N'th try: Pro Tools sound quality A.K.A

[[Benjamin]]

In the opinion of many, the Protools mixer, although perfectly functional, isn't optimal.

In my humble opinion:
<font size="2" face="Verdana, Arial">I've pointed out completely valid reasons a numer of times, showing why the mixers A/B listening tests* cannot be used as grounds for any decision regarding not making a dithered mixer or not trying alternative routes in mixer design. Lack of dithering might not be the only issue with the mixer design. The double precision maths sounds a little Voodoo to me, ARE YOU SURE IT'S THE RIGHT WAY TO GO? <font size="2" face="Verdana, Arial">Many of us will strongly argue that Protools in general sounds better at lower levels (recording, waveforms, playback, etc - whichever is the real culprit). Shouldn't it really be the other way around? WHY IS THIS?<font size="2" face="Verdana, Arial">Now - Dear Fellow Board Members - We've discussed theese things many a time before. Maybe this time if we try to stick to the topic? images/icons/smile.gif
* The mixer test: "Pro Tools Sound Quality???"
http://duc.d igidesign.com/cgi-bin/ubbcgi/ultimatebb.cgi?ubb=get_topic;f=2;t=002348 (http://duc.digidesign.com/cgi-bin/ubbcgi/ultimatebb.cgi?ubb=get_topic;f=2;t=002348)
and "Pro Tools Sound Quality (continued)???"
http://duc.d igidesign.com/cgi-bin/ubbcgi/ultimatebb.cgi?ubb=get_topic;f=2;t=002612 (http://duc.digidesign.com/cgi-bin/ubbcgi/ultimatebb.cgi?ubb=get_topic;f=2;t=002612)

footnote: Since the mixer issue recently reared it's ugly head again in the thread "WHY?"
http://duc.d igidesign.com/cgi-bin/ubbcgi/ultimatebb.cgi?ubb=get_topic;f=16;t=007885 (http://duc.digidesign.com/cgi-bin/ubbcgi/ultimatebb.cgi?ubb=get_topic;f=2;t=007885)

I wanted to focus on two aspects that wasn't really part of the original discussion in "WHY?"and bring them to discussion in their own right.

[Dave Lebolt]

For a Digidesign response, please see my reply on this topic back on the "Why" thread (as it relates to many issues brought up directly in that area of the forum).

I’d also like to answer the questions you bring up here:

- As to double-precision math being “voodoo,” I don’t think there’s any reason to believe this. This term refers to he fact that the developer/algorithm designer is using 48 bits (in the case of a 24-bit system) to represent the audio waveform. In our mixing system, we preserve the 48 bit word up and until you mix to the final output. At that point (when leaving the system to a DAC or digital output), the mix results are finally truncated or dithered directly to 24 bits.(Dithering is accomplished by the user with a post-fader insert dither Plug-In of their choice.)

Preserving the 48 bit result in this way allows channel faders to be moved downward without losing any of the resolution in the original signal (all 24 bits are preserved). It also yields more headroom at the mix bus (in our case, around 30 dB).

- As to the belief that Pro Tools sounds better at lower levels, I’m not sure why this would be. Proper attention to gain (and avoiding clipping on the master bus or at any point in between) would be crucial to get pristine mixing results, of course. If you believe that you hear these differences, I’d recommend performing a bounce test between two mixed example files (one higher and one lower in level). Compare the results by lining them up and inverting polarity to see if there are any audible differences at any normal listening level (see steps for this in my post on the “Why” thread if you wish). At any reasonable level (i.e., not down in the 24th bit or something ;-), the results should be identical.

[blairl]

Originally posted by Dave Lebolt:
In our mixing system, we preserve the 48 bit word up and until you mix to the final output. At that point (when leaving the system to a DAC or digital output), the mix results are finally truncated or dithered directly to 24 bits.(Dithering is accomplished by the user with a post-fader insert dither Plug-In of their choice.)

In my opinion there is absolutely nothing wrong with using double precision math. This is a good thing.

One question Dave. I'm not doubting the audibility of differences question, I just have a technical question on the above statement. You said that 48 bits is maintained throughout the mix process. I am assuming that you are referring only to the gain of the mixer since any plug-in or send will only accept a 24 bit input? Also with the 24 bit mixer only 26 channels of any kind can be used per DSP chip. After 26 channels two new mixers are created. One submixer and another mixer for the next 26 channels. Here's the question. When new mixer chips are used are the 48 bits maintained between the original mixer and the submixer or is it necessary to truncate or round the output of one chip to the submixer chip?

You mentioned that 48 bits is maintained until the DAC or digital output. You then mentioned that at this point the 48 bits are either truncated or dithered to 24 bits depending on whether or not the user inserts a post-fader dither plug-in. As far is I know there are no dither plug-in's that dither from 48 to 24 bits. Are you saying that if a master fader is created, these post-fader inserts maintain the 48 bit word length? I had always understood that any use of inserts forced the word length to be truncated to 24 bits. Is there an acception to this rule when using a master fader?

Back to the 48 to 24 bit dithering. Again I am not aware of any dither plug-in that dithers from 48 to 24 bits. Please correct me if I am wrong. The only case I know of is when Mixer B is used where all outputs are dithered from 48 to 24 bits. When I say outputs I mean anytime a post fader signal passes through the TDM bus or any final output. This leads me to the next question. I had read somewhere that Mixer B no longer works with version 5.1. Is this true? I just installed 5.1 last week and then went on vacation for a week so I am not in the studio to be able to verify this. If Mixer B no longer works with 5.1 would digidesign consider making it work? I know there are questions as to whether or not Mixer B had any audible difference but the specs sure look good when using a dithering mixer. For example your standard 24 bit mixer has a signal to ERROR ratio of -144 dbfs while your dithering mixer had a signal to ERROR ratio of -288 dbfs. Hey, this is even better than the Oxford's theoretical signal to ERROR ratio of -192 dbfs. Of course this is all a "numbers" game. Unfortunately many potential studio clients are caught up in this numbers game to the point that they are listening to misinformation being spouted off by salespeople and end users of other systems. By using the dithering mixer Pro Tools at least wins one numbers game hands down. ANYWAY........

[PeeTee]

"Back to the 48 to 24 bit dithering. Again I am not aware of any dither plug-in that dithers from 48 to 24 bits."

Waves L1+ Ultramaximizer dithers to 24-bit. From the L1 .pdf manual:

"New in the L1 is 48-bit (double precision)processing,offering improvements for all resolutions,and the opportunity to _dither to a 24-bit output_ for archives and DVD.The L1 offers superb requantization for all bitdepths,including 24,20 ,16,12,and 8-bit outputs."

I'm quite positive that any double-precision 48-bit plug-in dithers it's output to 24-bits. (All the Waves Renaissance plugs, McDSP etc)

[Dave Lebolt]

Gee, Blair, you sure do ask a lot of questions ;-). Folks, I’ll answer this round, but I’ve got to get on to other things (a lot to do around here):

- Re your questions about submixing, this is covered in detail already in Frederick Umminger’s post (from Digi engineering) on page 3 of the “Why?” thread.

- You mentioned 26 inputs as the maximum number mixed per DSP. This is the number for older DSP Farms (the “classic” Farms). The MIX cards allow 59 x 2 to be mixed together without any submixing.

- Your statements about dithering on the master output are correct. But when dither is applied to the 24 bit result, I believe the point is moot. The “dancing of the LSBs” that dither creates provides the same basic benefits you would get regardless, and all of this is happening (as I’ve said numerous times already), in an area that is down far below the thresholds of audibility. As we’ve said before, the audibility tests between the special “dithered mixer” that we created for DUC users to check out were inconclusive.

- Blair, you your post you mention “signal to ERROR” ratio. Folks, lest we get confused or carried away by this term or what it represents (I know I was ;-), the term “signal to error ratio” (no caps necessary) is used by DSP engineers to describe the quantization error common to all digital signal processing (not anything specific to Pro Tools). It’s pretty much like signal to noise ratio. You can see from the numbers mentioned that we’re starting down –144 dB. That’s a long, long way below the threshold of audibility as we understand it.

- There is not reason that I know of for the special mixer that we posted (and continue to provide) to stop working with v5.1x. Someone will check into this, and if there’s a problem with it, we’ll fix it and continue to provide it as before on the DUC.

[blairl]

That's great to hear of the new L1+. I was aware that the Waves plug-ins that use double precision math all dithered the output to 24 bits. Colin at McDSP said he uses Noise Shaping instead of plain dither. The reason they do this is because the TDM bus will only accept 24 bits, so the math of the plug-in is being dithered back to 24 bits before entering the TDM bus. If the master fader inserts maintain 48 bits then this is a revelation to me. I anxiously await a reply from Dave.

[PeeTee]

Dave, some simple questions:

If truncation is applied on each separate track, does a cumulative degradation to the final mix output signal occur? My understanding is that PT's mixer truncates from one insert to the next (unless you are using a dithering plug-in). Can that lead to a cumulative effect? Also, isn't lots of subsequent dithering bad?

How does truncation effect jitter (and vice versa)? Do they 'enhance' each other?

Also, remember that -144dBFS is the theoretical limit and not the actual limit of a 24-bit system. At -144dBFS, the molecules in your room are louder! We are actually talking about a "real-world" 20-22-bit system here when all paramaters of a DAW are taken into account (jitter, ad/da, monitoring, psu, transformer etc).

[blairl]

I guess Dave was responding at the same time I was writing the last response. Thanks for the input! Yes, I do understand that the signals are way down there and thus my not so serious "numbers game" analogy. Also, I don't mean to scare anyone with the signal to error ratio terminology. As Dave said, this is common to all DSP, not just Pro Tools. I was only pointing out that in this area Pro Tools beats the Oxford in the numbers game when using the dithering mixer. http://duc.digidesign.com/ubb/images/icons/smile.gif

[Sonsey]

Ok! This is the topic of concern that I had in the "Why" thread. While I understand that the mixer is 48bit my question is this... If I take two channels both recorded say 2 db below full scale (in this case a kick and a snare) and I add a Master Fader to the session the clip lights on the Master come on. Now according to my binary math that should be at a maximum of 48 bits. That should be allowable inside the mixer as a "full code word" (i.e all 1's). But I'm still getting the clip lights! So do the clip lights come on at ANY full code reading? And why do I (and others aparrently as Benjamin will attest) find that this degrades the sound quality? Hence we record/playback at lower levels to avoid this. Now I understand that this happens on ANY digital mixing system because you've only got so many bits, I'm just curious as to the hows and whys of when and where this happens on PT to produce the best possible sound.

Howard Sonnenburg
Atomic Productions

[F Umminger]

Sonsey-

I tried to explain this in my post on the "WHY" thread. If you combine two channels, each at -2 dB, then the sum will typically have peaks at 4 dB (= -2 + 6) and an average value of 1 dB (= -2 + 3) and hence will clip. This is more than a full code word which can only represent signals up to 0 dB.

This degrades the sound quality because you are getting hard clipping and thus distortion. That is why the clip lights are there to warn you.

I certainly hope no professional recording engineer would be mixing with the clip lights flashing and then blaming us for the bad sound. Avoiding clipping is absolutely basic to digital recording.

Pee Tee-

Truncation and jitter are completely independent. Nothing the mixer or any other plugin does can affect jitter. This can only be affected by your source of clock.

The mixer does not truncate individual tracks. Only the final mix is truncated, so there is no accumulation of error. If you use the faders on plugins to control gain, as some have suggested, then there _will_ be an accumulation of noise since the noise generated by the plugins will add together. However, this noise is usually going to be dwarfed by the original noise in the tracks.

With a chain of plugins there can be an accumulation of error/noise. It is difficult to say how much since it depends on what the plugin is doing, the quality of the plugin, and the gain-staging you use. With stupid enough gain-staging (lowering the gain 144 dB with one plugin and raising it 144 dB with the next) the error can be arbitrarily bad. With good gain-staging I would expect about 3-6 dB of error/noise to be added by a typical plugin. Since the error will usually be white you could check the noise floor yourself with a spectrum analyzer like SpectraFoo, although you won't be able to see anything until/unless the noise builds up to an amount higher than the original noise floor of your recording.

Anyone who has used a chain of guitar effects pedals will recognize that this kind of accumulation of noise happens with analog gear also http://duc.digidesign.com/ubb/images/icons/smile.gif.


------------------

Frederick Umminger
Digidesign Plugin Engineer

[blairl]

Originally posted by F Umminger:
With a chain of plugins there can be an accumulation of error/noise. It is difficult to say how much since it depends on what the plugin is doing, the quality of the plugin, and the gain-staging you use. With stupid enough gain-staging (lowering the gain 144 dB with one plugin and raising it 144 dB with the next) the error can be arbitrarily bad. With good gain-staging I would expect about 3-6 dB of error/noise to be added by a typical plugin. Since the error will usually be white you could check the noise floor yourself with a spectrum analyzer like SpectraFoo, although you won't be able to see anything until/unless the noise builds up to an amount higher than the original noise floor of your recording.

Anyone who has used a chain of guitar effects pedals will recognize that this kind of accumulation of noise happens with analog gear also http://duc.digidesign.com/ubb/images/icons/smile.gif.




My understanding is that truncation error is not the same as white noise. It contains upper-order harmonic distortion and is therefore very unpleasant, whereas white noise or dither is not objectionable and accepted as a good signal. So truncation = distortion and dither = noise. Thus the two different phrases "sinal to noise" and "signal to error". So unless the plug-in's apply dither at the output, a digital system cannot be compared to analog gear. If Pro Tools is used with Mixer B, the dithering mixer, then the signal to error ratio will start at around -288 dbfs. If the standard 24 bit mixer is used then the signal to error ratio starts out at around -144 dbfs and depending on how many tracks, plug-in's etc., the signal to error ratio can rise to around -126 dbfs or even higher. The contention of some designers is that most advertised 24 bit systems are actually not 24 bit systems at all. In the opinion of some a true 24 bit system needs to maintain a signal to error ratio greater than -144 dbfs with all channels functioning. Mixer B turns Pro Tools into such a system. Now dither noise is around -126 dbfs, but we can hear into the noise floor of dither. Anyway, does any of this matter? Can we hear truncation noise that is so far down there? This is where the argument starts to have trouble...

If anyone from digi engineering is interested, I picked the brain of Paul Frindle who is the main designer of the Sony Oxford. The communications can be found at the following link. Our conversation starts after the first couple of posts.

http://www.musicplayer.com/ubb/Forum3/HTML/000761.html

[This message has been edited by blairl (edited May 25, 2001).]

[F Umminger]

I view the distinction between signal-to-noise and signal-to-error as being somewhat pedantic. In either case we are talking about the difference between what the signal ideally should be and what it actually is. In the analog world that difference will usually be a white noise hiss. In the digital world it may sound different (i.e. distortion). In both worlds, it does not matter much what it sounds like unless it reaches an audible level.

Furthermore, if the source audio is broadband or of a reasonably high level in relation to the bit depth then the truncation noise will be essentially white noise. A reference is chapter 2 of "Digital Audio Signal Processing" by Udo Zolzer.

Even when the truncation error is not a smooth hiss it will usually have a white spectrum in real-world situations, and thus you will be able see the level of the noise in a spectrum analyzer like I suggested.

I would suggest that anyone who is seriously worried about or interested in these kinds of issues should spend some time experimenting with a spectrum analyzer, the signal generator plugins, a bit-reducer like Maxim or Lo-Fi, and some time-adjusters to boost gain.

Here is one set of experiments: Set up three signal generators to send three different frequencies to a mono bus, call it bus A. Create two aux tracks with input bus A and output bus B. One one aux track place an instance of Lo Fi and on the other place Time Adjuster. Create another aux track with input bus B and output to your speakers. Put a time adjuster on that. Flip the phase of the 1st time adjuster and confirm that you get complete cancellation with the output of Lo Fi. Now adjust the bit depth in Lo Fi; you then be hearing the resulting truncation error. Use the second time adjuster to boost the level if it is too low to hear. With my current set-up, the signal through Lo Fi is truncated to nothing at a bit depth of 4. For depths 5 through 13 I get varying amounts of distortion and by depth 14 the truncation noise is just a hiss. This is with the signal generators playing a chord (frequencies 300 400 and 500 hz), hiss will result sooner if the frequencies are less correlated. See what happens if you only use two or one signal generators or if you mix in various amounts of white noise (simulating the noise floor in your tracks). Try this with audio from real recorded tracks instead of from signal generators. Take a look at what you are listening to in a spectrum analyzer; when you can hear distortion you should also be seeing spikes in the spectrum and as the distortion turns to hiss you should see the spikes turn to a flat noise floor.



------------------

Frederick Umminger
Digidesign Plugin Engineer

[[Benjamin]]

Thank you Dave,

I've been reading through your earlier "essays" (again) in the other threads dealing with sound quality and math's, and I'll admit that your arguments hold water (you have some patience.. http://duc.digidesign.com/ubb/images/icons/smile.gif -do you curse us in your spare time?). I'm not really an antagonist in this issue, it's just that I believe there's room for improvement.

1) My reference to Voodoo had little or no foundation, what I was trying to say was this:
-Okay, maybe dither or lack thereof isn't what we're hearing, what else could it be?

2) As we all know, "lower levels" can mean just about anything, and if there's an issue somewhere in this regard, it could be just about anywhere, so:
-When I was talking about "the sweeter sound with lower levels", I'm not actually refering to attentuation or mixer operations, I'm refering to recording at lower levels, or possibly gain reducing tracks using AS.
I'll also submit that I've gone back into sessions and re-bounced them with the master fader attentuated 3 additional dB's below any clipping (-3dB peaks in other words), and experienced slight sound quality improvements when comparing results in mastering. (Bouncing without conversion, 24bit/44.1k) I've been paying close attention to this during 6 months and three albums.

3) So from my perspective, headroom/clipping is where I'd look. On occasion I've seen clipping in Peak that wouldn't light up the indicators in Pro Tools. I could verify this in both Peak and Pro Tools examining the waveforms.

4) There's another, more remote, possibility aswell. Thomas Lund and Soren Nielsen at TC in Denmark has shown that distortion can occur at or below 0dBFS. A recent paper on this can downloaded from AES, it's preprint # 5251,from the 109th. It deals primarily with mastering and consumer playback, so it doesn't apply 100%, but since we use AD/DA's with any digital audio it's not an impossible lead.I would also assume that the fenomena described by Thomas Lund and Soren Nielsen could apply to internal conversion, quantisation and filtering processes aswell.


An excerpt from the paper:
"A sine tone at 0dBFS is often believed to be the maximum level
obtainable from
a digital medium. Therefore it is typically the maximum level digital
filters and
analog circuitry in consumer equipment is aimed at reproducing.

As we have showed in previous papers, inter-sample peaks may be
considerably
higher than 0dBFS."

------------------
www.hypersonic.se (http://www.hypersonic.se)

[F Umminger]

Speer-

Frederick, you wrote:
"The mixer does not truncate individual tracks. Only the final mix is truncated, so there is no accumulation of error. If you use the faders on plugins to control gain, as some have suggested, then there _will_ be an accumulation of noise since the noise generated by the plugins will add together."

Are you saying that the faders (and then I assume also the pan pots) now have dither? Since when? In one of the few post Digi ever made on the subject, it was clearly stated that the faders and pan-pots truncate back to 24-bit. What's the deal?<font size="2" face="Verdana, Arial">No, the mixer does not dither. Nothing in what I said implies that it does. What I meant is that the truncation by the mixer occurs after tracks are mixed at full precision, and therefore happens only once no matter how many tracks you mix (until you mix more than 59 and a sub-mixer is created). The truncation noise generated by the mixer has a peak value of -144 dB whether you mix one track or 59.

When you change gain via plugins then you get either truncation or dither noise for each track and this will accumulate.

blairl-

If the standard 24 bit mixer is used then the signal to error ratio starts out at around -144 dbfs and depending on how many tracks, plug-in's etc., the signal to error ratio can rise to around -126 dbfs or even higher.<font size="2" face="Verdana, Arial">This is not true, for the reason I stated above. Or, at least, if the signal-to-error ratio does rise that high it is not due to the mixer but rather due to the errors being fed into the mixer by plugins.

[ca4]

Originally posted by F Umminger:
Speer-

When you change gain via plugins then you get either truncation or dither noise for each track and this will accumulate.



So when someone prefers to mix via plugins what they are actually doing is adding more noise that mask the 48 -> 24 bit error better?

It is becoming more and more apparent to me that ProTools is a very well designed system and that we are discussing if the 48 -> 24 bit truncation is audible or not.

I have also learnt a lot from these discussions (and from blairl:s discussion with Paul Frindle) which has made me more confident not only with ProTools but also with digital audio.

I would be perfectly happy if "Mixer B" could be made to work with ProTools 5.1+ and available for download. I personally think this is audible on some material.


[This message has been edited by ca4 (edited May 26, 2001).]

[[Benjamin]]

Thank you Dave,

I've been reading through your earlier "essays" (again) in the other threads dealing with sound quality and math's, and I'll admit that your arguments hold water (you have some patience.. http://duc.digidesign.com/ubb/images/icons/smile.gif -do you curse us in your spare time?). I'm not really an antagonist in this issue, it's just that I believe there's room for improvement.


1) My reference to Voodoo had little or no foundation, what I was trying to say was this:
-Okay, maybe dither or lack thereof isn't what we're hearing, what else could it be?

2) As we all know, "lower levels" can mean just about anything, and if there's an issue somewhere in this regard, it could be just about anywhere, so:
-When I was talking about "the sweeter sound with lower levels", I'm not actually refering to attentuation or mixer operations, I'm refering to recording at lower levels, or possibly gain reducing tracks using AS.
I'll also submit that I've gone back into sessions and re-bounced them with the master fader attentuated 3 additional dB's below any clipping (-3dB peaks in other words), and experienced slight sound quality improvements when comparing results in mastering. (Bouncing without conversion, 24bit/44.1k) I've been paying close attention to this during 6 months and three albums.

3) So from my perspective, headroom/clipping is where I'd look. On occasion I've seen clipping in Peak that wouldn't light up the indicators in Pro Tools. I could verify this in both Peak and Pro Tools examining the waveforms.

4) There's another, more remote, possibility aswell. Thomas Lund and Soren Nielsen at TC in Denmark has shown that distortion can occur at or below 0dBFS. A recent paper on this can downloaded from AES, it's preprint # 5251,from the 109th. It deals primarily with mastering and consumer playback, so it doesn't apply 100%, but since we use AD/DA's with any digital audio it's not an impossible lead.I would also assume that the fenomena described by Thomas Lund and Soren Nielsen could apply to internal conversion, quantisation and filtering processes aswell.


An excerpt from the paper:
"A sine tone at 0dBFS is often believed to be the maximum level
obtainable from
a digital medium. Therefore it is typically the maximum level digital
filters and
analog circuitry in consumer equipment is aimed at reproducing.

As we have showed in previous papers, inter-sample peaks may be
considerably
higher than 0dBFS."

------------------
www.hypersonic.se (http://www.hypersonic.se)

[blairl]

Some of the points I bring up here are not intended to raise concern over the quality or usability of Pro Tools as a mixer. I have heard mixes done on Pro Tools that meet or beat mixes done on SSL's, Neve's, Oxford's, Capricorn's et al. I think that anyone who has read any past posts of mine knows that I am a big fan of mixing entirely within Pro Tools. I think it's a wonderful and inovative tool. I also think that it's the most well known DAW and therfore the most widely scrutinized DAW. I have seen many unfair comparisons made that make no sense at all. When I bring up things like the difference between signal to noise ratio and signal to error ratio, these are finite things that have little or no bearing on how a final product sounds mixed entirely within Pro Tools. I don't think I am making any pedantic statements. I'm certainly not just spouting off information that I don't understand just to spout it off. This is a topic that I have been interested in for quite some time. I thought it might interest the designers of Pro Tools to see what the designers of other well respected gear think about these finite things so that they might be examined, taken into consideration, implemented or disregarded.

In the end I think the numbers game is something that too many people don't understand and something that is taken too seriously.

[blairl]

For those of you who don't know how I feel about Pro Tools, I'll copy a post I made a while ago. It pretty much sums it up.
-------------------------------------------

OK, Some people happen to like the sound of analog circuits and for this reason use analog consoles. Some people like digital consoles. I understand that this is all a matter of taste. I'll just post another viewpoint.

Many people have commented on the sound of the Pro Tools mix bus. There are two things to consider here. First is quantization error. Some people believe that something that may cause "thin" sounding mixes is this distortion or quantization error. If you mix entirely within Pro Tools then the mixer maintains the quantization at 48 or 56 bits and then rounds it to 24 bits at the final stage. This produces a quantization error at this level which is about -144 dbFS. I'm only talking about the faders. The different plug-ins and effects sends are another story. (This may be why using plug-ins that dither or noise shape the output seem to sound so good, the dither eliminates the quantization error. Some of these are Waves and McDSP). Anyway -144 dbFS is way down there and it's arguable that it won't interfere with the final output. However if you decide to run direct out to an analog console where each channel bypasses the internal 2 bus then something different happens. Each and every output is truncated. Instead of doing all of the calculations and saving them for one final truncation from 48 to 24 bits at final output, direct out mode requires that each and every output be truncated. Every time a signal is doubled the level will rise by 3 db, so for a 64 track session running out to an analog console the quantization noise floor will be at -126 dbFS instead of -144 dbFS if you were to mix within Pro Tools. Granted -126 dbFS is also way down there, but people seem to be afraid that the more channels used in Pro Tools, the worse it sounds. I'm just saying that this cumulative quantization distortion that may be affecting the sound is actually worse if you use direct out mode to an external analog or digital console than if you were to just use the interal Pro Tools mixer. If you were to use various stems instead of the full 64 channel out then the quantization noise would be less, but still more than a completely internal Pro Tools mix. The only way that this wouldn't be a factor at all is if you were to leave all of the faders at 0 and not do any automation at all within Pro Tools. As long as the faders remain at 0 then no quantization distortion will be introduced. But that makes a full blown Pro Tools mix plus system a prety extensive hard disk recorder. You can buy a RADAR or Mackie, Tascam etc. for much less. I understand however that Pro Tools would have an advantage over these because of the editing and plug-in factors, but unless you use plug-ins that dither their ouput, quantization distortion will also be introduced.

Another thing that people think may cause cumulative sound degradation within Pro Tools is the headroom. The more channels added, the more the mix bus is taxed. Some people think that with many channels this may be too much for the Pro Tools mix bus which may be causing some overload. There are ways to deal with this. Roger Nichols explained it best:
As with any time you mix multiple tracks together there can be headroom problems. If you are using an analog console you can get into a situation where you have to pull the master fader down because of the addition of all of the individual tracks. To get the master fader back up, you have to pull down all of the individual faders an equal amount to enable you to get the master
fader back up where it belongs. The exact same thing happens in Pro Tools or any digital console. If every track you have is
recorded with dense information slamming the top of the meters, then you can't have all of the track faders at zero for your mix. You will have to pull them down to some reasonable level so that the mixed signal is at a reasonable level.
1) Use the 24 bit mixer. It provides more headroom for large track counts.
2) Don't pull the master fader down more than 10 dB. If you need to pull it lower, then trim all of your track faders.
3) watch for overs on the master fader. If there are no overs, then there is no headroom problem.
I have been mixing Pro Tools sessions with 48 to 120 tracks of instruments and have never had a problem that Couldn't be remedied by one of the above points.

This is a great tip to keep your Pro Tools mixes sounding great.

Another consideration is just the fact that multiple A/D and D/A conversions can degrade the sound quality. Even Fletcher who is a strong proponent of analog said this:

The projects that I have worked on that have been stored in the digital domain, once it gets there, I prefer to stay there. As far as I can tell, no possible good can come from leaving the digital domain, converting to analog, adding a piece of processing or two, then going back to the digital domain.

Back to my original point. Some people happen to like the sound of analog circuits and for this reason use analog consoles. We can see opinions all over the place that claim analog is superior to digital. It's my opinion that so many people that get into the business are not using their ears, but only listening to the opinions of others. When they read all over the place that analog is superior and digital sucks then some people just go along with it. I just want people to use their ears and decide for themselves. So just for the sake of another view I though I'd remind people that there are also some pro digital arguments out there. When George Massenburg was asked about his view on the SSL 9000J and the AMEK 9098i he said the following:
Overall, I want to say that anyone spending this kind of money on large frame analog desks in the year 2000 should have their heads examined. (I should put it far more kindly, of course, but I'm late for class.) Neither of these consoles really does resetability effectively, and I believe in my heart and my mind that mixing in the next few years is going to DEMAND instant, RELIABLE resettability. I like digital desks...Frankly, most who complain about digital desks not having 'feel' or 'sound' that analog desks supposedly do are babies who've grown used to Pablum audio for too long...

George Massenburg was asking an engineer in the Japanese music market about consoles in Japan where the SSL 9000 reigns supreme:
David, I have a question. In my opinion the Oxford is by far the best digital console I've heard and the most capable that I've interfaced with. Why...don't the Japanese get it? Why are 9000's still the only choice? Even Sony Music is getting custom-built analog desks (and alot of them at that) and avoiding Oxford altogether. Frankly, I'm stunned with the complete lack of sensitivity to sound, and the over-emphasis on "what he's doing". You've heard mainstream Japanese pop music on the radio; much of it is pathetic...

Oli P who posts a lot on the DUC had this experience:
I just got a test CD from mastering of an album I mixed in various rooms (SSL and Euphonix and at home on PT), and the PT mix was probably the best sounding.

He further explained his techniques and comparisons by saying:
The mix bus on an SSL has a lot of headroom, so if you compare your PT mix bus (on two faders on the SSL) with sendning all the tracks individually to the SSL faders, you won't do PT justice. The level out from PT will seem veak compared to the level you are used to push the SSL with.

The real comparison is to bring a 24bit master from PT to mastering along with an SSL master, and compare what you can get the final result to sound like.

Many people have mixed great sounding albums all within Pro Tools and have not experienced a sound quality degradation that has been refered too. One comparison that many may be familiar with is a test that was done at Mick Guzauski's studio where all other things being equal he and others there could not identify a difference between the Pro Tools mix bus and the Sony Oxford mix bus.

Roger Nichols has also mixed successfully within Pro Tools. When questioned about the quality of the Pro Tools mix bus he said:
I have used the Oxford many times also. And the Capricorn. Mixed albums on both of them. Sometimes from Sony 3348-HR, sometimes from ProTools feeding each channel
directly. The reason for doing that is because of the EQ and compression and 7000 inputs you get with a large console. At home I am using 2 Sony DMX-R100s.

I got out my SIM machine and Audio Precision and made sure that there was no difference
between what came out of the ProTools mix bus and the other digital consoles. The data
was the same...

Also there is a difference between listening to ProTools in your home studio and then
listening through an Oxford in a multi-million dollar control room. The mixes will ALWYS sound better in the tuned control room with $40,000 monitors. You cant tell me that the guy was in the Oxford control room looking at his ProTools screen and tried mixing the record with a mouse in ProTools while listening to the expensive monitors above the Oxford. Only I would do something like that.

Also in the listening process there is the level matching to deal with. If you listen to a mix from ProTools and then listen to the mix through the Oxford... If the Oxford is 0.1dB louder, it will sound better. Louder always sounds better.

If you would rather, you can mix your
records your way, afraid of bottom bits getting cut off when you turrn the level down 6dB,and that you must mix your ProTools through an Oxford. I, on the other hand, armed with the correct information about how this stuff works, will avoid the bad things and work with the good things and continue to make great sounding records...

Roger

Even the infamous Bob Ludwig has had some pleasant things to say about Pro Tools as you can read in the digi propoganda:

Although Ludwig was initially skeptical of projects mixed entirely within Pro Tools, he has recently warmed up to the idea. "Over the last couple of months, I've had projects come in completely in Pro Tools, as opposed to mixed on a (Neve) Capricorn or (Sony) Oxford, and the quality was so good I couldn't believe it," he beams. "It led me to believe that it certainly can be done."

Anyway this is way too long so I'll quit. I'm just saying that many very respectable people do not have a problem with digital consoles or the Pro Tools mix bus. Again I do understand that some people just happen to like the way analog consoles sound, everyone has an opinion and that's great. I just want people to use their own ears and make their own decisions, like many of you here have already done. The idea that analog is superior to digital and that digital sucks just doesn't seem like a valid argument to me. They are just different. And many people are mixing successfully within Pro Tools.

[Sonsey]

Frederick...Thanks for your reply. I went back and re-read your post in the "Why" thread. Somehow I missed it the first time ...my bad :-) Explained a lot. Many thanks. Just for the record I mix in Pro-Tools everyday and love it. I was just curious about the internal workings.

Howard Sonnenburg
Atomic Productions

[blairl]

Originally posted by F Umminger:
blairl-


quote:
--------------------------------------------------------------------------------

If the standard 24 bit mixer is used then the signal to error ratio starts out at around -144 dbfs and depending on how many tracks, plug-in's etc., the signal to error ratio can rise to around -126 dbfs or even higher.

--------------------------------------------------------------------------------

This is not true, for the reason I stated above. Or, at least, if the signal-to-error ratio does rise that high it is not due to the mixer but rather due to the errors being fed into the mixer by plugins.

That's right. When I did the test the higher errors didn't show up until I started using plug-in's that don't dither. The more I added the higher the errors became. However, it is the inserts on the Pro Tools mixer that force the plug-in's to return a 24 bit word length. This is even the case with the dithering mixer. If all inserts were able to maintain a 48 bit path until final output just like the faders on the mixer do then that would be something else. The posibility of the entire mixer maintaining a 48 bit path including inserts and sends is something I have showed interest in in the past. Some plug-in's do dither from 48 to 24 bits to avoid this issue. So if the dithering mixer is used with only plug-ins that dither then we have a setup that maintains a signal to error ratio of somewhere around -288 dbfs. Does any of this matter though is the real question. Even the standard mixer with non dithering plug-in's being able to maintain a signal to error ratio greater than -120 dbfs is a great spec and as mentioned by Dave and others it is beyond our personal dynamic range.

[This message has been edited by blairl (edited May 27, 2001).]