You must log in or register to post. Registration is free! [ Register ]

The Recording Project Forum Index -> Signal Processing

FAQ Search Photo Album!
 dBu vs dBFS
View previous topic :: View next topic 
Post new topic  Reply to topic
Author Message
RedStone Offline
Giant Member


Joined: 27 Nov 2004
Posts: 1431
Location: British Columbia, Canada


PostPosted: Mon Feb 27, 2006 9:37 am    Post subject: dBu vs dBFS Reply with quote

I was reading some stuff on the differences between dBu and dBFS (It always confused me) - so, I thought I'd share ...

-24dBFS (FS means full scale - the digital peak limit) = 0dBu or dBv (the analog way of measuring a sound signal)

dBu is the same as dBv where u means 'unloaded' and v referres to voltage - dBu is the modern abbreviation

0dBu is equal to 0.775 Volts at 600 ohms

0dBu = -24 dBFS

So, if you are using analog processing, and the highest peaks in your digital recording reach up to -3dBFS, and your analog gear is rated to clip at +24dBu, this means that you have ZERO headroom to work with!

-3dBFS = 21dBu, which will mean some heavy saturation .... and If you clip your analog gear - it won't work anymore ... which is why plenty of headroom in the gear's circutiry is important (6dB or more above 0dBFS = +30dBu clip rating) - although this kind of headroom is expensive

Otherwise, you should be mixing at very low levels while using any analog gear. If your gear is rated to clip ANY lower than 24dBu (not good), I would think that it is dangerous to push the levles of your digital stuff above -6dBFS

-6dBFS = +18dBu -

Pushing the levels of analog gear that is rated to clip at lower than +24dBu will induce even heavier saturation. The minimum recommended level you should be peaking at for digital audio is -10dBFS - I suppose so that you keep a good actual amount of RMS power in the mix

I guess this is why you should use attenuators with analog gear eh? one at each end - the first brings the level down so as to not blow the gear, and the other brings it back up to Full Scale goodness?

Ryan


Last edited by RedStone on Mon Feb 27, 2006 1:18 pm; edited 1 time in total
Back to top
View user's profile Visit poster's website
skod Offline
Giant Member





PostPosted: Mon Feb 27, 2006 12:24 pm    Post subject: Reply with quote

Well, unfortunately, you are still a little bit confused. You have part of the answer, but not all of it. The problem is that measurements referred to in decibels are all relative- and only some of the measurements you mention can actually be compared at all.

0dBfs is actually only meaningful internal to a piece of gear. There is absolutely *no standardization* with respect to precisely what signal voltage will result in a full-scale value inside anything, nor is there any standardization with respect to what voltage will come out of something that has a full-scale sample in its D/A. I have one set of converters that produce fullscale saturated samples (0dBfs) with an input level of +26.8dBu (16.9VRMS), and another set that go to 0dBfs with an input level of only +8.6dBu, or 2.08VRMS.

The AES has attempted to standardize 0dBfs as +24dBu (+21dBV, or 12.27VRMS) for pro gear, but very few of us can afford hardware that actually pays any attention whatsoever to the full set of AES specs- and even the stuff that tries probably doesn't get it right that often. The stuff we actually use will vary all over the place. So stating that 0dBfs = +24dBu (and it ought to be plus 24, not minus 24, BTW) is just not accurate in the general case.

The important thing is that 0dBfs is the law _inside your gear_, whatever voltage reference the designers of your gear selected. Go over that, and you have clipping. But there is no general-case way to state what absolute level (whether expressed in dBu, dBm, dBV, dBv, dBr, or rms or peak-to-peak voltage) will cause any given piece of gear to convert a fullscale sample, or how that will then interact with the next piece of gear down the pipe: it simply varies all over the map. So the answer is not only somewhat harder than you were led to believe, it also varies all over the place...

Here is essentially the seminal accessible-to-the-non-nerd article about not pushing 0dBfs (regardless of what voltage it takes to get you there!): http://www.tcelectronic.com/media/nielsen_lund_2000_0dbfs_le.pdf, for what it is worth. In it, you can see how samples can get away from you even with pseudo-"conservative" metering. The way music is squashed all to hell these days, I think more nasty things happen in the digital domain than in the analog....

_________________
Scott Griffith, Scratchpad Studio/Earfull Sound
the Nerd formerly known as Skippy
Back to top
View user's profile
RedStone Offline
Giant Member


Joined: 27 Nov 2004
Posts: 1431
Location: British Columbia, Canada


PostPosted: Mon Feb 27, 2006 12:56 pm    Post subject: Reply with quote

wow ... ok
that's awesome ... so the key is to find out how each piece of gear pumps out the signal - if nothing is regulated, then ya - it will surely be all over the map!
I guess in the race for wealth and cheap products, they decided that standardization just wasn't that important ... huh ...

This was just me trying to figure out how to read a VU meter properly and relate it back to dBFS - then I got into this whole other netherworld ...

Say I have a VU meter on an analog piece of gear that is showing 0dBu, it won't necessarily translate to acutal -24dBFS because of the lack of voltage regulations of the signal leaving the gear??

hmmm ... I can see how anyting over +24dBu can produce a saturated 0dBFS sample - Doesn't it just means that the AD converters are designed with a lot of headroom? check the specs ... I'm almost sure that they will have a clip point above 24dBv ...

As for the gear that produces a 0dBFS sample with an input of +8dBv - the clipping point must be pretty low (like +10dBv or less) - so when it converts the sample - it reads the signal at +8dBv as "clip", saturates it and ends up converting to 0dBFS - I'd stay well away from that converter ... (I think)

Does that make sense?
Back to top
View user's profile Visit poster's website
Mosfet Offline






PostPosted: Mon Feb 27, 2006 1:03 pm    Post subject: Reply with quote

Also, analog clipping is NOTHING like digital clipping, and can actually have a VERY pleasing sound, as well as offer a sort of "limiting" of the signal. So, to say that "and If you clip your analog gear - it won't work anymore", is TOTALLY off base! Some analog gear does not produce very pleasing tonality while clipping, while some offer a very nice sound for certain applications. I have saturated the hell out of eq's and got very nice sounding results!


"Headroom" is a dangerous word to use. It is far too often misused! To start, there is what I call the "theoritical headroom", or more specifically, the maximum Peak level a peice of gear can produce. Then there is the "usable headroom", or more specifically, the maximum RMS level the gear can produce and still sound acceptable! The difference between these will vary with every peice of gear. What it all means is that you could very well have far exceeded the maximum peak level the gear can work at, but you can still increase RMS level at the expense of clipping. Whether that clipping sounds good or not depends on any number of factors. Usually, if the peak is caused by a fast transient (which usually means it was also high frequency!) you can clip it MUCH farther, thus increasing RMS level than you can say a sound with a very slow transient (which usually means a slower level). This has to do with the ears ability to actually HEAR the distortion! The longer the duration of clipping, the better you will hear the distortion. Short lived clipping can often go undetected by the ear! This is why in the analog world, you can often clip the hell out of a snare drum, or other fast transient percussion peices and not hear audible distortion. If anything, the sound will actually "mellow" because the high frequencies are topped out, yet more low frequency information is contributing to the RMS. Think a sort of frequency limiter. The longer the signal is in clipping, the lower the frequency that you will hear distortion applied to.

We could discuss this a LOT, but I think I have covered some basic concepts that you can work with in a production environment. I concern myself FAR less with watching meters (except digital peak meters!!!! I watch those like a hawk!) and more with having the analog gear produce the type of sound I want. If that means pushing it into clipping, then it will clip! If the clipping on the analog gear is causing clipping on a A/D converter, I will find a way to reduce the signal somewhere between the cool sounding analog clipping and the A/D converter.

Redstone, I casually follow some things that you say on the BBS. Please don't take what I am going to say next as me not liking you, or even trying to embarrass you.

But............

I have noticed that you often misunderstand "involved" topics in audio, but post like you have the facts. What I don't see in your posts is you posting them as more of a "question". If you are 100% SURE of your facts, that is okay, but I find you to be confused about things a little too often.

Maybe you should back off a bit on offering "advice" on how to do things or how to interrpret audio production until you have a better handle on it. To the average newbie, your initial post could have been taken as "fact" based upon how you presented it, yet, a guy like skod comes along and pretty much blows holes on the technical aspects of almost everything you said, and a guy like me comes along and blows holes in the more "artistic" aspects of what you said. While me and skod would certainly agree that exceeding 0dbfs on a A/D/A is VERY bad, that appears to be about where our agreement with your post ends.

Again, I am just suggesting that you back off a bit. Maybe start posting in more of a "hey, I was figuring things this way, am I high or something?" type of way. It sort of gives the opportunity for those that kind of know what is going on to come along and give you the good information without having to tell you that you are wrong/confused.

Peace.

_________________
My wang is porn stah big in metric numbers!
Back to top
View user's profile
Mosfet Offline






PostPosted: Mon Feb 27, 2006 1:09 pm    Post subject: Reply with quote

Also, I should point out something about "headroom".

Headroom is simply the amount of signal above your "average" signal that you can go before clipping. You CREATE headroom through gain staging. As I posted above, there is the theoretical headroom, and the usable headroom. If the gear handles clipping gracefully, and/or the signal that is causing clipping is short lived, your usable "headroom" is going to be MUCH higher than the theoretical headroom.

Again, headroom is just the difference between your average level and the level you can achieve before clipping. If you need more, you simply turn down something BEFORE the clipping stage. It is that simple.

_________________
My wang is porn stah big in metric numbers!
Back to top
View user's profile
RedStone Offline
Giant Member


Joined: 27 Nov 2004
Posts: 1431
Location: British Columbia, Canada


PostPosted: Mon Feb 27, 2006 1:17 pm    Post subject: Reply with quote

fair enough ... I get excited easily by reading and learning ... and trying (haven't got to the trying stage yet)

Most of what I post is theory anyway -- not real world experience ... something that I hope to get more of in a real studio - not just a home based thing ---

I will try to re-word things ... I'm not really pompus - just excited Smile
Back to top
View user's profile Visit poster's website
skod Offline
Giant Member





PostPosted: Mon Feb 27, 2006 1:40 pm    Post subject: Reply with quote

RedStone wrote:
Say I have a VU meter on an analog piece of gear that is showing 0dBu, it won't necessarily translate to acutal -24dBFS because of the lack of voltage regulations of the signal leaving the gear??


Exactly. It could mean anything from something like -30dBfs on some gear that has lots of design margin (I won't call it headroom here), to possibly even hard fullscale clipping on some other gear (the consumer-grade stuff that is designed to a -10dBV reference level and is working hard to even acommodate *that*). It'd be great if everybody agreed, but the painful fact is that they don't- notwithstanding the efforts of the AES to impose some order.

Worse yet, VU meters are an approximation for real-world program material: the ballistics of the meter do a great deal of averaging of the readings. For the purposes of discussionh ere, let's talk simple steady-state sine waves, like from a test tone/lineup oscillator- just so that we can leave VU ballistics out of it. That's one good thing about real digital metering: you can get real peak readings, without any averaging, and that is something you can _never_ get from a mechanical metering device....

RedStone wrote:
hmmm ... I can see how anyting over +24dBu can produce a saturated 0dBFS sample - Doesn't it just means that the AD converters are designed with a lot of headroom? check the specs ... I'm almost sure that they will have a clip point above 24dBv...


Maybe, maybe not. As an example, the analog portions of my RME converters have a great deal of real "headroom" over what they can handle in the digital domain: they can crank along with fullscale samples and still have a great deal of analog swing and bandwidth left without contributing to the distortion number. Here are their published specs for the switch-selectable voltage references they support:

Input/Output level at 0 dBFS @ Hi Gain: +19 dBu
Input/Output level at 0 dBFS @ +4 dBu: +13 dBu
Input/Output level at 0 dBFS @ -10 dBV: +2 dBV

Their reference level can be switched to support -10dBV, +4dBu, and "even hotter than that just in case" levels. Internally, they use +-24v supplies, and from looking at the circuit design, I can make easily myself believe that the analog buffer amps on either side of the converters can handle +26dBu signal swings with ample design margin to spare. But the bottom line has to be "so what": as long as I can't get the analog circuitry to distort when they are driving the A/D block up to 0dBfs, or reconstructing 0dBfs signals from the D/A on the output side, I'm fat, dumb and happy!

RedStone wrote:
As for the gear that produces a 0dBFS sample with an input of +8dBv - the clipping point must be pretty low (like +10dBv or less) - so when it converts the sample - it reads the signal at +8dBv as "clip", saturates it and ends up converting to 0dBFS - I'd stay well away from that converter ... (I think)


You can't really say. If I were doing the design, for example, I'd want to have a great deal of design margin in the analog support circuitry on the A/D side (perhaps 12dB at a minimum), just to be absolutely certain that my analog electronics didn't provide the weak link in terms of voltage swing, slew rate, and overall bandwidth. I'd want to have absolutely perfect output from the analog blocks right up to the point that the signal hit the A/D converter and it said "Fuck, this is way too hot!".

The point is that the clipping points of the analog blocks that surround the converters is not spec'd, and is usually not even all that interesting for the putposes of this discussion: if you have enough bandwidth and clean voltage swing that the A/D can't go there (because it would be over 0dBfs), you're in good shape.

However, whatever box you plug into the _outputs_ of that circuitry may care a lot: your converter may be able to swing a perfectly clean +19dBu at its outputs, but if the next block in the chain goes tits-up with analog clipping at +12dBu, you have a problem.

_________________
Scott Griffith, Scratchpad Studio/Earfull Sound
the Nerd formerly known as Skippy
Back to top
View user's profile
RedStone Offline
Giant Member


Joined: 27 Nov 2004
Posts: 1431
Location: British Columbia, Canada


PostPosted: Mon Feb 27, 2006 2:39 pm    Post subject: Reply with quote

skod wrote:
Quote:
a test tone/lineup oscillator-


OK ... making sense .... so assuming you can tweak the system to pump out, digitally, whatever is read on the digital meter - using gain staging? (I assume that this is the art of using attenuators?) to make it so that If it says OdBFS, that is what's leaving. (using whatever equipment people use to test that sort of thing .. I don't know - the test tone and oscillator?)

So you send an actual full scale 0dB to an analog box that can handle a signal "cleanly" up to around +19dBu - if your full scale signal is a true 0dBFS, then your gear will start to saturate the signal at that level - how much exactly depends on the quality of the components? Since you've checked out and know your gear, you can assess how much you can push it without - blowing it -

So there are no regulations saying that even though the box can handle +19dBu, that the components will not produce some saturation before that critical point as well - or how much saturation will be produced?

Mosfet wrote:
Quote:
"Headroom" is a dangerous word to use. It is far too often misused! To start, there is what I call the "theoritical headroom", or more specifically, the maximum Peak level a peice of gear can produce. Then there is the "usable headroom", or more specifically, the maximum RMS level the gear can produce and still sound acceptable!


I misused headroom to mean the point between the peak of the signal and the destruction point of the analog box. So theoretical headroom is the amount between the peak of the signal and the clipping point of the analog gear - which won't "clip" and distort right away (I was thinking that the clip point was the point of explosion, when it is more like the point of saturation? ...) - so you have to push and push above this clipping point for it to blow ....? (is this soft clipping?)

Maybe for what I am trying to describe, which is more theoretical than anything, 'headroom' would be the amount of clean signal you can process before saturation begins ... the higher the clip point, the higher you can drive it without saturation. If a peice of gear can withstand saturation at higher dB levels, would the end result have more RMS power?

One more .... is it safe to assume that you can saturate a signal more with good results using analog gear in digital audio since the digital medium won't saturate the signal more (as opposed to analog tape) ....
Back to top
View user's profile Visit poster's website
skod Offline
Giant Member





PostPosted: Mon Feb 27, 2006 5:41 pm    Post subject: Reply with quote

RedStone wrote:
OK ... making sense .... so assuming you can tweak the system to pump out, digitally, whatever is read on the digital meter - using gain staging?


Nope- there's the problem. 0dBfs for one system may represent a completely different _voltage_ level than another system, but is is always the same _bits_. Let's say I have a cheapie soundcard input that is set up for a reference level of -10dBV, and I crank my test tone oscillator up until it hits 0dBfs on my digital metering- the sample values are entirely saturated (arithmetically speaking), and can't go even one bit further. I measure the voltage it takes to do this, and it comes out to be +2dBV (1.25 VRMS, or +4.22 dBu). Then, I walk over to my RME converters with inputs that are set for a reference level of +4dBu, and crank up the oscillator until I hit 0dBfs there: and measure the voltage. And I get 3.46 VRMS, +13dBu by their specs)- pretty much double the voltage.

But the _bits_ being output by the converters are exactly the same: 0dBfs results in the same string of bit values (in the case of 16-bit two's complement numbers, the max positive value is 0111111111111111, and the max negative value is 10000000000000000), completely regardless of the fact that one started at a different voltage level than the other. So the bottom line is that *the bits don't care*: it is the analog electronics that come before the A/D, or after the D/A, that do.

RedStone wrote:
(I assume that this is the art of using attenuators?) to make it so that If it says OdBFS, that is what's leaving. (using whatever equipment people use to test that sort of thing .. I don't know - the test tone and oscillator?)


Nope. Bits is bits. You may need to use attenuators (or makeup gain) if the levels don't match- but the conversion process moves you into a different domain, the digital domain. Record something with the cheapie soundcard at 1.25VRMS, and then play it back through the RMEs and it comes out at 3.46VRMS. There's no gain in there, in the analog sense- this happens because the two converters disagree as to what the reference level is.

RedStone wrote:
So you send an actual full scale 0dB to an analog box that can handle a signal "cleanly" up to around +19dBu - if your full scale signal is a true 0dBFS, then your gear will start to saturate the signal at that level - how much exactly depends on the quality of the components? Since you've checked out and know your gear, you can assess how much you can push it without - blowing it -


Nope. Bits is bits. Your full scale 0dBfs could get swallowed without a problem by some hardware, and not by others. Let's say we're using the cheapie soundcard to play back, and we get 1.25VRMS out from our 0dBfs digital signal. And we run it into, let's say, a real old Pultec analog EQ. And we look at a hypothetical meter hooked up deep into the guts on the Pultec (it's gotta be hypothetical, because real Pultecs don't have meters....), and it tells us that the very most we can get out of those converters is -18dBm (or thereabouts)- _nowhere close_ to its limits, and frankly well out of the sweet spot for getting the best signal to noise. Why? The cheapie soundcard can't drive it: it can't swing enough voltage (and, realistically, can't put out enough _power_) to properly drive that old-style, 600-ohm terminated transformer load.

Let's take the other example- we run the same stream of bits into the RME converter, and then we let it drive the input on my cheapo Teac cassette deck. And the RME is humping out a nice healthy 3.46 VRMS, which is about 12dB hotter than the max level the cassette can handle- so it is clipping like _mad_, since we are seriously overdriving it. We'd have to put an attenuator in there to keep from crunching the input.

RedStone wrote:
So there are no regulations saying that even though the box can handle +19dBu, that the components will not produce some saturation before that critical point as well - or how much saturation will be produced?


Saturation is maybe the wrong word here. Arithmetic saturation (in the digital sense) is when the bits have reached that min or max value I wrote above. Once you get there, you can't go any further, so whatever value gets produced is just flat _wrong_: digital conversion just flat can't go to 11. My whole point is that there is no general-case correlation between a specific voltage level and the arithmetic saturation points for a d/a or a/d conversion process. Each piece of gear has an operating level that it will work the best at, and whatever level that takes, that's where you run it. You can't drive a Pultec with a transistor radio (in anything resembling an optimal way) without supplying some makeup gain from somewhere, and you had best not drive a low-voltage-reference unit with a high-voltage-reference source, unless you have an attenuator handy! But none of that analog behavior has anything material to do with the bits themselves.

Saturation has another completely unrelated meaning in the analog world, having to do with the behavior of magnetic devices (tape and inductors like transformers) when they are overdriven. This analog saturation can produce a very gentle and downright pleasant nonlinearity- we like the way transformer-based preamps and power amps sound when overdriven. So let's stay away from using "saturation" when discussing digital processing: use "digital distortion", or "two's complement error" instead. There's nothing whatsoever pleasing about getting into arithmetic saturation in a digital system.

RedStone wrote:
I misused headroom to mean the point between the peak of the signal and the destruction point of the analog box. So theoretical headroom is the amount between the peak of the signal and the clipping point of the analog gear - which won't "clip" and distort right away (I was thinking that the clip point was the point of explosion, when it is more like the point of saturation? ...) - so you have to push and push above this clipping point for it to blow ....? (is this soft clipping?)


No, it's much more subtle than that. For digital, passing the point of arithmetic saturation (trying to go above 0dBfs) is the point at which whatever is going through the signal chain certifiably sounds like hammered dogmeat in a puddle of stale piss: there's very little chance of missing it. Analog electronics can be run well into clipping (let's generically call this "nonlinearity"), and sound _great_ while doing it: as a matter of fact, the worst sound in the world is an electric guitar that has _not_ experienced some of theis wonderful nonlinearity when being made... (;-). Destruction very seldom enters into it!

However, headroom _does_ indicate something useful, if you are trying to optimize your signal path for maximum quality and minimum noise. You want to run at the highest possible level _for each stage of the chain_ before you begin to experience distortion from overdrive, because that means that you are using all the clean level you can get above the noise floor. If you are running some chunk of gear with the peaks 40dB below the max it can cleanly handle, that means you have boosted the contribution that that unit is making to your overall noise floor by approximately 30-something dB (handwave, handwave).

"Soft clipping", on the otherhand, is usually a euphemism for some digital processing that is done when the gear realizes that you are going to crash into the hard limits of arithmetic saturation, and tries to save your signal from sounding like hammered dogmeat in a puddle of stale piss. Instead, it usually just omits the piss. (;-)

RedStone wrote:
Maybe for what I am trying to describe, which is more theoretical than anything, 'headroom' would be the amount of clean signal you can process before saturation begins ... the higher the clip point, the higher you can drive it without saturation. If a peice of gear can withstand saturation at higher dB levels, would the end result have more RMS power?


Well, sort of: I think I can agree to that point. Headroom is the difference between where you ideally want to run and where you have to be careful not to run. The definition is "the difference between nominal and overload". You don't want run any given block of gear at too low a level (level you could have tolerated without incurring noticeable/objectionable nonlinearity), because if you do, you will have brought your noise floor up higher than it had to be- but by the same token, if you run it too hot, you will experience nonlinearity on the peaks.

So how hot is too hot? The ears tell the tale, and every chunk of gear has its own characteristics. There is no general answer: you simply gotta play with it.

RedStone wrote:
One more .... is it safe to assume that you can saturate a signal more with good results using analog gear in digital audio since the digital medium won't saturate the signal more (as opposed to analog tape) ....


In the digital domain, 0dBfs isn't just a good idea: it's the law. (;-). In the analog domain, there are very few laws- just your ears. So a good general rule is to leave some additional headroom in the digital side. If you are recording at 24 bits, there is no reason whatsoever to push your peaks on even accurate digital metering up above maybe -6-8dBfs. I personally track at about -12dBfs these days, mostly. But in the analog domain, you have to figure out the behavior of each piece of gear- and be ready to _slam_ that good ol' Pultec when you are tracking guitar... (;-)

Analog tape is the best example of nonlinearity being used in a pleasing manner. Pushing enormous levels onto tape is now all the rage. The signal you get back usually doesn't look much like what you put in, on a scope, but it sure does sound interesting...

_________________
Scott Griffith, Scratchpad Studio/Earfull Sound
the Nerd formerly known as Skippy
Back to top
View user's profile
Mosfet Offline






PostPosted: Tue Feb 28, 2006 12:08 am    Post subject: Reply with quote

skod has explained it all pretty good.

I have to add this. "I" define "headroom" as the difference between my "average" level I am working at, and the point that I can hear audible distortion (in analog) or 0dbfs in digital. I increase this amount of headroom if there are things happening in the audio that cause peak levels to hit the clipping stage, and decrease the headroom if the audio levels are very stable.

For example, I would probably allow a LOT of headroom on an inexperienced drummer! Why? Simply because his kick drum hits can be as much as 18 db different in volume from one hit to the next. Hell, he might have just ONE hit in the whole freakin' song that is 24 dB louder than his softest hit even!!! Grrrrrrrr....(bad drummers!)

On the other hand, a distorted electric guitar is seldomly going to need 24dB of headroom! Hell, 6dB is generally more than enough! A distorted electric guitar generally will not all of a sudden JUMP UP in volume. It is a fairly consistent signal.

So, headroom is all a matter of allow enough difference between the average level and the level where the audio JUST starts to clip audibly.

In digital, like skod says, it is a LAW! Trying exceed 0dbfs is going to be ugly! But, modern matering techniques allow for a certain amount of samples in a row to clip and you will not hear the distortion. That number of consecutive samples in the clipping stage that will produce audible distortion seems to be debated right now. Some say 8 samples in a row, others say 12 (assuming 44.1 sample rate).

So, it is possible to even clip in digital and get away with it possibly!

But like skod said earlier, in analog, you can drive the circuit straight to hell with clip lights showing and still get some fantastic sound, and increase overall level! Think of the clipping as "limiting". Some analog gear does it MUCH better than other. I have throw eq's, compressors, mixer busses into some serious clipping and had great sounding audio come out the other side. KILLING analog tape with HUGE signal is a well established production approach in big time studios.

So, in the end, don't get too caught up in "headroom". Yes, you should pay attention to it, and as skod explained, it is good to optimize your gain stages to lower the noise floor as much as possible. The more musically dynamic the production, them more reason you need to optimize gain stages and allow enough headroom to achieve the dynamics you want. Punk rock on the other hand probably doesn't need that much attention to headroom! Smile

_________________
My wang is porn stah big in metric numbers!
Back to top
View user's profile
skod Offline
Giant Member





PostPosted: Tue Feb 28, 2006 9:06 am    Post subject: Reply with quote

Mosfet wrote:
In digital, like skod says, it is a LAW! Trying exceed 0dbfs is going to be ugly! But, modern matering techniques allow for a certain amount of samples in a row to clip and you will not hear the distortion. That number of consecutive samples in the clipping stage that will produce audible distortion seems to be debated right now. Some say 8 samples in a row, others say 12 (assuming 44.1 sample rate).

So, it is possible to even clip in digital and get away with it possibly!


Yes, and that is a source of some consternation in the mastering world. For the production side of things, setting an absolute limit of 3 or 4 consecutive samples is a very, very good idea IMNSHO, even for rock. I personally prefer none, but then I don't generally record rock any more. Then, your client sends your mix off to the mastering lab, and in order to sound as loud as everybody else does, they squeeze and limit and crank and mung and squeeze and squeeze and _wreck_ your mix until it has strings of 20 or 30 fullscale samples all over the place. And, on *their* equipment, it may actually sound pretty good.

Problem is: sticking out long strings of fullscale samples will create different results in different equipment, in these days of DSP reconstruction filters in even the cheapest consumer gear. And it will sound _radically_ different on different units as a result- and this is not a golden ears phenomenon. It can be about as subtle as a flung brick. Rush "Vapor Trails" is the first commercial release I ever paid for and then immediately retired (except for doing demos to illustrate the problem) because of that phenomenon... On one of my players, all that digital clipping rings like a bell in the DSP reconstruction filter, and just puts a nasty raspy high-end schmutz all over everything. Downright unlistenable. But then my cheapie little JVC desktop stereo here in the office handles it just fine, and it just sounds badish (instead of sounding I-wanna-slit-my-wrists bad, like it does on the good player).

Whoever mastered that was dancing with the devil, and did not do the band any justice. Anyway, I choose not to participate in the loudness wars: if I get even one "over", I'd prefer to go back and retrack it. Your mileage may vary, but unless you have a bunch of really deaf speedmetal kids for clients in your studio, do something unique and record whatever they play as cleanly as you can. There's always somebody further down the production pipeline who will be willing to squash the living, breathing bejeezus out of it for them, and you can avoid that creepy unclean back-of-the-neck feeling. (;-) The only way to win the loudness wars game is not to play!

_________________
Scott Griffith, Scratchpad Studio/Earfull Sound
the Nerd formerly known as Skippy
Back to top
View user's profile
RedStone Offline
Giant Member


Joined: 27 Nov 2004
Posts: 1431
Location: British Columbia, Canada


PostPosted: Tue Feb 28, 2006 10:58 am    Post subject: Reply with quote

skod
Quote:

There is no general-case correlation between a specific voltage level and the arithmetic saturation points for a d/a or a/d conversion process. Each piece of gear has an operating level that it will work the best at, and whatever level that takes, that's where you run it.


With reference to your RME converter (a key word that I overlooked) – when it says 0dBFS @ +19dBu – does that mean on that specific setting, a digital peak of 0dBFS will translate into a analog signal of +19dBu at a pre-defined voltage? When you consider the RMS power of the music and the design of the converter, can you actually drive the analog components to a higher dB level? And is output voltage constant? (Within the confines of a single converter, or a single piece of fully analog gear)

sorry about the questions ... but damn, you know your stuff

Quote:
Saturation has another completely unrelated meaning in the analog world


ahhh ... so we were talking about two different things at the same time (my fault since I didn't specify) - I was talking about the saturation point of fully analog gear – I have heard tube saturation (mostly on my old Traynor tube amp) and extreme tape saturation on some old tapes of my dad’s band (recorded too hot on 2-inch) ... so when I talked about an analog unit’s saturation point, I meant the point where it starts to saturate a signal with its own characteristics of tube, or transistor saturation goodness (I still think that this would be an acceptable way to measure actual headroom since if you know where it will start to saturate, you know how much ‘room’ you will have in reality, or in theory before saturation in the analog domain. For example, if you don't want any saturation on a few of the boxes in an all-analog chain, either to avoid over saturation, or to control the tonal qualities of saturation within the chain ...

But, would you look for analog saturation when purchasing AD/DA converters – or would you be looking more for voltage ratings and a good chunk of theoretical headroom so that you get a nice clean conversion? - and then make sure the voltage of the rest of your analog gear matches up, or else use (attenuators – I obviously didn’t know what they were) to make them match?

I had no idea about all of the goings-on of the AD converter (VERY useful!)

We at least agree that “digital saturation” doesn't really serve any appropriate use as a term - and limiting can be completely self defeating to music (who wants a dynamic range of 3dB or less – not me!). I was listening to Jimmy Eat world the other day and had to turn it right off ... I found it was fatiguing my ears, even at lower volumes...

Mosfet: Speaking of limiting, to limit 3 or 4 consecutive samples at 44100hz, how could that translate into a release time for the limiter, and how much difference would that be for 48khz or 96khz? I guess I am wondering about when peaks of a digital signal cross the threshold, how can you define what "3 or 4" samples will be using your ears?

I try to rely on my ears when limiting as much as possible (but I also watch the gain reduction to make sure I'm not plowing the mix) - I usually aim for a few dB - but I mostly just test to find a good non-plowed level that is still transparent to my ears - which, at this point, might still be too "digitally saturated" ha ha.
Back to top
View user's profile Visit poster's website
Mosfet Offline






PostPosted: Tue Feb 28, 2006 12:03 pm    Post subject: Reply with quote

I personally don't push anything I master that hard. I ALWAYS set my digital limiter to .1dB below full scale. While on normal CD players, you can't really hear the distortion artifacts on many modern recordings, I can hear them easily on my DAW and studio monitors. It is a place I just don't want to go! I don't see ANY need to EVER master something higher than around -10 RMS. That is more than double the RMS that was happening about 10 years ago!
_________________
My wang is porn stah big in metric numbers!
Back to top
View user's profile
RedStone Offline
Giant Member


Joined: 27 Nov 2004
Posts: 1431
Location: British Columbia, Canada


PostPosted: Tue Feb 28, 2006 1:07 pm    Post subject: Reply with quote

good point -- I leave it at 0.2dB below 0 -

10 RMS eh ... I'll have a look see at my mixes and see where I'm at -
wait a second ... I don't have any RMS levels in my programs ... (I don't think)
Acid, Vegas, soundforge (I think SF has VU simulated something or other ... blah)
maybe I'm using the wrong software

Or can you judge RMS by looking at the thicker part of a waveform below the highest peaks and match that against a dB scale to the left or right ... (it's like that in SF anyway)
Back to top
View user's profile Visit poster's website
skod Offline
Giant Member





PostPosted: Tue Feb 28, 2006 9:26 pm    Post subject: Reply with quote

RedStone wrote:
With reference to your RME converter (a key word that I overlooked) – when it says 0dBFS @ +19dBu – does that mean on that specific setting, a digital peak of 0dBFS will translate into a analog signal of +19dBu at a pre-defined voltage? When you consider the RMS power of the music and the design of the converter, can you actually drive the analog components to a higher dB level? And is output voltage constant? (Within the confines of a single converter, or a single piece of fully analog gear)


Yup. 0dBfs at the D/A will produce a +19dBu output signal (6.9VRMS, +16.78dBV...) at the analog outputs. And that's *all* it will produce, today, tomorrow, or whenever. So it is not as if it will produce +19dBu now, and +24dBu later. The reference level set by the selector switch directly governs the maximum analog output level you can get- and, by association, directly governs how much input level can be tolerated before the converters experience arithmetic saturation and go crunch. You want +19dBu, you got it. Period.

Now, keep in mind that the actual analog support electronics inside the RME converter box can go a lot further than that quite cleanly- but they _won't_, because they don't have to. The scaling factor is fixed by making that selection. Their job is to approximate perfection as closely as possible.

RedStone wrote:
sorry about the questions ... but damn, you know your stuff


Thanks for the kind words! I guess that I've just been a nerd-without-portfolio for a long time. And it doesn't hurt that at various times I've made a living designing audio and DSP gear. There's nothing like having to live with this crud full-time to give you a wierd view of it...

RedStone wrote:
ahhh ... so we were talking about two different things at the same time (my fault since I didn't specify) - I was talking about the saturation point of fully analog gear – I have heard tube saturation (mostly on my old Traynor tube amp) and extreme tape saturation on some old tapes of my dad’s band (recorded too hot on 2-inch) ... so when I talked about an analog unit’s saturation point, I meant the point where it starts to saturate a signal with its own characteristics of tube, or transistor saturation goodness (I still think that this would be an acceptable way to measure actual headroom since if you know where it will start to saturate, you know how much ‘room’ you will have in reality, or in theory before saturation in the analog domain. For example, if you don't want any saturation on a few of the boxes in an all-analog chain, either to avoid over saturation, or to control the tonal qualities of saturation within the chain ...


Yup. Analog gear starts to color the signal at some point with increasing drive, and it often sounds good. It's in meshing the world of digital (which is very black-and white, and not tremendously flexible) with the nice, happy, flexible, no-rules world of analog where we get problems. It's made even worse by the fact that the nomenclature is screwed up all to hell: digital arithmetic saturation (a strictly numerical phenomenon), inductor magnetic saturation, tape magnetic saturation, tube transconductance saturation, and transistor minority-carrier saturation are all totally different things that use the same _word_ in different contexts. What a freakin' pisser.

In any case, analog nonlinearity (coloration) usually sets in in a very subtle way. Small increases of signal above some arbitrary "threshold" will begin to get increasing levels of nonlinearity, and a little bit more can usually sound even better. Of course, you can go too far, and at some point even the best gear goes from sounding "hey, that's neat" to "turn it down before the smoke comes out". But these analog effects can be very subtle, and can be used in a very creative way- like your tube or tape saturation. On the other hand, there's *nothing* subtle about digital nonlinearity: when you get there _at all_, it sounds like shit.

RedStone wrote:
But, would you look for analog saturation when purchasing AD/DA converters – or would you be looking more for voltage ratings and a good chunk of theoretical headroom so that you get a nice clean conversion? - and then make sure the voltage of the rest of your analog gear matches up, or else use (attenuators – I obviously didn’t know what they were) to make them match?


Exactly the latter. In the case of converters, you are looking for perfection, not coloration. The only job of the converters (and the analog electronics that surround the actual converter hardware, inside the same box: the input and output buffer amps and whatnot) is to be perfect at all times. _Then_, you mung with the signal in the analog domain either before it gets there or after it comes out. So your converter box puts out +19dBu at 0dBfs, like my RME, and the next block in the chain starts to crunch unpleasantly when it sees more than +16dBu on its inputs? Then put a 3dB attenuator between them (at a theoretical minimum: I'd probably go with at least a 9dB attenuator to stay well away from crunch, but only experience with the gear will tell).

I have an old buddy who has a big box of nasty old matching transformers that he puts between the output of his Apogee converters, and before anything else, just so he can get that sound of actual magnetic saturation (damn- that word again!) and hysteresis from the iron cores in the transformers: a lesser and more subtle form of the effect you get from the iron in tape heads, and the saturation (Not again!) of the magnetic tape as it crosses the head gap. He lets the Apogees hit those transformers as hard as they can, and then he pads down (attenuates) the output if necessary to whatever level he needs to drive the next block. In the analog domain, there are more tricks and hacks than you can shake a stick at: but in the digital domain, 0dBfs is just the law, and there's no more fun to be had there...

RedStone wrote:
We at least agree that “digital saturation” doesn't really serve any appropriate use as a term - and limiting can be completely self defeating to music (who wants a dynamic range of 3dB or less – not me!). I was listening to Jimmy Eat world the other day and had to turn it right off ... I found it was fatiguing my ears, even at lower volumes...


Yup. The arithmetic saturation of the conversion process is just not useful to us as a creative tool- all we can do is avoid it like the plague.

RedStone wrote:
Mosfet: Speaking of limiting, to limit 3 or 4 consecutive samples at 44100hz, how could that translate into a release time for the limiter, and how much difference would that be for 48khz or 96khz? I guess I am wondering about when peaks of a digital signal cross the threshold, how can you define what "3 or 4" samples will be using your ears?

I try to rely on my ears when limiting as much as possible (but I also watch the gain reduction to make sure I'm not plowing the mix) - I usually aim for a few dB - but I mostly just test to find a good non-plowed level that is still transparent to my ears - which, at this point, might still be too "digitally saturated" ha ha.


Well, probably not. There are software packages like the Waves limiters that allow you to digitally squeeze the life out of the material without actually slamming quite so hard into running out of numbers. Software like that makes it possible to really mash it without quite so much hard clipping as numeric saturation implies- but I still don't like the way they sound...

There are digital metering setups that allow you to look for consecutive strings of fullscale samples. But the bottom line still is "if it sounds good, it is good." That much hasn't changed with the advent of digital. However, the more you squash it, the more imperative it is to listen to it on many pieces of gear. The cheaper the better, generally speaking. And if you really are smashing it right up to the numerical limits, some pieces of consumer gear are going to reward you with the sound of hammered dogmeat floating in stale piss. And I think everybody can agree that that should be avoided...

_________________
Scott Griffith, Scratchpad Studio/Earfull Sound
the Nerd formerly known as Skippy
Back to top
View user's profile
RedStone Offline
Giant Member


Joined: 27 Nov 2004
Posts: 1431
Location: British Columbia, Canada


PostPosted: Wed Mar 01, 2006 3:24 pm    Post subject: Reply with quote

*breaths*
Glad I'm somewhat clearer on the issue ... there is a veil of mud still, but I'll figure it out eventually hehe

smashing avoided ... definatley ...
I even stopped using compression for a while because I was over using it - I'm slowly weening myself back on to it
Back to top
View user's profile Visit poster's website
NEGTEN84 Offline
Newbie


Joined: 04 Jun 2014
Posts: 1


PostPosted: Wed Jun 04, 2014 1:32 am    Post subject: Reply with quote

I am new....and need help: I have a presonus FIREPOD and below are the specs...

Preamp
Bandwidth.................................................10Hz to 50kHz Preamp Input
Impedance................................................1.3k Ohms
Instrument Input Impedance .......................1M Ohms
Preamp THD ............................................. <0.005%
Preamp EIN ............................................. -125dB
Preamp Gain ..............................................54dB
Preamp Send Output Impedance ..................51 Ohms
Preamp Return Input Impedance.................. 10K Ohms
Line Trim................................................... +/-20dB
Line Input Impedance..................................10k Ohms
TRS Output Impedance................................51 Ohms
TRS Main Outputs Impedance...................... 51 Ohms
TRS Cue Outputs Impedance ...................... 51 Ohms
Headphone Output.......................................150mW/Ch 20Hz-20kHz Phantom Power ..........................................48V +/- 2V
Power Supply ............................................ Ext line Transformer, Internal Switching
Analog to Digital Converters ....................... 24-bit / up to 96khz
ADC Dynamic Range .................................. 107db
DAC ..........................................................24-bit / up to 96kHz
DAC Dynamic Range ..................................110db IEEE1394
Speed ....................................................... 400mbps


I have been reading this post... so in context what out of these specs should I care about? I appreciate any and all wisdom you can impart!
Back to top
View user's profile
Display posts from previous:    View previous topic :: View next topic 
Quick Reply      
Use this form to post a quick reply without having to go through the normal reply process.    
     
Post new topic   Reply to topic    The Recording Project Forum Index -> Signal Processing All times are GMT - 7 Hours
Page 1 of 1
Jump to:  


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You cannot attach files in this forum
You can download files in this forum