The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
tell me what you think you know about dBs Studio Monitors
Old 18th September 2011
  #1
Lives for gear
 

tell me what you think you know about dBs

Recently read an AES paper that certainly appears to make a noob error.

Help me out here and tell me where I am wrong.
(And I know there are lots of you who enjoy doing that)

In the digital domain we have dBFS
the reference is maximum digital value
(eg 111111111111111111111111) and goes down from there to lower binary values ending at all 0's with a dBFS value relating to the number of bits used (eg 16 24 32 binary).

There is no way to relate any dBFS to a voltage as there is no voltage in digital. And in analog the absolute voltage is set by the d/a converter not the digital value.

In the analog domain we have things like dBu/dBv dBV dBm etc. They all relate to a specifiec voltage in the analog domain and can be compared to others in the same domain if you consider the different references. EG +4dBu compared to -10dBv is not 14 dB higher but about 12 dB because of the different references.
[dBV reference = 1V; dBu reference about 0.775 volts]
But since both references are in the same type of units eg voltage we can calculate the difference.

In the physical domain we have dBSPL for sound pressure.

There is NO way to convert between the dB types in different domains.

Yet that is exactly what the AES paper did!. It claimed that things in the analog domain were higher in terms of dBFS.
NONSENSE! There is no correlation between dBFS and dBu or dBV. With this error the AES paper merrily claims that there is an alleged intersample peak. Clearly only if you confuse digital and analog domains and mix units that do not relate.

bash away !
Old 18th September 2011
  #2
All this is kind of irrelevant unless you can point us to the relevant paper. Everything else you say is audio 101 basics. Maybe the AES did state something daft; maybe there's a massive typo. Maybe you've just misunderstood it. But without the paper to hand, who possibly knows?!
Old 18th September 2011
  #3
Lives for gear
 
DaveUK's Avatar
. . .
Old 18th September 2011
  #4
Lives for gear
 

Quote:
Originally Posted by psycho_monkey View Post
All this is kind of irrelevant unless you can point us to the relevant paper. Everything else you say is audio 101 basics. Maybe the AES did state something daft; maybe there's a massive typo. Maybe you've just misunderstood it. But without the paper to hand, who possibly knows?!
nelson and lund 109th AES convention is just one example

everybody who claims there are intersample peaks has confused the digital and analog domains or is totally unable to speak english
Old 18th September 2011
  #5
Lives for gear
 
bogosort's Avatar
Quote:
Originally Posted by oldeanalogueguy View Post
There is no way to relate any dBFS to a voltage as there is no voltage in digital.
Well, the ADC uses a reference voltage to determine the value of dBFS. The convertor's word length then determines the value of each bit, e.g., if Vref = 5 volts, and we're using 16-bit words, then each bit is equal to about 75 microvolts and dBFS is a little less than 5 volts.
Old 18th September 2011
  #6
I think maybe a reading of Don Quixote might serve some people better perhaps?
Old 18th September 2011
  #7
Lives for gear
 
Ben B's Avatar
 

Play a 1kHz test tone out of your DAW at -18dBFS or -20dBFS (depending on the calibration standard used) and measure the voltage. It should be 1.228 V RMS. Send a tone into your DAW from your mixer at 1.228 V RMS. It should read -18dBFS or -20dBFS (again, depending on the calibration standard used).

It's all about how converters are calibrated. dBFS doesn't directly relate to voltage, unless some common reference is established (which it is).

In the same sense, you can state that dBFS values don't correspond to any specific SPL. But, to cite one example, it is common in film mixing to adjust your monitoring system so that pink noise played at -20dBFS RMS will result in 85dBSPL (C-weighted).

-Ben B
Old 18th September 2011
  #8
Lives for gear
 

Quote:
Originally Posted by Ben B View Post
Play a 1kHz test tone out of your DAW at -18dBFS or -20dBFS (depending on the calibration standard used) and measure the voltage. It should be 1.228 V RMS. Send a tone into your DAW from your mixer at 1.228 V RMS. It should read -18dBFS or -20dBFS (again, depending on the calibration standard used).

It's all about how converters are calibrated. dBFS doesn't directly relate to voltage, unless some common reference is established (which it is).

In the same sense, you can state that dBFS values don't correspond to any specific SPL. But, to cite one example, it is common in film mixing to adjust your monitoring system so that pink noise played at -20dBFS RMS will result in 85dBSPL (C-weighted).

-Ben B
you can set 0dBFS to any voltage value in a/d/a converters
so since it is not unique you cannot use any dBFS values and start claiming there are peaks in the analog domain. there are only peaks if you do a bad design of the d/a.
Old 18th September 2011
  #9
Lives for gear
 

Quote:
Originally Posted by bogosort View Post
Well, the ADC uses a reference voltage to determine the value of dBFS. The convertor's word length then determines the value of each bit, e.g., if Vref = 5 volts, and we're using 16-bit words, then each bit is equal to about 75 microvolts and dBFS is a little less than 5 volts.
the a/d/a circuits can be designed to any voltage values
so you cannot match them up to dBFS and validly claim that there are any intersample peaks.

the only way to have problems is to badly design the d/a.

THERE ARE NO INTERSAMPLE PEAKS !!
it is nonsense spouted by people who confuse the digital and analog domains.
Old 18th September 2011
  #10
teo
Lives for gear
 
teo's Avatar
 

Quote:
Originally Posted by oldeanalogueguy View Post
There is no way to relate any dBFS to a voltage as there is no voltage in digital. And in analog the absolute voltage is set by the d/a converter not the digital value.
Of course you can relate dbFS and voltage, given a specific calibration. On my system, 0dbFS = +16 dbV
Old 18th September 2011
  #11
Lives for gear
 
stinkyfingers's Avatar
 

intersample peking duck-sauce heh
Old 18th September 2011
  #12
Lives for gear
 
O.F.F.'s Avatar
 

Quote:
Originally Posted by oldeanalogueguy View Post
In the digital domain we have dBFS
the reference is maximum digital value
(eg 111111111111111111111111) and goes down from there to lower binary values ending at all 0's with a dBFS value relating to the number of bits used (eg 16 24 32 binary).
All 1s is 0dBFS but all 0s is effectively -infinitydBFS (ie no signal at all) regardless of bit depth.
The bit depth just determines how many discrete steps the signal can assume between those two points (For 16bit that is around 65 000 and for 24bit around 2.5 million) and, once past 16bit, a largely theoretical noisefloor.
Old 18th September 2011
  #13
Lives for gear
 

Quote:
Originally Posted by O.F.F. View Post
All 1s is 0dBFS but all 0s is effectively -infinitydBFS (ie no signal at all) regardless of bit depth.
The bit depth just determines how many discrete steps the signal can assume between those two points (For 16bit that is around 65 000 and for 24bit around 2.5 million) and, once past 16bit, a largely theoretical noisefloor.
only in the digital domain

the analog voltage range depends on the d/a circuit

and i can set the a/d so that the noise floor is near all 0's even with 32bit binary and a lot of the range will be above the noise
Old 18th September 2011
  #14
Lives for gear
 

Quote:
Originally Posted by teo View Post
Of course you can relate dbFS and voltage, given a specific calibration. On my system, 0dbFS = +16 dbV
nonsense

you can relate dBu and dBv exactly
you cant relate dBFS to any voltage
because everybody can relate it to a different voltage
which makes it meaningless

tell me what dBFS value = -11dBSPL
Old 18th September 2011
  #15
Lives for gear
 
Ben B's Avatar
 

Quote:
Originally Posted by oldeanalogueguy View Post
you can set 0dBFS to any voltage value in a/d/a converters
I guess you didn't read my post all that carefully.

Quote:
Originally Posted by oldeanalogueguy View Post
so since it is not unique you cannot use any dBFS values and start claiming there are peaks in the analog domain. there are only peaks if you do a bad design of the d/a.
I never mentioned anything about intersample peaks in my post. I haven't really read much of what you've written on that particular topic.

-Ben B
Old 18th September 2011
  #16
Lives for gear
 
Ben B's Avatar
 

Quote:
Originally Posted by oldeanalogueguy View Post
nonsense

you can relate dBu and dBv exactly
you cant relate dBFS to any voltage
because everybody can relate it to a different voltage
which makes it meaningless

tell me what dBFS value = -11dBSPL
Again, this all depends on calibration standards, which DO exist in our industry.

-Ben B
Old 18th September 2011
  #17
Lives for gear
 
Ben B's Avatar
 

Besides, "-11dBSPL" is below the threshold of hearing. Why would you choose this number?

-Ben B
Old 18th September 2011
  #18
Quote:
Originally Posted by Ben B View Post
Besides, "-11dBSPL" is below the threshold of hearing. Why would you choose this number?

-Ben B
I believe that is the theoretical SPL of one hand clapping.
Old 18th September 2011
  #19
Lives for gear
 
Ben B's Avatar
 

Quote:
Originally Posted by theblue1 View Post
I believe that is the theoretical SPL of one hand clapping.
I see. This is getting very Zen.



-Ben B
Old 18th September 2011
  #20
Lives for gear
 

Are you claiming intersample peaks are not real? (they are)

Or are you claiming they can be minimized or eliminated through good d/a design? (also true)

You have to set your reference somewhere. All it is saying is for example if you set 0dbfs to 10V it is possible that after d/a you will have signals greater than 10V which is a result of intersample peaks and can cause distortion or other unpleasantness.
Old 18th September 2011
  #21
Lives for gear
 

heh

i have no clue, but i'm positive 12.42 posts per day in the digital World results in severe clipping and harsh and unpleasant distortion...

Old 18th September 2011
  #22
teo
Lives for gear
 
teo's Avatar
 

Quote:
Originally Posted by oldeanalogueguy View Post
nonsense

you can relate dBu and dBv exactly
you cant relate dBFS to any voltage
because everybody can relate it to a different voltage
which makes it meaningless

tell me what dBFS value = -11dBSPL
What are we talking about, a 1k tone at -11db SPL, because if we are talking pink noise it's more complicated. Anyway...

If you tell me how your room is calibrated (both converters and speakers) I could tell you. Actually, I couldn't, because I'm never been into calibrating monitors at specific levels, don't need to.
But the point is that, given the information on your room above, I think it can be done. You do have a room, right?
Old 19th September 2011
  #23
Lives for gear
 
audiogeek's Avatar
 

Quote:
Originally Posted by Ben B View Post
Besides, "-11dBSPL" is below the threshold of hearing. Why would you choose this number?

-Ben B
LOL, glad someone else caught this too. What exactly does -11dBSPL sound like, anyways?
Old 19th September 2011
  #24
Lives for gear
 

Quote:
Originally Posted by johnnyc View Post
Are you claiming intersample peaks are not real? (they are)

Or are you claiming they can be minimized or eliminated through good d/a design? (also true)

You have to set your reference somewhere. All it is saying is for example if you set 0dbfs to 10V it is possible that after d/a you will have signals greater than 10V which is a result of intersample peaks and can cause distortion or other unpleasantness.
they are not real

they are nonsense spouted by people who confuse the analog and digital domains

peaks are real but only a problem with a badly designed d/a

i set my reference so the analog sample is way below all the digital "peaks" in binaryland. hence with my d/a there are NO INTERSAMPLE PEAKS. nto just because that term is meaningless but because the d/a does not go that high if it were meaningful.
Old 19th September 2011
  #25
Lives for gear
 

Quote:
Originally Posted by Ben B View Post
Besides, "-11dBSPL" is below the threshold of hearing. Why would you choose this number?

-Ben B
to see if anybody knew wht spl really was
and perhaps trap some blowhard who claims there is a relation to dBFS
Old 19th September 2011
  #26
Lives for gear
 

Quote:
Originally Posted by Ben B View Post
Again, this all depends on calibration standards, which DO exist in our industry.

-Ben B
and nobody uses them
or they all use different ones
or theyignore them because they cause problems and want better results than the standard allows
Old 19th September 2011
  #27
Lives for gear
 
Ben B's Avatar
 

Quote:
Originally Posted by oldeanalogueguy View Post
and nobody uses them
or they all use different ones
or theyignore them because they cause problems and want better results than the standard allows


-Ben B
Old 19th September 2011
  #28
Gear Maniac
 

Intersample peaks are related to digital level metering, yes? It's not about the complete set of samples that make up the audio data, it's about the samples that the metering pulls out from that data.

Imagine each "O" as a sample of audio data.

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

Now, the metering takes "samples" from the complete audio data. These are represented by the X's.

OOOOOOXOOOOOOXOOOOOOXOOOOOOXOO

The meter tells you what level the Xs are running at, but if a peak happens in between the "X" samples (an "intersample peak"), you won't know. So you could be clipping without it showing up.

The ADC samples at 44.1/96/whatever and the metering in your DAW samples at a much slower rate.
Old 19th September 2011
  #29
Here for the gear
 

Quote:
Originally Posted by oldeanalogueguy View Post
they are not real

they are nonsense spouted by people who confuse the analog and digital domains

peaks are real but only a problem with a badly designed d/a

i set my reference so the analog sample is way below all the digital "peaks" in binaryland. hence with my d/a there are NO INTERSAMPLE PEAKS. nto just because that term is meaningless but because the d/a does not go that high if it were meaningful.
Johnnyc has a few good (and correct) points. Let me try to bring it closer to the practical side of this theory.

The AES standard for calibrating adc's and dac's is a sinewave at 1kHz peaking at -18dBFs in the digital domain.

Many converters are calibrated this way when the come out of the factory.

Many studios are therefore calibrated to -18dBfs = +4dBu = 0dB VU. Only movie mixers have an official relation with dBSPL and Bob Katz has suggested a K-system where a dBSPL is linked (www.digido.com)

This would mean we could boost the sinewave with 18dB before the signal would clip in the digital domain. Nothing special yet.

Is this the paper you are refering to?
http://www.tcelectronic.com/media/le...per_aes109.pdf

A sinewave has no harmonics. A triangle wave for example, though, has harmonics and will cause intersample peaks when the audio is converted to analog. It is inherrent to the process of digital-to-analog conversion as the process is not linear. You will always need a lowpassfilter and/or decimation filter correcting the harmonics created by the process of digitization and it depends on the signal and on these filters how much it suffers from this process . It is also true that well designed dac's should take this into account and provide enough headroom to allow these peaks.

However, this never should never be a problem if you work with 'normal' levels, staying well under 0dBfs.

Only mastering engineers have to take it into account, as some poorly designed consumer dacs could have more of a problem with those signals.

All the best.
Old 19th September 2011
  #30
Gear Head
 

I prefer to set my converters so 0 dBfs = -16 Bels, or -18.420446688 Np, whichever best suits the music - it provides more headroom.
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump