The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
2buss comparison: Fatso vs. Oxford Dynamics Dynamics Plugins
Old 16th October 2006
  #61
Lives for gear
 
minister's Avatar
Hi Dave,

yeah...i am aware that Mastering Engineers and Non Mastering Engineers go up to DIGITAL FULL SCALE. but it is not for the reasons you are suggesting....

you may want to read up on the linearizing power of Dithering. with dither, there is NO difference between analog and digital audio at ANY levels, except clipping. bit depth sets the noise floor. has nothing to do with 'resolution'.

here is another thread on Gearslutz that might be of interest to you.

https://www.gearslutz.com/board/mastering-forum/69929-q-paul-frindle.html

here is another thread from PSW that explains some more about inter-sample peak distortion and signal reconstruction issues in DAW's....form the old George Massenburg forum...starts about 4 pages in.

http://recforums.prosoundweb.com/ind...18/0/48/11603/
Old 16th October 2006
  #62
Lives for gear
 
gsharp's Avatar
 

re: mastering I think they do it to be as loud as possible, no?

again, if what you say were true there would be a discernable degredation in quality bewteen the same material mastered to -3dBFS and -6dBFS.
Old 16th October 2006
  #63
The Distressor's "daddy"
 
Dave Derr's Avatar
 

Clipping within Daws

Minister

They definitely are trying to squeeze the last bit of performance out of the 16bits that CDs afford us, as well as make some records as loud as possible. Classical CDS sound much softer than rock records, but Mastering guys still bring the often momentary peak levels right up to 0dBfs so as not to lose dynamic range and quality.

gssharp - "discernable degradation" is not actually what Ive been talking about as much as theoretical degradation. With modern 24 bit converters, I have stated several times that you probably would not notice quality differences between full scale signals and signals recorded -20 to -30dBfs. But ideally, you would want to get the best performance out of your ADC by getting maximum resolution which occurs close to 0dBfs.

The thing that many people using Protools HD and other systems hear when they sum or process full scale signals is internal software clipping. We discussed how nice it would be to have channel input trims in the Daw mixers just like analog consoles. This would allow poorly designed busses, summers, and plugins to have headroom and not create nasty nasty digital clipping. A smart DSP guy would just add the input trim to the fader level value so there was only one mulitply per channel, but then again, if there was a poorly designed EQ or other plug in "pre-fader", it could clip! HD with its 24 bit NON floating point is very prone to clipping Full Scale signals as discussed a few pages back. Its a shame that poorly designed busses and software often clip perfectly recorded full scale signals, making the users wary of recording close to full scale in many cases... as I see here.
Old 16th October 2006
  #64
Lives for gear
 
minister's Avatar
hi dave,

yer missing my point. there is no greater 'resolution' of bits closer to 0. and there is no degredation of sound when you print a CD at -6 or -8. ...which is what i am taking you to suggest when you say they are squeezing every last bit of performance they can. CD's are mastered to the ceiling to compete with other LOUD masters.

also, if you are recording, at say, -3dBFS, and 0VU =-18dBFS, you are over-driving the mic pre.

Pro Tools fixed point dual 24bit (48 summing bit) mixer has more headroom than any analog desk.

seriously, do a google on dithering and linear effects and read those links...you are suffering from the common misconception about digital audio. and i know you won't have time for this, but check out Ken Pohlman's book on the Principles of Digital Audio.

(but i will say your analog stuff kicks ass! i love my distressors and can't wait to try a li'l freq!)
Old 16th October 2006
  #65
The Distressor's "daddy"
 
Dave Derr's Avatar
 

You are losing resolution if you print a CD below 0dBFS. A CD that peaks 6dB down is working at 15 bits at best. Please dont doubt this for a second. Not sure why this is so hard to understand actually.

Channel gain before an ADC is not standardized but I record with peaks near 0dBFS all the time, as do thousands of others. Just as in the old analog tape days, you have to make sure each stage is optimized and not clipping. Somethings wrong if you cant record up to 0dBfs and get a great sounding track. What you and your software do after that is another story.

Ive been designing converters and working with digital since 1986 at Eventide. In the 80's, I saw Stanley Lip****z present some pretty seminal papers on dithering as well as heard his demos on low level non linearities and how dither improved the sound on low signals. I also saw lots of BIG HAIR here in New Jersey. The converters back then were not so good, and even the DACS had Zero crossing THD caused by MSB errors, so a decaying source would crackle and pop when it faded to a really low level signal. Reverb tails were very revealing. Dithering helps hide this. But dithering does not bring back bits or resolution. NEVER EVER.

Again its a matter of definition and how converters work. Recording below 0dBFS does lose resolution although you may never hear it with modern converters. Its just a matter of how much resolution you are willing to lose on a given source. If you record a signal 90dB down and listen, there will be no doubt you are losing quality. Its pretty danged horrid! 60dB down and its better, but you are losing quality and resolution. 30dB down is better. 10 dB down is better still. Its just a matter of degree. But with any ADC, to be assured you are getting the full resolution it promises, you have to record up to 0dbFS. ITS THE LAW!

Thanks for your compliments on our products! Makes it all worth while.

Gotta GO. Life is short and theres just too many audio toys to play with.

Happy Recording gang
Old 17th October 2006
  #66
Lives for gear
 

Quote:
Originally Posted by Dave Derr View Post
You are losing resolution if you print a CD below 0dBFS.

....snipped for brevity....

But with any ADC, to be assured you are getting the full resolution it promises, you have to record up to 0dbFS. ITS THE LAW!
OMG!

As I have said so many times before (and so many originally doubted me) - whoever coined that darned term 'resolution' has done more irrevocable damage to our industry than almost ANY other single factor. The term 'resolution' is the technical equivalent of a virus, it spreads exponentially and cannot be ousted without inordinate levels pain and misery! Once disseminated, it is as though every unsuspecting and innocent victim of the concept requires the poison to be almost surgically removed - before we can go forwards once again.

There really isn't any point is there :-(
Old 17th October 2006
  #67
Lives for gear
 
minister's Avatar
hi dave,

yeah, we are all working hard to do good things and i can't even begin to comprehend how you do your great products. you clearly know what you are doing there...and it helps me do my job better.

Quote:
Originally Posted by Dave Derr View Post
You are losing resolution if you print a CD below 0dBFS. A CD that peaks 6dB down is working at 15 bits at best. Please dont doubt this for a second. Not sure why this is so hard to understand actually.
Mr. Paul Frindle (and many others, Dave Collins, Nika, Dan Lavry) know(s) FAR better than i...but, what you say is not hard to understand at all. but it is, in fact, wrong. what you are saying is at -6 we are at 15 bits; at -12 we are at 14 bits; at -18 were are at 13 bits; at -24 we are at 12 bits...et cetera... by your logic, fades would sound like total a**... but the truth is, that with dither, a 16 bit CD files operates just like analog at any level and it does not have to do with 'resolution' of decibel (or signal) level to bit depth. your fades go perfectly down to a benign noisefloor. with dither, the resolution of a system is far below the least significant bit. theoretically, there is no limit to the low-level resolution.

in other words, you mix does not 'sound worse' at -10dBFS with dither than it would at -.05dBFS.

today's etymology lesson : 'dither' comes from the middle english verb 'didderen' which means to shiver or tremble. and one of the first implimentations of it was in WWII, air pilots used it in their mecahnical computaters for trajectory calculations....the vibration redced the error of the 'sticky' mechanical parts.

and also, what you are recommending by recording near 0dBFS is pushing your mic pre's or compressors to +21 to +24dbU !! across an entire mix, is that a good idea? and standard analog practise? ...sems to me, in ye olde days you were going close to 0VU, not +20VU.

also, much better to record averageing -18dBFS (close to 0 VU) and make up the gain at the very end.... digital mixes will sound much much better. mine already have! and my masteing engineer can bring it up close to 0 and it does not collapse!
Old 17th October 2006
  #68
The Distressor's "daddy"
 
Dave Derr's Avatar
 

Hiya again Paul

Yup you have it right with the loss of bits.

As with analog, your quality does smoothly get worse as you get lower in level. Its imperceptable though because as you fade out, it's much softer in volume and the noise and distortion drop out of view along with the signal. As signals get louder and softer, theres not jumps and leaps in the sound quality at all, its a subtle thing and usually not perceptable at all.

The problem is, your converter noise floor stays the same whether you record peaks at 0dBFS, or -50dBfs. So if you adjust your monitor level to make up for say 2 bits of unused level (12dB), you have turned up the noise floor and distortion as well. Lets say you have 96 dB Signal to Noise with a standard CD, your distortion will usually also be close to 96 dB down. Now if you drop the 12 dB off the top (because you left headroom) and raise your monitor level 12 dB to hear at the same volume, you have brought that noise floor and distortion up to -84dB (96-12dB).

So in the above example, you end up with the noise and distortion 84dB down which matches the theoretical dynamic range of 14 bit Audio.... Because now it is 14 bit audio! This is why mastering guys and professionals who mix to 16 bit, always try to get right up to 0dBFs (or like -.1dBfs) to squeeze all the quality and dynamic range out of CDs as they can.

Dither smoothes over lumpy quantization steps and hides some nasty tones on low level signals, but it also does add noise. You improve nonlinear quantization errors by trading off a little bit of signal to noise. It doesnt give you better resolution or add meaningful accuracy (or bits) to your audio, but generally it does sound better way down low without "rocks", crackles, wierd tones, etc. They play all kinds of tricks to hide dither such as moving it out of band (usually above 10 - 20Khz), or using other masking methods, but ultimately, you are adding a form of noise to try to get rid of low level quantization steps and residuals.

Gotta run run run!
Old 17th October 2006
  #69
Lives for gear
 
gsharp's Avatar
 

Quote:
Originally Posted by Dave Derr View Post
Gotta run run run!
Hopefully to your local bookstore to pick up Nika Aldrich's or Ken Pohlman's books.
Old 17th October 2006
  #70
Gear Guru
 
u b k's Avatar
 

Quote:
Originally Posted by Dave Derr View Post
You are losing resolution if you print a CD below 0dBFS. A CD that peaks 6dB down is working at 15 bits at best. Please dont doubt this for a second. Not sure why this is so hard to understand actually.

the only thing i can think of to say to this is that i got a lot of older cd's that rarely hit 0dbfs, and i got a lot of cd's that hit 0dbfs all the time. many (most) of those former cd's sound better to me than the latter. you seem very attached to the theoretical aspects of the discussion when time and time again my experience shows me that converters like it when i take it easy on them. digital systems as a whole, ime, prefer quieter levels, including internal levels in a daw running plugs.

my understanding, limited as it is, is that a 24bit recording that peaks at -6db still has the benefit of the increased resolution of 24bit, as opposed to a 16 bit recording that peaks at -6db. you could, for instance, have a 4 bit recording that used the full scale of it's resolution and peaked at 0 all the time without clipping, but it'd sound like ass because the resolution is so limited. likewise, the beauty of 24 bits is that you can record quieter and still have objective improvements over a 16 bit recording with the same loudness.

all other things being equal, specifically converter quality.


gregoire
del
ubk
.
Old 17th October 2006
  #71
The Distressor's "daddy"
 
Dave Derr's Avatar
 

UBIK SAID: "A 24bit recording that peaks at -6db still has the benefit of the increased resolution of 24bit, as opposed to a 16 bit recording that peaks at -6db. "

This is absolutely TRUE! In fact, I agree with everything you said, UBIK. Protools HD especially has "GOTCHAS" when working with full scale levels. I predict they will someday move to 32 bit floating point, like many other DAWS.

24 bit Audio has 256 times the resolution of 16 bit audio. This is why its not a problem to record 10 - 20dB down on 24 bit converters as I have said. Its very very forgiving. It will still probably have better accuracy than 16 bit converters recorded at full scale. However, I think you will find your older CDs probably still peak at close to 0dBfs somewhere on them. They probably arent as compressed. Over compression is really such a fad these days. But Mastering Engineers have always tried to get as much level out of the medium they are working on as they can. Even on old medums like Vinyl, it was always a challenge to get the most level you could on them without getting "unmusical".

Just a stupid late night memory: I remember back in the 80's when I got a Jethro Tull album on one of my first CDs. It sounded really thin and just plain not like I remembered. After a year or so, I listened to the album on a record player a fellow employee had and was just thrilled with the sound. The mastering guy who did the CD did a really bad job of capturing the musicality of the album, and didnt temper the high end, or fatten the bottom as he should have for the CD. The older records definitely had soooo much more dynamic range than the overcompressed stuff thats popular today. THe mixdown guys slam the analog tape decks, and compress before slamming. Its LOUDER for sure. But sheesh, I agree that I miss the dynamics and realism of many older albums.

Its funny that now we have this really wide dynamic range medium, yet modern pop music does everything it can to get rid of dynamic level changes of songs. Ironic.
Old 17th October 2006
  #72
Lives for gear
 
minister's Avatar
Quote:
Originally Posted by Dave Derr View Post
Yup you have it right with the loss of bits.
hi dave,

this is the classic misconception.

oh well, i don't seem to be getting through to you. if you have time, check out the links i gave you, and pick up a copy of nika aldrich's and ken pohlman's books on digital. also, google paul frindle and read some of his posts. and, yes, there is a white paper by lipschitz on dither...if you want it, i have it.

i understand that, in essence, the application of dither amounts to adding low-level random noise to the signal. the introduction of a low-level noise dither can linearize a digital audio system and increase the signal resolution well below the quantizer step size. the noise penalty is only 3dB -- which is why the dynamic range of a CD is 93.3 and not 96. but the signal-to-noise ratio is so high that you don't hear the noise on the LSB.

i also still maintain that recording as high as you are suggesting (well over +15 VU) is not a typical (or recommended) analog practise.
Old 17th October 2006
  #73
Lives for gear
 
minister's Avatar
Quote:
Originally Posted by Dave Derr View Post
Even on old medums like Vinyl, it was always a challenge to get the most level you could on them without getting "unmusical".
or so that the needle wouldn't jump out of the groove.
Old 17th October 2006
  #74
Lives for gear
 
WunderBro Flo's Avatar
guys, without giving my opinion here...I have one major suggestion:

as dave says and explains things really logically, and many other posts contradict him with at least partly-esoteric explanations, why not just PROVE those theories by making LISTENING EXAMPLES and measurements.

If someone says a 16bit track peaking at -24dB does not sound worse (has no smaller resolution) that the same track peaking at 0dB then go and prove it. Even better, run some sines through a track, bring the level down in 12dB steps and measure the resulting distortion figures. If you get more distortion on lowlevel files then dave is obviously right with his statement that lower level recordings lose bit-resolution resulting in distortion/degradation... if not, then the "esoteric" explanations are right and dither cures any flaws of bit quantisation and we can all save loads of harddiskspace by recording in 12bit.

all those theoretical battles mean nothing when nobody is up to prove them right or wrong. I could give you all explanations why the world is a disc, and no matter what you responded, I could always come up with an answer that says: but it is a disc.


just a suggestion. get real. I wanna see the results, not hear the theories.

Rock on!
Pat
Old 17th October 2006
  #75
Lives for gear
 
norman_nomad's Avatar
Quote:
Originally Posted by Paul Frindle View Post
whoever coined that darned term 'resolution' has done more irrevocable damage to our industry than almost ANY other single factor.
Yep.

And continues to be misused or undefined in the context of these conversation.

I think when people read "resolution" they think "fidelity" and more than likely they carry with them the connotations that accompany the word when used to describe digitized visual media.

from Dictionary.com, 7th definition from the top

resolution

1. <hardware> the maximum number of pixels that can be
displayed on a monitor, expressed as (number of horizontal
pixels) x (number of vertical pixels), i.e., 1024x768. The
ratio of horizontal to vertical resolution is usually 4:3, the
same as that of conventional television sets.

In this case, "resolution" connotes "fidelity" in terms of the allotment of pixiliation used to define an image... Dave pointed this out earlier with his digital camera analogy.

In a feat of transference, it now becomes easy to think of audio "resolution" as a measurement of "granularity" - more pixels, more little squares per inch, means a better (more accurate) image: more bits, more little steps per second, means a better (more accurate) sound.

But this is a poor comparison, because this doesn't describe "bits".

Bits measure a potential to capture changes in amplitude. 1 bit can capture roughly a 6db span of amplitude. You can not make this 1 bit more accurate by adding bits just as getting out a yardstick won't better measure the 1 foot span of a grade school ruler.

More bits = more dynamic range

Digital picture "granularity" as measured in dpi (dots per inch) is probably most akin to digital audio sampling frequency as both are a digital reconstruction of continuity and therefore it is probably best (if we need to compare the two) to use digital photo's "resolution" to describe sampling rate rather than bit depth.
Old 17th October 2006
  #76
Lives for gear
 
WunderBro Flo's Avatar
...but at the same time the bit resolution goes hand in hand with distortion. low bitrate means more quantization error means less true reproduction and more distortion, therefore lower quality of sound.

I still cannot see any explanation why this simple, basic rule of digital audio should not be valid.


Rock on!
Pat
Old 17th October 2006
  #77
Lives for gear
 

Quote:
...but at the same time the bit resolution goes hand in hand with distortion. low bitrate means more quantization error means less true reproduction and more distortion, therefore lower quality of sound.

I still cannot see any explanation why this simple, basic rule of digital audio should not be valid.
Because that quantization error is happening anywhere between 100 dB and 120 dB below full scale with most 24-bit converters, so in most cases peaking at, say, -12 dB, on a signal that is at 110 dB SPL in a room with ambient noise at, say, 40 dB...in other words, with a dynamic range of 70 dB...the ambient noise in the room is at -82 dB FS, so that quantization noise (which, by the way, is not distortion, it's random noise) is a nonissue. As long as you are recording at a level where the quantization noise is below the noise in your room you're not losing any "resolution" in terms of audio quality if you record at a lower level. Sure, you're not "using" ever last bit, but the way I look at it, one of the advanages of 24-bit recording is that in most cases you don't need to worry about "filling up" every last bit...so I'm still "using" those bits, but only to leave extra headroom so I don't distort my converters if the signal I'm capturing gets louder than I anticipate, and so I can run my analog gear at the level it's happiest at, since my converters are happier over a much wider dynamic range.

As far as levels on CD are concerned, that's an entirely different thing...sure, I'll push them up so they're "using" all the bits. But that doesn't mean I need to do the same thing when I'm tracking...

-Duardo
Old 17th October 2006
  #78
Lives for gear
 

Quote:
Originally Posted by norman_nomad View Post
Yep.

And continues to be misused or undefined in the context of these conversation.

I think when people read "resolution" they think "fidelity" and more than likely they carry with them the connotations that accompany the word when used to describe digitized visual media.

from Dictionary.com, 7th definition from the top

resolution

1. <hardware> the maximum number of pixels that can be
displayed on a monitor, expressed as (number of horizontal
pixels) x (number of vertical pixels), i.e., 1024x768. The
ratio of horizontal to vertical resolution is usually 4:3, the
same as that of conventional television sets.

In this case, "resolution" connotes "fidelity" in terms of the allotment of pixiliation used to define an image... Dave pointed this out earlier with his digital camera analogy.

In a feat of transference, it now becomes easy to think of audio "resolution" as a measurement of "granularity" - more pixels, more little squares per inch, means a better (more accurate) image: more bits, more little steps per second, means a better (more accurate) sound.

But this is a poor comparison, because this doesn't describe "bits".

Bits measure a potential to capture changes in amplitude. 1 bit can capture roughly a 6db span of amplitude. You can not make this 1 bit more accurate by adding bits just as getting out a yardstick won't better measure the 1 foot span of a grade school ruler.

More bits = more dynamic range

Digital picture "granularity" as measured in dpi (dots per inch) is probably most akin to digital audio sampling frequency as both are a digital reconstruction of continuity and therefore it is probably best (if we need to compare the two) to use digital photo's "resolution" to describe sampling rate rather than bit depth.
I really shouldn't get into this stuff as I have far too much respect for the people in this conversation....

However - no - no - and no again.. to the idea that spatial digital picture resolution on a monitor with FIXED pixels has anything what so ever to do with digital audio signals. Try to find any dictionary definition of 'resolution' relating to digital audio!!

There is an element of truth in this - but it's not the whole story - and the rest of the story is so dramatically important. It's the single biggest misconception around because it's incomplete. If you believe this stuff it will mess you up forever and ruin your ability to understand what a digital audio signal actually is.

Please, for your own sakes, go and read what has been said publicly in plain language about all this - before the rot sets in. You owe it to yourself.. None of this is esoteric or hard to grasp - once understood it's simple.

It's so very very important - if you believe the digital picture / monitor analogy you cannot begin know just how important :-)


I am really trying to help here :-)

I never went to any college and I have absolutely no diplomas...
Old 17th October 2006
  #79
Deleted bd1be4f
Guest
This is quite interesting. Here we have the creator of some of the finest analog gear (Mr. Derr) at odds on theory with the creator of some of the finest digital gear (Mr. Frindle). Personally, I have to side with Mr. Frindle on this (I still love ya and your products Dave!).

To boil it down to its most basic form, if what Dave is saying were true (that essentially the lower the level you record in digital, the lower the sonic quality), it would basically mean that digital as a capture medium for audio is fatally and irrevocably fukked, far moreso than say, analog tape, where the price paid for lower recorded signals is simply greater relative tape hiss. It would mean that essentially a goodly portion of the dynamic range in digital is nearly or totally unusable.

This would mean that digital could in NO way be considered an acceptable medium for recording, say, classical music, which may have a dynamic range that goes from near silence to full scale fortissimo passages. However, anyone who has listened to a good digital classical recording knows this to not be the case, and that the quiter portions of the recordings do not sound in any way degraded or inferior to the louder portions. If they did, we should immediately abandon digital as a recording medium.

You do not need to "maximize the bits" when recording in digital to obtain optimal results! Further, as Paul has pointed out elsewhere in his essential treatise on summing and mixing ITB, to attempt to process full or near full scale digital signals in the digital domain without allowing for proper headroom will result in illegal signals that degrade the sound. I know, because since I read Paul's information and started recording and mixing ITB more conservatively level-wise, the difference (for the better) has been major. So thanks Paul!
Old 17th October 2006
  #80
AFAIK, digital photography has a simular "resolution=bits" paradigm. (apart from the fixed bits difference)
My girlfriend is a painter (autonomous abstract) and uses digital processes, including photography, and colour and texture is about everything for her, so we did a bit of research. Not that we're specialists, but still..

first there's the lens. One could say the equivalent is the analogue part of a converter.
then there is the AD unit, the chip that captures the image and makes it into bits, same as in an audio converter.
Third there is the format in which the digital data is stored, e.g. "compressed". In digital pictures there's "Raw". But one Raw is different from another Raw.

Then, more bits mean better? Because of more resolution? Not really. Yes more bits do mean finer imaging, so that would be a valid assumption, but there is more to it. The way the pixels are aligned, how the subtle differences in a beam of light going into the camera are translated, that is where the quality of an digital image is also very much decided upon. Just like photographs where points are merged, flow over in each other. So ONLY more bits is not always a "better", meaning "more lifelike, more analogue" image. This is especially noticable in colours. (Funny enough, altough I'm a male and she's obviously not, we do see simular colours, when comparing and discussing. The individual difference, between people's ears, or eyes is another can of worms. I'd guess we're going towards a number of different "experiences", (not the "Creative" products ) but more like warm or detailed, or... regardless to the technology available, where a common ground is found, a compromise, what is pleasant for most ears.

IMO a AD or DA converter is a "translator", between two radically different worlds, and not a mirror, just capturing and duplicating. I do hear a difference in my humble setup, between hot level (just about clipping) and something recorded at lower levels. Especially when doing a mix. (stem mixing OTB) However, because it's not state of the art (but not bad either IMHO), I cannot make any defining comments.
Don't mind my ramblings too much, however; it's just the two cents of a mere mortal. heh

Question for you digital gods... What about this 1 bit technology? Would that change the perspective on the "resolution" debate?

(camera is a Leica digilux 2 into Photoshop (duh) on a Powerbook, and the audio chain is RME Fireface into passive controller (Elma switches and resistors) into K&H monitors) Anyone tried Aperture software yet?
Old 17th October 2006
  #81
The Distressor's "daddy"
 
Dave Derr's Avatar
 

LOL

Eek Zboy, Where did this come from? I have tremendous respect for Paul and havent really seen us at odds... except I tend to use the word "resolution" like many others to characterize dynamic range and linearity.... IE the ability to "resolve" teentsy details.

My assertions which mostly concern getting maximum quality from ADCs, are "Digital 101." To get the full dynamic range and detail advertised from any Analog to Digital converter, you must use the full dynamic range.... right up to 0dbfs. Again, this is simple, and irrefutable. Its also the reason we go for higher "definition" converters, so that we dont have to get so scarily close to going "OVER", and therfore, leaving some headroom below 0dbfs will not be damaging or even noticeable.... especially when it all ends up on 16bit CDs.

What you are recording, or what you and your equipment do after the conversion is not whats at issue. Its mainly just how A to D conversion works that I am discussing. "MIXING IN THE BOX" PROBLEMS ARE COMPLETELY IRRELEVANT TO MY ASSERTIONS. (OF COURSE YOU HAVE TO LEAVE HEADROOM. SHEESH!)

The thing that kind of started me on this was someones assertion that to record with PEAKS 30dB below full scale going into an ADC was ok, or even preferable. As far as the actual conversion process.... THIS IS ABSOLUTELY FALSE! I mean sheesh, if your peaks are -30dB down, where is your average level? -50dB down? Paul has never encouraged putting really low levels into ADCS, and his many papers are clear on this.

One other thing, the linearity that modern digital affords us blows away the linearity analog tape gives us. The dynamic range of digital audio so greatly overshadows the dymamic range of "raw" analog tape that its not even funny. Whether you like tape or digital is not the issue again. I consider digital much better for recording Classical becase of its huge dynamic range and ability to resolve low level detail. I recently got a recording of Stravinsky's RITE OF SPRING from a friend that raises the hair on my tongue. heh It makes me wanna blow my speakers up! However, I do have to say that analog tape with Dolby SR is still reallllly SWEET and "MUSICAL".

By the way, I think the "1 bit technology" nomer is really misleading. The equivilent output is NOT 1 BIT quality! Audio is sampled at an extremely high rate at low res, and stored as a single bitstream, which when filtered on playback, puts out a slower, but much higher res signal, IE 20 bit equivilent. Because DSD itself samples at like 2.8Mhz, it also offeres extended frequency response... which I really like the idea of. Delta Sigma AD converters which are pretty much the main technology these days, sample at really high speeds, and then mathematically filter and convert the high speed, low bit rate(depth) data to lower speed, higher bit rate data. Its kind of a conversion of "quantity vs quality", with the mathematical ability to take a large quantity of lower res data, and turn it into much higher quality but slower rate data. Delta Sigma DA converters kind of work the opposite direction and put out a higher sample rate, interpolated version of the data. This allows you to use less steep analog filtering and is a blessing to designers. I was lucky to be able to design-in some of the earliest Delta Sigma Converters back around 1990 at Eventide. There is some question as to whether linear phase digital filters are pleasing... but thats another discussion.

Incidently, I think the old TC2290 Delay box essentially used a 1bit DSD-type method. Perhaps TC was WAYYEEEE ahead of their time. Anyone know about this, for sure?
Old 17th October 2006
  #82
Lives for gear
 
norman_nomad's Avatar
Quote:
Originally Posted by Paul Frindle View Post
I really shouldn't get into this stuff as I have far too much respect for the people in this conversation....

However - no - no - and no again.. to the idea that spatial digital picture resolution on a monitor with FIXED pixels has anything what so ever to do with digital audio signals. Try to find any dictionary definition of 'resolution' relating to digital audio!!

There is an element of truth in this - but it's not the whole story - and the rest of the story is so dramatically important. It's the single biggest misconception around because it's incomplete. If you believe this stuff it will mess you up forever and ruin your ability to understand what a digital audio signal actually is.

Please, for your own sakes, go and read what has been said publicly in plain language about all this - before the rot sets in. You owe it to yourself.. None of this is esoteric or hard to grasp - once understood it's simple.

It's so very very important - if you believe the digital picture / monitor analogy you cannot begin know just how important :-)


I am really trying to help here :-)

I never went to any college and I have absolutely no diplomas...
Paul - please reread my post!

I'm making a case for how digital photo descriptors can be MISUSED to describe digital audio.

I'm trying to illustrate how the spurious line of reason might easily develop in the mind of the audio engineer by following the dichotomy of the thought process!

I'll make it easy: DIGITAL PICTURE TERMINOLOGY SHOULD NOT BE CARELESS APPLIED TO DIGITAL AUDIO AS THEY ARE NOT OFTEN IN CONGRUENCE.

Old 17th October 2006
  #83
Quote:
Originally Posted by Dave Derr View Post
........
By the way, I think the "1 bit technology" nomer is really misleading. The equivilent output is NOT 1 BIT quality! It is mathematically molded into some semi-equivelent higher bit rate stream, IE 16 - 24 bit equivilent. Delta Sigma AD converters which are pretty much the main technology these days, sample at really high speeds, and then mathematically filter and convert the high speed, low bit rate data to lower speed, higher bit rate data. Its kind of a conversion of "quantity vs quality", with the mathematical ability to take a large quantity of lower res data, and turn it into much higher quality but slower rate data. Delta Sigma DA converters kind of work the opposite direction and put out a higher sample rate, interpolated version of the data. This allows you to use less steep analog filtering and is a blessing to designers. I was lucky to be able to design in some of the earliest Delta Sigma Converters back around 1990 at Eventide. There is some question as to whether linear phase digital filters are pleasing... but thats another discussion.
I think you're adressing my point so..

I understand what you're saying, and this is obviously how this is done most often now. However, did you see the new Tascam and especially Korg products? It might be that soon DAWs will process DSD signals. The Korg software seems to allow storage of DSD recordings, on a regular computer, the tascam has a harddisk now, and Pyramix seems to be able to process DSD. (For processing in a Logic or Protools DAW still conversion to PCM, in case of the Korg, or Tascam, but I think it is a step in the right direction. Still happy with what I'm working with though. heh)
https://www.gearslutz.com/board/remote-possibilities-in-acoustic-music-and-location-recording/89381-new-korg-portable-dsd-wdsd-recorders.html
https://www.gearslutz.com/board/product-alerts-older-than-2-months/89510-tascam-dv-ra1000hd.html

Without going into a full scale debate about DSD vs PCM, which has been done here before, and going OT, I would like to hear you guys' ideas about DSD, considering hot vs. moderate signal levels. It would seem to me the same idea is valid, or isn't it?
Old 17th October 2006
  #84
Lives for gear
 
norman_nomad's Avatar
Quote:
Originally Posted by Dave Derr View Post
LOL

The thing that kind of started me on this was someones assertion that to record with PEAKS 30dB below full scale going into an ADC was ok, or even preferable. As far as the actual conversion process.... THIS IS ABSOLUTELY FALSE! I mean sheesh, if your peaks are -30dB down, where is your average level? -50dB down? Paul has never encouraged putting really low levels into ADCS, and his many papers are clear on this.
Then again, I'll ask. How do you interface outboard with this paradigm?

I'll, for example, record my blaring Marshall at -2dbfs peak -8dbfs average (not a lot of dynamics in the blaring Marshall).

Now I want to stem it out to Fatso. According to you I want to maximize my D/A resolution by pushing to as close to -0dbfs as possible, so I raise the fader in my DAW up 1.9db and now my peak output will be -0.1dbfs going into the Fatso.

Most analog likes to see +4dBu, so how do I accommodate this... stick a pad on my output?

If I lower the fader in my DAW to output a lower bit depth file, won't I be defeating the purpose of recording hot in the first place?

What do you suggest?

Old 17th October 2006
  #85
The Distressor's "daddy"
 
Dave Derr's Avatar
 

Norman Nomad: "Now I want to stem it out to Fatso. According to you I want to maximize my D/A resolution by pushing to as close to -0dbfs as possible, so I raise the fader in my DAW up 1.9db and now my peak output will be -0.1dbfs going into the Fatso."

D/As are by far more forgiving in many aspects, and you probably cant output full scale from a D/A into lots of gear! That is unless you use a pad or something, as you mentioned. Most any Daw these days does a good job of rounding off and dithering an attenuated digital signal that DACs use. So I agree with you totally. ALso, remember that if are attenuating 12 dB going into a 20bit DAC, you still have friggin great 18bit quality coming out, all things being equal. If youre going into a 24 bit DAC, and are attenuating 20dB, you still have better than 20 bit resolution. This is better than most analog quality! And dont forget, you are dealing with an already recorded source and theoretically can always go back and re-stage your gain. But try telling a singer he has to re-sing a part because you recorded his vocal peaks 60dB down and its gnarly sounding. Once you recorded a track well, chances are anything else can be fixed.

My main issue was getting the most out of your initial AD conversions, and clarifying that the AD itself is not particularly happier with low level inputs. I personally have mixed digital tracks that were recorded 20 - 30 dB down on 16 bit, and they were noticeably inferior. On one track in particular, there was noise and grit problems, especially after it was brought up with compression.
Old 17th October 2006
  #86
Deleted bd1be4f
Guest
Quote:
Originally Posted by Dave Derr View Post
LOL

Eek Zboy, Where did this come from? I have tremendous respect for Paul and havent really seen us at odds... except I tend to use the word "resolution" like many others to characterize dynamic range and linearity.... IE the ability to "resolve" teentsy details.
Unless I've read the contents of the thread wrong, I got the impression that you and Paul are at odds with respect to your assertions regarding dynamic range and bit "usage".

Quote:
My assertions which mostly concern getting maximum quality from ADCs, are "Digital 101." To get the full dynamic range and detail advertised from any Analog to Digital converter, you must use the full dynamic range.... right up to 0dbfs. Again, this is simple, and irrefutable. Its also the reason we go for higher "definition" converters, so that we dont have to get so scarily close to going "OVER", and therfore, leaving some headroom below 0dbfs will not be damaging or even noticeable.... especially when it all ends up on 16bit CDs.

What you are recording, or what you and your equipment do after the conversion is not whats at issue. Its mainly just how A to D conversion works that I am discussing. "MIXING IN THE BOX" PROBLEMS ARE COMPLETELY IRRELEVANT TO MY ASSERTIONS. (OF COURSE YOU HAVE TO LEAVE HEADROOM. SHEESH!)
So then your point is simply with respect to limitations of converter design and NOT any actual sonic degradation of lower signals in digital in general? In other words, let's say you bounce an already digitized signal to another track in your DAW at a much lower volume, say even 30db less. Do you agree that the resultant new file is not of a lower quality sonically, even though it is now much lower in level?

With respect to the mixing in the box comments, I wasn't implying that it was part of your assertions, I was simply including it as an aside, and to thank Paul for his wisdom which has helped me immensely.

Quote:
The thing that kind of started me on this was someones assertion that to record with PEAKS 30dB below full scale going into an ADC was ok, or even preferable. As far as the actual conversion process.... THIS IS ABSOLUTELY FALSE! I mean sheesh, if your peaks are -30dB down, where is your average level? -50dB down? Paul has never encouraged putting really low levels into ADCS, and his many papers are clear on this.
This being the case, again you are speaking more about problems or limitations of converter design, rather than strictly a problem with signal level with respect to digital in general, no?
Old 17th October 2006
  #87
Lives for gear
 
norman_nomad's Avatar
Quote:
Originally Posted by zboy2854 View Post

This being the case, again you are speaking more about problems or limitations of converter design, rather than strictly a problem with signal level with respect to digital in general, no?
My point exactly... and to quote myself

Quote:
Originally Posted by norman_nomad
If what you're trying to say is that an ADC will perform better near 0dbfs because it specs better (distortion, s/n, frequency reponse, etc), then I'm willing to agree with the assertion. If what you're trying to say is that an ADC will sound better near 0dbfs simply because you're utilizing more bits then I don't agree.
Old 17th October 2006
  #88
Lives for gear
 
macr0w's Avatar
 

Hey Dave, I EFFIN love my FATSO.
Old 18th October 2006
  #89
Lives for gear
 

Quote:
Originally Posted by norman_nomad View Post
Paul - please reread my post!

I'm making a case for how digital photo descriptors can be MISUSED to describe digital audio.

I'm trying to illustrate how the spurious line of reason might easily develop in the mind of the audio engineer by following the dichotomy of the thought process!

I'll make it easy: DIGITAL PICTURE TERMINOLOGY SHOULD NOT BE CARELESS APPLIED TO DIGITAL AUDIO AS THEY ARE NOT OFTEN IN CONGRUENCE.
I didn't mis-read your post and I'm sorry if in your haste you thought that I had - please don't shout at me :-( And I completely understand and agree with the content of your other posts - before 'resolution and pixels' got thrown into the subject :-(

This is the passage I was refering to:

"Digital picture "granularity" as measured in dpi (dots per inch) is probably most akin to digital audio sampling frequency as both are a digital reconstruction of continuity and therefore it is probably best (if we need to compare the two) to use digital photo's "resolution" to describe sampling rate rather than bit depth"."

This is also misleading, because although it seems quite logical, the pixels on the screen are fixed and constant and do not go through the reconstruction process..

So whereas on a screen, the spatial resolution is quantised permanently by the number of dots per inch etc. the freq/phase response of the digital audio signal is absolutely not. I.e. (very fortunately) in digital audio we do not suffer from discrete freqs with gaps in between or signal timings which go in steps of the sample clock, with nothing in between etc..

This is another conceptual 'virus' I'm afraid.

The truth is that in a properly designed digital system there is NO level quantisation and there is NO time quantisation in the SIGNAL after it has been decoded :-)

So even ignoring the current debate over level quantisation and distortion - the analogy you made with screen resolution is the main factor that underlies the other big misconception in our industry - that higher sampling rates are required to produce 'high resolution' audio. The broken thinking in this misunderstanding is that more points in time are required to more accurately reproduce audio signals.

Of course this is just as misguided as the current argument that level quantisation steps cause distortion.

I was just trying to head off this discussion as well - before it started to add yet more confusion.

But I see I have failed - as further on the the thread I note we are now (very predictably) arguing about quantisation of freq and phase - as well as level and distortion!!

Please understand (and this is absolutely not an insult) that over the last couple of decades I have had so many scores of discussions on this subject for such a massive chunk of my life, that I know exactly how the discussion will develop and even at what point it will become 'multi-dimensional'. It has all been said before in a multitude of threads..

Sorry - please forgive me:-(

And BTW - please please don't get me on the subject of DSD again! :-)
Old 18th October 2006
  #90
Lives for gear
 
norman_nomad's Avatar
Quote:
Originally Posted by Paul Frindle View Post
I didn't mis-read your post and I'm sorry if in your haste you thought that I had - please don't shout at me :-(

What I am refering to is this quote:

"Digital picture "granularity" as measured in dpi (dots per inch) is probably most akin to digital audio sampling frequency as both are a digital reconstruction of continuity and therefore it is probably best (if we need to compare the two) to use digital photo's "resolution" to describe sampling rate rather than bit depth"."

This is also misleading, because although it seems quite logical, the pixels on the screen are fixed and constant and do not go through the reconstruction process..

So whereas on a screen, the spatial resolution is quantised permanently by the number of dots per inch etc. the freq/phase response of the digital audio signal is absolutely not. I.e. (very fortunately) in digital audio we do not suffer from discrete freqs with gaps in between or signal timings which go in steps of the sample clock, with nothing in between etc..

This is another conceptual 'virus' I'm afraid.

The truth is that in a properly designed digital system there is NO level quantisation and there is NO time quantisation in the SIGNAL after it has been decoded :-)

So even ignoring the current debate over level quantisation and distortion - the analogy you made with screen resolution is the main factor that underlies the other big misconception in our industry - that higher sampling rates are required to produce 'high resolution' audio. The broken thinking in this misunderstanding is that more points in time are required to more accurately reproduce audio signals.

Of course this is just as misguided as the current argument that level quantisation steps cause distortion.

I was just trying to head off this discussion as well - before it started to add yet more confusion.

But I see I have failed - as further on the the thread I note we are now (very predictably) arguing about quantisation of freq and phase - as well as level and distortion!!

Please understand (and this is absolutely not an insult) that over the last couple of decades I have had so many scores of discussions on this subject for such a massive chunk of my life, that I know exactly how the discussion will develop and even at what point it will become 'multi-dimensional'. It has all been said before in a multitude of threads..

Sorry - please forgive me:-(

And BTW - please please don't get me on the subject of DSD again! :-(
Paul,

The info you provide is greatly valuable to me as an eager hobbyist in the field and I have no problem with you clarifying my half-witted analogies! heh

I understand exactly what you're saying as I've done plenty of reading.. and I was trying to speak to your exact point by making an abstract analogy that obviously didn't hit home!

My point about "continuity" was referring to Nyquist's thrust that perfect analog reconstruction can be performed with a sampling rate double that of the highest frequency to be represented.

More sampled steps on a continuous sine wave (aka 96khz sampling) are not necessary for your D/A to represent the original sine wav perfectly, just as more DPI, after a certain threshold, is not necessary to provide for a perfect representation of visual continuity.

So this was my half-witted analogy... ... to say that more isn't necessarily better; it can be just the same as less.

Hope this makes sense!
Post Reply

Welcome to the Gearslutz Pro Audio Community!

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


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

Advanced Search
Similar Threads
Thread
Thread Starter / Forum
Replies
XHipHop / Music Computers
20
Ghost Logic / Gearslutz Secondhand Gear Classifieds
0
benelli / Gearslutz Secondhand Gear Classifieds
0
Dutchmuzik / Gearslutz Secondhand Gear Classifieds
0

Forum Jump
Forum Jump