(re) Review of Sound Devices MixPre-3 II Multichannel Recorder Review - Advanced mode

[RexrothPigeon]

I have a 788t and a MixPre 10ii and while I like them both and they serve my needs well, the 788's pres sound better to my ears. I have not tested them, though, and I definitely do not lay any claim to having "golden ears" either.

They're both very good, useful tools for the right applications, without a doubt, and I have never had any regrets about buying/using them.

 

[Rja4000]

Let me first thank you for this thorough test and starting this very interesting discussion. I am new to this forum, but I am very interested in some questions regarding the performance of these portable mixer-recorders.

You mention the DPA 4006 and similar high quality microphones. This is a good example. At the reference level of 94 dB SPL this microphone has a 79 dB signal-to-noise ratio (with the noise measured A-weighted, with flat rating just 67 dB). From a listener's perspective, 94 dB SPL is quite loud. A microphone can, of course, encounter way louder sounds in a classical orchestra, but those are loud levels where the ear doesn't hear as finely as otherwise. A very quiet audience probably also creates about 30 dB SPL background noise. What I am getting to is that the noise of the microphone probably would mask any distortion that is more than 80 dB lower than the signal. Why would such a microphone deserve a better signal path than the MixPre-3 II offers?

Or, what means "better"? In the 1970s it became clear that harmonic distortion, SINAD and even intermodulation distortion measurements don't fully correlate with the listeners' subjective judgement. How can the musicality, the natural quality of such a device be properly measured? Transient response, thermal stability, power supply reserve and other characteristics might also be in play.

I am mentioning all these because of my most on-topic question: how come the MixPre-3 II has such a gorgeous sound? I have heard a lot of beautiful sounding sound equipment, and Sound Devices preamplifiers and ADC architectures have alway struck me as just purely beautiful sounding. Especially the MixPre (and I assume the virtually identical preamps of the 8-series) is virtually noiseless, and every little detail sounds natural and clean. How come a sound recorder with - based on the review - rather poor characteristics can sound so beautiful? My question is honest, I am amazed at the expertise and measurement methods you and Amir have applied, and I would be grateful for any advice regarding how to measure if my impressions are objectively correct.

Hello
And welcome to Audiosciencereview !

Your question, if I summarize, is:
"Why should I care about the not-so-good measurements, because anyway one can't hear such level of distortion and I think it sounds great".
Correct ?

Let's start by the first topic: can we hear such distortion ?

You're using DPA microphone as an example.
They've been quite active on marketing their actions for lowering distortion on their mini Microphones recently.
Let's read what they say about microphone distortion:

The basics about distortion in mics

Before considering the amount of distortion, we must define the various forms of distortion we meet in our equipment – and in this case – with a special focus on microphones. We must also consider the hearing system (the ears and brain), which sets the real limits for audibility.

www.dpamicrophones.com

In general, there should be no distortion in your microphones. However, there is. Measures like THD and difference frequency distortion measurements do not tell the full story. However, the numbers in the specs can be regarded as an indication of how "healthy" the design is. What is perceived by the ear is rather complex because the ear produces distortion itself. Despite the numbers, the more distortion in your microphone, the muddier and more unclear the sound gets in your recording.

That's exactly the idea here.
The bad measurements are a SYMPTOM of a design that didn't get enough attention.
Probably (my reading) because the company thought spending more on marketing would give more return on investment than on engineering.

Nowadays, there is NO excuse for low engineering quality, something poor measurements gives a warning about.
Specialized chips give better performance at much lower cost, without need for in depth engineering of all details or painful and expensive components matching.

Will what we see here make good recording impossible ?
Of course not !
We are not at that level.
And the device, as I noted, has some very good performance for noise, a more immediately audible quality factor.

You may do excellent sounding recordings with much older and weaker hardware, if you know what you're doing and are careful in microphone selection, positioning and in gain setting.

Is that so that the distortion measured here will never be heard ?
We can not be sure.
There may be some cases where there will be more audible distortion artifacts.

So why to take the risk, if you have a choice (*), while other hardware don't show this kind of symptoms at similar or lower cost ?

That's what we do here:
If a device passes all measurements without issue, we know with good level of probability, that it won't affect negatively what you hear.
If not, we document it, and warn people about what we see.

After that, the informed decision is yours.


(*) "If you have a choice", because this device may be the best practical choice for quite a few situations.
And I'm sure that using a device that is not a good match for your functional, practical requirements is a perfect recipe for really bad results.

 

[andorrr]

Hello
And welcome to Audiosciencereview !

Your question, if I summarize, is:
"Why should I care about the not-so-good measurements, because anyway one can't hear such level of distortion and I think it sounds great".
Correct ?

Let's start by the first topic: can we hear such distortion ?

You're using DPA microphone as an example.
They've been quite active on marketing their actions for lowering distortion on their mini Microphones recently.
Let's read what they say about microphone distortion:

The basics about distortion in mics

Before considering the amount of distortion, we must define the various forms of distortion we meet in our equipment – and in this case – with a special focus on microphones. We must also consider the hearing system (the ears and brain), which sets the real limits for audibility.

www.dpamicrophones.com

That's exactly the idea here.
The bad measurements are a SYMPTOM of a design that didn't get enough attention.
Probably (my reading) because the company thought spending more on marketing would give more return on investment than on engineering.

Nowadays, there is NO excuse for low engineering quality, something poor measurements gives a warning about.
Specialized chips give better performance at much lower cost, without need for in depth engineering of all details or painful and expensive components matching.

Will what we see here make good recording impossible ?
Of course not !
We are not at that level.
And the device, as I noted, has some very good performance for noise, a more immediately audible quality factor.

You may do excellent sounding recordings with much older and weaker hardware, if you know what you're doing and are careful in microphone selection, positioning and in gain setting.

Is that so that the distortion measured here will never be heard ?
We can not be sure.
There may be some cases where there will be more audible distortion artifacts.

So why to take the risk, if you have a choice (*), while other hardware don't show this kind of symptoms at similar or lower cost ?

That's what we do here:
If a device passes all measurements without issue, we know with good level of probability, that it won't affect negatively what you hear.
If not, we document it, and warn people about what we see.

After that, the informed decision is yours.


(*) "If you have a choice", because this device may be the best practical choice for quite a few situations.
And I'm sure that using a device that is not a good match for your functional, practical requirements is a perfect recipe for really bad results.

Thank you for the thorough answer! I completely agree that a low SINAD value is often indicative of a poor design. But I would summarize my question differently:

How much (and which kind of) THD does realistically affect sound, and is this really the best indicator of what sounds best?

In this case, Sound Devices is a very respectable and high quality manufacturer. They are proud of their discrete transistor inputs and, for example, service everything they ever produced. Negligence is not very likely in their case. Their priorities, as far as I can tell:
1. extremely rugged, reliable and compact construction
2. very flexible redundant powering
3. extremely low equivalent input noise (while 2. is maintained)
4. very high CMRR (especially important with long-wired microphone connections)
5. more dynamic range and preferably lower noise and distortion than the recorded sources.

Now, discrete transistor input stages might reduce noise or can eventually provide better CMRR, but is likely to introduce more distortion. Historically, they have also built a lot of products with input transformers and optocouple-based analog limiters which are also infamous for their harmonic distortion and often improve the overall quality of a recording..

And there is this 32-bit float mode. In practice, it is of great value that you don't have to be afraid of clipping the audio. For example, if the orchestra is louder than during rehearsal, or, in a documentary film, if someone suddenly shouts during a generally soft conversation. But combining the signals of the two ADCs is troublesome business and is also likely to introduce distortion.

So, it is true that the distortion measurements (and, important to note: the specifications of the recorder) indicate worse-looking values than most state-of-the-art devices. If we suppose that the reason is not poor design, how can we measure whether the widespread subjective opinion that these recorders sound wonderful is objectively right?

I am mentioning this because in the 1970s some people (Jean Hiraga of L'Audiophile comes to mind) realized that THD and IMD are not enough to explain why some very well specified devices sound worse than, e. g. old valve amplifiers. They were trying to measure things like transient intermodulation distortion, thermal distortion and other things. I was mostly wondering if this or other measurements would be of any interest in audio interface measurement.

 

[RexrothPigeon]

Sound Devices is a very respectable and high quality manufacturer

Yes and no--it was acquired in 2021 by a multinational holding company.

 

[andorrr]

Yes and no--it was acquired in 2021 by a multinational holding company.

I don't know if that should affect quality. If so, that is bad news: the same company owns DPA Microphones, Austrian Audio, Wisycom, Harrison, SSL. Hopefully Audiotonix treats them well because otherwise there will be quite a large market gap in the pro audio industry.

Circumstantial data: the Scorpio is about 7 years old, and so are the MixPre II recorders. The previous generation (6-series) was brought out about 7 years before, and the 744T and 788T I think about 5-7 years before that. So if we don't see innovation in the next 1-2 years, we might conclude that something changed. My subjective impression (having been working with SD recorders for the past 15 years) is that the build quality is still the same.

 

[Rja4000]

Thank you for the thorough answer! I completely agree that a low SINAD value is often indicative of a poor design.

That's a good starting point.

in the 1970s some people (...) realized that THD and IMD are not enough to explain why some very well specified devices sound worse than ...(...)
They were trying to measure things like transient intermodulation distortion, thermal distortion and other things.
(...)
I was mostly wondering if this or other measurements would be of any interest in audio interface measurement.

it doesn't have to be that complicated.

We here think that below some threshold for some carefully selected measurements, we will not be able to hear the influence of a device.
We can then consider the device that is below the threshold for all those measurements to be 'transparent'.
Which is the ultimate goal in "Hi-Fi" - for "High Fidelity".

The good news is:
Devices reaching those thresholds are becoming more and more common.

Unfortunately, and that's a fact, this one is not one of those.

So I'm afraid, because of that, we won't be able to confirm here that the

subjective opinion that these recorders sound wonderful is objectively right

This will remain a subjective opinion.

Whatever the trust you have in that company, they missed this one.

 

[RexrothPigeon]

My subjective impression (having been working with SD recorders for the past 15 years) is that the build quality is still the same.

My subjective impression (having been working with SD recorders for the past 18 years) is that the build quality has slipped slightly and the sound quality has slipped less slightly.

 

[andorrr]

That's a good starting point.



it doesn't have to be that complicated.

We here think that below some threshold for some carefully selected measurements, we will not be able to hear the influence of a device.
We can then consider the device that is below the threshold for all those measurements to be 'transparent'.
Which is the ultimate goal in "Hi-Fi" - for "High Fidelity".

The good news is:
Devices reaching those thresholds are becoming more and more common.

Unfortunately, and that's a fact, this one is not one of those.

So I'm afraid, because of that, we won't be able to confirm here that the

This will remain a subjective opinion.

Whatever the trust you have in that company, they missed this one.

Thank you for your honesty. I was hoping to learn something new. You write about a certain threshold for measurements, but you didn't answer my question: where is that threshold? Keeping in mind that these exact measurements would be several orders of magnitude worse in the case of the best microphones, let alone loudspeakers?

About the carefully selected measurements: I do believe that what you praise performs exceptionally in every respect, measured or not. I also understand that you are convinced that the 1970s realization that harmonic and intermodulation distortion is not the most defining factor in sound quality, is false.

I admire your measurements and the work you put into them, even if I don't find your conclusions 100% convincing.

 

[Rja4000]

You write about a certain threshold for measurements, but you didn't answer my question: where is that threshold?

We typically consider that the minimum is to overperform CD quality in normal use (which is not playing at full level), so some margin is needed.
110dB for SINAD is typically the target for DAC.

Note that it may be overkill, but our goal is to set the bar high enough to cover all cases and bring enough confidence in transparency.

It's not a 'required' condition but it is for sure 'sufficient'.

We can set the bar high because more and more affordable devices outperform that condition.

The device under test here fails the conditions for distortion by a considerable margin.

Keeping in mind that these exact measurements would be several orders of magnitude worse in the case of the best microphones, let alone loudspeakers?

Do you mean that microphones can't reach those thresholds ?
I think you're wrong.

A microphone like the DPA 4006 is capable of dynamic range of over 120dB and the distortion level at, say, 94 db SPL is unpublished because it's not measurable, but most probably it's below -100dB at that level.
Noise is given for 15 dB SPL(A) at 94 db SPL. That's 79dB SNR.
We can measure distortion below noise level, and the actual distortion is said to be unmeasurable, so way below that level.

The device here has distortion levels at the level of the microphone noise at 94dB SPL. So way more distortion than the microphone itself.

And distortion, like noise -if not in the same way- is additive.

Commonly admitted rule of thumb says that you'd need your device to perform 10dB better than your microphone to not have a significant impact. If this is the case for noise, we are far from that for distortions, and especially for IMD, which is less subject to masking.

I also understand that you are convinced that the 1970s realization that harmonic and intermodulation distortion is not the most defining factor in sound quality, is false.

I specifically wrote that noise is usually more audible than distortion.
I don't believe the distortions are very audible.

The fact is that, in 1970, NO (or very few) device was performing at the level we set the bar today.

So the concern of people with measurements at the time was "How do I define with measurements which of the flaws I measure is less painful than others".

They also had much more limited measurement hardware: the level of measurement an amateur like me can do today is light-years ahead of what was achievable with the best equipment by then.

All in all, that allows us to say: why to bother sorting the flaws if we can make reasonably sure we don't have any ?
And we can.
And quite a few devices reach that level, even at modest price.

So do we want some manufacturers to continue not to perform as good, just because they have a reputation and nobody will challenge them ?
Or do we want to send them a message:
"Please do better. Because you can."
?

We do that for Hi-Fi equipment.
Why would we lower the bar for the equipment which is providing the very source material for the recordings we feed our Hi-Fi systems with ?

 

[RexrothPigeon]

Why would we lower the bar for the equipment which is providing the very source material for the recordings we feed our Hi-Fi systems with ?

well, to be clear, not a lot of music is recorded on Sound Devices recorders. They're in wide use in journalism, reality TV sound, etc. (I know a lot of tapers use them for live bootleg-style recordings, but I'm talking about professionally produced music.)
so even if they don't rise to hifi expectation standards, they are still great tools because of all their other functionalities (they really pack a lot in a small, durable package).

 

[Rja4000]

not a lot of music is recorded on Sound Devices recorders.

As I explained in the first posts, I see in a group about Classical music recordings quite a few people using those (or the bigger models) with expensive microphones.

they are still great tools because of all their other functionalities (they really pack a lot in a small, durable package).

I never said something different.
Functionality is the first thing you need to care about.

But I suppose you'll agree with me when I say that the same set of functionalities but with better measurements wouldn't hurt ?
And it doesn't have to come with a higher price.

 

[andorrr]

Thank you for describing the threshold! This sounds completely reasonable. I am not sure though if it applies to portable recorders, which have to (also) meet different criteria (apart from the already mentioned, temperature range now comes to mind as an obvious difference - indoor audio interfaces usually specifically prohibit freezing temperatures). In any case, there is no product that I know of that would reach the mentioned threshold while meeting these extra criteria. You did mention the different feature set, but these are partially differences in signal handling.

If there are a number of current devices that provide a very low noise, THD, IMD and jitter in lab environments, would it possibly be worth looking at the differences that exist between them? Temperature dependence, transient artifacts, aliasing, microphony (probably negligible), CMRR, mains power noise filtering and many other things could be interesting.

 

[andorrr]

As I explained in the first posts, I see in a group about Classical music recordings quite a few people using those (or the bigger models) with expensive microphones.


I never said something different.
Functionality is the first thing you need to care about.

But I suppose you'll agree with me when I say that the same set of functionalities but with better measurements wouldn't hurt ?
And it doesn't have to come with a higher price.

That is where I am not sure: I don't know if you can reach a gain-setting-free 142 dB input range with extreme temperature tolerances, high CMRR and very flexible powering in an incredibly small package while providing those better measurements _for the same price_.

 

[RexrothPigeon]

But I suppose you'll agree with me when I say that the same set of functionalities but with better measurements wouldn't hurt ? And it doesn't have to come with a higher price.

Of course, for sure. I wouldn't mind having a sonosax or a nagra...but they definitely come with a higher price. What similarly priced tool with equivalent functionality and durability would you recommend? (genuine question--there may exist something I don't know about!)

As I explained in the first posts, I see in a group about Classical music recordings quite a few people using those (or the bigger models) with expensive microphones.

whoa, for professional release? that's surprising. I've been in all the main sound devices fora for ages and I haven't seen much of this, mostly amateurs or hobbyists on the music side, and pros on the journalism/TV sound side.

I never said something different.

I wasn't arguing against you

 

[andorrr]

well RJA's criticism was more related to distortion levels, not really noise (sinad is the sum of both), and as we see the dynamic range is quite a bit larger than it needs to be but the distortion is the bugger.
.
in general, noise being below the level of mic self noise does not make it inaudible... as noise is additive. in regard to "why would a microphone need a better signal path than the mix pre provides" is because 94db/ 1Pa is an arbitrary measurement number. In reality microphones regularly find themselves in 120 dB environments in loud concerts and close mic'ing isntruments and vocals

"How can the musicality, the natural quality of such a device be properly measured?"

it cant, which is why this site is centered on objective, not subjective aspects, which are measurable

"Sound Devices preamplifiers and ADC architectures have alway struck me as just purely beautiful sounding. Especially the MixPre (and I assume the virtually identical preamps of the 8-series)"

these are actually quite different. even in the 7 series the 788 preamps are completely different than the 702. none of the 8 series use 32 bit float so the front end of the gain structure is almost certainly more old-school. I have a 702 here ill put on the bench. my guess is despite its age it measures better than the modern 32-bit zoom and mixpres. And thats not even a 'great' preamp by SD's own standards: https://www.sounddevices.com/788t-microphone-preamplifiers/

"how come the MixPre-3 II has such a gorgeous sound?"
Does there exist and good comps of mixpres vs other equipment with all other factors being equal? i find these very hard to come by. even attempting them myself in the field its almost impossible to duplicate exact mic angles and spacing, all other factors being equal. And in a live venue a few feet of separation between rigs can sound different as well

"How come a sound recorder with - based on the review - rather poor characteristics can sound so beautiful?"

because the dynamic range of a field recording is inherently limited. A good example would be the dozens of recordings produced at the recent Phish concerts at the sphere. No expense or amount of back-breaking lugging was spared to bring in the best recording equipment possible. Yet at the end of the day the sound in the venue, as state-of-the-art as it is, is not far from that 94 dB SPL you mention above (rare for rock concerts). And if you could isolate the crowd noise and use that as a noise floor youd find that we are working with maybe 50-60 dB max of dynamic range here. The same reason why tapers were able to make excellent recordings with the venerable sony D6 cassette deck for a decade, or alternatively why people swoon over their vinyl collection which struggles to hit 60 dB of noise floor.

i think RJA's point was not that its not a good-sounding useful recorder, more so that it isnt up to modern standards that you find in interfaces costing as little as $100-$200 these days, which is what we would expect from a good design

That said they know their market, and this fits a niche. Everyone i know who has a mixpre or zoom these days are pretty satisfied with the recordings they bring home while lugging a fraction of the gear they had to two decades ago. You used to need a dedicated pre pushing a high level balanced signal into a dedicated ADC then a reliable recorder to take the digital stream, and a big brick of lead to power it. Modern designs leverage the modern ADCs which have specs that didnt exist twenty years ago, and do not require as sophisticated of a supporting circuit architecture to get the most out of it. Thus instead of amplifying signals unnecessarily, we are taking the most direct path to a chip that doesnt want or need signals upwards of 10+V to perform their best.

but we shouldnt be comparing SD performance to stuff that was plenny good enough 20 years ago. We should be comparing it to the performance their competitors are achieving in the modern day. IMO it would be better if we were using studio-quality gear for field recording vs equipment that is "good enough for field recording". Same reason i archive all my recordings in 24 bit despite no human being able to hear the difference between that and a 16 bit recording

The 8-series has the same kind of 32-bit float mode with double ADCs. According to Sound Devices, the preamp is also virtually the same. I assume that the 1dB noise difference is due to an even better power supply in the 8-series, but the THD value is the same for a reason. This is actually a pretty cool thing: Sound Devices apparently packs practically the same audio quality and reliability in the $1000 device as in the $10000 device, just less of it and with way less features (inputs, DSP effects, workflow-optimizing features, data redundancy, knobs and buttons). I admit, I like the products of this company, while I am sure if I was going for a rack-based solution I would obviously choose a different sound interface.

 

[RexrothPigeon]

I am sure if I was going for a rack-based solution I would obviously choose a different sound interface.

Yeah, of course. The use-case of SD recorders is one that requires portability and durability.

 

[andorrr]

Of course, for sure. I wouldn't mind having a sonosax or a nagra...

whoa, for professional release? that's surprising. I've been in all the main sound devices fora for ages and I haven't seen much of this, mostly amateurs or hobbyists on the music side, and pros on the journalism/TV sound side.

I wasn't arguing against you

The current Nagra recorder has an EIN of -117dBu and a 0.1% THD, both significantly worse than SD. Sonosax seems to be out of the game.

 

[chelgrian]

Of course, for sure. I wouldn't mind having a sonosax or a nagra...but they definitely come with a higher price. What similarly priced tool with equivalent functionality and durability would you recommend? (genuine question--there may exist something I don't know about!)

whoa, for professional release? that's surprising. I've been in all the main sound devices fora for ages and I haven't seen much of this, mostly amateurs or hobbyists on the music side, and pros on the journalism/TV sound side.

I wasn't arguing against you

90% of location classical recording is having the right mic in the right place, I doubt you’d see many of anyone in fora because there simply aren’t many people doing it.

What gets used somewhat depends on space I see quite a lot of people with a 2u setup with some sort of multiway mic preamp and a recorder from the likes of JoeCo.

There are also camps which will use RME interfaces and a laptop or Merging interfaces and a laptop.

 

[Rja4000]

That is where I am not sure: I don't know if you can reach a gain-setting-free 142 dB input range with extreme temperature tolerances, high CMRR and very flexible powering in an incredibly small package while providing those better measurements _for the same price_.

I think it could be quite "easy" if they were giving a bit more attention to engineering.

Let's have a look at IMD plot vs level:

My interpretation is that you see this "hill" of increased distortion mainly because, when they switch mic preamp and ADC they duplicate the behavior you see at the max level: you see a steep distortion increase when level gets closer to max.
This is duplicated close to gain switching level.

Now imagine that the individual preamp/ADC behaves like this, with almost no saturation at highest level ?

Don't you think that we would also get a much better profile, even by combining the 2 preamps/ADC ?

OK, I'm also an engineer, so I know it's not that simple.
But you get the idea.
There is room for improvement.


Note: This is for illustration only, don't compare values. Bottom plot is IMD+N while above plot is IMD only.

EDIT:
I processed the THD+N plots and here is what we could hope and what we get.
We see the noise performance is pretty good, but the distortion is taking over from -35dBu to -20dBu and at levels above -3dBu.
At highest levels, we may say we won't go there (?), but between -35dBu and -20dBu, that's for sure not true.

 

[andorrr]

Don't you think that we would also get a much better profile, even by combining the 2 preamps/ADC ?

As far as I know, combining preamps/ADC is far from trivial. The issue goes back, I believe, to the 1970s-1980s and submarine digital sonar developments. There is a gain matching issue and a timing issue. I would think both these can raise the distortion level in the critical ranges.