Sound Devices MixPre-6 II (second generation)

The place for questions about shooting with Blackmagic Cameras.
  • Author
  • Message
Offline
User avatar

rick.lang

  • Posts: 12926
  • Joined: Wed Aug 22, 2012 5:41 pm
  • Location: Victoria BC Canada

Re: Sound Devices MixPre-6 II (second generation)

PostSat Mar 07, 2020 9:35 pm

I still do adjustments within the MixPre-6 II so the levels I’m monitoring look decent. I know you’ll do that too better than I can. But I’m often in category 3 in your list and something like category 1. So I try to be somewhat skilled at it all, but then I can relax because it’s 32bit. Especially true now that Resolve fully supports 32bit.
Rick Lang
Offline

robedge

  • Posts: 584
  • Joined: Tue Dec 24, 2019 1:24 am
  • Location: U.S.
  • Real Name: Rob Edge

Re: Sound Devices MixPre-6 II (second generation)

PostSat Mar 07, 2020 10:24 pm

rick.lang wrote:I still do adjustments within the MixPre-6 II so the levels I’m monitoring look decent. I know you’ll do that too better than I can. But I’m often in category 3 in your list and something like category 1. So I try to be somewhat skilled at it all, but then I can relax because it’s 32bit. Especially true now that Resolve fully supports 32bit.


I’m just think it’s good that Paul Isaacs, who is as authoritative on the subject as one can be, has drawn a line between the hype around 32-bit and the reality. He has also confirmed advice that I’ve received from New York’s two most important vendors of location sound gear.

Like Paul Isaacs, I think that 32-bit is especially interesting for recordings of wide dynamic range sound effects. Post-processing is less of an issue because sound effect recordings are usually fairly short and are made in the expectation that there may be a lot of post-processing anyway. Post-processing manipulation is the whole point of recording sound effects at 192kHz, which is increasingly common.

What I want to find out now, by trying 32-bit, is exactly what it does to workflow and processing/rendering times during editing. A video that I saw yesterday says that the impact on processing/rendering is exponential, not linear, and substantial. However, the person who made the video did not talk about how powerful/fast his computer is.

Have you edited 32-bit recordings? Do you know what the impact is compared to 24-bit?
Video: Pocket 4K, Fujinon MK 18-55mm, Leica primes 35-135mm, Nikkor 55mm macro

Stills: Leica M3/M6/M240; Mamiya 7II; Arca Swiss 4x5/8x10

Sound: Schoeps omni & supercardiod mikes; DPA miniature omni mikes; Sound Devices 702T/MixPre-3 v.2
Offline
User avatar

rick.lang

  • Posts: 12926
  • Joined: Wed Aug 22, 2012 5:41 pm
  • Location: Victoria BC Canada

Sound Devices MixPre-6 II (second generation)

PostSat Mar 07, 2020 10:43 pm

Audio file sizes are a third larger so that takes more resources and signed 32bit floating point takes more processing time than 24bit binary, but I’m afraid I’ve not done a controlled comparison. I could have told you a few decades ago when I was working directly with those things, but processors have changed greatly since then and it’s possible that there’s little impact on the processing end by the CPU given the much more capable designs we have now. I don’t think it’s exponential, but it will take a little longer since you’re handling that binary value anyway as the mantissa and shifting it left or right by the value of the signed exponent. Shifts are a lot faster as we are talking about the theoretical shift of roughly 750 dB in Waithe direction, but the shift of likely 1 or 2 bits only for most sounds we might encounter. Given everything else going on in the production of the video, the larger audio file size is the larger impact.
Rick Lang
Offline

robedge

  • Posts: 584
  • Joined: Tue Dec 24, 2019 1:24 am
  • Location: U.S.
  • Real Name: Rob Edge

Re: Sound Devices MixPre-6 II (second generation)

PostSat Mar 07, 2020 10:58 pm

rick.lang wrote:Audio file sizes are a third larger so that takes more resources and signed 32bit floating point takes more processing time than 24bit binary, but I’m afraid I’ve not done a controlled comparison. I could have told you a few decades ago when I was working directly with those things, but processors have changed greatly since then and it’s possible that there’s little impact on the processing end by the CPU given the much more capable designs we have now. I don’t think it’s exponential, but it will take a little longer since you’re handling that binary value anyway as the mantissa and shifting it left or right by the value of the signed exponent. Shifts are a lot faster as we are talking about the theoretical shift of roughly 750 dB in Waithe direction, but the shift of likely 1 or 2 bits only for most sounds we might encounter. Given everything else going on in the production of the video, the larger audio file size is the larger impact.


Sound Devices’s Paul Isaacs at 24:30 of the video above: “24-bit is gonna be more than good enough for most applications”.

Here’s the second video that I’m talking about. The person who made it uses 32-bit because he’s a drummer and he makes videos at shows like NAAM. He says that he can render 4K video on his laptop without difficulty. His comments on 32-bit rendering time are at 02:20. He’s talking about very significant differences from 24-bit, which would be really felt with long recordings. I’m going to find some time to test this myself over the next few days:


Video: Pocket 4K, Fujinon MK 18-55mm, Leica primes 35-135mm, Nikkor 55mm macro

Stills: Leica M3/M6/M240; Mamiya 7II; Arca Swiss 4x5/8x10

Sound: Schoeps omni & supercardiod mikes; DPA miniature omni mikes; Sound Devices 702T/MixPre-3 v.2
Offline
User avatar

rick.lang

  • Posts: 12926
  • Joined: Wed Aug 22, 2012 5:41 pm
  • Location: Victoria BC Canada

Sound Devices MixPre-6 II (second generation)

PostSat Mar 07, 2020 11:09 pm

You know he doesn’t know what he’s talking about when he says 32bit audio requires “a lot more space” than 24bit. It’s simple math, a third more! Then he claims a 10 minute audio render in 24bit requires more than 20 minutes in 32bit. That’s not exponential and I wasn’t clear if he’s rendering 24bit output from 32bit input or something else.

If there are modern theatre projectors that handle 32bit or if the DCP specification includes DCP with 32bit audio, then you don’t have to convert anything. If the output is 24bit, you may need to shift the occasional amount of values one or two bits in real world usage.

At least this is my assumption how 32bit is used and converted to 24bit audio which I’ve done.

I suppose he’s not wrong about the results he says he gets but it’s not clear what else might be happening.
Rick Lang
Offline

robedge

  • Posts: 584
  • Joined: Tue Dec 24, 2019 1:24 am
  • Location: U.S.
  • Real Name: Rob Edge

Re: Sound Devices MixPre-6 II (second generation)

PostSat Mar 07, 2020 11:14 pm

rick.lang wrote:You know he doesn’t know what he’s talking about when he says 32bit audio requires “a lot more space” than 24bit. It’s simple math, a third more! Then he claims a 10 minute audio render in 24bit requires more than 20 minutes in 32bit. That’s not exponential and I wasn’t clear if he’s rendering 24bit output from 32bit input or something else.

If there are modern theatre projectors that handle 32bit or if the DCP specification includes DCP with 32bit audio, then you don’t have to convert anything. If the output is 24bit, you may need to shift the occasional amount of values one or two bits in real world usage.

I suppose he’s not wrong about the results he says he gets but it’s not clear what else might be happening.


Hi Rick,

Sorry, but you completely misheard what he said about rendering times. Listen again. And yes, 33% more data is indeed “a lot more data” to store, not that it advances the discussion to complain about his use of language.

As someone who owns a MixPre recorder with 32-bit float, I’m just interested in finding out the facts about where 32-bit is useful, where it isn’t and what the costs are. Sound Devices’s Paul Isaacs, and my NY sound shop, have answered the first two questions, now I’m working on an answer to the third.
Video: Pocket 4K, Fujinon MK 18-55mm, Leica primes 35-135mm, Nikkor 55mm macro

Stills: Leica M3/M6/M240; Mamiya 7II; Arca Swiss 4x5/8x10

Sound: Schoeps omni & supercardiod mikes; DPA miniature omni mikes; Sound Devices 702T/MixPre-3 v.2
Offline
User avatar

rick.lang

  • Posts: 12926
  • Joined: Wed Aug 22, 2012 5:41 pm
  • Location: Victoria BC Canada

Sound Devices MixPre-6 II (second generation)

PostSun Mar 08, 2020 12:54 am

Thanks for catching my bad read on that portion on the rendering times which went from “one and a half or two minutes for audio to twenty to thirty minutes.” That’s big alright. I was only remembering the “twenty to thirty minutes” thinking the difference was 10 minutes or 50% whereas it was really two minutes becoming thirty minutes. Sorry about that folks.

He does mention he used Adobe Premiere and may be related to how he defines his timeline and so on. It’s good we try and get a handle on these impacts with Resolve 16.2 which now supports 32bit audio.

A 33% increase in audio storage isn’t “a lot more” for someone shooting long video of which only a few percent is audio information, but that’s subjective. If you’re only shooting audio it’s a lot more, but when you’re shooting and storing both audio and video, it’s not a concern to me. So different perspectives of course result in different conclusions.
Rick Lang
Offline

robedge

  • Posts: 584
  • Joined: Tue Dec 24, 2019 1:24 am
  • Location: U.S.
  • Real Name: Rob Edge

Re: Sound Devices MixPre-6 II (second generation)

PostSun Mar 08, 2020 1:22 am

rick.lang wrote:Thanks for catching my bad read on that portion on the rendering times which went from “one and a half or two minutes for audio to twenty to thirty minutes.” That’s big alright. I was only remembering the “twenty to thirty minutes” thinking the difference was 10 minutes or 50% whereas it was really two minutes becoming thirty minutes. Sorry about that folks.

He does mention he used Adobe Premiere and may be related to how he defines his timeline and so on. It’s good we try and get a handle on these impacts with Resolve 16.2 which now supports 32bit audio.


I use Final Cut, Logic and iZotope RX/Ozone, all of which are fine with 32-bit audio.

I have to do some tests of my own. If the fellow in the video four posts up is right, rendering time is a significant factor in deciding when/if to use 32-bit. If anybody on this forum has actual experience with editing 32-bit audio, comments on that experience would be more than welcome.

If rendering times are long, it’s one thing to use 32-bit to record sound effects, which is two of the three examples Paul Isaacs talks about in the video above, but it’s another altogether to use it for long form performances/documentaries, which Isaacs clearly doesn’t recommend. Maybe OK for hobbyists with lots of time on their hands, not so fine for people who are paying for sound, or indeed for hobbyists who value their time.

I understand Isaacs’s point about novices, but I really wonder whether someone who doesn’t know how to set gain properly knows how to place a microphone. It’s not like setting gain is rocket science. In my view, the novice argument for 32-bit is badly flawed, and what’s likely to happen is a bunch of people who don’t know what they’re doing using 32-bit as a crutch. What they do on their own time is their business; the question is whether they think that other people should pay for their lack of basic knowledge.
Video: Pocket 4K, Fujinon MK 18-55mm, Leica primes 35-135mm, Nikkor 55mm macro

Stills: Leica M3/M6/M240; Mamiya 7II; Arca Swiss 4x5/8x10

Sound: Schoeps omni & supercardiod mikes; DPA miniature omni mikes; Sound Devices 702T/MixPre-3 v.2
Offline
User avatar

Robert Niessner

  • Posts: 2858
  • Joined: Thu Feb 21, 2013 9:51 am
  • Location: Graz, Austria

Re: Sound Devices MixPre-6 II (second generation)

PostSun Mar 08, 2020 10:05 am

I have quite a bit of experience of working with 32-bit now. Just recently finished a short DCI 4K commercial for cinema where we recorded the audio with the SD MixPre-10 II in 32-bit float mode. I haven't found that rendering times increased exponentially. For me 32-bit didn't have any significant impact on rendering times.

I have also been working with iZotope RX7 Advanced for months on repairing the fu..ed up surround mix of a full feature without having the stems. Everything has been done in 32-bit.
The only big increase I have found is when I was working on a 192kHz 24bit project (recording of an orchestra) - that is 4 times the data of 48kHz 24bit and that takes a big hit on rendering times.

What I found to be really great with using the 32bit mode of the MixPre is that on small sets you can have the sound guy concentrating on booming the sound without the need to constantly fiddle with the levels.
So I didn't understand why Paul Isaacs downplays this...
Robert Niessner
LAUFBILDkommission
Graz / Austria
---------------------------------------------------------------------
Blackmagic Camera Blog (German):
http://laufbildkommission.wordpress.com

Read the blog in English via Google Translate:
http://tinyurl.com/pjf6a3m
Offline

robedge

  • Posts: 584
  • Joined: Tue Dec 24, 2019 1:24 am
  • Location: U.S.
  • Real Name: Rob Edge

Re: Sound Devices MixPre-6 II (second generation)

PostSun Mar 08, 2020 10:21 am

Robert Niessner wrote:For me 32-bit didn't have any significant impact on rendering times.


Thanks, good to know. I’ll be doing some tests this week using RX, Logic and Final Cut, hopefully with the same result. Your experience suggests that the fellow who made the video above, despite being able to render 4K UHD, hasn’t got enough processor power/RAM to handle 32-bit. We also don’t know how much data he’s driving at a time. What computer/processor and how much RAM were you using?
Last edited by robedge on Sun Mar 08, 2020 10:34 am, edited 1 time in total.
Video: Pocket 4K, Fujinon MK 18-55mm, Leica primes 35-135mm, Nikkor 55mm macro

Stills: Leica M3/M6/M240; Mamiya 7II; Arca Swiss 4x5/8x10

Sound: Schoeps omni & supercardiod mikes; DPA miniature omni mikes; Sound Devices 702T/MixPre-3 v.2
Offline
User avatar

Robert Niessner

  • Posts: 2858
  • Joined: Thu Feb 21, 2013 9:51 am
  • Location: Graz, Austria

Re: Sound Devices MixPre-6 II (second generation)

PostSun Mar 08, 2020 10:33 am

robedge wrote:Thanks, good to know. I’ll be doing some tests this week using RX, Logic and Final Cut, hopefully with the same result. What computer/processor and how much RAM were you using?


Rob, my main workstation is now in its 9th year:
Intel i7-980 sixcore 3.33 GHz with 48 GB RAM, Asus P6X58D-E Deluxe mb, Nvidia GTX780 with 6 GB, LSI-9750-8i raid controller with a 35 TB Raid-5, 2 TB Samsung 850 SSD for System, 0.5 TB Crucial M4 SSD for render cache

Another great use case for me: setting up the recording of a huge orchestra where I can only estimate the levels before the rehearsal and have to leave the device unattended otherwise (the conductors hate if you do anything behind them during the rehearsal; and of course you can't do anything during the live performance). Of course ideally this kind of recording would be done with an audio crew on its own - but that is not always possible.
Robert Niessner
LAUFBILDkommission
Graz / Austria
---------------------------------------------------------------------
Blackmagic Camera Blog (German):
http://laufbildkommission.wordpress.com

Read the blog in English via Google Translate:
http://tinyurl.com/pjf6a3m
Offline

robedge

  • Posts: 584
  • Joined: Tue Dec 24, 2019 1:24 am
  • Location: U.S.
  • Real Name: Rob Edge

Re: Sound Devices MixPre-6 II (second generation)

PostSun Mar 08, 2020 10:40 am

Robert Niessner wrote:Rob, my main workstation is now in its 9th year:
Intel i7-980 sixcore 3.33 GHz with 48 GB RAM, Asus P6X58D-E Deluxe mb, Nvidia GTX780 with 6 GB, LSI-9750-8i raid controller with a 35 TB Raid-5, 2 TB Samsung 850 SSD for System, 0.5 TB Crucial M4 SSD for render cache


Positively ancient. I think that I know now why you and the guy who made the video above are experiencing different render times :)

Re your comment “I didn’t understand why Paul Isaacs downplays this”... Not sure he’s downplaying it, but he does regard 24-bit as the norm, and his comments about 32-bit may be based, in part, on the assumption that most MixPre owners aren’t running a professional video workstation. Most people don’t have 48GB of RAM, or a dedicated SSD render cache.

I may try to do my tests on a couple of different machines with different specs.
Video: Pocket 4K, Fujinon MK 18-55mm, Leica primes 35-135mm, Nikkor 55mm macro

Stills: Leica M3/M6/M240; Mamiya 7II; Arca Swiss 4x5/8x10

Sound: Schoeps omni & supercardiod mikes; DPA miniature omni mikes; Sound Devices 702T/MixPre-3 v.2
Offline
User avatar

rick.lang

  • Posts: 12926
  • Joined: Wed Aug 22, 2012 5:41 pm
  • Location: Victoria BC Canada

Re: Sound Devices MixPre-6 II (second generation)

PostSun Mar 08, 2020 4:15 pm

Thanks for the feedback of the orchestra recording, Robert. I was steeped in binary and floating point math fifty years ago on IBM mainframes among others. It wasn’t a big deal then (and was more efficient than packed decimal math that I believe IBM created) and our phones today have far more power than a 50 year old mainframe computer.

I think the test results reported in that video may be more attributed to using Adobe Premiere or some other settings as the gentleman suggested.
Rick Lang
Offline

robedge

  • Posts: 584
  • Joined: Tue Dec 24, 2019 1:24 am
  • Location: U.S.
  • Real Name: Rob Edge

Re: Sound Devices MixPre-6 II (second generation)

PostSun Mar 08, 2020 6:22 pm

I just tested 32-bit on a new Mac mini. The computer is maxed out with the important exception of RAM. The RAM is only 8GB because I'm installing my own, and this was an opportunity to see how 8GB of RAM would handle a 32-bit recording. The other thing worth noting is that the current Mac minis have Apple's T2 chip. This chip greatly improves performance of some functions. I doubt, but can’t say for sure, whether it is a significant factor in this test. The CPU is top of the line for the current mini.

I recorded a 32-bit 48kHz Wave file, single track, 2.5 minutes long, deliberately clipping the audio. I then opened iZotope RX, which is the first stop for most of my sound recordings, and imported the file. I reduced the gain to below 0dB in several places in the recording and then exported (“bounced”) the file as 24-bit, which is what I would do before import to Final Cut.

As far as I am concerned, the export was lightning fast. There certainly wasn't any noticeable delay. I'm not going to bother testing in Logic or Final Cut, because in a situation like this RX would be the first stop anyway, and I don’t edit sound in Final Cut. I'm also not going to bother testing on a computer with more RAM, given that 8GB worked fine. It's possible that long recordings would result in some delay, but I'm doubtful that it would be terribly significant. I don’t know what would happen with an appreciably slower CPU. I would suggest the obvious - that people test 32-bit with their specific computer system and workflow before using it.

The added workload is gain correction and an eventual 24-bit export. The gain correction has to be done so that it sounds natural in relation to the audio before and after. This is often not as simple a matter as it sounds. I would much rather get this right via proper gain setting in the first place. What other people do is their business, but I think that Sound Devices's/Paul Isaacs's examples constitute a correct analysis of the kinds of situations where 32-bit should be used.

Sound Devices has made a short video (see below) showing the 32-bit workflow. Consistent with Paul Isaacs's comments on 32-bit, the video shows an RX workflow, and the example is a sound effect of an automobile engine revving.

Note that in the video below there is a need for only one discrete gain correction. In a recording of a performance, or during a documentary shoot, there may be many corrections to make if gain has not been set properly. There’s a point where this becomes a discussion about expertise, professionalism and, in the paid world, cost-cutting (axe the sound mixer position), but those issues are probably not worth discussing here.

Video: Pocket 4K, Fujinon MK 18-55mm, Leica primes 35-135mm, Nikkor 55mm macro

Stills: Leica M3/M6/M240; Mamiya 7II; Arca Swiss 4x5/8x10

Sound: Schoeps omni & supercardiod mikes; DPA miniature omni mikes; Sound Devices 702T/MixPre-3 v.2
Offline

Brad Hurley

  • Posts: 1456
  • Joined: Wed Jan 17, 2018 7:42 pm
  • Location: Montréal

Re: Sound Devices MixPre-6 II (second generation)

PostMon Mar 09, 2020 11:40 am

I've been working with 32-bit floating point WAV files for three years now on a much lower-spec Mac Mini (2014 i5 with 8 gigs of RAM) and have never encountered problems. These are files recorded on my live sound mixer (QSC Touchmix), which records 32-bit floating point WAV into an external SSD or high-speed hard disk. I use it to record concerts when I'm doing the sound.

I still use the original MixPre 6 for location recording and have no plans to upgrade, but as someone who works on documentary I can think of a few times where the newer generation MixPre would have made life easier. One example was last year in Brittany where I was documenting a procession that has happened every year on the same date since the late 1600s. Part of the event took place inside a church, followed by a procession to a statue, and once the church service was over I ran down the path that goes to the statue and positioned myself to capture the procession as it went by. There was no way to predict the right levels on the MixPre (the procession involved about 75 people and included singing). I set the gain relatively low, relying on the low-noise preamps and my low-noise mics to give me room to bring up the levels in post, but even so the limiters came on when the loudest group of singers passed by. Once the procession drew near I was too busy capturing video to even glance at the levels on the MixPre.
Resolve 16.1.2 Studio, Mac Pro 3.0 GHz 8-core, 32 gigs RAM, dual AMD D700 GPU.
Audio I/O: Sound Devices USBPre-2
Offline

robedge

  • Posts: 584
  • Joined: Tue Dec 24, 2019 1:24 am
  • Location: U.S.
  • Real Name: Rob Edge

Re: Sound Devices MixPre-6 II (second generation)

PostMon Mar 09, 2020 2:32 pm

Brad Hurley wrote:I still use the original MixPre 6 for location recording and have no plans to upgrade, but as someone who works on documentary I can think of a few times where the newer generation MixPre would have made life easier. One example was last year in Brittany where I was documenting a procession that has happened every year on the same date since the late 1600s. Part of the event took place inside a church, followed by a procession to a statue, and once the church service was over I ran down the path that goes to the statue and positioned myself to capture the procession as it went by. There was no way to predict the right levels on the MixPre (the procession involved about 75 people and included singing). I set the gain relatively low, relying on the low-noise preamps and my low-noise mics to give me room to bring up the levels in post, but even so the limiters came on when the loudest group of singers passed by. Once the procession drew near I was too busy capturing video to even glance at the levels on the MixPre.


This is Paul Isaacs’s third example, where as a practical matter there’s no sound recordist present. In those cases, the choice between a limiter and 32-bit is a judgment call. I’d go with a limiter unless I think that the shot and perfect dynamic range fidelity are sufficiently important that fixing gain problems in post is worth my time or, if I’m paying a sound editor, worth writing a cheque for.

Isaacs’s example of a show where 32-bit might be useful is Carpool Karaoke. The recordist sets up the recorder and mikes, but is not in the car. It’s unclear whether Carpool Karaoke is actually recorded in 32-bit, but given the profile of James Corden and his guests, and the budget, it could make sense. However, it’s interesting that Jerry Seinfeld’s Comedians in Cars Getting Coffee was being recorded successfully long before 32-bit was available. There’s a lengthy, informative video on YouTube in which the Seinfeld show’s Sound Supervisor explains how they do their recordings, and there isn’t a word about 32-bit.

Interestingly, 32-bit is still not a feature in professional field recorders, including in Sound Devices’s own new top of the line, US$10,000 Scorpio.

Here’s the video about how it was done on Seinfeld’s show as of 2016:

Video: Pocket 4K, Fujinon MK 18-55mm, Leica primes 35-135mm, Nikkor 55mm macro

Stills: Leica M3/M6/M240; Mamiya 7II; Arca Swiss 4x5/8x10

Sound: Schoeps omni & supercardiod mikes; DPA miniature omni mikes; Sound Devices 702T/MixPre-3 v.2
Previous

Return to Cinematography

Who is online

Users browsing this forum: Teemu Kyytinen and 16 guests