Michael Nielsen wrote on Mon, 08 January 2007 19:59 |
So what happened that now it's better to "go easy" on the DAW? |
PaulyD wrote on Mon, 08 January 2007 22:51 |
A biggie though is 24-bit recording. Each bit in digital recording gives 6 dBFS of dynamic range, so a 24-bit system has 144 dBFS of internal dynamic range. The very best ADC's and DAC's have "only" about 120 dBFS dynamic range in their analog filters. Even accounting for the noise floor, that is more dynamic range than nearly anything most of us will ever record. So...the idea is to give up a digital bit or two to give the digital system some headroom. Like Paul Frindle says in the thread maxim referenced, this will give your plug-ins the headroom they need to do their math and the DAC the headroom it needs to reconstruct the final output samples without anything getting digitally flattened. It really works! Paul |
Michael Nielsen wrote on Tue, 09 January 2007 00:36 |
WOW. Thanks guys I'm gonna go read that other thread. |
compasspnt wrote on Wed, 10 January 2007 06:03 |
I like to call it "Zoom." |
J.J. Blair wrote on Wed, 10 January 2007 22:22 |
This is where the idea of recording at low levels comes in, because then the bus doesn't run out of headroom and leave you with that nasty digital distortion. Some of us have done experiments and found that if you do the same mix at near peak levels, and then do a -15 dB trim across the whole mix, the result of the latter is much better sounding. |
J.J. Blair wrote on Wed, 10 January 2007 23:05 |
Yeah, there's a huge diffrence, because the weak link is in the bus. You have to bring down the levels going TO the bus, by bringing down each track. |
compasspnt wrote on Thu, 11 January 2007 14:51 |
For the thousand and second time, FORGET that stuff about "using all of the bits." ESPECIALLY in 24 bit sessions. Record each track at reasonably lowered levels (let's say...-12 to -20 peaks). Then mix at a reasonably lowered level (let's say -6 to -10 peaks). As JJ just said, this will avoid overloading the digital mix buss, one of the major concerns in "digititis." This will allow your plug-ins more room to work properly, and avoid most overload therein (which is where much other terribleness can occur). As for the mic pre's, gain stage them properly to begin with. Must we point out again that no less than George Massenberg would often record even in ANALOGUE this way? Sometimes he would have meters on a tape machine peaking at -5 to -10 ON DRUMS. As long as you stay away from noise (and in digital, there is no real noise proplem at any level), you will be better off. Go forth and multiply. |
CHANCE wrote on Thu, 11 January 2007 16:07 |
You lost me. How would you lose dynamic range? Recording at lower levels, the relative differences between 0db and your highest level will still be the same yes? |
Kendrix wrote on Thu, 11 January 2007 17:28 | ||
I think not.- You are effectively reducing the "highest level" by recording at a lower volume. So, dynamic range is reduced. If you A to D convert into a 24 bits and you turn the input or trim down so that you record say 12db below where you'd be if you peaked at zero then you are reducing the dynamic range of the signal being converted by 12db. In doing so you are using 2 fewer bits than are available. The point is that the quality of the subsequently reconstructed waveform (after the D to A) does not suffer when you do this. Since virtually any real sound source hs much less than 144db dynamic range you really dont lose anything- however you gain the benefit of digital headroom when summing. |
compasspnt wrote on Thu, 11 January 2007 09:39 |
And surely no one in today's climate is concerned with KEEPING maximum dynamic range anyway! Seriously, if, instead of 144 dB DR, you have ONLY 132, yet at the same time you are virtually eliminating digital distortion, whilst reducing the "artifacts" commonly associated with digital recording, I think you're still doing "real good." |
Tomas Danko wrote on Thu, 11 January 2007 20:00 | ||||
Not really. The sound you record will live roughly somewhere below the incoming noise floor (ie background noise, microphone, preamp and so forth) up to the maximum peak of the sound in question. It will most definitely be able to live between 0 dBFS and, say, -144 dBu. Now, if you set your A-D converter to only peak at -12 dBu you will still be recording the entire dynamical span of that sound set lower within that 24 bit file. So practically speaking, if something had, say 80 dB of total dynamic range then you can decide to record it into 0 dBFS and downwards or you could record it into -12 dBu and downwards. Align the recordings by shifting the mantissa up or down (ie within a 32 bit floating point register) and you'll see it's all there regardless of how hot you printed it. The difference is that if you slammed the converter there are some penalties to be had regarding the analog front end of the converter, the decimation process in it. And even more so, downstream when you apply plug-ins and a mixing engine. That just about sums it up. Regards, Tomas Danko |
Kendrix wrote on Thu, 11 January 2007 20:29 |
I understand the distinction you are making between available dynamic range of the medium and the actual dyunamic range of the signal you place on that medium. I agree. However, if you turn the amplitude of an incoming signal down to zero you have zero dynamic range. If you fully modulate a signal having 100 db of dynamic range then you've got 100db. Modulate to 50% and you have reduced the range accordingly. So, doesnt it follow that by reducing the amplitude of the incoming signal so that it to sits at -12 verus zero you have reduced its dynamic range? That was my point. |
Tomas Danko wrote on Fri, 12 January 2007 00:12 | ||
The cool thing is that within a 24 bit system you can slide that 100 dBu-range recording up and down... say... 44 dBu down from the dreadful zero and still have full dynamics. Practically speaking that's not the most common thing to record, and also practically speaking those 44 dBu will be diminished somewhat due to the inherent s/n ratio of the converters. Still, lowering the maximum peak will not reduce the overall dynamics in a 24 bit digital recording system. |
Kendrix wrote on Fri, 12 January 2007 02:23 |
I sure hope this got us out of the "academic" doghouse |
compasspnt wrote on Fri, 12 January 2007 14:09 |
Thanks, rm. I have actually been considering doing just that, but to do it right is a massive job, somewhat akin to the books I am mired in writing right now. I will try to put it together though. |
cerberus wrote on Fri, 12 January 2007 18:18 |
protools may be a popular choice, but perhaps it isn't state of the art for sonics. for example: with other daws, it can be nearly impossible to clip anything internally. jeff dinces |
compasspnt wrote on Fri, 12 January 2007 23:40 |
Haris, read the thread http://recforums.prosoundweb.com/index.php/f/29/6490/ in its entirety. Pay especial attention to the last 9-10 pages. Then if you still have such questions, come on back and ask them. |
Thomas Lester wrote on Fri, 12 January 2007 21:03 | ||
And unfortunately, not having to pay attention to what you are doing has created a slew of so called engineers that have no clue about proper gain staging or how to load a mix bus. Quite frankly, they've dropped the "engineering" out of Audio Engineering. I'm not saying a floating point system is better or worse, I"m just saying that a great deal of incompetence has surfaced due to the modern DAW. |
cerberus wrote on Sat, 13 January 2007 03:12 |
i believe that we are better off making decisons by listening to the musical results of all our actions, not watching for whether a red light is on or off! that is bogus toil; and as devo has said "toil is stupid"! |
Quote: |
i think that lack of imagination is another reason for bad engineering... "by the book" "paint by numbers" .. sure i'll tell you what frequency is "presence".. but who cares if nobody is buying it? yeah, we have problems. |
Quote: |
fact: analog gpes to eleven. if we want to emulate analog-like responses, then our system has to go to eleven, all the way... no bottlenecks, none of that bullsh*t is necessary. |
Quote: |
why keep your signal below -6? you protools users are acting like chicken littles! what is so scary about the uppermost bit of a 24 bit recording? why should the ones and zeroes contained in the "most significant" bit be less valuable to us than the data from any of the other bits? bits=information... music, that is. |
Quote: |
a.f.a.i.a.c...it's all part of the music, once the signal passed the a/d converter, it would be a crime to throw any of it away. |
Buzz wrote on Sat, 13 January 2007 13:47 |
I'm just curious I have tried recording at the prescribed -18/-20bd to test this idea , at least in my DAW ( it uses a 64 bit fixed mix buss ) I can't tell a difference between -6 and -20 ??? , all the punch and clearity is still there to MY EARS !! ( and there are'nt the best out there but not bad either ! ) this is with plugins inserted etc, Anything to a 64 bit fixed mix buss ???? Later Buzz PS: I normally am @ -6/-10db recording levels |
Quote: |
If it's truly a floating point plugin, it will have a gain range of about 1500dB (at 32bit floating). There's about 512dB of headroom above 0dBfs, and about 1024dB going downwards. If the plugin is clipping, then it must not coded properly. I did plenty of tests in the ol' trusty SADiE system, and I couldn't clip any of their plugs. When it comes to fast transients, that's where floating point loses a bit in terms of math, but fixed point wins out a bit. Bob, can you give some examples of floating point plugs that clip? I've never seen this! I'm curious and worried! Cheers, JB |
maxdimario wrote on Thu, 18 January 2007 13:21 |
Since no one answered on the other thread I am posting this quote and would like to know what this is about..(regarding transients) I (and a friend of mine with good ears) have always noticed that drums lose their impact in digital.. In addition I have the feeling that the levels being too high and damaging the overall sound is NOT due to clipping but due to some dynamic process in the algorhythm of the DAW. it may be that lower signals are indeed processed at a different resolution...or using a different process... |
maxdimario wrote on Fri, 19 January 2007 17:55 |
I think it has to do either with the way faders work in daws or the summing algorhythms are DYNAMIC in some way. |
Jim W wrote on Tue, 06 February 2007 23:01 |
Thanks for the reply, I was actually just reading that thread you linked as well. My question is a simple one ( I think? )and was in reference to my signal as I'm recording it, ( ie: synth is being sequenced, now I want to get it into the computer as a peice of audio...) What I'm getting here is that it's actually better to not record at such a high level but I'm missing why simply reducing the channel faders level on a part which is recorded at a high level is any different? |
maxim wrote on Tue, 06 February 2007 21:38 |
does digital have a "sweet spot"? |
seriousfun wrote on Wed, 07 February 2007 19:30 | ||
(you probably know this) The sweet spot was always the zone between distortion and noise. With analog tape you hit distortion gradually as you pushed up the level of the signal hitting the tape, and noise was a fixed floor before that. With digital recording (the point at which the A/D does its math), distortion comes on suddenly, and is usually not useful or desirable. Luckily, noise is way, way, way below what it was with tape (and of a different character). The sweet spot is still there, but huge. This large sweet spot can be used for the purposes of good. For example, track within that sweet spot and shoot for the level where each individual element will sit in the mix. Far less trimming will be needed and (to great benefit, arguably) the math will be easier on the mix bus. |
maxim wrote on Wed, 07 February 2007 00:38 |
does digital have a "sweet spot"? |
garretg wrote on Fri, 09 February 2007 10:06 |
Presto: I think all the pros have tired of the endless debate in this thread... so I'll take a stab. |
presto wrote on Thu, 08 February 2007 16:53 |
^^ no worries mate. i guess we all work with what we have, in my case a poorly designed DAW (a Carillon PC running Cubase and Audiophile 2496) sorry to drag the standard here down |
Quote: |
i do reiterate however, that tracking at lower lower levels has worked for me, |
presto wrote on Fri, 09 February 2007 18:11 |
I thought this would have been ok, as it's really the possible accumulation of distortion of a large number of tracks (eg. a 30 track song with each track recorded at -2dB) that's the concern |
RKrizman wrote on Fri, 09 February 2007 19:47 |
Really, I thought Paul Frindle's whole thing was about intersample peaks. |
cerberus wrote on Wed, 07 February 2007 14:16 |
the main topic of this thread is a non-issue for floating point systems. jeff dinces |
cerberus wrote on Wed, 07 February 2007 20:16 | ||
the main topic of this thread is a non-issue for floating point systems. jeff dinces |
gatino wrote on Sat, 10 February 2007 02:17 |
can you elaborate on this or point me to info on the advantage of having my projects set to 32/64 bit floating point? my daw has the option of working in 32/64 bit floating point. |
Ashermusic wrote on Sat, 10 February 2007 11:41 |
Well that is what I would have thought also but some pretty high priced talent here, like Terry, seem to feel otherwise. |
cerberus wrote on Sun, 11 February 2007 22:44 |
hi carter; it appears to me and some others that conversion is a degenerative process. i think that the real world (hybrid style) is more about saving a buck than about making a better sounding record than <name your favorite album of all time>. |
compasspnt wrote on Mon, 12 February 2007 21:27 |
As much as I like and admire all of our august contributors, in my opinion, there is some misinformation floating about. I suggest recording at lowered levels in ANY digital system, float or fixed. That's what I do, when I use either one, FWIW. And yes, there is a standard the real world uses when the levels come out into a desk. It's ridiculous to attenuate your bass guitar track by 12 dB just so it hits the console properly. Do what you want to do. If red works for you, so be it. If there are consequences, then pay the piper. |
cerberus wrote on Sun, 11 February 2007 20:44 |
i heard some good records lately which i am sure they were recorded and mixed on tape. e.g. the last teenage fanclub album, the recent flaming lips... sigor ros.. mumm... |
compasspnt wrote on Mon, 12 February 2007 15:27 |
Do what you want to do. If red works for you, so be it. |
Andy Peters wrote on Mon, 12 February 2007 17:34 |
[At War With The Mystics is all computer, all the time. |
cerberus wrote on Mon, 12 February 2007 15:24 |
hi carter; if i were looking for increased distortion from analog, i would set the peak level to -0.2dbfs in a floating point environment and then dither to 24 bits with mbit+. then i would adjust the input gain to the analog device with a digitally regulated analog vca fader, not a potentiometer, unless i wanted its color. if i were seeking minimal distortion from analog i would do the same thing, but the gain change in the floating point environment would be for under 0vu before bit depth reduction, the amount of gain reduction would be an ear decision so perhaps it would be convenient to use an analog vca there too, although i would prefer not. the sample rate going out and back in would be 88.2khz, never 44.1khz. i might choose to record back to digital at 192khz if the option were available, and then convert the file back to 88.2khz offline. jeff dinces |
cerberus wrote on Sat, 10 February 2007 11:11 |
i trust that when terry is able to apply scientific method to prove his fascinating but questionable theory, he will. |
Ashermusic wrote on Tue, 13 February 2007 11:46 |
I am a composer who has learned to become his own engineer when necessary due to shrinking budgets but by no means am I a real engineer. I teach Logic Pro 7 at UCLA extension and I am a Logic Certified Trainer. I have a friend, another Logic Certified Trainer whop also has a degree in compure science,, who tells me he never uses pre-fader metering formixing because in a 32 bit float app like Logic the only thing that matters is that the Output is not clipped. He says that he did a number of tests with the chnnaels peaking and noitpeaking with the output not clipping and they they null. The engineer I regiularly use told me that I still would be better off observing traditional mix practices of keeping my levels under contriol because with third party plug-ins, D/A converter factors, etc. that you still get a better mix floating or not and it seems to me to be so as sometimes I was hearing doistortion even when the output was not clipped. So this is what i have been advising my students. If I am reading this thread correctly Jeff seems to be in agreement with my friend the other trainer while Terry et al seem to agree with my engineer. Nika's papers are frankly a little over my head. So bottom line, am I advising my students correctly? |
Thomas Lester wrote on Tue, 13 February 2007 17:41 | ||
Yes... you are teaching them correctly. And your fellow teaching bud is full of crap. You can't take the same signal, clip one and not clip the other and have it null. He completely made that up. If nothing else, the amplitude difference from raising the gain until it clipped will be off enough to keep it from null'ing. |
J.J. Blair wrote on Tue, 13 February 2007 17:48 |
Asher, I find the question to be more about the woeful headroom of the DAW's sereo bus. Null or not, that's a good reason to teach the practice. |
Ashermusic wrote on Tue, 13 February 2007 09:58 | ||
Hi J.J. So in your opinion, all Daws, fixed or float, have stereo busses that have lousy headroom? |
PaulyD wrote on Mon, 12 February 2007 21:06 |
If you can explain yourself in technical and mathematical terms |
PaulyD wrote on Mon, 12 February 2007 21:06 |
Why should Terry have to rehash that which has already been brilliantly spelled out by Paul Frindle and Nika Aldrich? Ever read Nika's book or any of the white papers he has posted on his web site? |
CWHumphrey wrote on Tue, 13 February 2007 17:59 |
My question is what is the next guy down the line saying to you about your work? |
cerberus wrote on Tue, 13 February 2007 23:57 |
in sound recording: it is better to record the full signal than to record only part of the signal. |
cerberus wrote on Wed, 14 February 2007 13:25 |
all of the extant 32 and 64 float daws have red lights only at the master fader. i hope that this forum can at least reconcile with that fact. jeff dinces |
MoreSpaceEcho wrote on Wed, 14 February 2007 09:56 |
hasn't it already been well established by nika, et al. that as long as you are recording at a level that allows for the entire dynamic range of the signal in question to be captured, you are in fact recording the "full" signal? |
cerberus wrote on Wed, 14 February 2007 13:25 |
tomas, i am sorry that i insulted your daw. daw is like religion to some people, so i think perhaps it is a mistake to diss them in public, i'm sorry to everyone. but i hope we rise above and feel the love... i feel this in the talented people i work with... and it seeps into their music. that is enough for me. i could overlook that they ruined their shit and now i've got to clean the mess up. 'coz i get paid. in such a case, i don't care what brand name daw is used on music i must hear for work, or even that i choose for entertainment, when i am paying... as long as it hasn't dicked with the sound. but. i said what i really hear already: in so many cases, i feel that the music got ripped off...i did the job anyway. but it builds up for me... i've spent thousands in good faith on daws that claimed sonic integrity; then discovered that they sounded like shite. every once and a while my pain comes out during technical discussions involving digital audio; and i am sorry if that burdens any of you. all of the extant 32 and 64 float daws have red lights only at the master fader. i hope that this forum can at least reconcile with that fact. jeff dinces |
Andy Peters wrote on Mon, 12 February 2007 16:34 | ||
At War With The Mystics is all computer, all the time. -a |
cerberus wrote on Tue, 13 February 2007 15:57 |
in sound recording: it is better to record the full signal than to record only part of the signal. in sound recording: whether you use the full signal or not: it is correct to archive the full signal. |
cerberus wrote on Tue, 13 February 2007 15:57 |
paul frindle's comments on reconstruction only apply where there is a converter in the chain. not for processing that takes place inside the daw. |
Nika Aldrich wrote on Sat, 02 October 2004 15:22 | ||||||||
I think that this is not true with respect to the analog world, and it is certainly not true with respect to the digital world. In analog, running in "as hot as possible" means running into saturation that may or may not be desirable. Certainly when doing classical recording of purist material this is generally avoided. When recording kick drums this is usually desired and "over" done, if we can put a valuation on it. In the world of digital we need to record in also at an optimal level. In the old days the digital converters were shoddy enough in the low end and had so little dynamic range that we wanted to take advantage of all of the dynamic range available by keeping the input levels as hot as possible. Things have improved, however, in the past 15 years, so we are no longer confronted with the problem of having prohibitively low dynamic range capabilities at the converters. We now effectively have all of the dynamic range that we can practically use, so we get to look at some other details that can affect the sound quality. For example, your analog outboard equipment (say, your preamps) were not meant to run as hot as digital's "as hot as possible" and when you push that equipment to those limits you are invariably getting a more distorted (and often undesirable) sound. Since we have more than enough dynamic range in the converters now it would be most advantageous to optimize the signals for the benefit of the analog outboard gear, as that is now the inhibiting factor of your sound quality. This is not meant to be a rule of thumb, for there are situations where this does not apply (snare drums, for example) but it is meant to address your question about the obsolescence of the "record as hot as possible" approach. There was indeed a time when this was valid - particularly with R-2R/SAR converters and 16 bit recorders. With today's converters and with 24 bits of recordable data the paradigm has shifted.
Well let's think about this. Turn your monitoring system up pretty loud - as loud as you listen when you really like the music. OK, how loud in SPL does 0dBFS equate to? That's a little difficult to figure out. We might have to guess. How loud in RMS is some good, hot music? OK, 0dBFS is probably about 12dB hotter than that. So lets say you monitor so that full-scale RMS is around 85dB SPL (which is pretty darned loud in my opinion). That means that full scale is 97dB SPL. Does that make sense? At the leve you set your monitoring system to 97dB SPL is equivalent to 0dBFS. Ergo, 85dB SPL is equivalent to -12dBFS, etc. Now your A/D and D/A converters give you around 108-120dB of dynamic range. This means that your converters have a noisefloor that is somewhere in excess of 108dB lower than full scale. Since full scale is 97dB SPL on a loud listening day for you, the noisefloor of the converters is lower than -11dB SPL. That is lower than many things: The noisefloor of your room The noisefloor of your microphone The noisefloor of your analog gear The threshold of human hearing Atmospheric noise - the noise of atoms colliding with each other in free space. Now, let's talk about a quiet passage in the music. Perhaps this quiet passage is played at 40dB SPL - 57dB lower than full scale. Yes, that does "suffer" from using not as many bits - granted. Using roughly 10 bits less than the full capabilities of the converter, it only uses 14 bits at the most. Admittedly, this sounds scary and awful. On the other hand, the noisefloor of the converter is still in the area of -11dB SPL, so we are still getting clean recording of everything above that mark, including your quiet passage. Your quiet passage uses fewer bits - but that's OK, because quiet passages have lower dynamic ranges than loud passages. That means that the difference between your signal level and the noise level of the various combined sources of noise (room, mic, preamp, converter noise, etc) is much smaller with quiet passages than with loud passages. Because of this, they NEED fewer bits, so we have nothing to worry about. For the sake of discussion, everything fed into a converter above its noisefloor is perfectly captured. If the converter's noisefloor is calibrated based on our settings above to be -11dB SPL then everything above that is captured accurately. And -11dB is so far below anything we're capable of hearing that you aren't losing anything audible below that. Does that help?
Yes, that addresses why you don't lose quality inside the system. My illustration above was talking about not losing quality going into or coming out of the system. I hope this all helps.
Err, that can of worms is pretty big and we should probably carefully just put it back on the shelf until we get the rest of it straight. Cheers! Nika. |
Nika Aldrich wrote on Sat, 02 October 2004 15:22 | ||
Err, that can of worms is pretty big and we should probably carefully just put it back on the shelf until we get the rest of it straight. Cheers! Nika. |
compasspnt wrote on Thu, 01 March 2007 13:48 |
People should, as always with regard to Internet Fora, try things out carefully for themselves, and find what works for them. |
cerberus wrote on Thu, 01 March 2007 16:21 |
i got a reply from paul frindle indicating to me ... |
Ashermusic wrote on Sat, 17 February 2007 02:17 | ||||
So what does that mean for Logic (32 bit float) users ? B |
cerberus wrote on Sat, 03 March 2007 17:53 |
waves native plug-ins in these formats: [rtas, vst, au, dx, mas] are 64 bit float with the exception of the L-series limiters, which are 48 bit fixed point double precision. tdm are all 48 bit fixed point double precision. none of them will clip internally. jeff dinces |
astroshack wrote on Sat, 03 March 2007 02:11 | ||||||
The relationship between recording levels and subsequent processing levels is somewhat relevant, given many here have mentioned gain staging and overhead within mix engines as the dominant reason for keeping record levels lower. It is less important in floating point systems, but (as Nika implies) is is still important. This is because calculation accuracy diminishes at levels where signals attempt to exceed full scale in float processing. This loss of accuracy is a good trade off when compared to the complete barfing which happens when fixed point systems attempt to exceed full scale, but it still gives inferior results compared to "always staying below full scale". The reason Nika didnt go into detail is the argument is mostly about "how much accuracy is lost?", especially when comparing very high resolution systems using 64 bit float or double precision fixed point. Depending on who is doing the arguing (and I have read many posts in various forums over the years from Nika and others on this topic) the loss of accuracy is considered a "good trade-off" or a "significant trade-off"....suffice to say, floating point processing can be a lifesaver in many situations. In other words, floating point systems do remove the worry of clipping within the processing chain, but you will still achieve best results if you never approach levels which would clip in a fixed point system. Of course, not all plugins within floating point hosts are also float based. Most DSP card solutions use fixed point - the only exception I know of is the UAD1 DSP card, which can use floating point because the processor is a GPU rather than a dumb DSP chip. So all of the UAD1 plugins use floating point. On the other hand, some native plugins use fixed point even when invoked in a floating point host - a good example is the Waves plugins, many of which are fixed point. Waves do this to maintain identical processing across their entire range of platforms (or, maybe, becuase they dont see any benefit in reconfiguring their fixed point algorithms for floating point). So Waves users need to carefully watch their "gain staging", even in floating point systems. The converse needs to be considered here: is there any loss in fidelity if moderately lower signals are used in floating point processing? The answer is "no". Therefore, the same concept applies to floating point systems as it applies to fixed point systems: lower levels are best in modern DAWs. Sean |
RSettee wrote on Thu, 15 February 2007 16:52 | ||||
What? Is Tarbox computer based? If it is, it really honestly has fooled me. The drum sounds that they've got over the years there have honestly been the biggest that i've ever heard. |
tom eaton wrote on Wed, 14 March 2007 06:28 |
but makes no sense at all in terms of audio.-tom |
tom eaton wrote on Wed, 14 March 2007 09:28 |
Yep.. zero on your analog multitrack is somewhere between -20 and -14dBFS depending on your converters. Looks awful puny on screen. |
garretg wrote on Wed, 14 March 2007 10:30 | ||
I think this is a very good point that daw manufacturers could/should address. I'm not sure about other programs, but Sonar at least has a way to vertically zoom audio tracks (can't remember the keyboard shortcut off hand)... it's very handy to get tracks so you can see them without actually gaining em up Only problem is if you accidentally zoom up a track so it's visually clipping, and forget what you did... then tear your hair out for an hour trying to figure out why you're seeing flat tops but not hearing them. Er, not that I've ever done that. -Garret |
daveseviltwin wrote on Tue, 03 April 2007 10:15 |
I find that when you record too low you pull up artifacts when you try to boost it. |
RedStone wrote on Wed, 04 April 2007 17:49 |
if you have too much noise in the chain, get rid of it at the source - |
Quote: |
Well, either those artifacts are IN your recordings or your DAW is terrible. Gain should in no way increase "artifacts." The entire argument in this thread is FOR recording at lower levels. An average level of -20dBFS for a 24 bit input signal is fine. -tom |
Quote: |
An average level of -20dBFS for a 24 bit input signal is fine. |
Quote: |
Gain should in no way increase "artifacts." |
tom eaton wrote on Wed, 04 April 2007 23:34 |
My Dorrough meters, for example, center on -20dBFS. Using 0VU= -20dBFS gives you 20dB of headroom over +4 from your console or preamp, and likewise gives your console or external gear a signal that can be handled realistically in the analog world on the back end of the DAW during an analog mix or analog processing in a digital mix. Some folks might say -18dB is a good average, that's fine, too. But for simplicity sake, +4dBm=0VU=-20dBFS works, sounds good and is portable. No one will scream at you when they have to mix your clipping tracks, and no broadcaster will reject your tracks for level reasons (ABC/Primetime has let me know that they're not happy when I've submitted tracks that RMS at -12... they want -20, so I give it to them). -tom |
Hitmaker wrote on Sat, 21 April 2007 15:45 |
Surprisingly , a very dynamic track can be more difficult to get to commercial levels than a more compressed one . Recent case-in-point ...a very dynamic track ( initial average RMS .. ~-25 dB )... |
Hitmaker wrote on Sat, 21 April 2007 18:45 |
I suspect your imaginings of that process are somewhat romantic ... Tracks are NEVER turned down .... IMO , the only reason mix engineers use a ME is because they can't achieve the type of controlled level gain , using the tools , experience , and monitoring environment that the ME can .... Anyhow , my rationale in this diatribe is to point out to mix engineers that they are part of a system , between tracking , and mastering , and mixes that are too soft may require more undesired artefact than they imagine ... |
I'd call an RMS level of -25dB low.
Hitmaker wrote on Sun, 22 April 2007 00:33 |
Hi , I'd call an RMS level of -25dB low. So would I .... but it still peaked near full-scale ... Very , very dynamic .... and calling for more fairly intense compression/limiting .... because of this ... Cheers .... |
If a track shows up HERE (in my studio where I do and have mastered records for years) with an RMS of -8
. So maybe you're right on that count, but not the way you meant it.
Can you explain to me how a mix that is "too soft" may require more "artefact" than one that is already pinned?
And the picture doesn't tell me anything about which file sounds better
Your position is, if I read you correctly, that you are able to make a louder master with fewer compromises from a mix that has a higher RMS value
Paul Cavins wrote on Sun, 11 February 2007 23:33 |
But seriously, I'm wondering how, in the real pre-DAW analog stage, your signal gets into the computer. Here is my deal. My little PTLE/iMac setup has the mic going into a Sytek MPX-4A pre, perhaps the Summit TLA-50 compressor, then into the Digi 002r. When I've tracked in the past, the Sytek was not turned up all that high, the comp not cranking out signal (to my knowledge), but the peaks were much higher than would be desired if I were to follow the advice in this thread, which I intend to do. What about people who like to "drive" their preamps to get a desired effect? Anyhoo, do you find yourself using an analog fader to diminish the signal, or do you have other ways of keeping it modest before the DAW. Please forgive the technical naivete revealed in this post- PC |
bewarethanatos wrote on Thu, 23 August 2007 08:58 |
Ok, thanks Tom. I still don't really understand how everyone else has such "an easy time" recording at lower levels to try this out. Myself and the producer I work with don't understand how people are able to drive the pre-amp to get the color they want and still stay at levels like -20dBFS. Are you saying most of you guys just turn the pre-amp up until you hit -20dBFS ITB? That's my main point of confusion. And I apologize for my brazen use of "dBVU." Like I said, still learning. I read about the different kinds of "dB" in the "Sound Reinforcement Handbook," but not having a previous knowledge of much of what they're talking about, needless to say I got pretty confused. And like you said, whether my examples were appropriate or not, I was simply trying to get my point across. |
bewarethanatos wrote on Thu, 23 August 2007 05:58 |
Ok, thanks Tom. I still don't really understand how everyone else has such "an easy time" recording at lower levels to try this out. Myself and the producer I work with don't understand how people are able to drive the pre-amp to get the color they want and still stay at levels like -20dBFS. Are you saying most of you guys just turn the pre-amp up until you hit -20dBFS ITB? That's my main point of confusion. |
CWHumphrey wrote on Thu, 23 August 2007 22:05 |
Unfortunately, Pro Tools meters continue to suck. |
compasspnt wrote on Sat, 01 September 2007 14:10 |
In the older XLR-out 888-24's, was the same "problem" existent, or did this start with those little multipin outputs? |
Bob Olhsson wrote on Sun, 09 September 2007 15:36 |
We've finally hit the point where a better sounding converter is no longer coming out every few months. I'm amazed that more studios aren't finally investing as Terry has in better multi track converters. The improvement they make can be jaw dropping. In my opinion much more so than a fancy mike preamp. |
Hitmaker wrote on Sun, 22 April 2007 03:33 |
Hi , I'd call an RMS level of -25dB low. So would I .... but it still peaked near full-scale ... Very , very dynamic .... and calling for more fairly intense compression/limiting .... because of this ... Cheers .... |
compasspnt wrote on Sat, 06 October 2007 15:18 |
Indeed, the term "dynamics" is on the way to becoming a negative one. Remember when we actually desired dynamic range...the more, the better? If everything is loud, then nothing is loud. Relatively. Volume is the only technical tool we have to work with. Let's not take that away from ourselves. |
PaulyD wrote on Thu, 11 October 2007 02:47 |
I was reading a post from Steve Berson on Brad Blackwood's forum where he was hipping everyone to SSL's new free X-ISM plug-in. Even if you're not interested in the plug-in, go to that web page and read! As I was reading it, it struck me that this clearly illustrates what causes the illegal reconstruction samples that Paul Frindle was talking about. It's inter-sample peaks. I now fully understand what Paul Frindle meant when he stated "this isn't a sample value issue, it's a signal value issue." It's all so clear now. Of course, this begs the question, then why don't DAW's have meters that indicate signal value rather than sample value? Perhaps it's CPU load. SSL does warn that the X-ISM plug-in is CPU hungry. Nevertheless, what a wonderful freebie! Paul EDIT: Something else just occurred to me: This must be part of why higher sample rate recordings sound "better." If sample rate is more than doubled, e.g. 44.1 KHz vs 96 KHz, it also more than doubles the possibility that sample value peaks are going to coincide with true signal value peaks, meaning the meters in a DAW are going to more accurately reflect true signal value. This perception will cause a recordist to reduce gain somewhere in the signal chain and yield cleaner results. Not to mention PCM white noise getting shifted into the ultrasonic range... I feel like a big window has just opened up to me...and it's so simple! |
mikey wrote on Sat, 29 March 2008 20:39 |
Quote attributed to Dan Lavry: ''you want to use most of the AD range,but not hit 0db(full scale)." |
mikey wrote on Sat, 29 March 2008 20:39 |
Quote attributed to Dan Lavry: ''If you work too far away from full scale you are wasting db's dynanic range without gaining anything'' |
mikey wrote on Sat, 29 March 2008 20:39 |
Quote attributed to Dan Lavry: ''Whenever possible use the internal clock for less jitter'' |
brubart wrote on Fri, 04 April 2008 12:53 |
But is clipping at 11kHz a problem? The lowest harmonic produced by clipping would be 33kHz, which is inaudible and would be filtered out by the reconstruction filter. |
brubart wrote on Fri, 04 April 2008 12:53 |
Intersample peaks might not be a problem. Here's my reasoning. Suppose you sample an 11kHz sine wave at 44.1kHz, or four samples per cycle. Suppose those samples occur at 45, 135, 225 and 315 degrees in the cycle. At those points, the instantaneous value of the sine wave is .707 its peak value, or 3 dB below the peak. I suspect that this is a worst-case scenario: the reconstructed intersample peak voltage is 3 dB above the measured sample levels at 11 kHz if you use a 44.1kHz sampling rate, and the samples fall at the right places in each cycle. So if the samples read 0 dBFS, the reconstructed analog wave might be 3 dB into clipping at the instant it reaches a peak. But is clipping at 11kHz a problem? The lowest harmonic produced by clipping would be 33kHz, which is inaudible and would be filtered out by the reconstruction filter. Also, if a broadband signal's level is reaching 0 dBFS, the 11kHz component of that signal is likely to be several dB below 0 dBFS, since instruments' spectra generally fall off at high frequencies. (A trumpet or cymbal might be an exception.) However, the I.M. distortion of several ultrasonic frequencies interfering with each other could be audible. Comments? brubart |
groundhog wrote on Mon, 26 May 2008 20:37 |
Don't know if this has been brought up ............. if you're working with solely virtual instruments is it still best to run with the lower levels ? I run it low even for that here , BUT lately I've been wondering if it's as necessary or even all that helpful with VI 's especially if no or a minimum of processing is being added . ?? |
compasspnt wrote on Thu, 23 August 2007 16:39 |
First get the sound you want from your microphone>preamp chain. Whatever that takes. Be sure you like it. (Some call it "colour." I personally think of colours as different shades of things I can see. I prefer to call it "good sound.") Next, do whatever you have to do (staying as sonically pure as possible) to reduce the level so that it does not go to digital at maximum. Some converters have pads. Some people run through another outboard hardware unit that has an output pot (and ideally, a good transformer!) Adjust the input-to-digi level to your desired amount, be it -20, -18, -12, etc. (Some call it "to taste." I personally think of taste as different flavours of things I can eat. I prefer to call it "good level.") Record good things. Collect your awards. That is the process. |
maccool wrote on Sun, 08 June 2008 16:20 |
I use an ISA428 pre with the internal ADC card. The ISA428 has no pad to modulate it's preamps' outputs before they hit the convertor. But since it's a 4-channel pre with an 8-channel ADC card, and an additional 4 analogue-ADC inputs, then I should be able to take the analogue signals from the ISA428's pre's, route them through "another outboard hardware unit that has an output pot (and ideally, a good transformer!" and back into the ADC, yes? |
Roger Langvik wrote on Sun, 08 June 2008 16:05 |
Isn't this what the "trim" pot does? |
maccool wrote on Sun, 08 June 2008 15:20 |
I use an ISA428 pre with the internal ADC card. The ISA428 has no pad to modulate it's preamps' outputs before they hit the convertor. My question is, what would you suggest I use to do that? Budget is definitely an issue, and I want to be sure that I won't degrade what I'm getting out of the pre's. |
maccool wrote on Sun, 08 June 2008 10:38 |
At the moment, I use the gain pots on the pre's to keep the signal down at a level which produces that -12dBfs peak in the DAW. These pre's have analogue peak level meters, not VU, and they indicate about -16dB to achieve that -12 in the daw. I simply want to explore and listen and try to hear any appreciable difference between tracking at -16 and then at -3 on the pre's level meters. |
garret wrote on Tue, 24 June 2008 15:46 |
...Here's an idea: How about getting an SM Pro iNano passive attenuator, and hooking it up to a channel insert (w/ TRS cables)? The channel inserts are post-preamp, pre-converter... |
Mark Kellen wrote on Sat, 13 December 2008 04:30 |
More perplexing (to me at least!) are plugs like an 1176 where the input controls the threshold or a saturation plug (DUY Tape, AC1, AC2...) that you may want to drive hard. Doesn't raising the inputs on these to get the desired effect wreak havoc with the headroom/potential internal clipping/etc. -- even if it is compensated for on the output? Thanks in advance for your help! |
compasspnt wrote on Sun, 14 December 2008 15:34 |
If recording @ 16 bit, it is even more important to keep these levels reasonably lowered. |
jetbase wrote on Tue, 23 December 2008 01:20 | ||
How so? I was under the impression that, at 16 bit, the levels need to be kept up, or was this simply in regards to interacting with plugins? |
Paul LaPlaca wrote on Tue, 19 May 2009 00:13 |
This is one of the most useful threads I have ever read... ...I've done one mix with reduced tracking and mix levels and it's mind boggling how much better it sounds. |
compasspnt wrote on Tue, 19 May 2009 07:03 |
But if staying all ITB, just use the best software you can get, and do it in stages, not pushing any one "unit" too hard. Hope this helps a bit. |
Paul LaPlaca wrote on Mon, 18 May 2009 21:13 |
If -20 is supposed to be treated as 0dbu, visually it's very difficult to see how individual elements fit into that range. Is there a plug in for the master fader with a HUGE meter? What's an affordable external meter? |
compasspnt wrote on Tue, 19 May 2009 22:09 |
http://www.colemanaudio.com/ |
compasspnt wrote on Tue, 19 May 2009 01:03 |
As for the final mix, personally I would never normalise anything. |
a crowley wrote on Wed, 20 May 2009 06:28 | ||
I have some mixes that are right where I want them , an ME is telling me that they sound fine in his room ( haven't heard them at his place ) and that if I feel no compression is needed then I don't need him for this and to since the volumes are a little low to go ahead and normalise the tracks . Instinctively , I believe that there's something wrong with doing that , however unless I do finish up in a room other than mine it will best be done ITB . Still , comprehending the benefits to staging is beyond me . |
Hank Alrich wrote on Mon, 13 July 2009 12:08 | ||||
If these tracks constitute an "album", normalization is not likely to yield a balanced track-to-track equivalent listening level. I wouldn't do what that ME is suggesting. |
oversampling wrote on Sat, 08 August 2009 04:48 |
A great thread with many useful information about digital recording But i simply don't understand why would somebody use analog meters. Why not just using digital ones and set them up properly? |
KB_S1 wrote on Thu, 15 October 2009 06:17 |
I recently attended a seminar type event hosted by Prism Sound in Edinburgh. Most of what he discussed was very much supportive of what is in this thread. The number of 'bits' you use is not of great importance. |
Geoff Emerick de Fake wrote on Sat, 26 December 2009 14:29 | ||
What does it mean "of great importance"? Does it mean 8 bit is as good as anything or that it doesn't matter as long as there are at least 16 significant bits? It's hard for me to believe that Prism sound could dismiss the importance of operational resolution. An abstract of this communication would be helpful. |
Ashermusic wrote on Mon, 09 August 2010 11:37 |
I just had a "discussion" with a friend:) I pointed out to him that virtually every good engineer I know recommends recording 24 bit at lower levels, i.e -12-018 dBs, for a number of reasons. He says that while he agrees that there is no loss when doing so he frequently records at hotter levels because if you record it low, there is only so much you can raise the playback level to i.e. send it out to a compressor and really smack it without adding plugins, which change the math. So he says, as long as you understand gain structure fully, as he does, and depending on what you know you want to do with it, it may be, and frequently is for him, a better idea to record it at -3 dBs rather than the -6-18 range that almost everyone else recommends. It is not passing my smell test but he says it is because I do not really understand gain structure. I thought I did Whaddya think? |
Ashermusic wrote on Sun, 08 August 2010 20:37 |
depending on what you know you want to do with it, it may be, and frequently is for him, a better idea to record it at -3 dBs rather than the -6-18 range that almost everyone else recommends. |
Geoff Emerick de Fake wrote on Mon, 09 August 2010 06:46 | ||
|
tom eaton wrote on Mon, 09 August 2010 21:02 | ||||
I believe he's talking about peaking at -3dBFS. Let's discuss. If you run your converter at +24dBm = 0dBFS, which is -20dBFS = 0VU = +4dBm, and you like to peak around -3dBFS, that means an analog level peak of +21. That's pretty hot. Would easily overload some of my compressors (Drawmer 1968 clips at +20). If you have a particular track that you want to hit a particular compressor that hot, go for it... that's the spirit of "whatever works." But as a general practice, I think it's a bad idea. If you run your converters -16dBFS=0VU... well, then you're talking about a -3dBFS peak of +17 in the analog world. Certainly generally usable. Unlikely you'd clip any real piece of analog gear there. |
KB_S1 wrote on Tue, 10 August 2010 10:49 |
A big part of the issue (to me) is what happens to the signal levels once they are in the DAW. Forget the issue of convertors altogether for now. If you have 40 channels of audio: all at -0.5dbfs: there will be mix bus issues. |
Quote: |
Also consider any plug-ins that are to be used (if you use them). If a process you want to use involves boosting a signal, you will first need to attenuate the signal, then process it. |
tom eaton wrote on Tue, 10 August 2010 14:16 |
Hey Jean Luc- A page or two ago on this very thread you'll find me suggesting folks use the Dorrough meters for this very reason. Not only can you see average and peak levels simultaneously, you can look at the dynamic range of your signal (and with the digital meter you can see your absolute headroom as well!). Great for tracking, mixing, mastering... |
Bob Olhsson wrote on Wed, 18 August 2010 14:59 |
Amazing how people listen with their eyes today! |
organica recording wrote on Fri, 20 August 2010 23:08 |
<If you break out an audio oscillator, feed it into all of your own A to D inputs at once so as to place maximum stress on the converter power supply and then sweep different frequencies and levels, you can hear which converters have enough headroom and which don't. An IM distortion test is even more revealing.> Must do this. |
Bob Olhsson wrote on Wed, 18 August 2010 14:59 |
Amazing how people listen with their eyes today! |
Bob Olhsson wrote on Wed, 18 August 2010 13:59 |
Amazing how people listen with their eyes today! If you break out an audio oscillator, feed it into all of your own A to D inputs at once so as to place maximum stress on the converter power supply and then sweep different frequencies and levels, you can hear which converters have enough headroom and which don't. An IM distortion test is even more revealing. |