Can this next generation of location recorders improve on the highly capable range that went before?
Sound Devices' superb line-up of MixPre portable recorders has only been in production for a couple of years, but the company have already given the entire product range a substantial make-over, with all-new and significantly more powerful and capable internal hardware. From a user's point of view, these three new 'second generation' machines (the MixPre‑3 II, MixPre‑6 II and MixPre‑10 II) all appear more or less identical to their forebears in terms of their physicality and core capabilities: the MixPre‑3 II is still an astonishingly compact unit with three Kashmir preamps and five recording tracks; the slightly larger MixPre‑6 II still has four preamps and eight tracks; and the flagship MixPre‑10 II model still has eight preamps and 12 tracks. The machines' user interfaces and control functions are also more or less the same across the range — but there are a few welcome enhancements and new features.
For example, a new automatic copying facility duplicates the data recorded on the internal SD card directly to an external USB thumb-drive inserted into the machine's USB-A port. And in addition to the standard factory mode, the limiters can now be switched into a customised operation, allowing the Ratio, Threshold and Release parameters to be adjusted (custom settings are remembered once set).
The upgraded hardware has also allowed the pre-roll record buffer to be increased to 10s on all models, and they all now support 192kHz recording. Frustratingly, though, there's still no option to record individual WAVs — as before, everything is stored as a polyWAV (unless using the Musician plug-in); when recording long concerts at high sample rates this results in lots of separate files which have to be joined together in a DAW. However, it's clear from looking at the sample-rate options (44.1, 47.952, 48, 48.048, 96 and 192 kHz) that these machines were designed primarily for video applications, in which the polyWAV format makes the workflow easier. The same underpinning role guided the decision to include synchronisable internal timecode generators in all three models, which will please videographers even more.
Impressively, the internal timecode generator is claimed to remain accurate to within half a frame over 24 hours (0.2ppm) and can maintain that accuracy over four hours, even when turned off and without batteries! Moreover, this timecode signal can be routed to an analogue output if required, making it available for synchronising other devices.
Another useful upgrade only found on the MixPre‑10 II is that its line outputs (on TA3 connectors) can now deliver a full +18dBu (the previous-generation model could only manage +10.8dBu), and this makes for much easier interfacing with other professional equipment. The MixPre‑3 II and -6 II models provide a smidgen under +8dBu at their unbalanced 3.5mm stereo line output sockets.
However, important though all of these enhancements and new functions obviously are, the headline feature is that all three models can be configured to capture audio in a 32-bit floating-point format (fixed-point 16- and 24-bit modes are still available too, of course). The floating-point mode, which involves a complete reboot of the machine as the internal hardware is reconfigured, provides a real-world dynamic range capability of an impressive 142dB! (See 'Floating Free' box.)
In practical terms this means that unexpectedly high peaks can be captured without distortion or the need for protective limiters (which are unavailable in the 32-bit floating-point mode anyway), and the wanted signal can always be retrieved by adjusting the replay gain in a DAW, or on the MixPre itself, after the event. In practice, this means that (given appropriate mic choice and technique) unskilled users will always end up with a high-quality recording, regardless of whether they accidentally peak the signal to +50 or -50 dBFS! On a related tangent, I was slightly amused to read in the specifications that the machine's D-A converters, which provide the stereo and headphone outputs with "32-bit precision", to deliver a 115dB (A-wtd) dynamic range, which equates to around 19 useful bits.
There's currently no equivalent in the new range of the previous 'M' models, but all the mkII machines do accept the optional 'Musician Plug-in' (currently available for a very attractive $49). As with the previous MixPres, this clever plug-in allows each machine to be reconfigured to operate as a 12-track multitrack device, complete with overdubbing and track bouncing, punch in/out, reverb and air EQ effects, a metronome, and all the other features and facilities existing users have come to appreciate.
There's also a free Ambisonics plug-in option for the MixPre‑6 II and -10 II (not for the MixPre‑3 II; it doesn't have sufficient input channels for ambisonic recording). This plug-in adds facilities for first-order (four-capsule) ambisonic recording at sample rates up to 192kHz, including automatic four-channel gain-linking, and both a stereo down-mix and full binaural decoding for monitoring (although the binaural mode is so processor-intensive that it's only available when working at base sample rates).
Specific mention is made in the manual of the Sennheiser Ambeo VR mic, but other A-format mic arrays can be used, and the A-format signals can be converted to B-format for recording, if desired, complete with options to correct for up, down and end-fire mic orientations. B-format signals can also be accepted from suitable mics, such as the Soundfield ST450 MkII, and the audio can be stored and replayed in either the FuMa (WXYZ) or AmbiX (WYZX) file formats. MixPre‑10 II users also have the option to record both B-formats simultaneously, along with the stereo and binaural decoded tracks — that should cover all post-production requirements in one go!
The Kashmir preamps really are special. They offer 76dB of gain, and an incredible 96dB maximum gain from input to recorded file, if necessary, with an impressive headroom margin to boot!
I was sent a MixPre‑6 II for this review but as all three models share essentially the same hardware and firmware, bar the few features exclusive to the MixPre‑6 II and MixPre‑10 II that I've already mentioned. They all ship with a wall-wart mains adapter (with a variety of different mains socket adapter plates), a USB-C to USB-C cable, an AA battery sled (four cells for the MixPre‑3 II and -6 II, and eight for the MixPre‑10 II). Each model also includes the same tripod mounting points on the top and bottom panels, complete with the removable anti-rotation pin for camera-mounting, and the Allen wrench is still tucked into a recess under the battery sled.
The menu encoder knob on the right panel (which doubles as the headphone volume) now has a removable knobbled 'tyre' to make it easier to operate, which is nice. (It's removable because its very close proximity to the 3.5mm headphone socket potentially prevents insertion of fat-bodied headphone plugs.)
The MixPre‑6 II review machine was running firmware version 4.01 and it appeared to be very stable, with no obvious bugs found during my time with it. It's just as delightful, straightforward and familiar to use as its predecessor, and it sounds just as good too. The Kashmir preamps really are special. They offer 76dB of gain, and an incredible 96dB maximum gain from input to recorded file, if necessary, with an impressive headroom margin to boot!
I try to avoid using limiters — I just allow more headroom if I'm concerned about unpredictable peaks, because the ambient acoustic noise-floor will almost always be higher than the recording system's noise-floor, even with 30dB of headroom margin. But for those who like the reassurance of limiters the facility on the MixPre‑6 II is pretty effective in its default factory setting, although it's still an 'all-or-nothing' function; limiters can't be allocated to individual channels. Switching to the custom mode allows the ratio to be adjusted between infinite, 20:1, and 10:1 settings; the release time between 50 and 1000 ms; and the threshold between -2 and -12 dBFS. I'm not convinced these options will prove particularly useful in the field, although I found reducing the release time to 50ms made its action a little less obvious on brief transients. Sensibly, the limiters are only available when recording in the 16 and 24-bit fixed-point modes — they're irrelevant in the 32-bit floating-point mode, since that captures the entire dynamic range capability of the preamps anyway.
Given the prominence given to the 32-bit floating-point mode feature in the marketing literature, I was a little surprised to find that this isn't the default operating mode — to employ it, the user has first to navigate through the 'advanced' operations menus, and since this mode is probably of most relevance to novice sound recordists who are yet to fully grasp the concepts and application of headroom and gain structure, this strikes me as a strange decision in a way.
But it does mean that the default mode will be one that's more likely to suit most professional users. Personally, I'm not convinced the floating-point option really offers a significant benefit for most recording situations, because fixed-point 24-bit recording is still a perfectly practical format when used with skill and care, and there remain some practical advantages in its greater compatibility and slightly smaller file sizes, particularly when recording long takes (because of the need for fewer file splits).
Still, I can see the attractiveness and convenience benefits of the 32-bit floating-point mode for some users, of course, and I think it's a technology that we're likely to see increasingly often in future location recorders and video cameras; Zoom already offer something similar in their new F6 recorder, for example. Sound Devices' bespoke system appears to be particularly sophisticated and well implemented, and even though not all DAWs can currently handle the files correctly I would expect those few manufacturers that aren't yet fully compatible with the format to issue updates to address this fairly quickly.
All of the interfacing and control features associated with the previous generation also still work with the new models, including the Wingman App and several USB remote-control surfaces such as the Novation LaunchControl XL and the Korg NanoKontrols. Overall, then, this second generation of MixPre recorders is very welcome and raises the bar for location recorders still further. Sound Devices could still improve the machines' appeal to the pro-audio and project studio market by adding a few obvious options and facilities, but that doesn't prevent the MixPre II recorders already being extremely impressive and attractive devices indeed.
The various Zoom models remain the closest competitors, especially the new F6 model, which offers a 32-bit floating-point record option.
Floating-point is not a new format; most DAWs have been using it in one form or another for several decades. It's popular because it's a very convenient number-crunching format for audio signal processing, and not least because it effectively removes the traditional dynamic-range limitations as the noise floor and clipping point are around 1500dB apart! So it really doesn't matter how quiet or loud the signals are inside the DAW, or how far you pull the faders back or push them up when mixing — it's almost impossible to lose the signal in noise or crush it against the ceiling.
In simple terms, the way it works is that the audio information is contained in a 24-bit chunk of data called the 'mantissa' which is effectively the same as the 24-bit fixed-point format we use for physical interfacing via, say, AES3 or ADAT. In addition, though, the relative amplitude of that audio data chunk can be scaled up or down using an 8-bit multiplier called the 'exponent', and 24+8 bits together make up the 32-bit floating-point format.
This is all well and good for signal processing inside a DAW, I hear you say, but how is the raw audio coded into this format in the first place inside Sound Devices' new MixPre recorders, given that standard A-D converters only produce 24 bits (and most can only code useful audio data onto 22 of those bits anyway!).
The answer is a very clever and entirely proprietary technique, for which a patent has been granted. It involves several preamp and A-D converter combinations, each arranged to have different signal sensitivities, and the digital outputs from the different A-D converters are analysed and the relevant data combined in the machine's DSP to create the required floating-point output.
In essence, the idea is that a quiet signal is captured by the most sensitive preamp and A-D combination, while the loudest signal is captured by the least sensitive, and a third (or fourth) preamp/A-D deals with mid-range signal levels. The machine's DSP monitors the outputs from all of the A-Ds to work out the actual signal level range (to generate the appropriate exponent value) and extracts the 24-bits of useful audio content by combining relevant data from the different A-Ds.
Other manufacturers have used broadly similar 'stacked A-D' schemes before, dating right back to the earliest days of digital audio in the 1970s, but perhaps the first significant commercial application in pro-audio was Neumann's D-01 digital microphone introduced in 2001, which uses two A-Ds with different sensitivities to digitise the capsule's output directly. However, the Sound Devices arrangement involves a fresh and very clever approach — specifically in its novel use of multiple preamp stages as well as multiple A-Ds — which, naturally, the company feels offers a significant advantage over all previous manifestations.
The practical upshot of Sound Devices' new technology is that the signal-conversion process can genuinely cope with a total dynamic range of 142dB — the limiting factors being the noise floor of the most sensitive preamp (at -128dBu), and the clipping level of the least sensitive (at +14dBu) — with the entire range being accessible all the time. Just to put that in context, a Sennheiser MKH50 with its pad switched in will start distorting heavily just before the MixPre's preamp at around 142dB SPL, while its self-noise will still be captured perfectly above the preamp's own noise-floor!
I've used much earlier generations of 'stacked-ADC' converters — back in the late '80s and early '90s when 20-bit converters were still a real engineering challenge — and I recall issues with moving noise floors causing a kind of noise modulation effect. For that reason I deliberately recorded some 'bongs' with the MixPre‑6 II, this being a well-known engineering signal for testing converter linearity. In my case, I close-miked a large crystal trifle bowl and tapped it with a spoon to create a sharp transient with a smoothly decaying ring. Listening critically to the recording I couldn't detect anything untoward, so it seems that Sound Device's new technology works very well indeed!
The 32-bit floating-point format used by the second-generation MixPres is IEEE 754, which is compatible with many current DAWs. However, it appears some DAWs still have a few wrinkles to iron out, particularly in the way they derive and display the waveforms of floating-point files, and unfortunately some still cannot work with floating-point files at all at the time of writing. Thoughtfully, then, all MixPre II models have a Remix/Rerecord facility, which allows existing floating-point files to be replayed and remixed inside the machine, with gain adjustments, if necessary, to create new fixed-point files that can be used in any DAW.
Pleasingly, Sound Devices' floating-point files work perfectly with Adobe Audition, Audacity, iZotope RX (I tried versions 5, 6 and 7, and they all worked fine, although I had to split the individual tracks from the polyWAVs), and Reaper 64. They also work in Pro Tools 12 up to a point; the audio sounds fine, but there's currently a problem with the waveform display which can show clipping even when the signal level is pulled back below 0dBFS. Sadly, Apple's Logic Pro X can't work with Sound Devices' floating-point files at all at present. Sound Devices' format also works with Adobe's Premiere Pro and Apple's Final Cut Pro X, although both suffer the same waveform clipping display issues as Pro Tools 12. Avid's Media Composer can't currently work with the floating-point format at all, and neither can DaVinci's Resolve.
While the introduction of the new MixPre MkII models might be considered to render the original versions obsolete, I don't think that's a fair assessment. The six first-generation models (MixPre-3 and -3M, -6 and -6M, and the -10T and -10M) all remain superbly capable, high-quality portable digital recorders in their own right, and will continue to provide reliable top-notch service for many years to come.
And that's just as well, because there's no upgrade option to convert the original models into their MkII counterparts — the internal differences are just too substantial to permit that. Nor can the original models be traded-in with Sound Devices in part-ex for the revised versions, although some retailers may be prepared to offer that facility.
However, Sound Devices have stated that they will continue to support the original models with any appropriate or necessary firmware updates arising from their development of the second-generation models, subject to the original hardware being able to support them.
- Preamps sound superb, with massive clean gain and headroom.
- Still has everything that came before... plus some added bells and whistles!
- Free Ambisonics plug–in option with binaural down-mix.
- Timecode generators on all models.
- 32-bit floating-point recording mode option.
- Limiters are still not selectable to individual channels.
- No alternative to polyWAV files in standard mode.
These second-generation location recorders improve on the already impressive first-generation models with new features and a few worthy upgrades, most notably a very sophisticated option to record in 32-bit floating-point format to capture a genuine 142dB dynamic range.