| Home |
| Search |
| Today's Posts |
|
#41
|
|||
|
|||
"TonyP" wrote
Of course, but have you tried recording at 16/44 and processing at 32/44 or 64/44? I don't know if I have or not. What's Pro Tools' internal processing depth? Or done a double blind listening test of a file recorded at 16/44 and one at 24/44 with the same card? Yeah, and there's not enough difference to care about if that's all I was gonna do with it. The benefit of the longer word really only became apparent to me on sessions with a fair amount of processing going on. "Lots of processing" includes things like mixing a lot of tracks, making any fairly healthy level changes, and/or applying plugs. Of course hard drive space is so cheap these days, it's really a non issue for most people now. Saving heaps of wasted bits doesn't cost you too much as long as the hardware can handle the overhead. I suppose, within reason. At some point the benefits become so small that practical realities like backup space/cost, processing time (when the client is paying by the hour) and plug-in limits become a higher priority than absolute limits of audiophility. -- "It CAN'T be too loud... some of the red lights aren't even on yet!" - Lorin David Schultz in the control room making even bad news sound good (Remove spamblock to reply) |
|
#42
|
|||
|
|||
|
On Thu, 30 Sep 2004 06:11:12 -0700, Ron B wrote:
I'm a newbie to digital recording, but not a newbie to music or electronics. So that being said, I'm looking for a simple DAW setup. I'll be using a laptop to record solo guitar for the most part. I'll either be going firewire or usb, but I digress.... In deciding on gear, since I'll mainly be doing solo direct in guitar work I want good quality audio. That seems fairly easy to do since at best it will be two inputs of stereo. But since CD quality has to be 44.1khz at 16 bits, what do you experienced people hear in terms of quality when recording at 24 bit 96khz? Since you'll have to dither that down to CD quality do you still prefer the higher sampling? Does the dithering process alter the sound quality? Enquiring mind wants to know.....and thanks for the replies...... Here's some rambling on the subject... I have experimented tracking a drum kit at 16/44.1k 24/44.1k and 24/96k. I found a noticable difference between 24/44.1 and 24/96, and could reliably identify which was which. The 24/96 sounded clearer and more open, there is no other way to describe it without hearing it. The overheads (two new 414s at the time) benefited the most. The pres were focusrite red 1, isa 220, and some Amek dual pre/compressor for the overheads. Other instruments that I have heard a difference with are stringed, like Dan Goong, autoharp, mandolin... basicly anything steel strung and bright and twangy. Also, metallic percussion has more life to it, try a triangle at both sample rates.... (and be careful, some crappy pres 'thud' (I don't know a better way to describe it, it's a bit like someone tapping on a piece of foam in time with the triangle, and sounds *worse* at 96k)) I used the Ameks A/D convertors for the overheads and other experiments, and monitored back through my Delta 1010. Doing the same with the 1010s A/D convertors sounded similar, but not quite as nice. When the 24/96 drum multitrack was converted to 24/44.1 it sounded worse than the 24/44.1 recordings. I used the Cubase SX src, which is probably not the best, but it's a hassle to use a different one on a multitrack project. Converting a 96k stereo mixdown of the drums with the Wavelab src sounded ok, no better than the 24/44.1 recording, certainly no worse. So, I do acousticy projects at 24/96, as they have less tracks, and it sounds better to me. I would rather do everything at 24/96, but the lower track count and the way some of my plugins don't work at 96k means I'm stuck with 24/44.1 for a while. If you are doing solo DI'd guitar, along with midi/sample based backing tracks, I would not worry about it, 24/44.1 is fine. Ron B |
|
#43
|
|||
|
|||
|
On Thu, 30 Sep 2004 06:11:12 -0700, Ron B wrote:
I'm a newbie to digital recording, but not a newbie to music or electronics. So that being said, I'm looking for a simple DAW setup. I'll be using a laptop to record solo guitar for the most part. I'll either be going firewire or usb, but I digress.... In deciding on gear, since I'll mainly be doing solo direct in guitar work I want good quality audio. That seems fairly easy to do since at best it will be two inputs of stereo. But since CD quality has to be 44.1khz at 16 bits, what do you experienced people hear in terms of quality when recording at 24 bit 96khz? Since you'll have to dither that down to CD quality do you still prefer the higher sampling? Does the dithering process alter the sound quality? Enquiring mind wants to know.....and thanks for the replies...... Here's some rambling on the subject... I have experimented tracking a drum kit at 16/44.1k 24/44.1k and 24/96k. I found a noticable difference between 24/44.1 and 24/96, and could reliably identify which was which. The 24/96 sounded clearer and more open, there is no other way to describe it without hearing it. The overheads (two new 414s at the time) benefited the most. The pres were focusrite red 1, isa 220, and some Amek dual pre/compressor for the overheads. Other instruments that I have heard a difference with are stringed, like Dan Goong, autoharp, mandolin... basicly anything steel strung and bright and twangy. Also, metallic percussion has more life to it, try a triangle at both sample rates.... (and be careful, some crappy pres 'thud' (I don't know a better way to describe it, it's a bit like someone tapping on a piece of foam in time with the triangle, and sounds *worse* at 96k)) I used the Ameks A/D convertors for the overheads and other experiments, and monitored back through my Delta 1010. Doing the same with the 1010s A/D convertors sounded similar, but not quite as nice. When the 24/96 drum multitrack was converted to 24/44.1 it sounded worse than the 24/44.1 recordings. I used the Cubase SX src, which is probably not the best, but it's a hassle to use a different one on a multitrack project. Converting a 96k stereo mixdown of the drums with the Wavelab src sounded ok, no better than the 24/44.1 recording, certainly no worse. So, I do acousticy projects at 24/96, as they have less tracks, and it sounds better to me. I would rather do everything at 24/96, but the lower track count and the way some of my plugins don't work at 96k means I'm stuck with 24/44.1 for a while. If you are doing solo DI'd guitar, along with midi/sample based backing tracks, I would not worry about it, 24/44.1 is fine. Ron B |
|
#44
|
|||
|
|||
|
First, I'd argue about the concept of sloppy level setting. If/when you
have converters of decent quality to allow for at least 16 bit recording (i.e. -96 dB noise floor) then you have 8 bits of headroom at the max (a theoretical impossibility), but more like 4 bits, but geez, doesn't that equate to 24 dB of headroom, which is the typical amount of a good console's analog output headroom? So what does "sloppy" mean in relation to converters? I'd like to see someone actualy define, in meaningful terms, just what this means. First, it can't mean the same thing to two different people because no one works exactly the same as another with exactly the same equipment. Second, there are no methods to measure a term such as "sloppy". It's all too easy to say someone was sloppy in their signal path setup, but in the worst of situations, sans digital converters, even The Boss' Nebraska became a seminal album on cassette. -- Roger W. Norman SirMusic Studio "TonyP" wrote in message u... "Arny Krueger" wrote in message ... Just to pick a nit, he makes a factual mistake when he specifically says that *all* "24 bit" converters have at least 16 bit resolution. They don't, a leading example being the popular M-Audio AP2496. If you check http://audio.rightmark.org/rus/test/...iophile-2496.h tml you'll see that the unweighted dynamic range in 24/96 mode is less than 95 dB. This is just one example of many 24/96 converters with less than 96 dB unweighted dynamic range. Which is what I've been saying for years when people claim 24 bit gives much more room for sloppy level setting. In the case of the AP2496 there is no real extra headroom only wasted bits. In the case of *MANY* 24 bit cards there is a gain of 1 bit extra resolution, 2 if you're lucky, and 3 only in the case of the very best cards available. However 95dB unweighted is quite adequate for tracking, and better than any analog recorder anyway. TonyP. "Appreachable" is a good word for this group too, with SO many people preaching their twisted versions of the audio gospel :-) |
|
#45
|
|||
|
|||
|
First, I'd argue about the concept of sloppy level setting. If/when you
have converters of decent quality to allow for at least 16 bit recording (i.e. -96 dB noise floor) then you have 8 bits of headroom at the max (a theoretical impossibility), but more like 4 bits, but geez, doesn't that equate to 24 dB of headroom, which is the typical amount of a good console's analog output headroom? So what does "sloppy" mean in relation to converters? I'd like to see someone actualy define, in meaningful terms, just what this means. First, it can't mean the same thing to two different people because no one works exactly the same as another with exactly the same equipment. Second, there are no methods to measure a term such as "sloppy". It's all too easy to say someone was sloppy in their signal path setup, but in the worst of situations, sans digital converters, even The Boss' Nebraska became a seminal album on cassette. -- Roger W. Norman SirMusic Studio "TonyP" wrote in message u... "Arny Krueger" wrote in message ... Just to pick a nit, he makes a factual mistake when he specifically says that *all* "24 bit" converters have at least 16 bit resolution. They don't, a leading example being the popular M-Audio AP2496. If you check http://audio.rightmark.org/rus/test/...iophile-2496.h tml you'll see that the unweighted dynamic range in 24/96 mode is less than 95 dB. This is just one example of many 24/96 converters with less than 96 dB unweighted dynamic range. Which is what I've been saying for years when people claim 24 bit gives much more room for sloppy level setting. In the case of the AP2496 there is no real extra headroom only wasted bits. In the case of *MANY* 24 bit cards there is a gain of 1 bit extra resolution, 2 if you're lucky, and 3 only in the case of the very best cards available. However 95dB unweighted is quite adequate for tracking, and better than any analog recorder anyway. TonyP. "Appreachable" is a good word for this group too, with SO many people preaching their twisted versions of the audio gospel :-) |
|
#46
|
|||
|
|||
|
Glenn Meadows has often said that the quality of the sound of the conversion
is based on the analog input. ****ty input, ****ty output. Marrying good converters to a good analog front end, like perhaps Prism, Mytek or Lavry or a number of others will produce superior results with the same converters that perhaps MOTU uses. When I look at the input from my RME or my MOTU, both have analog in the way. How not? -- Roger W. Norman SirMusic Studio "Ty Ford" wrote in message ... On Sat, 2 Oct 2004 08:20:35 -0400, Scott Dorsey wrote (in article ): TonyP wrote: Which is what I've been saying for years when people claim 24 bit gives much more room for sloppy level setting. In the case of the AP2496 there is no real extra headroom only wasted bits. In the case of *MANY* 24 bit cards there is a gain of 1 bit extra resolution, 2 if you're lucky, and 3 only in the case of the very best cards available. However 95dB unweighted is quite adequate for tracking, and better than any analog recorder anyway. This is not an argument against 24-bit converters. This is an argument against calling many of those devices 24-bit unless they actually have 24 bits of real data coming out. Many of the devices advertised out there as 24-bit, though, have twenty or so actual significant bits, and that's not shabby. They should be called twenty-bit converters. I know my Prism 20-bit box has at least 19 real valid bits and only one doubtful one, which I figure is pretty good. The Lavry stuff is at least in the same league. Some of the 20-bit soundcards don't even have 16 valid bits. --scott And at least as important, how do they actually sound? Regards, Ty Ford -- Ty Ford's equipment reviews, audio samples, rates and other audiocentric stuff are at www.tyford.com |
|
#47
|
|||
|
|||
|
Glenn Meadows has often said that the quality of the sound of the conversion
is based on the analog input. ****ty input, ****ty output. Marrying good converters to a good analog front end, like perhaps Prism, Mytek or Lavry or a number of others will produce superior results with the same converters that perhaps MOTU uses. When I look at the input from my RME or my MOTU, both have analog in the way. How not? -- Roger W. Norman SirMusic Studio "Ty Ford" wrote in message ... On Sat, 2 Oct 2004 08:20:35 -0400, Scott Dorsey wrote (in article ): TonyP wrote: Which is what I've been saying for years when people claim 24 bit gives much more room for sloppy level setting. In the case of the AP2496 there is no real extra headroom only wasted bits. In the case of *MANY* 24 bit cards there is a gain of 1 bit extra resolution, 2 if you're lucky, and 3 only in the case of the very best cards available. However 95dB unweighted is quite adequate for tracking, and better than any analog recorder anyway. This is not an argument against 24-bit converters. This is an argument against calling many of those devices 24-bit unless they actually have 24 bits of real data coming out. Many of the devices advertised out there as 24-bit, though, have twenty or so actual significant bits, and that's not shabby. They should be called twenty-bit converters. I know my Prism 20-bit box has at least 19 real valid bits and only one doubtful one, which I figure is pretty good. The Lavry stuff is at least in the same league. Some of the 20-bit soundcards don't even have 16 valid bits. --scott And at least as important, how do they actually sound? Regards, Ty Ford -- Ty Ford's equipment reviews, audio samples, rates and other audiocentric stuff are at www.tyford.com |
|
#48
|
|||
|
|||
|
You mistake concepts. 24 bits has a theoretical limit due to thermodynamic
noise, 96kHz has an absolute limit based on the ability of humans to hear. Neither one offers a significant advantage simply due to the technical specs. There is NO technology that supplants one's ability to hear and to determine what sounds good to them. If they are high dollar people that still doesn't mean that what they'd use has **** to do with your music or the music you record. Get passed the specs and start making music. Sort out the small stuff when people are loving an buying up all your product. Everything else is fluff. -- Roger W. Norman SirMusic Studio "Ron B" wrote in message m... (Scott Dorsey) wrote in message ... Ron B wrote: I'm a newbie to digital recording, but not a newbie to music or electronics. So that being said, I'm looking for a simple DAW setup. I'll be using a laptop to record solo guitar for the most part. I'll either be going firewire or usb, but I digress.... In deciding on gear, since I'll mainly be doing solo direct in guitar work I want good quality audio. That seems fairly easy to do since at best it will be two inputs of stereo. But since CD quality has to be 44.1khz at 16 bits, what do you experienced people hear in terms of quality when recording at 24 bit 96khz? Since you'll have to dither that down to CD quality do you still prefer the higher sampling? Does the dithering process alter the sound quality? Enquiring mind wants to know.....and thanks for the replies...... For sampling rate, it depends entirely on the converters. Some converters sound worse at 96 ksamp/sec than they do at 44.1 ksamp/sec. Some might sound better. Many folks record at 88.2 ksamp/sec because it's easier to SRC down to 44.1. That becomes a question very specific to the conversion hardware you're running. But the wider word length is always a good idea because it gives you more room to be sloppy about levels. --scott Thanks Scott. So I guess the next logical question is one of recommendation. is an MBox at 48khz tops better than a M-Audio Quattro at 96 khz (both at 24 bits). Anyone with experience with these? I would ask about the Stienberg System|4 but the posts on that say the drivers and firmware are really screwed up, to bad since it comes with Cubase. I also like the OmniStudio (usb) but in most cases would need to buy the software seperately ($$). Ron |
|
#49
|
|||
|
|||
|
You mistake concepts. 24 bits has a theoretical limit due to thermodynamic
noise, 96kHz has an absolute limit based on the ability of humans to hear. Neither one offers a significant advantage simply due to the technical specs. There is NO technology that supplants one's ability to hear and to determine what sounds good to them. If they are high dollar people that still doesn't mean that what they'd use has **** to do with your music or the music you record. Get passed the specs and start making music. Sort out the small stuff when people are loving an buying up all your product. Everything else is fluff. -- Roger W. Norman SirMusic Studio "Ron B" wrote in message m... (Scott Dorsey) wrote in message ... Ron B wrote: I'm a newbie to digital recording, but not a newbie to music or electronics. So that being said, I'm looking for a simple DAW setup. I'll be using a laptop to record solo guitar for the most part. I'll either be going firewire or usb, but I digress.... In deciding on gear, since I'll mainly be doing solo direct in guitar work I want good quality audio. That seems fairly easy to do since at best it will be two inputs of stereo. But since CD quality has to be 44.1khz at 16 bits, what do you experienced people hear in terms of quality when recording at 24 bit 96khz? Since you'll have to dither that down to CD quality do you still prefer the higher sampling? Does the dithering process alter the sound quality? Enquiring mind wants to know.....and thanks for the replies...... For sampling rate, it depends entirely on the converters. Some converters sound worse at 96 ksamp/sec than they do at 44.1 ksamp/sec. Some might sound better. Many folks record at 88.2 ksamp/sec because it's easier to SRC down to 44.1. That becomes a question very specific to the conversion hardware you're running. But the wider word length is always a good idea because it gives you more room to be sloppy about levels. --scott Thanks Scott. So I guess the next logical question is one of recommendation. is an MBox at 48khz tops better than a M-Audio Quattro at 96 khz (both at 24 bits). Anyone with experience with these? I would ask about the Stienberg System|4 but the posts on that say the drivers and firmware are really screwed up, to bad since it comes with Cubase. I also like the OmniStudio (usb) but in most cases would need to buy the software seperately ($$). Ron |
|
#50
|
|||
|
|||
|
Roger W. Norman wrote:
So what does "sloppy" mean in relation to converters? I'd like to see someone actualy define, in meaningful terms, just what this means. First, it can't mean the same thing to two different people because no one works exactly the same as another with exactly the same equipment. Second, there are no methods to measure a term such as "sloppy". It's all too easy to say someone was sloppy in their signal path setup, but in the worst of situations, sans digital converters, even The Boss' Nebraska became a seminal album on cassette. It means you can record a 20 bit signal and have 16 valid bits and also 4 bits of "headroom." You can record so the signal never goes above -24dB on the meter, and still have a full valid 16 bits for release. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
|
#51
|
|||
|
|||
|
Roger W. Norman wrote:
So what does "sloppy" mean in relation to converters? I'd like to see someone actualy define, in meaningful terms, just what this means. First, it can't mean the same thing to two different people because no one works exactly the same as another with exactly the same equipment. Second, there are no methods to measure a term such as "sloppy". It's all too easy to say someone was sloppy in their signal path setup, but in the worst of situations, sans digital converters, even The Boss' Nebraska became a seminal album on cassette. It means you can record a 20 bit signal and have 16 valid bits and also 4 bits of "headroom." You can record so the signal never goes above -24dB on the meter, and still have a full valid 16 bits for release. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
|
#52
|
|||
|
|||
|
"Scott Dorsey" wrote in message ... It means you can record a 20 bit signal and have 16 valid bits and also 4 bits of "headroom." You can record so the signal never goes above -24dB on the meter, and still have a full valid 16 bits for release. But you will only have 4 VALID bits for headroom, if your signal chain has a DNR of 120dB unweighted. ***VERY*** unlikely. If you use 4 INVALID bits for your signal (by throwing away 4 valid bits as headroom) then you only have 12 valid bits left for release. Some people here seem to think they can just use 4 invalid bits for headroom, which is an impossibility. The invalid bits are at the other end. The question remains whether you really need more than 14 valid bits (84dB DNR), or more than 12 dB headroom though. Especially if the final result will be highly compressed and clipped anyway, as is most pop music these days. TonyP. |
|
#53
|
|||
|
|||
|
"Scott Dorsey" wrote in message ... It means you can record a 20 bit signal and have 16 valid bits and also 4 bits of "headroom." You can record so the signal never goes above -24dB on the meter, and still have a full valid 16 bits for release. But you will only have 4 VALID bits for headroom, if your signal chain has a DNR of 120dB unweighted. ***VERY*** unlikely. If you use 4 INVALID bits for your signal (by throwing away 4 valid bits as headroom) then you only have 12 valid bits left for release. Some people here seem to think they can just use 4 invalid bits for headroom, which is an impossibility. The invalid bits are at the other end. The question remains whether you really need more than 14 valid bits (84dB DNR), or more than 12 dB headroom though. Especially if the final result will be highly compressed and clipped anyway, as is most pop music these days. TonyP. |
|
#54
|
|||
|
|||
|
"Carey Carlan" wrote in message
. 194 You've lost me again, Scott. I know mathematically what monotonicity means, but what are you correlating in the converter? Analog to digital? That's a pretty tough test to measure. Depending on the nature of the non-monotonicity, you'll have nonlinear distortion and/or modulation noise. On a really bad day the lack of monotonicity will be in excess of the quantization noise that the dither is supposed to handle, and you'll have dither failture. |
|
#55
|
|||
|
|||
|
"Carey Carlan" wrote in message
. 194 You've lost me again, Scott. I know mathematically what monotonicity means, but what are you correlating in the converter? Analog to digital? That's a pretty tough test to measure. Depending on the nature of the non-monotonicity, you'll have nonlinear distortion and/or modulation noise. On a really bad day the lack of monotonicity will be in excess of the quantization noise that the dither is supposed to handle, and you'll have dither failture. |
|
#56
|
|||
|
|||
|
TonyP wrote:
"Scott Dorsey" wrote in message ... It means you can record a 20 bit signal and have 16 valid bits and also 4 bits of "headroom." You can record so the signal never goes above -24dB on the meter, and still have a full valid 16 bits for release. But you will only have 4 VALID bits for headroom, if your signal chain has a DNR of 120dB unweighted. ***VERY*** unlikely. If you use 4 INVALID bits for your signal (by throwing away 4 valid bits as headroom) then you only have 12 valid bits left for release. Depends on whether there is any correlated information below the noise floor. On some systems, where the noise floor is mostly from the analogue side, there may be plenty of useful stuff there. On other systems there might not be. Some people here seem to think they can just use 4 invalid bits for headroom, which is an impossibility. The invalid bits are at the other end. How invalid are they and what's on them? That's the question. The question remains whether you really need more than 14 valid bits (84dB DNR), or more than 12 dB headroom though. Especially if the final result will be highly compressed and clipped anyway, as is most pop music these days. I dunno, I have heard some 14-bit audio that sounded pretty good, and I have heard some 24-bit audio that sounded really bad too. But to my mind, having additional dynamic range isn't a bad thing. Thinking you have additional dynamic range when you really don't is. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
|
#57
|
|||
|
|||
|
TonyP wrote:
"Scott Dorsey" wrote in message ... It means you can record a 20 bit signal and have 16 valid bits and also 4 bits of "headroom." You can record so the signal never goes above -24dB on the meter, and still have a full valid 16 bits for release. But you will only have 4 VALID bits for headroom, if your signal chain has a DNR of 120dB unweighted. ***VERY*** unlikely. If you use 4 INVALID bits for your signal (by throwing away 4 valid bits as headroom) then you only have 12 valid bits left for release. Depends on whether there is any correlated information below the noise floor. On some systems, where the noise floor is mostly from the analogue side, there may be plenty of useful stuff there. On other systems there might not be. Some people here seem to think they can just use 4 invalid bits for headroom, which is an impossibility. The invalid bits are at the other end. How invalid are they and what's on them? That's the question. The question remains whether you really need more than 14 valid bits (84dB DNR), or more than 12 dB headroom though. Especially if the final result will be highly compressed and clipped anyway, as is most pop music these days. I dunno, I have heard some 14-bit audio that sounded pretty good, and I have heard some 24-bit audio that sounded really bad too. But to my mind, having additional dynamic range isn't a bad thing. Thinking you have additional dynamic range when you really don't is. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
|
#58
|
|||
|
|||
|
"Scott Dorsey" wrote in message ... TonyP wrote: "Scott Dorsey" wrote in message ... It means you can record a 20 bit signal and have 16 valid bits and also 4 bits of "headroom." You can record so the signal never goes above -24dB on the meter, and still have a full valid 16 bits for release. But you will only have 4 VALID bits for headroom, if your signal chain has a DNR of 120dB unweighted. ***VERY*** unlikely. If you use 4 INVALID bits for your signal (by throwing away 4 valid bits as headroom) then you only have 12 valid bits left for release. Depends on whether there is any correlated information below the noise floor. On some systems, where the noise floor is mostly from the analogue side, there may be plenty of useful stuff there. I'm not sure what you mean here, can you tell me what you can get below the wide band noise floor on analog that you cannot also get on Properly Dithered digital? Wide band and narrow band noise floors are not unique to any system. Some people here seem to think they can just use 4 invalid bits for headroom, which is an impossibility. The invalid bits are at the other end. How invalid are they and what's on them? That's the question. Noise. If they have truly useful signal information, they aren't invalid! I dunno, I have heard some 14-bit audio that sounded pretty good, and I have heard some 24-bit audio that sounded really bad too. But to my mind, having additional dynamic range isn't a bad thing. Thinking you have additional dynamic range when you really don't is. Exactly my point to begin with. So many people here think they have heaps of head room just because the file size is 24 bits. TonyP. |
|
#59
|
|||
|
|||
|
"Scott Dorsey" wrote in message ... TonyP wrote: "Scott Dorsey" wrote in message ... It means you can record a 20 bit signal and have 16 valid bits and also 4 bits of "headroom." You can record so the signal never goes above -24dB on the meter, and still have a full valid 16 bits for release. But you will only have 4 VALID bits for headroom, if your signal chain has a DNR of 120dB unweighted. ***VERY*** unlikely. If you use 4 INVALID bits for your signal (by throwing away 4 valid bits as headroom) then you only have 12 valid bits left for release. Depends on whether there is any correlated information below the noise floor. On some systems, where the noise floor is mostly from the analogue side, there may be plenty of useful stuff there. I'm not sure what you mean here, can you tell me what you can get below the wide band noise floor on analog that you cannot also get on Properly Dithered digital? Wide band and narrow band noise floors are not unique to any system. Some people here seem to think they can just use 4 invalid bits for headroom, which is an impossibility. The invalid bits are at the other end. How invalid are they and what's on them? That's the question. Noise. If they have truly useful signal information, they aren't invalid! I dunno, I have heard some 14-bit audio that sounded pretty good, and I have heard some 24-bit audio that sounded really bad too. But to my mind, having additional dynamic range isn't a bad thing. Thinking you have additional dynamic range when you really don't is. Exactly my point to begin with. So many people here think they have heaps of head room just because the file size is 24 bits. TonyP. |
|
#60
|
|||
|
|||
|
TonyP wrote:
Kludge writes: Depends on whether there is any correlated information below the noise floor. On some systems, where the noise floor is mostly from the analogue side, there may be plenty of useful stuff there. I'm not sure what you mean here, can you tell me what you can get below the wide band noise floor on analog that you cannot also get on Properly Dithered digital? Wide band and narrow band noise floors are not unique to any system. Nothing. It's the same process, and with both analogue and digital there can be useful information below the broadband noise floor. The problem with digital systems is that they often have nonlinearity issues above the noise floor, and the nonlinearity at low level becomes significant before the noise does. The point I was making is that many converter units out there have very poor analogue front ends on them, and that a lot of the noise on inexpensive converters is from the analogue front end rather than from the ladder itself. With these systems, there is apt to be a high noise floor, but there might well be usable information below the noise floor. Some people here seem to think they can just use 4 invalid bits for headroom, which is an impossibility. The invalid bits are at the other end. How invalid are they and what's on them? That's the question. Noise. If they have truly useful signal information, they aren't invalid! Okay, let's say you have an early Flying Cow A/D, that has a huge spike at 60 Hz. It's basically a ten bit converter... the bottom ten bits just have 60 Hz noise on them. But let's say you take the output of that and put it into a notch filter and take the 60 Hz noise out... then you have something like fifteen or sixteen valid bits. If the noise is correlated and removable, the data isn't necessarily invalid. Likewise if you have a 60 dB broadband hiss, and you add to it music that is 3 dB below the noise level, you'll still be able to make the words out, because there is useful information still below the noise. In this case, it's the signal that is correlated and not the noise. I dunno, I have heard some 14-bit audio that sounded pretty good, and I have heard some 24-bit audio that sounded really bad too. But to my mind, having additional dynamic range isn't a bad thing. Thinking you have additional dynamic range when you really don't is. Exactly my point to begin with. So many people here think they have heaps of head room just because the file size is 24 bits. That's true. But the split between conversion and storage word length is ANOTHER whole can of worms. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
|
#61
|
|||
|
|||
|
TonyP wrote:
Kludge writes: Depends on whether there is any correlated information below the noise floor. On some systems, where the noise floor is mostly from the analogue side, there may be plenty of useful stuff there. I'm not sure what you mean here, can you tell me what you can get below the wide band noise floor on analog that you cannot also get on Properly Dithered digital? Wide band and narrow band noise floors are not unique to any system. Nothing. It's the same process, and with both analogue and digital there can be useful information below the broadband noise floor. The problem with digital systems is that they often have nonlinearity issues above the noise floor, and the nonlinearity at low level becomes significant before the noise does. The point I was making is that many converter units out there have very poor analogue front ends on them, and that a lot of the noise on inexpensive converters is from the analogue front end rather than from the ladder itself. With these systems, there is apt to be a high noise floor, but there might well be usable information below the noise floor. Some people here seem to think they can just use 4 invalid bits for headroom, which is an impossibility. The invalid bits are at the other end. How invalid are they and what's on them? That's the question. Noise. If they have truly useful signal information, they aren't invalid! Okay, let's say you have an early Flying Cow A/D, that has a huge spike at 60 Hz. It's basically a ten bit converter... the bottom ten bits just have 60 Hz noise on them. But let's say you take the output of that and put it into a notch filter and take the 60 Hz noise out... then you have something like fifteen or sixteen valid bits. If the noise is correlated and removable, the data isn't necessarily invalid. Likewise if you have a 60 dB broadband hiss, and you add to it music that is 3 dB below the noise level, you'll still be able to make the words out, because there is useful information still below the noise. In this case, it's the signal that is correlated and not the noise. I dunno, I have heard some 14-bit audio that sounded pretty good, and I have heard some 24-bit audio that sounded really bad too. But to my mind, having additional dynamic range isn't a bad thing. Thinking you have additional dynamic range when you really don't is. Exactly my point to begin with. So many people here think they have heaps of head room just because the file size is 24 bits. That's true. But the split between conversion and storage word length is ANOTHER whole can of worms. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
|
#62
|
|||
|
|||
|
|
|
#63
|
|||
|
|||
|
|
|
#64
|
|||
|
|||
|
"Scott Dorsey" wrote in message ... TonyP wrote: I'm not sure what you mean here, can you tell me what you can get below the wide band noise floor on analog that you cannot also get on Properly Dithered digital? Wide band and narrow band noise floors are not unique to any system. Nothing. It's the same process, and with both analogue and digital there can be useful information below the broadband noise floor. Exactly. The problem with digital systems is that they often have nonlinearity issues above the noise floor, and the nonlinearity at low level becomes significant before the noise does. I've not seen a digital system in over ten years that was worse than analog systems in this regard. But I'll accept that you may have. The point I was making is that many converter units out there have very poor analogue front ends on them, and that a lot of the noise on inexpensive converters is from the analogue front end rather than from the ladder itself. Yep, it's analog once again that is the problem. With these systems, there is apt to be a high noise floor, but there might well be usable information below the noise floor. As with all systems. If they have truly useful signal information, they aren't invalid! Okay, let's say you have an early Flying Cow A/D, that has a huge spike at 60 Hz. It's basically a ten bit converter... the bottom ten bits just have 60 Hz noise on them. But let's say you take the output of that and put it into a notch filter and take the 60 Hz noise out... then you have something like fifteen or sixteen valid bits. If the noise is correlated and removable, the data isn't necessarily invalid. Agreed, that's what I said. Likewise if you have a 60 dB broadband hiss, and you add to it music that is 3 dB below the noise level, you'll still be able to make the words out, because there is useful information still below the noise. In this case, it's the signal that is correlated and not the noise. Yep with any system. It's all a matter of knowing what the measurements really mean. Exactly my point to begin with. So many people here think they have heaps of head room just because the file size is 24 bits. That's true. But the split between conversion and storage word length is ANOTHER whole can of worms. Actually it was the point I was making all along, including processing word length, which can be, and usually is higher again. TonyP. |
|
#65
|
|||
|
|||
|
|
Linear Mode
