivers.

9.  I probably have the lowest playback and measurement capability of just about anybody.

 

What assertation did I make?

1.  It was said, "desertdome made the assertions that the SL caps we use here misrepresent what's on the soundtracks"

2.  I said, "Spectrum Lab is accurately measuring and displaying the audio content."

3.  I said, "Spectrum Lab is accurately measuring and displaying the aduio content. I am in no way saying it isn't.

4.  I said:

 

 

What are the two spectrum analysis techniques?

"There are primarily two common spectrum analysis techniques: Constant Bandwidth (CB) and Constant Percentage Bandwidth Analysis (CPB). CPB Analysis can be implemented with analog or digital signal processing, while CB analysis is usually implemented using the digital FFT (Fast Fourier Transform) technique." (from Brüel & Kjær's technical documentation for hand held analyzers)
What type of analyzer is Spectrum Lab?
Spectrum Lab is a fast Fourier transform (FFT) and therefore is a Constant Bandwidth analyzer (also know as linear power spectrum)

  • "Constant Bandwidth – linear rendering of bandwidth (FFT calculates filters with constant BW)" (from Sound Reinforcement System Design by D. G. Meyer 2011)
  • "The FFT generates frequency domain data that is linearly-spaced as a function of frequency, unlike our hearing, which primarily perceives frequency in a logarithmic manner" (from FFT Fundamentals by Rational Acoustics)

What type of analysis is presented by REW, Omnimic, data-bass.com?

These present the data as Constant Percentage Bandwidth (also known as logarithmic power spectrum or octave analysis)

  • Octave analysis shares some similarities with power-spectral analysis, but in spectral analysis all frequency bands (bins) occupy equal bandwidth. For example, in a 1000-point FFT that covers DC to 10 kHz, each bin occupies 10 Hz. The frequency resolution is linear over the entire range. Because of its log scale, octave analysis can show results over multiple octaves, allowing you to see and compare signal levels over a broad frequency range. (Analyze Signals Octave by Octave by Sam Shearman)

     

Can I test this myself?

  • If you have Omnimic, you can play actual content and switch the Spectrum/SPL view between FFT and RTA.

 

John, the author of REW, summed it up best in his post at gearslutz in 2013:

 

I'm afraid you are creating a great deal of unnecessary confusion, and mixing up the measurement of transfer function (which is what REW and other audio measurement systems do) and the various ways of displaying the spectral content of a signal, which are not at all the same thing.

 

 

REW measures the audio system transfer function, the plot it produces shows how the system changes signals passed through it - for example, if the frequency response is 10 dB lower at 10kHz than it is at 1 kHz that is because the system that was measured reproduces 10 kHz tones at 10 dB lower levels than 1 kHz tones. The transfer function is independent of the stimulus used to measure it, it is a ratio of the measurement to the stimulus. Various types of stimuli are used by different systems, from noise (white, pink or other spectral shapes) to sweeps (linear or logarithmic) with a variety of others also available, but ALL will produce the same frequency response for the same system unless they are broken. As it happens a logarithmic sweep, as used by REW and many other measurement systems as it has a number of useful properties, has constant energy per octave because the rate at which the sweep frequency changes increases exponentially during the sweep - the stimulus looks flat on an RTA display.

That brings me to signal energy displays. The frequency content of a signal (as distinct from the frequency response of a system) can be displayed in various ways, the two most common are spectrum analysers and fractional octave analysers, more commonly called Real Time Analysers or RTAs. Spectrum analysers show the energy in bins of constant frequency width - for example, the energy in each 10Hz span of a signal. On a spectrum analyser white noise, which is defined as having constant energy in any given frequency span, appears as a horizontal line. Pink noise has constant energy per octave - for example, it has the same energy in the 100 Hz span between 100 and 200 Hz as it has in the 1 kHz span between 1kHz and 2 kHz. On a spectrum analyser a pink noise signal shows a level that drops 3 dB per octave. RTAs show the energy in octaves or octave fractions, a 1/3 octave RTA display shows the energy in each 1/3 octave span. On an RTA white noise produces a line that rises at 3 dB per octave, pink noise produce a horizontal line.
 

Link to comment
Share on other sites

Mo/DD,

 

I appreciate the update. :)

 

I'll repeat my objections to the AVS thread posts, for clarification as to why I posted at all:

 

 

Because Spectrum Labs has a different slope than other measurements you can't directly compare the charts to room measurements or subwoofer measurements such as those on data-bass.com. In order to directly compare, the Spectrum Lab chart has to be rotated. Here is an example of The Incredible Hulk as posted on the forum at data-bass.com:

 

This^^ is simply untrue. I can directly compare REW FR traces with log and linear graphs, to SL spectrographs, and routinely do. The SL caps do not have to be rotated.

 

 

This has several implications
When rotated to match room/subwoofer response graphs, the Spectrum Labs graphs show that
    1    bass output peaks somewhere in the 40-50 Hz region and flattens or rolls off as you go down in frequency
    2    subwoofers with limiters can still accurately produce most of the recorded bass without have a "variable frequency response" since the tilted Spectrum Lab graphs match very well with a lot of the subwoofer measurements at data-bass.com - even at higher output levels.
    3    ported subwoofers are producing more of the original bass than thought which corresponds with most listener impressions at subwoofer GTG's
    4    studios are viewing a different representation of the audio than what Spectrum Labs presents

 

1) Bass output varies from mix to mix. It peaks wherever the mix dictates that it peak.

 

2) SWs with limiters cannot accurately portray their own native FR let alone WOTW at reference level playback.

 

3) This statement is so general I don't know what to say other than it states nothing specifically and then attempts to relate that to a few (unspecified) GTG's subjectivity. I can say that ported subwoofers, those that have been tested, do not produce more than the test results show because of spectrumlab.

 

4. C'mon. Maybe this is all just poor writing? You say SL portrays the content accurately, then you say what studios are viewing is different. So, is not the logical conclusion here to ask what the hell are studios viewing that's different from accuracy, why would they do that and, most importantly, what does that have to do with the MWB data-bass?

 

 

That is a Spectrum Lab graph which has a 3 dB/octave rising slope as you go down in frequency vs REW, Omnimic or any other measurement software. This is why some of the current movie graphs at data-bass using actual disc data show a rising tail down below 10 Hz. In the 3-4 octaves below 10 Hz there is an increase of 9-12 dB.
 

 

The SL spectrograph shows what's on the disc. There is no correction file required. There is no increase over actual content shown on the graph. There is no rising slope as you go down in frequency.

 

The SL spectrograph has nothing whatever to do with the REW log sine sweep trace on it's frequency response magnitude graph.

 

But, one can easily refer to the REW frequency response magnitude graph resulting from measurement of his subwoofer with the microphone placed at the seats to reasonably expect how well or poorly his subwoofer will reproduce the content of a specific soundtrack scene once the scene has been run through SL and it's spectrograph posted.

 

One can then also mic the scene himself and run it through SL and compare the spectrographs made from the SW out of his signal chain to the mic'd at the seats version spectrograph to assess how accurately his subwoofer is reproducing that scene. THIS would explain the subjectivity of any GTG instead of any other explanation I'm aware of.

 

Thus, the gist of the entire enterprise of the MWB forums and their content.

 

You can address these points and I will appreciate it.

 

Finally, can you post your in-room FR at the seats to show the flat-to-DC response?

Link to comment
Share on other sites

thanks for taking the time to write your response sme.

 

my understanding is that pink noise is generally perceived as sounding 'flat' and that is why it is used and not some other signal for calibrating sound systems.  rta's calculate energy based on octaves (or portions thereof), which with pink noise, produce a visually flat response.  the end result being a pretty good correlation between what we see on the display and what we hear with our ears.  (i know its not perfect because of fletcher munson curves/equal loudness curves, etc., but close enough.)

 

now take that same pink noise content and move over to spectrum lab.  the pink noise still sounds flat, but the display shows a decline of 3db/oct.  across the audible bandwidth, this amounts to 30db or so and that doesn't correlate at all with what we hear.  so, to bring it back into alignment, a 3db/oct adjustment is necessary.  it is all simply an attempt to best correlate what is displayed visually with what we hear.  maybe there is a better way?

Pink noise sounds flatter than white noise but it still far from sounding truly "flat".  Those who use pink noise for calibration may be accustomed to the sound and be able to identify by ear when the system is responding less than flat, but that's not the same thing as the pink noise itself sounding flat.

 

Again, the problem here is not with the analysis but with the interpretation.  The SpecLab spectrogram does not quantitatively show what we hear, and no 3 dB/octave adjustment is going to fix that.  The "better way" to see what we actually hear is to apply a proper psychoacoustic model based on ELCs and masking thresholds, however, this is quite complicated in practice.  There is no psychoacoustic model that we can all readily agree is correct.  Therefore, it's better to just display the data the way it is and leave it to the reader to formulate an appropriate interpretation.  Presenting the content without alteration is also useful to those who wish to assess what system capabilities are necessary to reproduce that content.

 

As mojave points out, there are different conventions for presenting the same data.  I, for one, strongly prefer the "spectral analyzer" view over the "RTA" because it's consistent with modern transfer function measurement and the Fourier transform.  It also depicts sine waves at the correct levels.  It does not depict pink noise as flat, which is correct because pink noise isn't flat.  One should not assume that a flat spectrogram will sound flat, nor should one assume that a flat RTA measurement will sound flat either.  To the extent that studios rely on the RTA view, it is because they are familiar with it, not because it's more correct.  More importantly, studios should be mixing to make it sound good and not "mixing to a flat RTA" if they have any competence.

Link to comment
Share on other sites

I have used both REW and Spectrumlab in direct comparison for years.

 

Below is a scene from The Amazing Spiderman II, where the 2 jetliners nearly collide. I set the mic near the main LP (this exercise was done quickly for this thread and none of it is critically tweaked, such as the mic position, FR at the mic position, sub calibration, etc.), ran the FR in REW and capped the result in both log and linear scales. I then ran the scene through SL using the same mic rig and mic position. Finally, I ran the scene through SL using the analog SW out jack as the feed.

 

At a glance, I can tell from this exercise that there is zero audible harmonic distortion at the seats, that the subwoofer system rolls off starting around 3-4 Hz, that the subwoofer system is calibrated pretty darned flat with the mains and that there is a wide dip in response centered at 60 Hz that needs to be dealt with. I can also tell that the subwoofer system under test is extremely capable at reference level playback of wide bandwidth source.

 

All of this is directly comparable to and verified by the REW data. The linear REW scale makes comparisons a bit quicker, but no really when you're used to making the comparison in REW's log scale... like thousands and thousands of times.

 

I could also tell by the excursion of the subwoofer drivers that the <20 Hz content is not exaggerated by SL's spectrograph of the scene. In fact, the drivers and my senses told me there was strong <20 Hz content before SL had a chance to render the scene.

 

Correct interpretation of data in any format is a given.

 

 

 

zVwniOO.png

Link to comment
Share on other sites

Bossobass, your exercise is well done except it has nothing to do with this conversation for two reasons:

 

1.  REW's linear and log displays of the frequency range have nothing to do with linear and log displays of energy content. They use the same words, but that it is.

 

2.  Comparing a REW sweep to a Spectrum Lab FFT of The Amazing Spider-Man 2 is using entirely different content for comparison and doesn't clarify anything. You said, "I can directly compare REW FR traces with log and linear graphs, to SL spectrographs, and routinely do. The SL caps do not have to be rotated." I have never seen evidence that you have done that.

 

To help you out I digitally captured a REW sweep using JRiver Media Center. This is the same sweep you used although the start and end frequencies may be different. I pressed Measure in REW and set start frequency to 2 Hz and end frequency to 120 Hz. I then pressed "Start Measuring". The output of REW was captured with JRiver and recorded to a WAV file.

 

Now you can directly compare REW FR traces with SL spectrographs. 

 

I opened the wav file of the sweep in Spectrum Lab. Why does Spectrum Lab show a difference of 11 dB from 2 Hz to 120 Hz? As you can see, a frequency response measured with REW is completely different than the exact same data shown in Spectrum Lab. Do you think the volume of the REW sweep is getting lower as it gets higher in frequency?

 

If I post this in the The New Master List of BASS in Movies with Frequency Charts thread at AVS will people say, "Look at that cool bass effect and how strong the content is down below 20 Hz!"?

 

post-42-0-53703300-1432844533.png

 

Just for reference, here is Omnimic Track 9 - Left Channel, Long Sine Sweep with a slower scroll speed.

 

post-42-0-36183900-1432844621.png

 

I'm afraid you are creating a great deal of unnecessary confusion, and mixing up the measurement of transfer function (which is what REW and other audio measurement systems do) and the various ways of displaying the spectral content of a signal, which are not at all the same thing. - John Mulcahy, author of Room Equalization Wizard

post-42-0-53703300-1432844533_thumb.png

post-42-0-36183900-1432844621_thumb.png

Link to comment
Share on other sites

Bossobass, your exercise is well done except it has nothing to do with this conversation for two reasons:

 

1.  REW's linear and log displays of the frequency range have nothing to do with linear and log displays of energy content. They use the same words, but that it is.

 

2.  Comparing a REW sweep to a Spectrum Lab FFT of The Amazing Spider-Man 2 is using entirely different content for comparison and doesn't clarify anything. You said, "I can directly compare REW FR traces with log and linear graphs, to SL spectrographs, and routinely do. The SL caps do not have to be rotated." I have never seen evidence that you have done that.

 

1.  This is entirely correct.

 

2.  Here I think you are confused by what Dave is trying to show.  Dave is showing that the difference between the two spectrographs, at the sub analog out and miced in-room, are consistent with the transfer function measured by REW.  Where his REW measurement shows a drop in output, the spectrograph miced in-room exhibits a similar drop relative to the sub out spectrograph.

 

Nevertheless, even with a 3 dB/octave "rotation" applied to both spectrographs, they would still be consistent with the REW measurement because we are comparing their difference and the 3 dB/octave rotations applied to each would cancel out.

 

With that said, I think you are still confused by the fact that REW uses a log sweep to measure the room transfer function.  You may not realize that REW de-emphasizes the 3 dB/octave roll-off in the measurement it takes as part of its process for finding the room transfer function.  What it does is convolutes the inverse of the transfer function of the log sweep.  The inverse looks like the same log sweep except reversed so that it starts at high frequency and goes to low frequency and with a 3 dB/octave roll-off that goes the opposite direction with higher frequencies at a higher level than the lower frequencies.  If the signal chain and room response is ideally flat, then REW will record exactly what is shown in SpecLab for the sweep, including the 3 dB/octave roll-off.  When REW multiplies the recorded sweep by the inverse, the time reversal effectively warps all the phases back to zero (again, we're assuming an ideally flat room response).  The 3 dB/octave roll-off toward the low frequencies in the inverse cancels out the 3 dB/octave roll-off toward the high frequencies in the recorded response.  Thus the transfer function measurement will show the flat response as it is: flat!

 

The RTA does essentially the same de-emphasis by virtue of using filters whose bandwidth doubles with each halving of frequency.  Note that REW could just as well use linear sweeps instead of log sweeps for its measurements, and the results would be the same.  If using a linear sweep, the inverse would still be  time-reversed but would be be flat in frequency.  Going a step further, it's actually theoretically possible to do transfer function measurements using any signal including live program material, provided that the signal covers the entire frequency range.  The key requirement is that the correct inverse for the test signal is used to recover the transfer function from the measurement.

 

Practically speaking, REW uses a log sweep for two reasons.  First, it has very good overall signal-to-noise ratio (SNR), and more effort is made to achieve higher SNR in the low frequencies.  We want higher SNR in the low frequencies because we want more frequency resolution there.  We may also encounter more noise in the low frequencies than the high frequencies.  The second reason for using log sine sweeps is because it avails the software of a nifty trick for recovering the levels of each harmonic distortion component.  It's kind of hard to explain here, but it's an extremely useful feature of log sine sweeps.

 

Practically speaking, the RTA uses pink noise to calibration for similar reasons, and also because it provides data for a limited number of frequency bins, compared to FFT analyses as used taken by REW and SpecLab.  Because pitch perception is logarithmic in frequency, it makes since to scale both the center frequencies and widths of the bins on a log scale.  It turns out that the SNR using pink noise is actually much worse than with log sweeps, but the RTA requires noise because of how it measures energy by basically integrating the signal level output from each filter over a short period of time (like 100 ms).  This is a consequence of what technology was available at the time the RTA was developed.

 

Does this make things more clear?  If need be, I could put together some graphics to try to better explain what's happening here.

Link to comment
Share on other sites

2.  Here I think you are confused by what Dave is trying to show.  Dave is showing that the difference between the two spectrographs, at the sub analog out and miced in-room, are consistent with the transfer function measured by REW.  Where his REW measurement shows a drop in output, the spectrograph miced in-room exhibits a similar drop relative to the sub out spectrograph.

 

With that said, I think you are still confused by the fact that REW uses a log sweep to measure the room transfer function. 

I'm not confused by what Dave is trying to show. I completely agree with his graphs. Like I've said many times, both types of graphs are accurately displaying the data - but in a different method. 

 

I discuss the fact that REW uses a log sweep in my original post at AVS. Your further explanation agrees with what the author of REW has explained in the REW manual and on forums. I agree.

Link to comment
Share on other sites

I'm not confused by what Dave is trying to show. I completely agree with his graphs. Like I've said many times, both types of graphs are accurately displaying the data - but in a different method. 

 

I discuss the fact that REW uses a log sweep in my original post at AVS. Your further explanation agrees with what the author of REW has explained in the REW manual and on forums. I agree.

 

Then I don't understand why there is disagreement here, or why there is any argument about whether or not SpecLab spectrographs need to be "rotated" or otherwise adjusted by 3 dB/octave.  (They don't.)

Link to comment
Share on other sites

Then I don't understand why there is disagreement here, or why there is any argument about whether or not SpecLab spectrographs need to be "rotated" or otherwise adjusted by 3 dB/octave.  (They don't.)

Then I must of understood some you said. The paper I referenced by Swen Müller and Paulo Massarani says, "Logarithmic sweeps have a pink spectrum, meaning their amplitude decreases with

3 dB/octave.." You said this is then de-emphasized by REW resulting in an flat transfer response and that an RTA does the same de-emphasis. An FFT does not do the de-emphasis, right? It shows the log frequency sweep with its amplitude decreasing over each increasing octave.

 

Do you agree that the Speclab maximum peak level chart shows the exact maximum SPL at every frequency using he dB reference on the right? What happens when you show the content up to 20 kHz? On the Omnimic sweep I posted earlier the level at 20 kHz is at something lik -54 dB (don't have the file right now).

Link to comment
Share on other sites

Then I must of understood some you said. The paper I referenced by Swen Müller and Paulo Massarani says, "Logarithmic sweeps have a pink spectrum, meaning their amplitude decreases with

3 dB/octave.." You said this is then de-emphasized by REW resulting in an flat transfer response and that an RTA does the same de-emphasis. An FFT does not do the de-emphasis, right? It shows the log frequency sweep with its amplitude decreasing over each increasing octave.

 

The FFT alone does not use emphasis or de-emphasis.  De-emphasis is specifically needed for recovering the correct transfer function from a measurement using a given test signal.

 

Do you agree that the Speclab maximum peak level chart shows the exact maximum SPL at every frequency using he dB reference on the right? What happens when you show the content up to 20 kHz? On the Omnimic sweep I posted earlier the level at 20 kHz is at something lik -54 dB (don't have the file right now).

 

Yes.  The sweep's contribution to any one frequency decreases by 3 dB/octave.  On the other hand, this diminishing contribution from higher frequencies is due to the speed of the sweep increasing even though the RMS level and total power of the signal stays the same throughout the sweep.

Link to comment
Share on other sites

You may not realize that REW de-emphasizes the 3 dB/octave roll-off in the measurement it takes as part of its process for finding the room transfer function.  What it does is convolutes the inverse of the transfer function of the log sweep. 

 

 

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

Link to comment
Share on other sites

It's a strange discussion wherein you are trying a circle jerk to erase your original points and claim that I'm confusing the discussion by bringing up your original points.

 

Once again, I'll try to get a response from you on your original thoughts, those which LTD02 agreed with:

 

 

Because Spectrum Labs has a different slope than other measurements you can't directly compare the charts to room measurements or subwoofer measurements such as those on data-bass.com. In order to directly compare, the Spectrum Lab chart has to be rotated.

 

 

This has several implications
When rotated to match room/subwoofer response graphs, the Spectrum Labs graphs show that
    1    bass output peaks somewhere in the 40-50 Hz region and flattens or rolls off as you go down in frequency
    2    subwoofers with limiters can still accurately produce most of the recorded bass without have a "variable frequency response" since the tilted Spectrum Lab graphs match very well with a lot of the subwoofer measurements at data-bass.com - even at higher output levels.
    3    ported subwoofers are producing more of the original bass than thought which corresponds with most listener impressions at subwoofer GTG's
    4    studios are viewing a different representation of the audio than what Spectrum Labs presents

 

 

That is a Spectrum Lab graph which has a 3 dB/octave rising slope as you go down in frequency vs REW, Omnimic or any other measurement software. This is why some of the current movie graphs at data-bass using actual disc data show a rising tail down below 10 Hz. In the 3-4 octaves below 10 Hz there is an increase of 9-12 dB.

 

 

I'm not at all confused by these ^^^ statements.

 

The contention is that spectrographs posted at Data-Bass show an increase in the 3-4 octaves below 10 Hz of 9-12dB. Further, that because of this alleged distorted view the spectrographs convey, ported subwoofers and subwoofers with aggressive limiters can produce most of the recorded content without showing a "variable frequency response".

 

This ^^^ is obviously directed squarely at comments I've made on the subject. No one else has ever made those comments that I'm aware of.

 

Those comments have nothing to do with spectrographs of REW sine sweeps, so let's please leave that part of it to LTD02 and JPC. ;)

 

Ported subwoofers do not reproduce an iota more than previously thought.

 

This one is simple. You have a measurement rig and signal chain that's flat to DC, according to your claims. If SL is exaggerating low end content by 16-30 times and your ported subwoofer can easily replay that content, a mic'd version of the ported subwoofer playing that content in-room will show no difference from the direct injected spectrograph of the same scene.

 

Here's a comparison of 3 subwoofers playing back a snip of Bass I Love You from Basstronics. One is a sealed sub with HPF around 7-8 Hz. One is an 18 Hz tuned ported sub cranked some +10-12dB hotter than the other subs (subjective preference anyone???) and one is a 20 Hz tuned ported sub that was once a highly touted ID darling which today is shown as haplessly attempting playback.

 

a9a92aab5434c3a9450c7e089ea3323f.gif

 

SubMersive vs Captivator vs SVS 20-39 PC... can you guess which is which?

 

Subwoofers with aggressive limiters have an infinitely changing frequency response as a MWB is played at reference level. ALL of the available data supports the thesis. This is also a simple test to run using SL, but honestly, these graphics and tests are a lot of work to show the obvious. If these GTG fellas were really serious about proving anything they would doo the work themselves to illustrate proofs for their various and sundry theories and subjective preferences. ;)

Link to comment
Share on other sites

as a result, in spectrum lab, when there is equal energy on the disc in the octave around 5hz, it will present 13db higher than the same amount of energy in the octave around 100hz (and another 10db higher than 1khz, and another 10db higher than 10khz, etc).

 

the tones that shredhead posted are showing equal energy per tone.  in spectrum lab, such content will appear to be at the same level.  we all agree on that, i hope.  of course the higher frequency tones will sound much louder, which is why we don't use this method to eq sound systems.  we all agree on that, i hope too.

 

the bottom line is that most folks who look at the spectrum lab captures are thinking that it is presenting data in the same way as the typical audio rta.  it is not.  the typical rta has the -3db/oct adjustment baked in, spectrum lab does not.  i'm not really sure that either could be called more or less "accurate", but one correlates with the way that we hear while the other does not.

 

To me the question is whether or not SL is showing an accurate representation of the energy of the frequencies in the content.  By energy I mean showing the relationship of how much amplitude (voltage) a certain frequency has with the time that the frequency is held.  As SME has already pointed out a few times, SL does not show how we hear sounds, it shows the content that you feed into it.  Not only does it show that content accurately but it agrees with several other spectrogram plugins for music production that I've used over the years.  SL's resolution can be tweaked ridiculously higher though. 

 

I think a lot of confusion about this has to do with the wavelengths of the frequencies in question.  Look at these 2 CEA tone bursts:

2c0ed77d3218931fab3dfe7f2e13de14.png

 

Is the energy of these 2 bursts equal?  Nope.  The amplitude of these bursts is identical but the duration of the waves is much different, therefore their energy is not equal and SL would accurately show this. 

 

As far as the REW sweeps go in SL... when REW does a log sweep, from 1-10Hz is run linear.  If it ran it logarithmic, it would be holding out 1Hz out the longest which would have taken out many people's drivers and amps by now.  So from 1-10Hz is going to show much different energy than the rest of the band (linear vs. log).  Combine that with the difference in how long a 15Hz wave is vs. how long a wave at 100Hz is and you're going to see a difference in energy in SL.  When showing a linear sweep in SL, there is less difference in levels from 1Hz to 120Hz but it will still show more energy for the low frequencies because they are longer waves, they are held for a longer time duration and therefore have more energy. 

a4c840c53bd6e7f4243f2d8cbfd55f32.png

 

Now when you input into SL 2 tones that have equal time duration and voltage, it shows you that the 2 tones have equal amounts of energy regardless of frequency.

78cd78e94d90dc9e35324eec665849d8.png

 

This is why SL is a great tool.  It accurately shows us what information is encoded on the disc.  When given a mic'd feed of your sub, it shows how much energy at what frequency your amp and drivers are reproducing.  That is why when you see a mic'd SL showing a hot pink tone at 5-10Hz, you can look over and see your drivers jumping out of the enclosures. 

Link to comment
Share on other sites

any thoughts on what's going on with this one?

 

cea 2010 burst from rew.  set to repeat.  left level constant.  started spectrum lab recording.  then i manually changed the frequency in real time.

 

it appears that the level is falling as frequency is increased.  this would seem to suggest that it isn't just the level of the source signal, but also its duration, that figures into the output level displayed in spectrum lab. (as a constant number of cycles at double the frequency contains half the time duration and so half the energy?)  

 

buVwNQF.png

 

edit: i see that shredhead posted something along this line.   

Link to comment
Share on other sites

More on changing FR in subwoofers due to limiting...

 

Besides the obvious changes seen in the power compression graphs Josh posts, wherein the huighest output sweep shows the compression imposed on the sub's native FR by the limiters, specLab can and does also show this low end compression during playback of difficult scenes (scenes with significant ULF).

 

Here's the PSA sealed 15 sub, the one Josh tested) vs the SubMersive. The native measured FRs of the 2 subs are posted above the SL graphs of the 2 subs, mic'd at the seats, both playing a clip of some Basstronics tune that has ULF content.

 

Note that the subs' FRs are normalized at 45 Hz, which is where the capped spectrographs show output the same within 1dB. If the subs were not changing native FRs, the output at 9 Hz would be within 3dB because the 2 subs' native FRs are within 3dB at 9 Hz before the clip is played.

 

Instead, we see the PSA sub is already compressing some 8dB or so at 17-18 Hz and is down around -12dB below the SubMersive at 9 Hz, suggesting 6-9dB of compression.

 

b6cd424e9ace5d7edd711abd8f748bea.gif

 

These compression estimates coincide with what Josh measured in the PSA sub with sine sweeps when you consider that complex dynamic signal is much more difficult to reproduce accurately.

 

If you back off the volume, the sub will produce the scene with less compression, or, with a different FR. With playback of MWB, the sub's FR will constantly change depending on what content it's asked to play back as the movie plays, as opposed to changing in a controlled manner when progressive sine sweeps at 5dB increase in level are the input signal.

 

The point here is that SpecLab spectrographs will show this phenomenon precisely if they are made in a controlled way to expose the phenomenon and carefully examined afterward.

 

This exercise also shows why the listening audience preferred the SM over the PSA subjectively and goes a long way toward understanding those subjective comments for those who weren't in the audience.

 

This exercise is easily repeatable with minimal setup.

Link to comment
Share on other sites

More on changing FR in subwoofers due to limiting...

 

Besides the obvious changes seen in the power compression graphs Josh posts, wherein the huighest output sweep shows the compression imposed on the sub's native FR by the limiters, specLab can and does also show this low end compression during playback of difficult scenes (scenes with significant ULF).

 

Here's the PSA sealed 15 sub, the one Josh tested) vs the SubMersive. The native measured FRs of the 2 subs are posted above the SL graphs of the 2 subs, mic'd at the seats, both playing a clip of some Basstronics tune that has ULF content.

 

Note that the subs' FRs are normalized at 45 Hz, which is where the capped spectrographs show output the same within 1dB. If the subs were not changing native FRs, the output at 9 Hz would be within 3dB because the 2 subs' native FRs are within 3dB at 9 Hz before the clip is played.

 

Instead, we see the PSA sub is already compressing some 8dB or so at 17-18 Hz and is down around -12dB below the SubMersive at 9 Hz, suggesting 6-9dB of compression.

 

b6cd424e9ace5d7edd711abd8f748bea.gif

 

These compression estimates coincide with what Josh measured in the PSA sub with sine sweeps when you consider that complex dynamic signal is much more difficult to reproduce accurately.

 

If you back off the volume, the sub will produce the scene with less compression, or, with a different FR. With playback of MWB, the sub's FR will constantly change depending on what content it's asked to play back as the movie plays, as opposed to changing in a controlled manner when progressive sine sweeps at 5dB increase in level are the input signal.

 

The point here is that SpecLab spectrographs will show this phenomenon precisely if they are made in a controlled way to expose the phenomenon and carefully examined afterward.

 

This exercise also shows why the listening audience preferred the SM over the PSA subjectively and goes a long way toward understanding those subjective comments for those who weren't in the audience.

 

This exercise is easily repeatable with minimal setup.

 

 

Hmmm.  Those speclab captures look familiar.   :)   And just to be clear, these caps are of a single PSA and 3 submersives.  

 

I really need to get off my lazy @ss and do some capping of my system.  I've wanted to compare a single submersive vs duals to see where each starts to compress with some popular scenes.  When the home theater ULF score thread over at AVS starting getting popular I thought it would be a great idea to try and back up those ratings with speclab captures.  Pick a couple scenes that have content centered at the different frequencies in the ratings and see when the system starts distorting / compression.  But it never really caught on and real life got in the way for me.

 

Anyway, thank you everyone for the hard work and interesting posts.  Maybe someday I'll have something interesting to add.

Link to comment
Share on other sites

Hmmm.  Those speclab captures look familiar.   :)   And just to be clear, these caps are of a single PSA and 3 submersives.  

 

I really need to get off my lazy @ss and do some capping of my system.  I've wanted to compare a single submersive vs duals to see where each starts to compress with some popular scenes.  When the home theater ULF score thread over at AVS starting getting popular I thought it would be a great idea to try and back up those ratings with speclab captures.  Pick a couple scenes that have content centered at the different frequencies in the ratings and see when the system starts distorting / compression.  But it never really caught on and real life got in the way for me.

 

Anyway, thank you everyone for the hard work and interesting posts.  Maybe someday I'll have something interesting to add.

 

Man, I'm so glad you popped in. B)

 

This folder of caps is a treasure trove for me, but none of the scenes are labeled beyond the movie title and none of the subs are specific, as in the single PSA vs tri-SMs.

 

I've been able to identify enough of the scenes and it's obvious that it wasn't a single SM, but I preferred not to guess that sort of detail and really hoped you might pop in and add those missing details.

 

Your scroll speed is too fast but I'm able to squeeze the graphs and matching the frequency scale is simple enough as well. In the case of the argument in this thread that ported subs can play most scenes in MWB because of an errant interpretation of the spectrographs, your folder was just the data needed.

 

Many thanks again for the giant pile of work and for making it available to me. ;):)

Link to comment
Share on other sites

Hmmm.  Those speclab captures look familiar.   :)   And just to be clear, these caps are of a single PSA and 3 submersives.  

 

I really need to get off my lazy @ss and do some capping of my system.  I've wanted to compare a single submersive vs duals to see where each starts to compress with some popular scenes.  When the home theater ULF score thread over at AVS starting getting popular I thought it would be a great idea to try and back up those ratings with speclab captures.  Pick a couple scenes that have content centered at the different frequencies in the ratings and see when the system starts distorting / compression.  But it never really caught on and real life got in the way for me.

 

Anyway, thank you everyone for the hard work and interesting posts.  Maybe someday I'll have something interesting to add.

Mike?? Man good to see you here. Pretty cool stuff with the PSA and Submersive. Dave told me about that but I didn't realize that was you. I was actually thinking of buying some ID subs and doing the same thing. It would be great to see where these limiters kick in and distort the FR. Anywho glad you're ok and sorry about that bitch life. ;)

Link to comment
Share on other sites

Dunno what "bitch life" refers to. Prolly don't wanna know. :huh:

 

But, yes, by all means run the tests and post the details of your system.

 

Compression is real (that's why it's one of the tests in the ground plane regimen), it obviously changes your FR at the seats, content and playback level-dependent, and it will show up in SL.

 

Regarding the other claim...

 

3    ported subwoofers are producing more of the original bass than thought which corresponds with most listener impressions at subwoofer GTG's

 

 

This GIF isolates the Captivator vs the Submersive. Regardless of the number of subs or the playback level, as I've suggested for many years now, based on the early ground plane test results from Ilkka, Sir Edward and Slarti, ported subs will only produce noise below tune, where the driver is said to "unload", which should tell all you need to know. Despite posting the data and arguing the point at least a hundred times over the years, people continued to state their "usable output below tune". The quoted claim is yet another way to say the same thing.

 

I believe that the animation shows beyond reasonable doubt that the ported sub is not producing anything but noise below tune. If the actual content were dropping at a rate of -10dB per decade... and the ported sub is cranked +12dB hotter, which it is... the spectrograph would look much different than it does.

 

78cf488beb7c7b5ccf6856cb86e9f47a.gif

Link to comment
Share on other sites

Hmmm.  Those speclab captures look familiar.   :)   And just to be clear, these caps are of a single PSA and 3 submersives.  

 

I really need to get off my lazy @ss and do some capping of my system.  I've wanted to compare a single submersive vs duals to see where each starts to compress with some popular scenes.  When the home theater ULF score thread over at AVS starting getting popular I thought it would be a great idea to try and back up those ratings with speclab captures.  Pick a couple scenes that have content centered at the different frequencies in the ratings and see when the system starts distorting / compression.  But it never really caught on and real life got in the way for me.

 

Anyway, thank you everyone for the hard work and interesting posts.  Maybe someday I'll have something interesting to add.

 

Hey dude :D Get that car all squared away from the last trip to DC? 

Link to comment
Share on other sites

any thoughts on what's going on with this one?

 

cea 2010 burst from rew.  set to repeat.  left level constant.  started spectrum lab recording.  then i manually changed the frequency in real time.

 

it appears that the level is falling as frequency is increased.  this would seem to suggest that it isn't just the level of the source signal, but also its duration

 

Yes.  The integration time of an FFT is something you have to take into account when you are using SpecLab (like any other FFT-based application).  Wider frequency bins=lower integration time=ability to pick up transient phenomena better in 'y' axis on waterfall, at expense of resolution in 'x axis'.  Narrower bins=higher integration time=less transient resolution, more frequency resolution in 'x direction' less so in 'y' direction on the waterfall. 

 

Try adjusting your parameters for the FFT and use different bin widths.  Your graphs will look different.  Bosso likes more freq resolution.  I like a hybrid, with 1Hz-wide bins, to pick up some transients.  Nube uses 0.5Hz-wide bins.

 

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'.

 

JSS

Link to comment
Share on other sites

Yes.  The integration time of an FFT is something you have to take into account when you are using SpecLab (like any other FFT-based application).  Wider frequency bins=lower integration time=ability to pick up transient phenomena better in 'y' axis on waterfall, at expense of resolution in 'x axis'.  Narrower bins=higher integration time=less transient resolution, more frequency resolution in 'x direction' less so in 'y' direction on the waterfall. 

 

Try adjusting your parameters for the FFT and use different bin widths.  Your graphs will look different.  Bosso likes more freq resolution.  I like a hybrid, with 1Hz-wide bins, to pick up some transients.  Nube uses 0.5Hz-wide bins.

 

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'.

 

JSS

 

Perfectly well said, Doc. B)

 

He's also severely clipping the waveform graph. ;)

Link to comment
Share on other sites

Yes.  The integration time of an FFT is something you have to take into account when you are using SpecLab (like any other FFT-based application).  Wider frequency bins=lower integration time=ability to pick up transient phenomena better in 'y' axis on waterfall, at expense of resolution in 'x axis'.  Narrower bins=higher integration time=less transient resolution, more frequency resolution in 'x direction' less so in 'y' direction on the waterfall. 

 

Try adjusting your parameters for the FFT and use different bin widths.  Your graphs will look different.  Bosso likes more freq resolution.  I like a hybrid, with 1Hz-wide bins, to pick up some transients.  Nube uses 0.5Hz-wide bins.

 

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'.

 

JSS

 

You have no idea how good it is to see you post this information. Case in point is the E110 subwoofer. I shut it down (DSP kicked in massively) with less than 20 seconds of studio music at a decently loud level when I demo'd it for the first time. CEA testing shows great performance, but CEA is short bursts. Thanks again for making your post. 

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.


×
×
  • Create New...