|
|
Post by popmann on Oct 23, 2014 13:06:38 GMT -6
Well, setting aside that both of those guys mix almost exclusively 44.1 dance music....they both mix on HD Accel systems....which was my point. Those are not floating point software mixers. Now, Mick also has a Sony Oxford, which is even less resolution...it's 32bit fixed....to HD's 48bit fixed. I didn't look far enough to see if he does any analog summing or anything, because that's not my point....they're not using floating point native mixers. Hey Pop, sorry man, it's very clear to me that you certainly know your stuff, respect, but im having a hard time wrapping my peanut around what you're saying, how does it fit into all of this? I think i missed something? but i'm very curious to understand this. Can you black n white your point for me 8/ lol To oversimply, there are three kinda of mixers in the world, and for the purpose of my point, I mean the thing that takes channels, set the levels, pans and sums them to stereo. To oversimply--ignore the EQ or compression and such that may be used. The MIXER. Analog. Digital. Software (floating point, processing done by your general purpose CPU) digital. These do not behave the same at the job above, IME--setting levels, panning summing. "ProTools" is not a software mixer. You understand that, right? Yours is....Dave's is not. Dave's is a piece of $30k digital hardware that he controls with his computer. See their own documentation: www.rcc.ryerson.ca/media/TECHNICALWHITEPAPERProTools48-bitMixer.pdfPrior to that (HD circa 2000'ish) no one would mix in a computer. Fillipetti won a Grammy for his digital mix of Hourglass in 97 because it was literally one of the first to EVEN be mixed with a digital mixer and not sound of ass. And those 02Rs combined a 32bit fixed path for the channel with a greater accumulator chip (aka...summing).... The misconception that your ProTools shares anything beyond a UI with a TDM system is literally the reason they have the marketshare they have. I'm not going to digress into a Avid slam....cause that's not the point--just to say your ProTools mixer shares only a UI with Dave's. DaveP does not mix on a software mixer. You can test this yourself with an interface's 56bit mixer. Sum it out there. See if it nulls. I have. It doesn't. |
|
|
|
Post by tonycamphd on Oct 23, 2014 13:10:42 GMT -6
Thanx, does this apply with the Avid NativeHD cards? It seems things have changed no?
|
|
|
|
Post by popmann on Oct 23, 2014 13:13:25 GMT -6
Well, setting aside that both of those guys mix almost exclusively 44.1 dance music....they both mix on HD Accel systems....which was my point. Those are not floating point software mixers. Now, Mick also has a Sony Oxford, which is even less resolution...it's 32bit fixed....to HD's 48bit fixed. I didn't look far enough to see if he does any analog summing or anything, because that's not my point....they're not using floating point native mixers. Not to be argumentative, but based on his bottom profile in this article they say " the vast majority of his work is done 'in the box' and he states , "Also, now that Pro Tools is 32-bit floating-point, and some plug-ins also are 32-floating point, there no longer is that headroom bottleneck as the tracks get fuller and louder. ". I've seen him in other videos talking about how much he likes the automation in Pro Tools 11. Looks like we're going to hear some 32 bit floating point hit songs in the future if what he says is true. Yes we will. Or...you will, maybe, considering you've read deeply into his site. For the record--the summing in HDX is 64bit floating point...and done on the FPGA-not the computer CPU. I've never used it....so, I hold no opinion as to where that falls on the spectrum.... |
|
|
|
Post by popmann on Oct 23, 2014 13:14:24 GMT -6
Thanx, does this apply with the Avid NativeHD cards? It seems things have changed no? Yes. In Avid's world, things have changed. |
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Oct 23, 2014 13:27:06 GMT -6
i find jimwilliams Bob Olhsson exchange very interesting as my plan is to be using my future(as of now lol) Avid HD 16x16 converters as simple I/O's to my console, i would like to get my conversion I/O settings in both directions set as optimal as possible for achieving the best results. Same to me....as of yesterday, my brother and me had a late night talk about restaurating my console for going optional OTB mixing again next year and our technical options/converters/interfaces for the future...all that stuff you contemplate after one guy comes from the 12hrs. dayjob and the other one had 12 hrs. of installing software on a DAW machine, lol.... As far as i understood, you are best off using what the manufacturer tells you at which levels his converters are most transparent. But...the (i admit - a few years older already) ADs i have really don't sound perceivably worse if i hit them harder, as long as i don't hit (digital!) clipping with the hardest transients... In the end - are we forced to measure the specs at different levels with a loop run to get best recording level experimentally? Are we talking the same, if e.g. the metric halo guys recommend traditional -12dB as recording level and jimwilliams says that the analog stage in modern converters is nearly out of the equation i.e. transparent enough to record at higher levels? Which levels are we talking anyway, peak, rms,...is this why these statements seem to contradict each other? |
|
|
|
Post by scumbum on Oct 23, 2014 13:49:55 GMT -6
i find jimwilliams Bob Olhsson exchange very interesting as my plan is to be using my future(as of now lol) Avid HD 16x16 converters as simple I/O's to my console, i would like to get my conversion I/O settings in both directions set as optimal as possible for achieving the best results. Same to me....as of yesterday, my brother and me had a late night talk about restaurating my console for going optional OTB mixing again next year and our technical options/converters/interfaces for the future...all that stuff you contemplate after one guy comes from the 12hrs. dayjob and the other one had 12 hrs. of installing software on a DAW machine, lol.... As far as i understood, you are best off using what the manufacturer tells you at which levels his converters are most transparent. But...the (i admit - a few years older already) ADs i have really don't sound perceivably worse if i hit them harder, as long as i don't hit (digital!) clipping with the hardest transients... In the end - are we forced to measure the specs at different levels with a loop run to get best recording level experimentally? Are we talking the same, if e.g. the metric halo guys recommend traditional -12dB as recording level and jimwilliams says that the analog stage in modern converters is nearly out of the equation i.e. transparent enough to record at higher levels? Which levels are we talking anyway, peak, rms,...is this why these statements seem to contradict each other? Yes their answers seem to contradict , and as much as I'd love to see Jim and bob in speedos , spray on fake tans , body slamming each other in a wrestling ring .........I don't want to instigate an argument . But -1 vs -12 is a big level difference . Maybe it is all dependent on the converter like you said . |
|
|
|
Post by scumbum on Oct 23, 2014 13:54:06 GMT -6
Well, setting aside that both of those guys mix almost exclusively 44.1 dance music....they both mix on HD Accel systems....which was my point. Those are not floating point software mixers. Now, Mick also has a Sony Oxford, which is even less resolution...it's 32bit fixed....to HD's 48bit fixed. I didn't look far enough to see if he does any analog summing or anything, because that's not my point....they're not using floating point native mixers. Not to be argumentative, but based on his bottom profile in this article they say " the vast majority of his work is done 'in the box' and he states , "Also, now that Pro Tools is 32-bit floating-point, and some plug-ins also are 32-floating point, there no longer is that headroom bottleneck as the tracks get fuller and louder. ". I've seen him in other videos talking about how much he likes the automation in Pro Tools 11. Looks like we're going to hear some 32 bit floating point hit songs in the future if what he says is true. What confuses me is that LE was always 32 bit floating point . So that made it superior sounding to older TDM , HD systems ? |
|
|
|
Post by popmann on Oct 23, 2014 14:20:03 GMT -6
In as few words as I can use: floating point mixers sound different than fixed bit. ...in any discussion of subjective things, you might find someone who loves some quality about floating point mixing. But, you will not find anyone north of your Winer "digital is digital" crew who think they sound the same. But, yes--if you are to believe Avid's current marketing, their ten year old LE systems sounded better than the $30k TDM mixers that they USED to claim sounded better....and have been used now for 14 years to mix lots of records. You are correct in that assessment of Avid's marketing line. I believe the above link I provided is not on their site any longer.  It IS, however their white paper....explaining why their 48bit fixed TDM mixer is superior to their past and then current software floating point mixers. It's not surprising they've removed it now. |
|
|
|
Post by popmann on Oct 23, 2014 14:29:12 GMT -6
On converter levels....look, if you aim for what your converters are spec'd at....let's take the Burl's -12 or the RME's -13dbfs....EQUALS.....0db VU....understand that a VU is a poor peak meter. So, shooting for -12 WILL mean there are SOME momentary peaks at -3 or -5....maybe slightly higher, once one does an after the fact analysis. It's not contradictory, because it's not really talking about the same thing. Jim won't say record a distorted guitar (least dynamic range of anything) at -3dbfs RMS....at least I don't THINK he will....and I don't think BobO will swear no track he cuts ever peaks above -12dbfs....
Mud wrestling aside....they're really saying not as different things as it might appear.
|
|
|
|
Post by Bob Olhsson on Oct 23, 2014 16:28:04 GMT -6
I've always thought LE sounded better than a TDM system provided you dithered the output. When I first considered replacing a MIX system with LE I shot out a bunch of stuff and le always sounded as good or better.
|
|
|
|
Post by Bob Olhsson on Oct 23, 2014 16:29:56 GMT -6
I generally aim recordings and mixes at -10 with occasional peaks higher. Too bad mastering clients won't accept those levels!
|
|
|
|
Post by wiz on Oct 23, 2014 17:08:31 GMT -6
I generally aim recordings and mixes at -10 with occasional peaks higher. Too bad mastering clients won't accept those levels! well, you are gonna love me BOB!! ps I sent you a message cheers Wiz |
|
|
|
Post by scumbum on Oct 23, 2014 17:09:22 GMT -6
I've always thought LE sounded better than a TDM system provided you dithered the output. When I first considered replacing a MIX system with LE I shot out a bunch of stuff and le always sounded as good or better. I should have been more specific with my previous question regarding having to use dither , I'm using LE 7.4 . Does your answer still apply that you don't need dither when running out converters with untouched audio , no fader , or gain changes ? I know when bouncing to disk , when your session is 24bit48k and you bounce to disk to a 24bit48k stereo file , you put a 24bit dither on the master fader . So how about just going out the converters using LE , you don't need the 24bit dither ? Thanks for the help , Dither is the one thing that always really confuses me . |
|
|
|
Post by Bob Olhsson on Oct 23, 2014 17:41:06 GMT -6
If it's untouched, no. If there's even a 0.1 dB. gain change, you need to dither.
The cost of not dithering is around ten db. of low level distortion as the bottom bit behaves like a chattering noise gate. Dither makes the cut-off sound like hiss. It is not covering up the distortion, it's actually preventing it.
|
|
|
|
Post by scumbum on Oct 23, 2014 20:01:19 GMT -6
If it's untouched, no. If there's even a 0.1 dB. gain change, you need to dither. The cost of not dithering is around ten db. of low level distortion as the bottom bit behaves like a chattering noise gate. Dither makes the cut-off sound like hiss. It is not covering up the distortion, it's actually preventing it. Great , thanks for the tips , |
|
|
|
Post by jeromemason on Oct 23, 2014 21:07:44 GMT -6
Good thread, I just read the entire thing, I think it's good to do these things because they do matter, and anything the wiser can do to help people that just may not know is great. The polarity discussion is underrated, I always use a pencil tip on the center of the cone and watch how it moves to the kick and the whole kit when I flip the phase on the OH's, snares or kicks, it's not loose change but it does the same thing. It's a habit too, when I go to studios I can't freaking help myself, I always do that, I'm sure the manager or owner is thinking I'm about to jab a hole in it or something but no one has ever said anything to me, just looked at me as if I was crazy or something.
|
|
|
|
Post by popmann on Oct 23, 2014 21:17:01 GMT -6
Keep in mind with that req....if there's a panner on a mono channel, you are changing the volume.
When I sum out to the RME, I have to keep that in mind....the 1.5db difference in pan law between the -4.5 I use to mix in Cubase and the -3db of Totalmix....
|
|
|
|
Post by wiz on Oct 23, 2014 21:39:00 GMT -6
Keep in mind with that req....if there's a panner on a mono channel, you are changing the volume. When I sum out to the RME, I have to keep that in mind....the 1.5db difference in pan law between the -4.5 I use to mix in Cubase and the -3db of Totalmix.... Hey popmann what is your set up..can you describe it or post a picture, you gots me curious... what mixer are you using etc? cheers Wiz |
|
|
|
Post by popmann on Oct 24, 2014 7:02:04 GMT -6
The one referenced for summing is an RME Multiface. I have a cheap Berhinger line mixer for cue feed.
I refuse to "go back to analog" when the problem isn't digital, it's software code. I know from my decade with the DPS24. I've gotten to where I can kinda get "close enough" to the Akai that other aspects pushed it over the edge (double sample rate, distortions of varying flavor, HPF wherever needed)....but, the fundamental sound, just pushing the faders up and panning things out is different in floating point software. The level handling is greater too, in terms of clients sending me 24+ tracks all clipping or near--and that not clipping the master....
If we assume RME has stayed with 56bit RISC chips in their newer interfaces, you can test it yourself. However many "software outs" Totalmix gives you (middle row), you can buss things out there. Click the "loop back" button and you can arm a track in the DAW (assigning it no bus output) and record it. Change the assignments back to stereo....record it from inside the app. Compare. For greatest effect, IME, you need mono feeds--kick, vocal, bass....mainly, kick and bass. Thus reminding anyone trying to make sure you do any pan law offsets...and reminding Scum that there's a gain stage/change being done whether he touches the fader or not.
I've known for years about this....I keep trying new versions of software thinking maybe it was a kind of "wink wink....when CPUs get fast enough, we'll implement the better mixer code"...but, year after year....version after version, they didn't. Many just went analog summing. That's fine--that will fix it, really at ANY level...my $99 cue mixer will "give better center channel" than software. Good luck calibrating that little POS for recalls....but, I point out to say it's easy/cheap to solve the problem. But, to really SOLVE the problem! one has to admit that digital isn't the flawed tech you're band aiding with a string of resistors and makeup amp....digital can do better....I'm amazed that no one with Apollos are doing this--that should be 56bit SHARC code AND they can insert plugs on the busses!
|
|
|
|
Post by Bob Olhsson on Oct 24, 2014 7:20:03 GMT -6
The problem is that back in early 8 bit MIDI sampler days dither was made optional because the truncation distortion sounded better than the hiss. This myth of dither legitimately being optional persisted and way too many "developers" who took a little Unix in college and only know enough to patch code libraries together still believe this nonsense. We shouldn't ever need to think about it but we are at the mercy of well-intentioned people who really know nothing about DSP. There are plug-ins that read bit depth and will tell you if numbers are being crunched or not.
|
|
|
|
Post by matt on Oct 24, 2014 8:15:11 GMT -6
I am building a mix for a new song right now, and while I am aware of recommended levels such as -12dbfs, I always make everything a little crunchy and then end up backing off a bit. Red - bad, green - good. I always have to fight gain-creep and the "make everything louder than everything else" thing. I probably record hotter than I need to, as well, always fighting my preconception that it must be loud, or it won't Rock. I do try and avoid clipping, though. Mostly. I sure could use a year or two as an assistant in a well-staffed studio built around PT. And an SSL or Neve board, while I'm wishing.
I have much to learn, thanks to everybody for the comments. Very useful stuff.
|
|
|
|
Post by swurveman on Oct 24, 2014 9:39:39 GMT -6
After some reading, I found this from Avid concerning why they moved from 24-bit processing/48-bit fixed summing to 32-bit floating-point processing/64-bit floating-point summing Q: Why did Avid switch from 24-bit processing/48-bit fixed summing to hardware that offers 32-bit floating-point processing/64-bit floating-point summing, and what does it mean for me? A: There are several reasons for the transition. A 64-bit floating-point mix bus provides more than 1,000 dB of headroom, which is more than enough to handle the huge track counts that Pro Tools|HDX can deliver. Also, by moving the insert paths to a 32-bit floating-point format, Pro Tools|HDX offers much more dynamic range for plug-in processing, making it nearly impossible to clip the plug-ins, while also being able to handle greater than 24-bit audio file formats. Many of you have told us that even with dual-precision plug-ins processing at 48-bit, the path between inserts on Pro Tools|HD was still limited to 24-bit — thus, limiting any gains. With Pro Tools|HDX, all data streams are maintained at full 32-bit floating point and then summed in the DSP mixer at 64 bits. Our beta customers told us that they could hear and appreciate the difference. |
|
|
|
Post by svart on Oct 24, 2014 10:05:09 GMT -6
I don't really have a pony in this race, and I haven't read through it all, but I can tell you that it's all about bit depth and how it changes through the system.
It's all about truncation. If you come in at 24 bits, go up to 32 bits, you don't just gain more signal, you have the same signal, just with extra empty bits that do nothing.
Now put that into a plugin/fader process that does a lot of math (multiplication and division) and your resulting number is now TOO BIG to be represented by 32 bits, so it must be cut down to size.
It must be truncated.
Truncation can be done by any number of processes. You can simply chop off the extra bits, or you can round them up/down, or you can apply complex algorithms to them.
There are lots of ways to truncate, but each has it's pros and cons.
Some are fast but imprecise while some are extremely slow but super precise. It's all about the tradeoffs and how much CPU power you can use to do the math.
Now, here's the noodle breaking part..
So in the example above, you just added a bunch of bits that do nothing to your digitized signal. Now, if you multiply that word by another word that has had the same thing happen, you now have an even more watered down mathematical output. And then you truncate somehow, further eating away at your original signal.
Imagine taking a gallon of red water and mixing it with a gallon of clear water. Now you have two gallons of pink water. You need one gallon, so you pour half of the pink water out and take your gallon of pink water.
Same idea.
So each and every process you do ITB will eat some of your signal away, with some doing it more than others.
So while I don't know anything about PT faders and their code, I can say that if someone hears a big difference, then it's probably some bad coding choices. Or just bad design/application choices.
OR
The other thought is that the master fader needs to do math to sum ALL of those other channels, so it's resulting math output would be HUGE.. But it must also be fast enough that the user doesn't feel the latency. So there MUST be heavy and fast truncation happening (remember when I said fast but imprecise?).. Offline bounces can have much higher precision because they aren't expected to be "real time" and can be much slower.
So yeah, I see how someone can hear a difference.
|
|
|
|
Post by tonycamphd on Oct 24, 2014 10:27:39 GMT -6
good post svart , what happens if you start your session as a 32bit float? I'm doing all my stuff at 96/32 float in PT10 currently(32 for no other reason than "it goes to 11"), i'm planning on very little happenings ITB, in the meantime i'm trying to wrap my head around all this stuff, really great having Pop, JW, Bob O, and other guys like yourself around here with your big ole noggins! lol Thanx for sharing this stuff fella's! |
|
|
|
Post by svart on Oct 24, 2014 11:38:14 GMT -6
good post svart , what happens if you start your session as a 32bit float? I'm doing all my stuff at 96/32 float in PT10 currently(32 for no other reason than "it goes to 11"), i'm planning on very little happenings ITB, in the meantime i'm trying to wrap my head around all this stuff, really great having Pop, JW, Bob O, and other guys like yourself around here with your big ole noggins! lol Thanx for sharing this stuff fella's! Not really sure. I'm not much of a digital engineer, but I do know the basics. Most of what I know I've picked up from the digital guys I've worked with over the years. Since what I do generally breaks down into high speed sampling, I know that a lot of the same concepts are shared between RF and audio, at least in the digital domain. The digital guys are always grumbling about figuring out the sweet spots in truncations and stuff. Sometimes it's handled in software, and sometimes we do DSP in hardware(FPGA) to take care of operations that need super high speed computation. It's always a balancing act. |
|
It IS, however their white paper....explaining why their 48bit fixed TDM mixer is superior to their past and then current software floating point mixers. It's not surprising they've removed it now. 
