Become a Site Supporter and Never see Ads again!

Author Topic: New Sound Devices MixPre II-series coming?  (Read 88156 times)

0 Members and 1 Guest are viewing this topic.

Offline Gordon

  • Trade Count: (22)
  • Needs to get out more...
  • *****
  • Posts: 11780
  • Gender: Male
    • my list
Re: New Sound Devices MixPre II-series coming?
« Reply #30 on: August 29, 2019, 09:03:07 PM »
This is a bit worrying though.

What are my media options?

We recommend our SAM-32SD card. Additionally, most reputable manufacturers’ cards (such as SanDisk or Delkin) that meet or exceed class 10 speeds are acceptable.

They still only have one single SD card listed as approved. Their own.
https://www.sounddevices.com/mixpre-series-approved-media-list/

Same for gen 1!  I've had no issues with their card or a Sandisk though....
Microtech Gefell M20 or M21 > Nbob actives > Naiant PFA > Sound Devices MixPre-6 II @ 32/48

https://archive.org/details/fav-gordonlw

https://archive.org/details/teamdirtysouth

Online voltronic

  • Trade Count: (40)
  • Needs to get out more...
  • *****
  • Posts: 4104
Re: New Sound Devices MixPre II-series coming?
« Reply #31 on: August 29, 2019, 09:03:38 PM »
I'll just leave this here.  Rather curious, since the "II" versions must have been in development at the time this post was written:

https://taperssection.com/index.php?topic=190161.msg2297397#msg2297397
I am hitting my head against the walls, but the walls are giving way.
- Gustav Mahler

Acoustic Recording Techniques
Team Classical
Team Line Audio
Team DPA

Offline Paul Isaacs

  • Trade Count: (0)
  • Taperssection Regular
  • **
  • Posts: 96
Re: New Sound Devices MixPre II-series coming?
« Reply #32 on: August 29, 2019, 10:59:50 PM »
I know this contradicts stuff I've said before, but since being involved in the development of the 32-bit float MixPre-II, I'm a convert and am happy to admit I was wrong when I said 32-bit float has little relevance in a field recorder.

Having said that, for the majority of sound recording applications and for those who know how to gain stage, 24-bit has always been way more than enough to capture high fidelity recordings when combined with high quality, low noise/low distortion microphones, preamps and ADCs. That hasn't changed. So then, why the big deal with 32-bit float files ...

1) The first thing to note is that a 32-bit float file is a vast container that is essentially impossible to overload/clip. With 24-bit fixed-point file format, if you exceed 0dBFS (digital word = all 1's), distortion is burned into the recording and cannot be undone; with 32-bit float you can exceed 0dBFS by hundreds of dB and your audio will not incur digital overload distortion. Import that 32-bit float file into most DAWs and you have undistorted audio and associated waveform. The waveform may look clipped, but simply drag down the gain to normalize to below 0dBFS and you have a waveform that is no longer flat top!
2) When a 32-bit float file is combined with an ultra low noise, wide dynamic range input stage, you can record beyond the dynamic range (quietest to loudest sounds) offered by the highest quality microphones.
3) 32-bit float offers precision representation equally across the entire dynamic range. With traditional 24-bit fixed-point, super quiet signals may typically be just loud enough to have the lowest 4-8 bits (of the 24 available) be active - not great resolution at all. With floating point, even those super low level signals are represented by a full array of bits, thus maintaining accuracy at all levels. When you now normalize this low level signal back to 0dBFS, the precision is still there.

Here is a paper we put together regarding 32-bit float ...
https://www.sounddevices.com/32-bit-float-files-explained/

Paul

Offline jerryfreak

  • No PZ
  • Trade Count: (31)
  • Needs to get out more...
  • *
  • Posts: 6205
  • The plural of anecdote is not data
Re: New Sound Devices MixPre II-series coming?
« Reply #33 on: August 30, 2019, 02:48:54 AM »
I know this contradicts stuff I've said before, but since being involved in the development of the 32-bit float MixPre-II, I'm a convert and am happy to admit I was wrong when I said 32-bit float has little relevance in a field recorder.

Having said that, for the majority of sound recording applications and for those who know how to gain stage, 24-bit has always been way more than enough to capture high fidelity recordings when combined with high quality, low noise/low distortion microphones, preamps and ADCs. That hasn't changed. So then, why the big deal with 32-bit float files ...

1) The first thing to note is that a 32-bit float file is a vast container that is essentially impossible to overload/clip. With 24-bit fixed-point file format, if you exceed 0dBFS (digital word = all 1's), distortion is burned into the recording and cannot be undone; with 32-bit float you can exceed 0dBFS by hundreds of dB and your audio will not incur digital overload distortion. Import that 32-bit float file into most DAWs and you have undistorted audio and associated waveform. The waveform may look clipped, but simply drag down the gain to normalize to below 0dBFS and you have a waveform that is no longer flat top!
2) When a 32-bit float file is combined with an ultra low noise, wide dynamic range input stage, you can record beyond the dynamic range (quietest to loudest sounds) offered by the highest quality microphones.
3) 32-bit float offers precision representation equally across the entire dynamic range. With traditional 24-bit fixed-point, super quiet signals may typically be just loud enough to have the lowest 4-8 bits (of the 24 available) be active - not great resolution at all. With floating point, even those super low level signals are represented by a full array of bits, thus maintaining accuracy at all levels. When you now normalize this low level signal back to 0dBFS, the precision is still there.

Here is a paper we put together regarding 32-bit float ...
https://www.sounddevices.com/32-bit-float-files-explained/

Paul
Paul have you addressed the buffering issues in the new units? the current versions of mixpres crash at high channel/bitrate loads with most of the SD cards on the market, despite them testing above proper U3 sequential read rates (10x what is required at 8 ch of 24/192)

https://taperssection.com/index.php?topic=191314.0

 
Unable to post or PM due to arbitrary censorship of people the mod doesn't like. Please email me using the link in my profile if you need to connect

Online voltronic

  • Trade Count: (40)
  • Needs to get out more...
  • *****
  • Posts: 4104
Re: New Sound Devices MixPre II-series coming?
« Reply #34 on: August 30, 2019, 06:58:27 AM »
3) 32-bit float offers precision representation equally across the entire dynamic range. With traditional 24-bit fixed-point, super quiet signals may typically be just loud enough to have the lowest 4-8 bits (of the 24 available) be active - not great resolution at all. With floating point, even those super low level signals are represented by a full array of bits, thus maintaining accuracy at all levels. When you now normalize this low level signal back to 0dBFS, the precision is still there.

Paul,

Thanks for checking in here.  The benefits of recovering levels exceeding 0 dBFS are clear to me.  What is not as clear is your item 3 quoted above.  Several months ago, Samuel Green from Zoom made the same assertion at NAB when commenting on the F6.  This caught my attention then, and you reminded me of it again.  Are you saying that in FP recording, the full spectrum of bits "floats down" to the very low signal?
I am hitting my head against the walls, but the walls are giving way.
- Gustav Mahler

Acoustic Recording Techniques
Team Classical
Team Line Audio
Team DPA

Offline Paul Isaacs

  • Trade Count: (0)
  • Taperssection Regular
  • **
  • Posts: 96
Re: New Sound Devices MixPre II-series coming?
« Reply #35 on: August 30, 2019, 09:47:25 AM »
3) 32-bit float offers precision representation equally across the entire dynamic range. With traditional 24-bit fixed-point, super quiet signals may typically be just loud enough to have the lowest 4-8 bits (of the 24 available) be active - not great resolution at all. With floating point, even those super low level signals are represented by a full array of bits, thus maintaining accuracy at all levels. When you now normalize this low level signal back to 0dBFS, the precision is still there.

Paul,

Thanks for checking in here.  The benefits of recovering levels exceeding 0 dBFS are clear to me.  What is not as clear is your item 3 quoted above.  Several months ago, Samuel Green from Zoom made the same assertion at NAB when commenting on the F6.  This caught my attention then, and you reminded me of it again.  Are you saying that in FP recording, the full spectrum of bits "floats down" to the very low signal?

Essentially yes. What I'm saying is there is a lot more resolution and range for very low signals too. It's not quite as simple as all bits 'floating down' since 32-bit float is not like fixed-point in that it comprises of a mantissa and exponent to represent huge +ve numbers and -ve numbers way beyond what 24-bit fixed point offers. I suggest you read the section on 32-bit float in our paper and follow the link to IEEE 754 if you're interested in knowing more. https://www.sounddevices.com/32-bit-float-files-explained/

Offline gewwang

  • Trade Count: (4)
  • Needs to get out more...
  • *****
  • Posts: 6251
Re: New Sound Devices MixPre II-series coming?
« Reply #36 on: August 30, 2019, 11:03:45 AM »
Paul,

Has Sound Devices management ever considered maybe giving the early adopters of these products some sort of incentive for upgrading to the next version? Or rewarding them for being the early adopters and providing valuable feedback towards the development of the next version?

Online Gutbucket

  • record > listen > revise technique
  • Trade Count: (15)
  • Needs to get out more...
  • *****
  • Posts: 15700
  • Gender: Male
  • "Better to love music than respect it" ~Stravinsky
Re: New Sound Devices MixPre II-series coming?
« Reply #37 on: August 30, 2019, 12:57:26 PM »
3) 32-bit float offers precision representation equally across the entire dynamic range. With traditional 24-bit fixed-point, super quiet signals may typically be just loud enough to have the lowest 4-8 bits (of the 24 available) be active - not great resolution at all. With floating point, even those super low level signals are represented by a full array of bits, thus maintaining accuracy at all levels. When you now normalize this low level signal back to 0dBFS, the precision is still there.

Paul,

Thanks for checking in here.  The benefits of recovering levels exceeding 0 dBFS are clear to me.  What is not as clear is your item 3 quoted above.  Several months ago, Samuel Green from Zoom made the same assertion at NAB when commenting on the F6.  This caught my attention then, and you reminded me of it again.  Are you saying that in FP recording, the full spectrum of bits "floats down" to the very low signal?

What I'm saying is there is a lot more resolution and range for very low signals too. [snip]

What I remain to be convinced of is that 32bit floating point somehow improves "resolution" of low level signals through the A>D conversion.  This claim seems to contradict Nyquist–Shannon sampling theorem.

The SD 32-bit-float-files-explained paper does not address this.  It concerns manipulation of a signal after it is represented as a 32-bit float format.

musical volition > vibrations > voltages > numeric values > voltages > vibrations> virtual teleportation time-machine experience
Better recording made easy - >>Improved PAS table<< | Made excellent- >>click here to download the Oddball Microphone Technique illustrated PDF booklet<< (note: This is a 1st draft, now several years old and in need of revision!  Stay tuned)

Online voltronic

  • Trade Count: (40)
  • Needs to get out more...
  • *****
  • Posts: 4104
Re: New Sound Devices MixPre II-series coming?
« Reply #38 on: August 30, 2019, 01:29:57 PM »
3) 32-bit float offers precision representation equally across the entire dynamic range. With traditional 24-bit fixed-point, super quiet signals may typically be just loud enough to have the lowest 4-8 bits (of the 24 available) be active - not great resolution at all. With floating point, even those super low level signals are represented by a full array of bits, thus maintaining accuracy at all levels. When you now normalize this low level signal back to 0dBFS, the precision is still there.

Paul,

Thanks for checking in here.  The benefits of recovering levels exceeding 0 dBFS are clear to me.  What is not as clear is your item 3 quoted above.  Several months ago, Samuel Green from Zoom made the same assertion at NAB when commenting on the F6.  This caught my attention then, and you reminded me of it again.  Are you saying that in FP recording, the full spectrum of bits "floats down" to the very low signal?

What I'm saying is there is a lot more resolution and range for very low signals too. [snip]

What I remain to be convinced of is that 32bit floating point somehow improves "resolution" of low level signals through the A>D conversion.  This claim seems to contradict Nyquist–Shannon sampling theorem.

The SD 32-bit-float-files-explained paper does not address this.  It concerns manipulation of a signal after it is represented as a 32-bit float format.

This is why I asked my question; I had read the SD document already, and did not find the answer there, nor can I find any other documentation from other sources that address the low-level signal resolution.

Then there is the aspect of this that DSatz spoke of at length on the F6 thread: the internal noise floor.  If 32-bit float files truly do record low signals with better resolution as a result of more bits being available, the self-noise will still limit how low you can go.  Perhaps this scheme will still show its benefits at typical low signal levels in a recording.

Lastly, I think all of us are really waiting to hear are raw samples of very wide dynamic range music recorded in both fixed and float-point.  I want to hear just how transparently the ADC transitions between its segments, or if you can hear it working like you do a limiter.  So far, all we have are dialogue samples from both companies.
I am hitting my head against the walls, but the walls are giving way.
- Gustav Mahler

Acoustic Recording Techniques
Team Classical
Team Line Audio
Team DPA

Offline EmRR

  • Trade Count: (5)
  • Taperssection Member
  • ***
  • Posts: 779
    • ElectroMagnetic Radiation Recorders
Re: New Sound Devices MixPre II-series coming?
« Reply #39 on: August 30, 2019, 01:47:03 PM »
Yep, what DSatz said.

Elsewhere, I've reported there are ways to overdrive the input circuitry of an F8n without hitting digital zero, in fact with line input trim set to -10, it squarewaves at about -5.5dBFS.

I don't know about the SD units.  If 32 bit float were to mean what's somewhat implied by the marketing, I'd want the analog input circuit before the AD to have at least 6dB more headroom than the AD, and that rarely if ever happens.  32 bit float is still a good thing for removing dither point penalty with drastic gain changes in post. 

Playback devices tend to have similar bottlenecks, with analog output circuitry having no headroom beyond max converter output, so a really hot brick walled playback source will make a lot of players sound different from one another because of the distortion artifacts, with some players going into obvious blatant distortion. 
Mics: DPA 4060 w/MPS 6030 PSU/DAD6001/DAD4099, Neumann KM 131, Oktava MK 012, Sennheiser MKH 105, MKH 20, MKH 30, MKH 40, MKH 800 TWIN
Recorders: Zoom F8n, Sony MZ-R50

Offline morst

  • I think I found an error on the internet; #UnionStrong
  • Trade Count: (2)
  • Needs to get out more...
  • *****
  • Posts: 5950
Re: New Sound Devices MixPre II-series coming?
« Reply #40 on: August 30, 2019, 02:03:05 PM »
What I remain to be convinced of is that 32bit floating point somehow improves "resolution" of low level signals through the A>D conversion.  This claim seems to contradict Nyquist–Shannon sampling theorem.
Nyquist says it's a perfect representation, I don't really care how many bits it takes. If it's a quiet signal, then 4-8 bits should be plenty. The rest of the 32 will be a detailed retelling of all the circuit noise beneath.
https://toad.social/@morst spoutible.com/morst post.news/@acffhmorst

Offline datbrad

  • Trade Count: (1)
  • Needs to get out more...
  • *****
  • Posts: 2295
  • Gender: Male
Re: New Sound Devices MixPre II-series coming?
« Reply #41 on: August 30, 2019, 02:31:23 PM »
Application is key to the benefits of even 24 bit. For example, when recording a concert sound reinforcement system, the following section from Rane says it all.

"For sound systems, the maximum loudness level is what is achievable before acoustic feedback, or system squeal begins. While the minimum level is determined by the overall background noise. It is significant that the audio equipment noise is usually swamped by the HVAC (heating, ventilating & air conditioning) plus audience noise. Typical minimum noise levels are 35-45 dB SPL (sound pressure level), with typical loudest sounds being in the 100-105 dB SPL area. (Sounds louder than this start being very uncomfortable, causing audience complaints.) This yields a typical useable system dynamic range on the order of only 55-70 dB -- quite different than unit dynamic ranges."

IMO, nature, Foley, measurement, and other applications where it's possible for dynamic ranges exceeding the 120db range of human hearing to be recorded, 32bit should be able to make it easier, since you wouldn't have to actually experience the highest peaks and lowest signal down into the noise floor to set gain in actual application.

For those of us taping bands off the PA, 65db dynamic range is all we have to harness. 24 bit provides more than enough margin to capture that range and give the taper a wide sweet spot for level setting.

These machines are awesome, but within the subject of 24 bit recording, we've beaten the topic of 192khz sampling rate to death on this forum, and I haven't read any posts that credibly argue it's value for concert recording. The need for 24/96 is even questionable in that application, IMO.

« Last Edit: August 30, 2019, 11:56:10 PM by DATBRAD »
AKG C460B w/CK61/CK63>Luminous Monarch XLRs>SD MP-1(x2)>Luminous Monarch XLRs>PMD661(Oade WMOD)

Beyer M201>Luminous Monarch XLRs>PMD561 (Oade CMOD)

Offline Paul Isaacs

  • Trade Count: (0)
  • Taperssection Regular
  • **
  • Posts: 96
Re: New Sound Devices MixPre II-series coming?
« Reply #42 on: August 30, 2019, 02:53:28 PM »
If it's a quiet signal, then 4-8 bits should be plenty.

For recording concerts, I agree. But if your work is high end sound fx or sound design for instance, you may want to capture very low level sounds for other purposes.
One example off the top of my head - recording delicate droplets softly hitting a surface, followed by a super loud thunderclap. I decide I want to just take the raindrops and manipulate them as the basis for a new sound effect. In that case, I would want to amplify or normalize those raindrops first so I can easily hear them. The lack of bit resolution will become apparent at this point because each sample may only be represented by 4 bits (16 amplitude steps). 32-bit float alleviates this. Of course, you also need a wide dynamic range input stage to pull this off ... the MixPre-II's have a 142dB dynamic range with Mic EIN of -130dBV.

Offline mjwin

  • Trade Count: (0)
  • Taperssection Regular
  • **
  • Posts: 73
  • Gender: Male
Re: New Sound Devices MixPre II-series coming?
« Reply #43 on: August 30, 2019, 04:39:40 PM »
One example off the top of my head - recording delicate droplets softly hitting a surface, followed by a super loud thunderclap. I decide I want to just take the raindrops and manipulate them as the basis for a new sound effect. In that case, I would want to amplify or normalize those raindrops first so I can easily hear them. The lack of bit resolution will become apparent at this point because each sample may only be represented by 4 bits (16 amplitude steps). 32-bit float alleviates this. Of course, you also need a wide dynamic range input stage to pull this off ... the MixPre-II's have a 142dB dynamic range with Mic EIN of -130dBV.

That's a good example, and something which I was doing myself a month or so back.  Trying to guess levels made me realize just how great a dynamic range my (now vintage) mixpre & low-noise mics could capture.  While waiting in the 45dBSPL night time rural ambience for an overhead thunderclap (with the mixpre gain set to clip at 130dBSPL) I mused that even at this low gain,  if the world went suddenly silent, the zero bit of the 24 bit convertor would still be rattling away about 25dB below the noise floor of my (quiet) microphones!

A very satisfying thought. 

Of course, it was imperative that I set the gain correctly for this once-in-a-lifetime direct overhead strike (which didn't happen)  And  it's the freedom to not worry at all about the gain setting which is the real benefit of the 32bit FP system. You just slide the output window around to wherever you need it!

As for signal resolution when amplified in post production, the world is simply too noisy. Even in my hypothetically silent world, If the zero bit is already 4 bits (24dB)  below the noise floor, adding more bits doesn't help.

Online Gutbucket

  • record > listen > revise technique
  • Trade Count: (15)
  • Needs to get out more...
  • *****
  • Posts: 15700
  • Gender: Male
  • "Better to love music than respect it" ~Stravinsky
Re: New Sound Devices MixPre II-series coming?
« Reply #44 on: August 30, 2019, 06:59:38 PM »
One example off the top of my head - recording delicate droplets softly hitting a surface, followed by a super loud thunderclap. I decide I want to just take the raindrops and manipulate them as the basis for a new sound effect. In that case, I would want to amplify or normalize those raindrops first so I can easily hear them. The lack of bit resolution will become apparent at this point because each sample may only be represented by 4 bits (16 amplitude steps). 32-bit float alleviates this.

Please explain the part I've bolded, which is what is in question here. It's the same 4 bits of information either way. The number of bits required to losslessly sample and reproduce that sound is determined solely by the dynamic range of the sound in question per Nyquist-Shannon.  What is different is how those 4 bits are represented in the different mathematical formats.  Yes the size of a 32bit floating point container is astronomically larger than a 24bit fixed container, which allows for shifting the bits of interest up or down in range to a but the 4 bits within that container that completely capture the very limited dynamics of the quiet sound in question are the identical.  Again, per Nyquist-Shannon, the basis of all digital audio.

Quote
Of course, you also need a wide dynamic range input stage to pull this off ... the MixPre-II's have a 142dB dynamic range with Mic EIN of -130dBV.
 
Yes exactly. It's the analog stage and ADC which are the constraints on dynamic range, not the storage file format.

Thanks for joining in the discussion here Paul, your presence is very much appreciated.. and SD gear is highly regarded for good reason. I just want to make sure we are rooted in reality and not chasing unicorns.
musical volition > vibrations > voltages > numeric values > voltages > vibrations> virtual teleportation time-machine experience
Better recording made easy - >>Improved PAS table<< | Made excellent- >>click here to download the Oddball Microphone Technique illustrated PDF booklet<< (note: This is a 1st draft, now several years old and in need of revision!  Stay tuned)

 

RSS | Mobile
Page created in 0.119 seconds with 44 queries.
© 2002-2024 Taperssection.com
Powered by SMF