Become a Site Supporter and Never see Ads again!

Author Topic: New Sound Devices MixPre II-series coming?  (Read 31122 times)

0 Members and 1 Guest are viewing this topic.

Offline justink

  • Trade Count: (4)
  • Taperssection All-Star
  • ****
  • Posts: 1869
  • Gender: Male
Re: New Sound Devices MixPre II-series coming?
« Reply #120 on: September 09, 2019, 10:38:11 PM »
Mics:
DPA 4028 (Subcardioid)
DPA 4023 (Cardioid)
DPA 4018V (Supercardioid)
Earthworks TC25 (omnis) 

Pres and a/d's:
Grace Design Lunatec V3 (Oade ACM)
Edirol UA-5 (bm2p+ Mod)

Recorders:
Edirol R-44 (Oade CM)
Sony PCM‑M10

Offline AB52

  • Trade Count: (0)
  • Taperssection Regular
  • **
  • Posts: 153
Re: New Sound Devices MixPre II-series coming?
« Reply #121 on: September 10, 2019, 12:04:08 AM »
I think I am going to order one.  Who else has better customer service than sound devices?  That has to be factored into a purchase.
I just do not understand how this much technology can be had for such a low price?
I wonder how good it really is for music recording?  I guess I will find out

Offline aaronji

  • Site Supporter
  • Trade Count: (8)
  • Needs to get out more...
  • *
  • Posts: 2719
Re: New Sound Devices MixPre II-series coming?
« Reply #122 on: September 10, 2019, 04:26:36 AM »

Thought this might be of interest to some of you ...

https://www.sounddevices.com/low-signal-32-bit-float/

Paul

Very interesting. Would it be possible to post the ISOs from this, pre-fader? Also, what is the third gain stage mentioned on the webpage? It says, "trim gain set to minimum (6 dB), fader gain set to -50 dB, master left (L) gain set to -40 dB". What is the "master left"? On the first generation, there are only trim and fader (as far as I can tell).

Thanks for your participation on the forum, Paul. It is great to have you contributing here...

Offline Gutbucket

  • record > listen > revise technique
  • Trade Count: (14)
  • Needs to get out more...
  • *****
  • Posts: 13653
  • Gender: Male
Re: New Sound Devices MixPre II-series coming?
« Reply #123 on: September 10, 2019, 09:40:25 AM »
^Post fader stereo mix gain, left channel only as the signal is mono. 



Thought this might be of interest to some of you ...

https://www.sounddevices.com/low-signal-32-bit-float/

Paul

Yes, thank you for your participation here Paul.  Very much appreciated! 

That 24bit file is a classic example of insufficient application of dither, resulting in audible noise modulation. As such, it highlights a bug in the 24bit mode operation of this unit.  If the unit was setup properly we should be hearing hearing decay into obviously audible dither noise in this >100dB amplified example, rather than into audible modulation noise.
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<<

Offline ts

  • Trade Count: (68)
  • Needs to get out more...
  • *****
  • Posts: 3240
  • Gender: Male
Re: New Sound Devices MixPre II-series coming?
« Reply #124 on: September 10, 2019, 09:56:01 AM »

Thought this might be of interest to some of you ...

https://www.sounddevices.com/low-signal-32-bit-float/

Paul

that's cool!

Yes, it is very cool indeed. I can see a stealther in front of a stack benefiting from this for obvious reasons or as the post said, someone recording fighter jets or thunder claps or making a huge error in setting levels. But a guy like me, who records pretty much the same kind of music from the taper section all the time in the same venues and sets levels the same all time, I really don’t see the need for it. Someone correct me if I’m wrong. Thanks.
Beyerdynamic CK930>Sound Devices MixPre 6

Offline Gutbucket

  • record > listen > revise technique
  • Trade Count: (14)
  • Needs to get out more...
  • *****
  • Posts: 13653
  • Gender: Male
Re: New Sound Devices MixPre II-series coming?
« Reply #125 on: September 10, 2019, 11:21:26 AM »
Not wrong in questioning your actual need for it, though perhaps in premature consideration of the example as cool.   The example exposes incorrect setup of the recorder by SD for 24 bit recording mode operation.

Internally, the machine uses 32-bit floating-point representation of the signal downstream from the clever ADC operation covered by the patent.  From the example above it seems that when in 24-bit recording mode the machine truncates the internal 32-bit float representation without applying proper dither first, prior to storage in the 24 bit fixed-point format. 

« Last Edit: September 10, 2019, 11:23:00 AM by Gutbucket »
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<<

Online jerryfreak

  • Trade Count: (29)
  • Needs to get out more...
  • *****
  • Posts: 3344
  • team WOOz
Re: New Sound Devices MixPre II-series coming?
« Reply #126 on: September 10, 2019, 02:06:23 PM »
aka Digital Fuckery
team schoeps, dpa, benchmark, and zoom mostly... subject to change without warning

Offline Gutbucket

  • record > listen > revise technique
  • Trade Count: (14)
  • Needs to get out more...
  • *****
  • Posts: 13653
  • Gender: Male
Re: New Sound Devices MixPre II-series coming?
« Reply #127 on: September 10, 2019, 02:39:12 PM »
In other words, the primarily thing the posted SD example illustrates is what happens when 24 bit recording is done incorrectly, rather than the improvement 32 bit floating-point storage is capable of.   

Done correctly, the only audible difference would be the signal dropping into dither noise around the least significant bit level of the 24-bit file.  The 32-bit floating-point file should have a very similar sounding noise floor profile, just somewhat lower still.. defined by the noise-floor limits of the analog input stage / ADC stage rather than that of the 24-bit fixed-point container.
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<<

Offline dogmusic

  • Trade Count: (0)
  • Taperssection Member
  • ***
  • Posts: 830
Re: New Sound Devices MixPre II-series coming?
« Reply #128 on: September 10, 2019, 02:44:00 PM »
In other words, the primarily thing the posted SD example illustrates is what happens when 24 bit recording is done incorrectly, rather than the improvement 32 bit floating-point storage is capable of.   

Done correctly, the only audible difference would be the signal dropping into dither noise around the least significant bit level of the 24-bit file.  The 32-bit floating-point file should have a very similar sounding noise floor profile, just somewhat lower still.. defined by the noise-floor limits of the analog input stage / ADC stage rather than that of the 24-bit fixed-point container.

Are you saying that this is a hardware problem in the new units?
"The ear is much more than a mere appendage on the side of the head." - Catherine Parker Anthony, Structure and Function of the Human Body (1972)

"That's metaphysically absurd, man! How can I know what you hear?" - Firesign Theatre

Offline Gutbucket

  • record > listen > revise technique
  • Trade Count: (14)
  • Needs to get out more...
  • *****
  • Posts: 13653
  • Gender: Male
Re: New Sound Devices MixPre II-series coming?
« Reply #129 on: September 10, 2019, 03:49:08 PM »
If the example accurately represents how the new units operate in 24bit mode, it represents an implementation error on SD's behalf with regard to these recorders.  I don't know if it is correctable via firmware or not. 

Is it a big deal?  Not in a practical sense for us or anyone using the recorder who isn't increasing gain by several orders of magnitude in post..  basically because of this:

Though if I ever screw up a piano recording this badly so that I need to do a >100 dB boost, I'll just leave the business quietly.

That ugly quantization noise which should not be heard at all in the 24 bit sample is in so low in level that in any reasonable real-world situation it effectively won't matter.

However, I do consider it a big deal for two more-fundamental reasons:
First SD is a highly reputable company.  A misleading example endangers their well-earned reputation and corporate good-will, even if only the technically minded folks notice the error.  This is aggravated when the claim is used specifically for marketing the product.  I'm not accusing SD of intentionally being misleading, only of making a mistake, drawing a false conclusion from it and leading customers to the same false conclusion.

Secondly, it represent a technical failure to implement a well-established fundamental rule of digital audio: Whenever reducing the bit-depth of a digital signal, always apply appropriate dither prior to truncation in order to avoid quantization noise artifacts.  Why has SD failed to implement this?
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<<

Offline Paul Isaacs

  • Trade Count: (0)
  • Taperssection Regular
  • **
  • Posts: 88
Re: New Sound Devices MixPre II-series coming?
« Reply #130 on: September 10, 2019, 03:51:15 PM »
The main point that these extreme examples are trying to demonstrate is that 32-bit float offers better accuracy and lower noise than 24-bit fixed point for super low level signals. Even if we did apply dither in 24-bit mode (which we don't), they'd still be less accurate and noisier than 32-bit float since you'd only have a few lower bits to encode the signal (e.g. 4 bits only gives you 16 amplitude levels), but in 32-bit float you have the full 24-bits (the mantissa). The dither might obscure the steps, but at the expense of more noise. The 32-bit float always maintains full 24-bit accuracy at lower levels.

Anyway, as I say, this example is mainly to make the point that no one need worry recording super low or super high levels with 32-bit float (with the wide dynamic range input stage of MixPre-II) and there are applications where that helps instill confidence.

To your point that we're trying to mislead people, certainly not intentionally - we're trying to educate people to the differences between fixed point and floating point because we have been getting many questions on this subject.  Re dither, we chose to not do internal dither because as you say, in the 24-bit real world with relatively normal gain staging, you'll never ever hear the lack of it.
« Last Edit: September 10, 2019, 04:14:26 PM by Paul Isaacs »

Offline Gutbucket

  • record > listen > revise technique
  • Trade Count: (14)
  • Needs to get out more...
  • *****
  • Posts: 13653
  • Gender: Male
Re: New Sound Devices MixPre II-series coming?
« Reply #131 on: September 10, 2019, 05:19:48 PM »
Yet it doesn't demonstrate better accuracy.   Additionally, it raises issues which are more concerning than the one it attempts to address.  For those who understand how digital audio works, this serves to diminish confidence in SD rather than inspire it.

Even if we did apply dither in 24-bit mode (which we don't) [..]

Okay. Confirmation no dither is being applied in 24-bit mode.  Why aren't standard best practices with regards to digital signal manipulation being followed?  If I were cynical, I might suggest such a compromisation of 24bit storage performance was being done to make 32-bit floating-point storage seem more of a practical improvement than it otherwise represents (which is what the example does).  But I'm not cynical and surmise that the engineers at SD have concluded this level of quantization noise is so inconsequential in practice it is not worth addressing, even if doing so would be the correct design approach. Then marketing folks offer an example of this bug as proof of low level "resolution" improvement provided by the shiny new feature.  Create a problem just to turn around and fix it.  SD can't have it both ways and remain truthful.

Quote
[..] they'd still be less accurate and noisier than 32-bit float since you'd only have a few lower bits to encode the signal (e.g. 4 bits only gives you 16 amplitude levels), but in 32-bit float you have the full 24-bits (the mantissa). The dither might obscure the steps, but at the expense of more noise. [..]

Noisier yes, because dither is intentional injection of noise at that low-level.  But not "less accurate".   You have yet to explain this repeated claim of accuracy other than simply restating it.

Dither is the intentional randomization of the least significant bit.  It randomizes the signal at the lowest stored value to prevent quantization noise from occurring, rather than obscuring it.  It does not obscure "steps" in level by burying them in noise.

Quote
[..]The 32-bit float always maintains full 24-bit accuracy at lower levels.

If you were to amplify further, enough to hear the noise-floor present in the 32-bit floating point file, the least-significant 4 bits above the noise-floor would similarly encode just 4 bits of dynamic range, not 24 bits worth.  You cannot digitally fabricate what was not there originally, nor would one want to.

[Note: bolding in the quoted portions above is mine]
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<<

Offline justme

  • Trade Count: (0)
  • Taperssection Regular
  • **
  • Posts: 140
  • Gender: Male
Re: New Sound Devices MixPre II-series coming?
« Reply #132 on: September 10, 2019, 06:00:36 PM »
Well if you keep on lowering the in signal too much you will at some point hit the noise floor. But the noise would still have a 24 bit resolution if 32-bit float is used although there will not be no wanted signal left to amplify.

A friend of mine is developing pre amps with extremely low noise floor.
-175dBV/root(Hz)

With an input level of -100dBV there’s still 75 left until you hit the floor.
A 32 bit float would keep on delivering in that situation while a fixed bit recorder would not.

I wish I didn’t own the first generation MP6 because then i would have bought the mixpre 6 II in an instant. 

Offline Gutbucket

  • record > listen > revise technique
  • Trade Count: (14)
  • Needs to get out more...
  • *****
  • Posts: 13653
  • Gender: Male
Re: New Sound Devices MixPre II-series coming?
« Reply #133 on: September 10, 2019, 06:05:03 PM »
To your point that we're trying to mislead people, certainly not intentionally - we're trying to educate people to the differences between fixed point and floating point because we have been getting many questions on this subject. [..]

The problem is that people are being mislead, even if unintentionally.  Paul, if you have the leverage, please try to arrange for the the technical people at SD to brief the sales, marketing and website folks so that everyone is on the same page.  I'm simply attempting to act as messenger about this disconnect, not as a character assassin.  Thanks for your patience in the wading to the technical waters on this here.

Quote
[..] Re dither, we chose to not do internal dither because as you say, in the 24-bit real world with relatively normal gain staging, you'll never ever hear the lack of it.

^Except when it is amplified dramatically and presented as an example of why the new 32-bit floating-point storage format feature is better!  It either matters or it doesn't.  As a company, which do you wish to claim?  This is why I suggest getting the technical SD folks in the room with the less technical folks at SD to help them answer the questions they are getting from customers.  Otherwise SD's credibility begins to fall into question.  At best it indicates a lack of internal communication and unified vision at the company.

My suggestion- Tell people this is really more about a storage format change than any change in recorder performance.  It's all about switching to a format capable of storing the last little bit of actual dynamic range throughput of which the front end of the recorder was already capable, yet was just shy of when previously recording to a 24-bit fixed-point format.  Not quite as sexy sounding, yet completely truthful and does not diminish the actual functional improvement of the new feature.


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<<

Offline noahbickart

  • phishrabbi
  • Site Supporter
  • Trade Count: (32)
  • Needs to get out more...
  • *
  • Posts: 2292
  • Gender: Male
  • So now I wander over grounds of light...
Re: New Sound Devices MixPre II-series coming?
« Reply #134 on: September 10, 2019, 10:09:05 PM »
Gutbucket:

I'm sympathetic to what you're saying. If I understand correctly, you're saying the following: SD decided not to implement dither when writing the 24bit data, based on the assumption that most users would be able to get decent levels so as to obviate the need for it. You, however, argue that adding the dither wouldn't have hurt, and because it could have helped (for example when people don't in fact get decent levels- precisely kind of situation exaggerated in the graphic comparisons posted above) they should have implemented dither in the first place. Having not done so, however, you feel that it's wrong to market the 32bit float as a solution to a self-implemented problem. Correct?

On the other hand, I'm also sympathetic to SD. Do any of SD's competitors implement dithering of 24bit signals? Sonosax? Zoom? Tascam? There must be some cost, in time or money, to implement the dither! They made the decision not to implement dither based on the reasonable assumption, that normal users would adjust gain to get decent levels. That doesn't seem so far fetched to me, but, then again, I'm a religion professor, not an engineer.   :spin: :rockets:

My read of all of this is as follows: Long after making the 24bit no dither decision, SD decided to tweak the product (version II) to allow for a novel 32bit float option, which just so happens to have a side effect of solving the very problem they decided not to anticipate initially. So the marketers decide to run with it, saying, essentially, "look at us, our gear solves even very remote problems." Given the outright bullshit that much of their competition employs, I don't agree with giving them crap for it. I too get excited when I outsmart myself, and realize after the fact that my decisions were even better than I thought they were before I did what I did. (Buying extra toilet paper when it was on sale comes immediately to mind).

They're not lying. It actually does make sense to use the 32bit float option, unless you need endless recoding time, and thus can't afford the large file sizes 32bit float creates.

I have to say, my interest is peaked. I'm really tempted to sell my mixpre6 (make an offer tapirs!!!!) and I'll get the new one for not much money. I would go to a show, set up my gear with 32bit float (@ 48khz because I can't hear the difference with 96 or 192 anyway), seal the bag, and stash it under my seat. When the lights go down I'd press record through the wingman app, and press stop (or not) at set break. Foolproof, no need to look at or even expose my bag at all during the show.

I remember clearly making the move from cassette to DAT, and realizing that I never ever had to worry about a tape flip again. This seems like a step forward in a similar way, never having to worry about levels.

Then again, having run the Mixpre6 for a few years, it's really easy to guess the ballpark gain setting needed, and 24bit files are likely more than enough, given the noise floor at a rock concert, and I'm already really happy with both the product and the tech support....
« Last Edit: September 10, 2019, 10:29:28 PM by noahbickart »
Recording:
Capsules: Schoeps mk41v (x2), mk22 (x2), mk3 (x2), mk21 & mk8
Cables: 2x nbob KCY, 1 pair nbob actives, GAKables 10' & 20' 6-channel snakes, Darktrain 2 & 4 channel KCY and mini xlr extensions:
Preamps:    Schoeps VMS 02iub, Naiant IPA, Sound Devices Mixpre6 I
Recorders: Sound Devices Mixpre6 I, Sony PCM m10

Home Playback: Mac Mini> Mytek Brooklyn+> McIntosh MC162> Eminent Tech LFT-16; Musical Fidelity xCan v2> Hifiman HE-4XX / Beyerdynamic DT880

Office Playback: iMac> Grace m903> AKG k701 / Hifiman HE-400

 

RSS | Mobile
Page created in 0.117 seconds with 57 queries.
© 2002-2020 Taperssection.com
Powered by SMF