Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Matt]

So I calibrated my system to reference volume level using the method described here:
http://wiki.jriver.com/index.php/Volume#Reference_Level_Calibration

This happens at -18.5 dB down from full level when using the Steinberg that mojave lent me.

But at this zero point, it's really loud when listening to pop music.  I played a song from a modern album (Fun. - Some Nights, 100% peak level, -14dB Replay Gain).  At reference level, it was pushing right around 100dB at the listening position.  This is "get me in trouble with my wife" loud.  And honestly it's almost a little painful that loud.

Is this to be expected?

[mojave]

I just posted this in the build thread:

The pink noise isn't correct and is too hot. It is clipping when using SPAN and just one channel at a time. Interestingly, SPAN includes Bob Katz's K scales, too.

Also, the pink noise used for the WOW test tones and others is bandwidth limited. I actually tested a couple test discs back in December but I haven't had time to post anything yet. The WOW LFE test tone are 10-80 Hz with more output in the higher frequencies and the the speakers are 500-2000 Hz. The DVE LFE test tones are 20-120 Hz with equal output across all frequencies and the speaker tones are 400-4000 Hz with the output peaking in the center. 

I have the AIX Blu-ray calibration disc and the 5.1 Audio Toolkit, but haven't checked them yet.

The JRiver test tones are a lot hotter than when I used the WOW disc. I don't have it with me, but can double check tomorrow.

[Matt]

I did my desk speakers, which are a 2.1 set of studio monitors.

They yield basically the same results (ie. near 100dB at the listening position).

Have I been listening to music too quiet all these years

[Matt]

ReplayGain is built around -20dB Pink Noise.  This is the reference clip:
http://replaygain.hydrogenaudio.org/proposal/ref_pink.wav

So since our Pink Noise analyzes to 0.0dB, just like the Replay Gain reference clip, it means we're matching what it defines as the standard.

Replay Gain is a widely adopted standard, so I can't imagine it's built around the wrong reference level?

[mojave]

Below is the WOW pink noise and the JRiver pink noise for the front left channel. Huge difference!

I then downloaded the Blue Sky test tones. There is a hyperlink in the Instructions section that is sort hard to see. It was also a huge difference from the JRiver pink noise.

[Matt]

What if you play the reference Replay Gain file (which is -20dB pink noise)?

The only thing that really matters is the RMS number.

The graph shows that WOW is filtering the pink noise to remove low and high frequencies (although after which it's not actually pink noise).

[Matt]

Bob Katz's file here also matches the definition of Replay Gain and our noise (within 1 dB):
http://yabb.jriver.com/interact/index.php?topic=76608.msg522396#msg522396

[Matt]

The frequency difference is because some calibration files use narrow-band pink noise:
http://yabb.jriver.com/interact/index.php?topic=76608.msg522396#msg522396

We might consider the same.

However, that won't change the dB level.  So we have to get to the bottom of what's the "correct" reference level.  I'm stumped, so would appreciate any ideas.

[mojave]

Bob Katz's file here also matches the definition of Replay Gain and our noise (within 1 dB):
http://yabb.jriver.com/interact/index.php?topic=76608.msg522396#msg522396

I went to download it and realized I already did so on Dec. 10th when I was comparing Reference Level test tones. Below are both the Hydrogen Audio and Bob Katz samples.

RMS
JRiver = 0.3
HydrogenAudio = -10
WOW = -18.5
BlueSky = -19.4
Katz = -19.7
Digital Video Essentials = -33 (this is supposed to be calibrated at 75 dB)

I've created the JRiver test files on two computers and they are both identical and over 0 RMS.

There are several reasons that filtered pink noise is used when calibrating the speakers:
1.  Most SPL Meters are fairly accurate within the unfiltered range
2.  Speakers might roll off or be measured off axis. By limiting the frequencies being measured, there is a greater chance of consistency among all speakers

[mojave]

However, that won't change the dB level.  So we have to get to the bottom of what's the "correct" reference level.  I'm stumped, so would appreciate any ideas.

I received book Master Handbook of Acoustics for Christmas. I think it shows how a signal with the same digital level will have varying acoustical levels depending on bandwidth. I'll have to check it tonight.

[Matt]

I received book Master Handbook of Acoustics for Christmas. I think it shows how a signal with the same digital level will have varying acoustical levels depending on bandwidth. I'll have to check it tonight.

It could, but pink noise is special.  A little frequency slice or the whole spectrum should measure as the same level.

The basic issue is that the two (three if you count ours) reference -20dB pink noise clips are a different volume than your WOW clips.

[Matt]

I went to download it and realized I already did so on Dec. 10th when I was comparing Reference Level test tones. Below are both the Hydrogen Audio and Bob Katz samples.

RMS
JRiver = 0.3
HydrogenAudio = -10
WOW = -18.5
BlueSky = -19.4
Katz = -19.7
Digital Video Essentials = -33 (this is supposed to be calibrated at 75 dB)

I've created the JRiver test files on two computers and they are both identical and over 0 RMS.

Did you use build 18.0.108 to make the files?  It's the first to use the -20dB measure.

Also, what's the Replay Gain (which is similar to an RMS) for the files?

Let's focus on the JRiver and Katz files.  The Replay Gain is within about 1dB for the JRiver, Katz Pink, and Katz Narrow Band files on my machine.

They all have a Replay Gain near 0dB, which means they also align with the Replay Gain calibration file.

[Matt]

I moved this to the public board, hoping somebody could help us get this solved.

To recap, the volume of our -20dB pink noise (requires 18.0.108, coming soon), Replay Gain's reference file, and Bob Katz's pink noise clips are all about the same volume.

But that volume is much higher than the reference clips mojave has on his WOW calibration disc.

[mojave]

I am using build 18.0.108 to create the files.

I just found out that the option "Process independently of internal volume" does not disable Loudness.   I had the volume almost all the way down so loudness was kicked in at maximum.

I now show:

RMS
-14.3 for JRiver
-23 for Katz limited
-20 for Katz full (which still rolls off in the low frequencies)
-20 for AIX Records Calibration Blu-ray
-19.2 for Disney's World of Wonder Blu-ray

Replay Gain
0 for JRiver
1.36 for Katz limited
-.78 for Katz full

Using the K-20 metering scale in SPAN, the Katz full is at 0 RMS, Katz limited is -3 RMS, and JRiver is at 5.5 RMS.
Note:  I am using the JRiver two channel pink noise.

I can't get Analyze Audio to work with the HydrogenAudio file. I even downloaded it twice. It keeps saying that the status is queued.  

Edit: I added the AIX and Disney WOW Blu-ray RMS results.

[Matt]

I can't get Analyze Audio to work with the HydrogenAudio file. I even downloaded it twice. It keeps saying that the status is queued.  

That's because it results in precisely 0.0000000 dB, which due to an internal detail of how numbers are handled in the library, is the same as empty.

[mojave]

I added results from AIX and Disney's WOW Blu-rays.

[Matt]

I now show:

RMS
-14.3 for JRiver
-23 for Katz limited
-20 for Katz full (which still rolls off in the low frequencies)
-20 for AIX Records Calibration Blu-ray
-19.2 for Disney's World of Wonder Blu-ray

Something doesn't make sense here.

Testing with SPAN, looking at RMS values, I see:
-21.0 for JRiver
-23.0 Katz limited
-23.0 Replay Gain

I can't explain the 2 dB difference yet, but it's nowhere near the 6dB difference you're seeing.  Make sure all DSP processing is disabled if you're testing with a VST, and set the volume to 100%.

[Matt]

-23dB seems correct to me.

A full amplitude sine wave at any frequency will read -3.01 dB.

So -20dB less would be -23dB.

[mojave]

A full amplitude sine wave at any frequency will read -3.01 dB.

I used the DBFS+3 metering in SPAN to compensate (bottom right of VST).

I think I figured out my problem.  I had ZoneSwitch on. When doing the WOW clips, I made sure I just configured output format for 7.1. In fact, I checked it quite a few times while playing clips. However, the JRiver clips were going to a different zone that I had set for mono*. This was causing increased levels.  Sorry for the trouble.

With DBFS+3 I now get -21 and with DBFS I now get -18 with the JRiver clips. 

*I have a window on the east side of my office and the sun is too bright in the morning. I don't have a blind, so I slide a little whiteboard back in forth in front of the sun. Last week it pushed my speaker cables together at the back of the amp and I blew out one channel of my amp (Hafler 9130). If it wasn't for JRiver, I would have replaced the amp by now. 

[Matt]

Alright, we're getting to the bottom of it.

I'm also switching our Pink Noise generator to use the Larry Trammell method, described here:
http://home.earthlink.net/~ltrammell/tech/newpink.htm

This creates essentially laboratory grade pink noise, and makes the levels a bit more consistent.  So I think we'll have perfect -20dB pink noise next build.

I'll also look at making clips with a narrow-band filter.

[Matt]

If you use a narrow band to set the reference level, I wonder what should be on the subwoofer line for the 5.1 / 7.1 files?

[edit: we'll do 25-150 pink noise at the same RMS, of course down -10dB for proper calibration]

[mojave]

I mentioned this in my first post in this thread:

Also, the pink noise used for the WOW test tones and others is bandwidth limited. I actually tested a couple test discs back in December but I haven't had time to post anything yet. The WOW LFE test tone are 10-80 Hz with more output in the higher frequencies and the the speakers are 500-2000 Hz. The DVE LFE test tones are 20-120 Hz with equal output across all frequencies and the speaker tones are 400-4000 Hz with the output peaking in the center. 

I think 150 Hz is too high since many sealed and bandpass subs rolloff in the upper end of their range. I think 25-80 Hz would provide better results.

[Hilton]

Band limited pink noise 500hz to 2K mains/surround and 40-80Hz LFE is really best used for level matching of speakers because it helps to avoid the problem frequencies that cause spikes in meter reading from room modes (natural standing waves from room size) Once levels are matched and DSP/EQ is applied then you can use full band pink noise to set reference levels and re-check level matching. Though full bandwidth pink noise is best measured with an RTA as SPL meters dont always measure full bandwidth pink noise accurately because of filters and mic limitations.

Just to throw a spanner in the works... (ive been doing lots of research and testing on this)

From a Dolby Professional Decoder manual...

------------------------
For film work, test noise at reference level should produce an SPL of 85 dBC for each
of the main front channels (Left, Center, Right) and 82 dBC for each Surround
channel . The lower Surround level is specific to film-style mixing rooms.

For television work, test noise at reference level is typically set to produce an SPL
ranging from 79 to 82 dBC for each of the main five channels. The lower reference
level for television reflects the lower average listening levels preferred by the
consumer (typically 70 to 75 dBC).

For music mixing, each speaker channel should be set to the same SPL (just as in
television mixing). There is no standard practice for reference levels for music
mixing. Some engineers prefer to mix louder than others do, but if the levels between
channels are correct, the overall level is not as crucial.

When mixing for television or music in very small mixing rooms (for example,
remote recording trucks), the Surround channel is generally set 2 dB lower than the
front channels. This takes into account the short distance to the Surround speakers.

Experience has shown that this setting makes the sound heard by the mixer more
accurately reflect the sound heard in the home environment.

------------------------

What does this mean??? Well the surrounds may be calibrated too high by 3db depending on how the studio decided to set their levels based on the above advice.
I've noticed in some recording I need to drop surrounds by 3db to get a better balance. This explains that!

And there's also this from the same manual...

-----------------------
Subwoofer Calibration
The ideal test noise for subwoofer calibration should be band-limited pink noise,
lowpass filtered at 120 Hz. The DP564 outputs band-limited pink noise (20 to 120
Hz) on the Subwoofer output. To properly calibrate the subwoofer, a real-time
analyzer (RTA) is required. If an RTA is not available, you can approximate the
settings with an SPL meter.
When using an RTA, proper calibration requires setting the LFE channel signal to be
sent to the subwoofer, within its typical bandwidth of 20-–120 Hz, 10 dB higher (as
measured by the RTA) than the main channels.  

If an RTA is not available, setting the subwoofer channel 4–-6 dB higher, as measured
by an SPL meter, provides an approximate level. For example, set the subwoofer
channel to 89 dBC when the Center channel measures 85 dBC.

------------

Again more chances for variation in the studio calibration which might explain the 4-6db variation in subwoofer output for some recordings. sigh!


For your pink noise to be able to create more accurate measured levels it needs to be recorded with a 6db crest factor to keep the peaks closer to the RMS value.
Here's a reference to crest factor to do with power handling and a nice set of graphs to illustrate.
http://www.doctorproaudio.com/doctor/temas/powerhandling.htm

The difference in pink noise calibration levels discussed above is probably related to different crest factors in the pink noise recordings.

Basically there has been a lot of fudging and variation in the mastering of content over the last 30 years (including test tones and measurement techniques) so you have to at some point trust your ears with some well known material! And you will always have to make fine adjustments due to the fudge factor in mastering techniques... So i wouldnt get too hung up on differing pink noise readings.

[Hilton]

Also your difference in levels of 2db between JRiver and the others is to be expected because the band limited pink noise has less overall energy.

Technically if using band limited pink noise for reference levels you should be measuring with a meter @ 83db to actually reach the 85db reference level with actual movies.

[hulkss]

To add to what was said below:

To measure the subwoofer channel when fed from the LFE channel (not re-routed bass from other speakers).

This reference (page 2) http://www.film-tech.com/warehouse/manuals/TN99051701.pdf says:

With band limited 20 -120 Hz pink noise at the same level as the main speakers, the subwoofer should measure 5.5 dB higher using a "wide band" spl meter. This means a SPL meter with a flat response setting (not C or A scale) which are filtered.

Since a meter with a flat measurement setting is generally hard to find and expensive, Matt could be nice enough to generate a special subwoofer test signal with the inverse of C scale filtering to compensate and allow the use of common C-scale meter readings. The sub SPL reading still needs to be 5.5 dB higher than the mains.

[bobkatz]

Bob Katz's file here also matches the definition of Replay Gain and our noise (within 1 dB):
http://yabb.jriver.com/interact/index.php?topic=76608.msg522396#msg522396

The SMPTE has revised the RP200 standard several times due to political fights between SMPTE, Dolby, Tom Holman (who gave up on it after meeting up with resistance), so the real definition down to the dB of the level of the pink noise and the averaging method of the SPL meter requires an expert to tell you. I'll simplify it by saying, "hey, you're working at home, you're not in a theatre", and so if you use the pink noise signal which either JRiver or I supply, feed it to all channels but only play one channel at a time, and if you set it to 83 dB per channel, C weighted, slow, with the meter located at the listening position, you'll be darn close enough. If your speakers are wide range, high headroom speakers with high power amplifiers, located at 9 feet or so or greater from your ear, and if you are using some semblance of an X-Curve, and if you like your sound effects loud, then the 0 dB volume control position will be accurate about 80% of the time. Other times you'll have to turn it down, 1, 2, 3, 4, or more dB.

Non-movies, e.g. music DVDs and concerts, are rarely normalized to the motion picture standard, and will likely be louder than the movie, and you will have to attenuate them.

Music CDs are the worst offenders. If you like Snoop Dogg, he'll have to be attenuated as much as 15 dB or more to keep you from going deaf. This still doesn't mean your calibration is off, it only means that there's no standard and it's the wild wild west out there for music.

I've written extensively on the subject and there are chapters in both my books if you'd like to learn more. There's also some relief on the way, I think within the next two years, and I'll tell you about it then :=).

[bobkatz]

The frequency difference is because some calibration files use narrow-band pink noise:
http://yabb.jriver.com/interact/index.php?topic=76608.msg522396#msg522396

We might consider the same.

However, that won't change the dB level.  So we have to get to the bottom of what's the "correct" reference level.  I'm stumped, so would appreciate any ideas.

Holman advocates the narrow band pink noise to accomodate cheap microphones, variations in room acoustics and subwoofer level, and so on. He has a good point. Using a calibrated RMS digital meter, he adjusted his narrow band pink noise to yield -20 dBFS RMS (relative to the RMS of a sine wave at full scale digital). On a good day, even with my subs set just right, I still measured as much as a 1 to 2 dB difference between his narrow band PN and a wideband test signal measuring -20 dBFS on the same RMS meter. So I shrugged my shoulders and decided on using the test signal that was most consistent with my previous measurement methods I'd used over the past 5 years. And that yielded a pretty good correlation with motion picture playback in my room, since I like my motion picture effects loud and I am using a variation on an X-Curve. If you are not using some kind of X-curve, your speakers will be too bright and they will REALLY bite!

When the wife comes in the room, level has to come down about 2-3 dB  :-).

In the end, if you, me, and the lamp-post agree within 3 dB as to where we like to set the playback level of DVD or Blu-Ray XYZ, then I think we are doing pretty well, don't you?

[hulkss]

Bob,

I like the preferred "X-curve" on page 24 of this article:
https://docs.google.com/file/d/0B97zTRsdcJTfY2U4ODhiZmUtNDEyNC00ZDcyLWEzZTAtMGJiODQ1ZTUxMGQ4/edit?pli=1&hl=en

Sounds great to me when set as a target response in Audiolense.

Brad

[bobkatz]

If you use a narrow band to set the reference level, I wonder what should be on the subwoofer line for the 5.1 / 7.1 files?

[edit: we'll do 25-150 pink noise at the same RMS, of course down -10dB for proper calibration]

If you don't have an FFT-based measurement system that can actually look at the level in the individual bands, then you have to use one of the fudges like a narrow band test signal for the sub. It's all a fudge, and the Dolby recommendation of 6 dB is based on their pink noise which uses an averaging meter and it has to be their pink noise generator. If you don't have the actual Dolby noise and an actual averaging meter, then all bets are off.

My SPL meter is RMS reading, and I don't have immediate access to Dolby's test signal. So I use gaussian pink noise and an RMS meter and the RP-200 standard from, what, 1995 or so? When Holman actually had enough influence to get them to go for 83 dB and RMS, which is technically absolutely correct. Then all the sound engineers got into a fuss because they had VU meters and VUs are averaging and they were reading it all wrong. So they eventually went back to 85, but cheated by using an averaging meter and their own pink noise generator which no one else has. K.K. Proffitt at Jamsync in Nashville went through the trouble of measuring both, getting the bandwidths right, and found that Dolby's 85, measured their way with the latest standard, is actually 84 the correct way, so you're only 1 dB off to begin with :-).

For the LFE I don't trust any of the fudges such as Dolby's 6 dB recommendation using narrow band pink noise limited to 120 Hz and an SPL meter. I prefer to go for the old official 10 dB up and I do it using the FFT, with 1 octave smoothing and bars instead of curves, and looking at one of the octave bands that are within the subwoofer's bandwidth, e.g. 60 Hz. That is, set the LFE to produce 10 dB more than the 60 Hz reading of the main channel. Seems to work out right, and it correlates more exactly with the intentions of the original "10 dB up" standard.

How does that sound?

[hulkss]

No doubt the +10 dB with FFT is the right way to do it.

For those using JRiver convolution and Audiolense:
In Audiolense set-up there is a box to check to get LFE at +10 dB if you do not use a surround processor for bass management. (or built in JRiver bass management)

[bobkatz]

Bob,

I like the preferred "X-curve" on page 24 of this article:
https://docs.google.com/file/d/0B97zTRsdcJTfY2U4ODhiZmUtNDEyNC00ZDcyLWEzZTAtMGJiODQ1ZTUxMGQ4/edit?pli=1&hl=en

Sounds great to me when set as a target response in Audiolense.

Brad

I have to try it. I'm stubbornly resisting having a progressively increasing bass response as it sounds conceptually to me like a built-in smile curve. I'm more inclined intellectually to "flat to 1 (2 or 3) kHz and then tilting to 10 dB down at 20k."

But when I get Audiolense going, changing the target will be a piece of cake so I will definitely experiment with both options!

[Matt]

Wow, lots of great information is showing up in this thread!

From our side, here are some of the steps we're taking:
1) We're switching to a higher quality pink noise generator (Larry Trammell method)
2) We'll make full band and narrow-band pink noise calibration files @ -20 dBFS with the 'Create Test Clips...' tool
3) We're adding RMS reporting in decibels for the overall level and level of each channel in DSP Studio > Analyzer

There are still a few open questions:
1) Are you supposed to get each speaker to deliver 83 dB with -20 dBFS pink noise, or should you get 83 dB with all speakers playing said pink noise?  I would expect that you'd target with all speakers playing, otherwise different speaker arrangements would be at different volumes.
2) What about the mention above that Dolby recommends surrounds to be -3dB from the mains?  Currently we target the same level from all speakers (and +10dB for the subwoofer), and this calibration underpins our DSP and mixing algorithms.
3) Using a consumer level sound level meter (like the common Radio Shack meter I have), what's the best tone for calibrating the subwoofer level?  Historically we've used pink noise with a high pass and adjusted for the +10dB standard calibration.  But above people are talking about fudge factors, inverse C-scaling, etc.

Thanks for all the help so far.

( ps. Bob, I don't remember our paths crossing before, but I noticed the other day we were listed next to each other in the acknowledgements for ReplayGain: http://wiki.hydrogenaudio.org/index.php?title=ReplayGain_1.0_specification#Acknowledgements )

[Mikkel]

Hi Matt,

There are more qualified people to answer your questions than me but I'll throw in a few comments anyway.

1) Yes, 83db with -20 dbFS with only one speaker running at a time is the standard.

2) Depends on whether DTS and other mixing standards claim. Anyway, surround-channels need gain adjustment to compensate for lower listening volume levels according to THX (http://www.thx.com/consumer/thx-technology/thx-loudness-plus/) so for must of us the same level would probably work fine unless a DSP setting would actually compensate for volume adjustments.

3) Between 40-120hz I think any Radio Shack meter will do fine - if you can accept perhaps +/- 1.5db variance. You say tone... I'd think pink noise band limited to the range 45-120hz as mentioned above would be fine?

EDIT: I forgot to mention that I had a feature request to setting the reference volume: I would suggest to make a "Set reference volume"-button which
a) automatically switches JRiver to internal volume
b) automatically sets the internal volume to -40db
c) asks the user to set the amp/receiver to for instance 0db
d) plays back a pink noise tone on only one speaker (at a time)
e) asks the user to manually turn up the internal volume until it reaches 83db. Hitting "finish" would make this the reference volume in JRiver.

Oh, btw an ability to adjust the loudness reference is still on my wish list :-) (because of the lack of mixing standards in the music industry).


Best regards,
Mikkel

[Hilton]

Just a comment on the Dolby vs THX reference levels... They cant really be compared as the playback equipment including source, decoding, speakers, mastering and speaker placement and levels are different.  You can either do a dolby implementation or a THX implementation, they sound quite different. Or you can bias your setup to either Dolby or THX and fudge it between the 2 which is what most people do.

If you have THX dipoles you would certainly need more level in surrounds as levels drop away faster the more you reduce volume from reference.  Monopole (direct radiating) speakers I prefer for most of my hi-res multi channel audio but some of it is too clinical and would probably benefit from a more dispersed sound for music listening, where as movies almost always sound more enveloping with a dispersed/reflected sound from dipoles compared to direct monopoles.

In anycase, I dont think we would benefit from trying to put THX cert type settings into JRiver at this time until the baseline is at least right for Dolby setups which most of us probably have.

Im with Bob on the sub/LFE setup.  I never ever trust pinknoise and RMS meters for the sub and LFE (even narrow band LFE noise) which is why I use sweeps, and individual tones from 20hz to 200hz with FFT analysis to check for funky cross over problems and room modes before i even think about level matching. Almost all common household rooms will have problem room mode frequencies in the 30-100hz range (standing waves and nulls) which makes RMS pink noise useless because your going to have upto 10db room mode peaks and 10db holes between 20-300hz or so. The meter wont be able to distinguish a room mode peak or a null or a crossover problem. Only careful measurement with proper instruments can properly set LFE/sub crossover and levels if you want the most accurate result.

As a shortcut you can do as Bob suggested and just choose 1 or 2 frequencies below or above your room mode resonant frequency for a quick setup, as long as you can test at least 1 frequency that's in the crossover region of both the sub and your mains.

Matt for the 3db surround thing from dolby, thats for mastering in a studio so when we play it home it sounds right for the closer distances we normally sit to our surround speakers at home and it also makes it sound right in the larger cinema. No need to have a different level for surround. Its just good to know that if something doesnt sound right its probably because the engineer didnt follow the advice in the studio so we just have to turn surrounds down.  This can be done quickly with a parametric 3db cut to all surround speakers which is what I have now for those occassions.  What I really would like is to be able to have hot keys assigned to individual DSP's check boxes and then be able to program them into the MCE remote! This way the DSPs become more like quick settings in a conventional processor instead of needing at minimum a mouse!


Bob the 1-2db acoustic variation on the meter, even though the electrical RMS signals say they are the same, is because acoustic energy doesnt correlate to electrical energy.  There is more acoustic energy in wideband PN at -20dbfs compared to band limited PN at the same -20dbfs.

Oh and Matt while I have your attention... I posted this question in my hi-res audio setup thread... Can you have a look as im having some Sub level issues with some 2 channel material and Im assuming its the way I have JRSS/room correction configured or the way you route bass in JRSS between 2ch and multichannel signals.

"How does JRSS handle bass management these days with LFE and room correction in effect?  I read several threads here discussing this particular topic but it was unclear to me how LFE and bass routing works in JRSS between 2 channel and multi channel?!?!?

Thanks for any advice!" from > http://yabb.jriver.com/interact/index.php?topic=77137.0

[Mikkel]

@Hiltonk: Great post and some very good points. As you point out there is no easy way to calibrating the subwoofer SPL (except if using REWs RTA or some room calibration software like Audiolense etc.).

Depending on how far Matt is willing to go implementing an real time analyzer into JRiver could be a solution. Since people using the reference volume setting already are required to have a SPL-meter it is a small step to also connect the SPL-meter to the line-in of the soundcard. It would allow JRiver to perform its own analysis on SPL, dips and peaks by applying 1-3 octave smoothing.


Best regards,
Mikkel

[Hilton]

P.S. Im no one special like Bob... But I worked in a studio for 3 years straight out of school 20 years ago and then moved into IT engineering and did live mixing work for pub bands for 10 years and have a few muso friends that I still share my passion for music with.  So I know enough to be dangerous, but am by no stretch an expert!

Matt: I read the sony cinema calibration PDF that someone posted here above and noticed when calibrating the cinema, they cross over the L C R to the optical sub processing and then sum it (with -3db for centre) to the sub through a low pass filter for a fuller front sound stage.  Does JRiver do something similar with bass management for 2.1 and 5.1?

[Matt]

1) Yes, 83db with -20 dbFS with only one speaker running at a time is the standard.

This is a little messy, because in my mind there are two steps to setup:
1) Get relative speaker levels calibrated (simple method is DSP Studio > Room Correction, advanced method is Convolution).  For this it doesn't really matter what target level you use since the relative volume between speakers is the only thing you're adjusting.

2) Set the overall reference volume level (which is used when displaying the volume and calculating Loudness)

For #2, it's easiest if there's a single file you play and adjust the volume in the program.

If the standard is 5.1 @ 83 dB for each speaker with -20dBFS pink noise, maybe we should make a calibration file that plays on all channels where you would target 91dB (6 incoherent 83dB sources added together)?

[Hilton]

My personal files that I made for my testing only have signal on each individual channel running for 1min each. It took me several hours to put them all together.
i.e. 1min for L @ -20db all other channels silent and then rotate through all the channels.

If I need longer I just put it on repeat current.  Even my test tones from 20-200hz and sweeps are done this way. It makes it much quicker and easier to setup and flick back and forth between channels and tones.  My Wav files and Blu Ray image i made are both the same.

Im not sure that running all 6 at once would be very helpful for calibration.

[bobkatz]

I think that Tom Holman's book "5.1 up and running" (or some title like that) should be the baseline for all discussions. May I suggest, Matt, that in your "copious free time", you check out Tom's book, which I consider to be the authority on the subject, and my responses below reflect Tom's book.

Wow, lots of great information is showing up in this thread!

From our side, here are some of the steps we're taking:
1) We're switching to a higher quality pink noise generator (Larry Trammell method)
2) We'll make full band and narrow-band pink noise calibration files @ -20 dBFS with the 'Create Test Clips...' tool
3) We're adding RMS reporting in decibels for the overall level and level of each channel in DSP Studio > Analyzer

There are still a few open questions:
1) Are you supposed to get each speaker to deliver 83 dB with -20 dBFS pink noise, or should you get 83 dB with all speakers playing said pink noise?  I would expect that you'd target with all speakers playing, otherwise different speaker arrangements would be at different volumes.

SMPTE, Dolby, Holman and the rest all advocate individual channel calibration. So, unequivocably, the answer must be EACH main channel with only one channel at a time playing! There is a pseudo-European standard which has been discredited by Holman that calls for all speakers, and goes for an SPL which is the RMS sum of 83/each speaker. There are several reasons to reject an all-speaker calibration. The most importatnt is, as you move up from 5.1 to 7.1 and then 10.1, the headroom of the system REDUCES! By keeping the standard at 83 dB per speaker no matter what the format, headroom goes up as the number of channels increases, when you master program material to a consistent loudness to the ears. Yes, you are absolutely correct that different formats will produce different sound levels when the number of channels go up if the same pink noise signal is fed to all loudspeakers at once...  but that is self-modifying on the production (mixing) end, because the mix engineer goes for a consistent loudness at his ears. Thus, the more the number of channels in the format, the more headroom, but not the more level in the actual program material which is produced. Yeah, 7.1 and 10.1 sound a bit louder than 5.1 because it is clearer and has more headroom, not because of the level calibration.

2) What about the mention above that Dolby recommends surrounds to be -3dB from the mains?  

Again, Holman makes this quite clear in his book. The -3 dB is strictly the theatre standard and NOT the home standard. It's a historical anomaly, that has to do with making the calibration somewhat compatible with the old mono surround of Dolby Surround. If the film industry had it to do all over again, with separate channel surrounds, then they would also be adjusting all channel levels equally.  However, it is quite clear with THX and other organizations that ALL main channels in the home must be adjusted to the same SPL! This brings up a logical question and here is the answer: When a theatrical film is translated to DVD or bluray, the surround channels in the program material are brought down 3 dB to compensate for the home calibration, so all will be ok, I promise!
 

Currently we target the same level from all speakers (and +10dB for the subwoofer), and this calibration underpins our DSP and mixing algorithms.

Which is good! As long as the user measures each channel individually when calibrating the SPL in the home

3) Using a consumer level sound level meter (like the common Radio Shack meter I have), what's the best tone for calibrating the subwoofer level?  Historically we've used pink noise with a high pass and adjusted for the +10dB standard calibration.  But above people are talking about fudge factors, inverse C-scaling, etc.

It really is a fudge and I do not recommend it. The variables going into that fudge are so variable it's a wonder the subs get set right at all. To repeat, I do not recommend it. Mention it en passant as a workaround and recommend that people get an FFT-based system and do it right as soon as possible, if they are serious about calibrating the LFE level. Same for the bass management levels. How do they expect to get that right if they don't have an accurate measurement system? So why should they then resort to a simple flat sound level meter and a fudge when they already have the tools available to get it right.

Thanks for all the help so far.

( ps. Bob, I don't remember our paths crossing before, but I noticed the other day we were listed next to each other in the acknowledgements for ReplayGain: http://wiki.hydrogenaudio.org/index.php?title=ReplayGain_1.0_specification#Acknowledgements )

Wow! Matt, my discovery of JRiver is directly thanks to Mitch, one of your avid users and your best fan. I got here through six degrees of separation but I'm glad to be here. The things I continue to learn...  I've never been directly connected with the ReplayGain organization but I have influenced so many of them over the years and I am closely connected with one of the developers of the ReplayGain standard in other groups that I am involved in. I guess it seeps in via osmosis  :-).

[mojave]

2) Set the overall reference volume level (which is used when displaying the volume and calculating Loudness)

For #2, it's easiest if there's a single file you play and adjust the volume in the program.

If the standard is 5.1 @ 83 dB for each speaker with -20dBFS pink noise, maybe we should make a calibration file that plays on all channels where you would target 91dB (6 incoherent 83dB sources added together)?

Once my relative levels are set, I just playback the -20dBFS limited bandwidth pink noise in the front left speaker and use that to set Reference Level in JRiver. Since I already set the relative levels (using convolution) the other speakers should all be correct.

[bobkatz]

This is a little messy, because in my mind there are two steps to setup:
1) Get relative speaker levels calibrated (simple method is DSP Studio > Room Correction, advanced method is Convolution).  For this it doesn't really matter what target level you use since the relative volume between speakers is the only thing you're adjusting.

2) Set the overall reference volume level (which is used when displaying the volume and calculating Loudness)

For #2, it's easiest if there's a single file you play and adjust the volume in the program.

If the standard is 5.1 @ 83 dB for each speaker with -20dBFS pink noise, maybe we should make a calibration file that plays on all channels where you would target 91dB (6 incoherent 83dB sources added together)?

I think that doing a quick SPL level check with all channels playing will reveal any accidental polarity inversions or phase shifts and will give you some kind of comfortable feeling, but individual channels is the way to go. Even if you use a convolution engine like AudioLense and adjust SPL of one speaker only after that, I would still listen at the end to things like coherent (mono) pink noise to the LF and RF and see if the center image is focused and precised, then switch to the center speaker only and compare the virtual positions, things like that just to confirm that the alignments really came out right. Personally, even if you use a convolution engine, I would play at least band-limited pink noise out of each speaker as a confirmation. "Never turn your back on computers".

BK

[mojave]

Matt, can you have JRiver generate a test clip that has 30 seconds of each channel before moving to the next channel? Also, instead of going in channel order I like it when it circles the room:  Left, Center, Right, Right Surround, etc.

[Mikkel]

Hi Matt,

As you mention there is no particular reason to use 83db for speaker level adjustments.
But when setting the reference volume only one speaker should be played - all other speakers will match that level because of before mentioned speaker level adjustment.

Sorry for the confusion.

Best regards,
Mikkel

[Matt]

Thanks for all the help everyone.  I think you've answered my questions.

Matt, can you have JRiver generate a test clip that has 30 seconds of each channel before moving to the next channel? Also, instead of going in channel order I like it when it circles the room:  Left, Center, Right, Right Surround, etc.

We'll do one better.  We'll add each channel as a track to Playing Now so you can flip through them just like you'd flip through any tracks.  Coming soon.

[BartMan01]

Is this to be expected?

In my experience, yes.  Running reference level movie audio in a home environment is typically LOUD - I find the same in Theaters that run at reference levels, the loudest passages are really too loud.  A lot of it depends on the source, but I usually find myself between -8 to -12 below reference when watching movies.  Even at -8 the wife almost always asks me to turn it down when watching together.