It’s good to know your gear, and I’ve been a real big fan of Sound Devices’ MixPres because they are generally free of the analog corner-cutting you find in most prosumer stuff. This should be celebrated, because that’s the way the market should be. Companies generally gut analog stages then add all kinds of bells and whistles that no one actually needs or cares about longer than 15 minutes after removing shrink-wrap. SD cannot comment on component-level decisions because, understandably, this is proprietary knowledge they’d rather keep to themselves so to not undercut their profit margin. So allow me…
A/D converter 
detail of two microphone preamps 
D/A converter 
aux input and dedicated a/d, and headphone amp 
another view of aux input and headphone amp 
clocking circuit 
disrobed 
underside of master clock circuit
Internal photos from a mixpre-6 (1st gen), showing two preamp channels, headphone amp and aux input, ADC, and DAC. They indeed feature a discrete front end with dual transistor pairs. On my 10T these are great low-noise Hitachi 2SA1084, which I was surprised to see because they’ve long been discontinued, but in this particular unit they’re using ZTX789.
What is really nice, however, are the super low-noise/low-distortion OPA1662 op amps (marked OUQI) acting as buffers after the transistor pairs. There are two per preamp, doubling the balanced differential signal from the mic all the way to the ADC, AK5578, an 8-channel chip. This means that for the MixPre-6, with 4 preamps*, there are 2 ADC channels per mic input. These are summed together in the ADC itself, which further reduces noise and increases dynamic range. They could easily have used one dual op amp (something bland like NJM4850 or NJM2068) to sum the incoming differential signal to mono, then used a cheaper ADC with four mono channels (or two stereo ADCs, like the Zoom H5). But they didn’t, the signal is double balanced (e.g., Millennia) through low-distortion parts all the way home, which is why these things eat batteries for breakfast.
*The mixpre 10 has two of these 8 channel ADCs
DAC is fantastic, and uses OPA1662 for aux output, and the headphone amp is a super snappy XR8052, with loads of output current, which is why it can drive anything.
The only place they seem to have cut corners is on the aux input, which uses the MC33178 (a kinda blah low-power opamp) and AK5359 ADC, same as Zoom H5/6 with rather disappointing dynamic range and noise specs. This can be substituted with AK5381, which has *slightly* better dynamic range (I did this in my wife’s H5, which I completely overhauled), but this is no longer manufactured and can only be found, at some risk, on eBay. The MC33178 could easily be substituted with OPA1662, or (if battery life is the concern here) OPA1692, a new audio-optimized opamp from Texas Instruments (successor to the 1662) designed specifically for battery-powered applications. Perhaps they’ve improved this in the gen 2 series, I don’t know, I haven’t popped the hood on one. They probably figure it doesn’t matter much because folks will mostly use plugin lav mics or other lesser-than-hifi sources. Just know that if you’re running signal from a nice external preamp or stereo feed from another mixpre as a backup, this is the one audio “bottleneck” (which is to say, it’s *fine*, but Zoom H5 quality).
All the ADC/DAC chips have ample power supply filtering (big capacitors), skimping on this is another common space-saving measure in most cheap gear. All the analog stuff appears to be running on +/-5V rails (10V rail-to-rail). +/-15V rails (30V) would have been even better, allowing even snappier response in analog stages (higher slew rate, for one), but they’ve done so much else WAY better than average, it doesn’t really matter. Most chips in the Zoom H5, for example, are running on a single-ended +5V rail which, plus the low-end ADC, makes for noise, slew rate limiting, and higher distortion.
Anyway, here’s to good sound! Cheers, sound devices!
Windowpane Audio 
Thank you for this article. Do you have any info about MixPre 6 II?
LikeLike
Nah, haven’t popped one open, or even tried one, but I imagine it’s a very similar design, if not identical, with perhaps a more robust DSP chip to handle all the streams of 32-bit audio.
LikeLike
I have opened MixPre 6 II and have made some pictures. It would be awesome if you could comment and say something good or bad 🙂 Is second edition better or worse? 🙂
https://www.dropbox.com/sh/wwrqlgw2ax4mubx/AAC1rsOrrdY-iXiz4IMXQv6pa?dl=0
LikeLike
Looks largely the same, actually, with the exception of the ESS converters, which makes sense. They were using AK before, but there was that big factory fire so they probably had to change their supply chain and therefore their board layouts too. But looks like they’re still using the OPA1662 throughout, which is a nice chip for power saving situations. Looks like they’ve swapped out the NOS Toshiba transistors in the preamps for something else, but the basic arrangement looks the same. I mean, I don’t believe for a moment that they’d make their analog circuitry worse on subsequent revisions, and if anything the 2nd gen mixers are a step up from the first because of the 32-bit functionality.
LikeLiked by 1 person
Thank you.
LikeLike
Did you try putting film bypasses on the 47uf electrolytics blocking P48 DC on the mic inputs? If so, how much benefit? Do you know which past models used transformer coupling on the inputs and if it sounded better, at the cost of older, lower spec. A/D?
LikeLike
Hah, no, I’ve not messed around with this. Actually, I sold the MixPre 6, but still use the 10T for pretty much everything since all my recording is on location. It sounds really excellent, truly, and *maybe* I could eke out some tiny improvements with better caps, or film bypasses, but honestly the risk of mucking something up (because those boards are TIGHT) so far outweighs any potential benefits that I’m just not interested in bothering. Even just getting the boards out of the enclosure is a feat of heroism. Anyway, these are improvements that are so small, simply moving a microphone a couple of inches would blow them out of the water. It makes more sense if you have tons and tons and tons of coupling capacitors in a signal path, but there are only capacitors at the input, and they’re Nichicon caps (not “audio-grade”, but better than some no-name thing you might find in a Scarlett or Mackie), I don’t think there would be any discernible difference.
As for whether the transformer coupled recorders sound “better”? That is so subjective. I have an old MixPre, like, the OG version with Lundahl input transformers, and I modded it, of course, and it sounds beautiful, squeaky clean, nice clear preamps with tons of gain for any mic. Does it sound better? Tough comparison because there’s no AD converter, so it’s still running into another device afterwards. Old SD recorders have much less in terms of dynamic range compared to these new models, esp. the 32-bit recorders, so “better” can easily become a question of whether the old limiters were up to snuff, or you set the gain levels properly, than any sort of “vintagey” color lended by the xformers. I have an old NAGRA IV-S, and those preamps definitely have some color, but are still definitely still on the clean side, plus they used liquid-filled/glass-sealed coupling capacitors that should have an operational life longer than my nervous system…
Anyway, SD makes awesome gear, and the new MixPres are such an amazing value for the cost/size, you really have to go ultra-high end to notice real benefits, especially in sound quality. Even the new Zoom F3 is kind of awesome…
LikeLike
Why not a review NAGRA IVs VS ZOOM F3?
LikeLike
Hah, excellent suggestion! I love both!
LikeLike