Universal Audio Apollo X16 Review

[beeface]

Rack mounting and power brick? What?

Where to put the power brick?

As someone who works in the AV industry, I can assure you that this is incredibly commonplace. The power bricks usually get mounted to the side of the rack using cable ties or hook & loop strips.

 

[dfuller]

As someone who works in the AV industry, I can assure you that this is incredibly commonplace. The power bricks usually get mounted to the side of the rack using cable ties or hook & loop strips.

Can confirm - or, in studios, they often just hang out the back of the rack where nobody can see them.

 

[sam_adams]

I can't speak to product plans, but it's worth noting why we like Thunderbolt.

...

We do make USB devices like our Apollo Solo USB and our Apollo Twin USB for people who just want a single desktop device.

Dave, could you have Santa leave an X16 in my Christmas stocking this year? I've been a good boy. I promise!

 

[Francis Vaughan]

Thunderbolt has always seemed to me to be a more Macintosh thing, whilst most PCs are stuck with USB. This may be okay because most studios prefer Mac but you should provide options and also cater more towards PC users as well.

It is a shame Thunderbolt gets painted as a Apple/Mac thing. The one sentence summary is that that its is a couple of lanes of PCI with all the neat smarts invented for Firewire for connectivity (hot reconnection, reconfiguration, clock management, isochronous guarantees etc). There is nothing all that is specifically Apple about it. It isn't hard to find PCs that have Thunderbolt, but it is impossible to find a Mac without.
In terms of utility in building anything slightly complex it beats USB totally. It is just that most consumer level users will never need its capabilities.
The level of backward compatibility possible is remarkable as well. I have an ancient Apogee Duet - the original 400Mb/s Firewire version I use now as a desktop DAC. It lives on the end of a Thunderbolt-2 to Firewire converter, which itself lives on the end of a Thunderbolt-3 to Thunderbolt-2 converter, and despite deprecation of support by Apogee, the entire thing works perfectly in Catalina.
Providing access via a PCI lane does raise interesting security issues, so it is perhaps not all roses. USB devices can't initiate transfers into your machine.

 

[Blake Klondike]

This is a review and detailed measurements of the Universal Audio Apollo X16 Interface (ADC, DAC and DSP). It was kindly sent to me by the company for testing and costs US $3,499. The X16 is the longest requested interface for us to measure going back to when I had barely started measuring audio electronics. Interest was driven by seemingly impossible distortion and noise specifications. So it is very fortuitous that we finally get to test it.

The X16 is designed for rack mounting:

View attachment 94384

The only computer interface is Thunderbolt ()TB). It was hell getting TB capability on my PC motherboard. Once there, I had to buy a cable which at $40 (from Apple) for a 2.5 foot version, was outrageous. Then again the interface is running at ultra high-speed with very low latency so the passive cables are short by definition. This made it hard for me to see the display on the unit so I have no feedback on it. However, I did use the extensive software package/control that came with the unit, to the tune of 600 megabytes!

As the name indicates, you have 16 channels in and out:

View attachment 94385

DB-25 connectors are used so if you don't have such break out "snakes," you need to buy them as I did. They are commonly available. Just make sure the mating is correct. Mine was a no name $50 one for the ADC portion which worked fine. For the DAC functionality I used the additional monitor out so didn't need to buy that snake.

As you see, a hefty external power supply is used, reminding me of the original Xbox power supply. Not that you would want to mess with replacing it but note that it mates with a 4 pin, XLR like DC connector to the unit.

I ran into a major stumbling block in testing the X16 in that it only exposes an Audio interface for 64 bit applications. The Audio Precision software is 32 bits so would not see it at all. My Roon player did as did Adobe Audition. If you are going to use this for hi-fi applications, be sure your app is 64 bits. My solution to testing was using the AES input on the X16 for most of the measurements. I provide a dashboard readout using Thunderbolt by playing the test tone using Roon player.

This review is focused on electronic performance of the X16. There is tons and tons of functionality in the X16 that I won't touch on, nor am I qualified to do so. Please seek out other reviews for that functionality which is likely responsible for good bit of cost of this unit.

Lots of programming capability exists to set up the unit. Per recommendation from UA, I set the unit to 24 dBu output. The default is 20 dBu, with the extra voltage left for headroom. Fortunately performance is just as good at 20 dBu.

FYI I reviewed the measurements with UA and they are fine with the results.

Apollo X16 Measurements: DAC
Let's start with our usual dashboard pushing bits using Thunderbolt:
View attachment 94387

Note that I have heavily reduced the output level to our standard 4 volt output. We get stellar distortion rating of -135 dBFS (20 dB below threshold of hearing). Being designed for higher output voltage, the SINAD which includes noise and distortion, is excellent but slightly below the best we can get on our desktop DACs:

View attachment 94388

If we untie the wings from the X16 however to push higher output level, it reaches the same heights:

View attachment 94389

SINAD is now clocking at 120 dB. Note that this is driven by the sum of the Audio Precision analyzer noise and X16. If you subtract the former, you get close to what the company has specified (123 dB THD+N).

Performance is same for AES/EBU digital input:

View attachment 94390

I ran into another common bug in Audio Precision software in that it has some kind of buffer underrun which drops bits. This causes the measurements to get corrupted. To avoid this, I had to turn off averaging to get a single snapshot so ignore the more busy noise floor of the FFT. As you see, performance is identical so we will use AES from here on.

Here is our dynamic range:
View attachment 94391

Company spec is 127 dB but it is with a-weighting.

Distortion is kept to a minimum especially in low frequencies where there is usually a rise:
View attachment 94392

Intermodulation distortion vs level naturally varies a bit depending on the output voltage:

View attachment 94393

Linearity is absolutely nailed:

View attachment 94394

Filter response is typical but with excellent attenuation:

View attachment 94395

As you see, the bottom fell out of the graph!

Jitter and spurious tones are of course kept under control regardless of input type:
View attachment 94396

Finally, a very clean THD+N versus frequency despite its much wider 90 kHz bandwidth:
View attachment 94397

Apollo X16 Measurements: ADC
For the target market, analog conversion is everything so let's see our dashboard there:

View attachment 94398

This places the X16 high on the ladder but not the top:

View attachment 94399

Sweeping the input level shows that this is about the best we can get:

View attachment 94400

Notice the very low noise level (SINAD above is dominated by distortion). You can see this in better dynamic range than SINAD:

View attachment 94401

Here is a drill down of a 1 kHz tone relative to RME ADI-2 Pro (non-FS version):

View attachment 94405

As you see, it is the third harmonic which holds the X16 back (in red). On others it tends to be lower.

At lower than max input, the X16 competes at the top of the class:
View attachment 94406

I had no luck trying to get the ADC to operate at higher sample rates even though I changed it to higher values. Likely an operator error. So this test of frequency response is not that meaningful:

View attachment 94407

Finally, here is our THD+N versus frequency:

View attachment 94408

The much better results once again indicates a clean response above the audio band.

Conclusions
It would have been a watershed event if we had tested the X16 when it came out as it would left all of our high-fi DACs in the dust. In the interim period those companies razor focused on best measured performance and given us incredible performance. Still, the X16 if left to produce its high output, is able to keep up with them and of course provide incredible amount of additional functionality to say nothing of 16 channels input and output.

The high output of X16 should allow you to drive amplifiers such as Purifi and Hypex without their input buffers and get better measured performance than you can with high-fi DACs. But do keep in mind the complexity of using Thunderbolt for interface/programming and need for 64-bit player application.

It is my pleasure to put the Universal Audio Apollo X16 on my recommended list.

------------
As always, questions, comments, recommendations, etc. are welcome.

Man, this garden of ours is refusing to give up this year. Went to the greenhouse expecting everything to have died down but instead, found a great bounty of peppers (Shishido, Bell and Banana), cucumbers, and tomatoes:
View attachment 94412

Made a tuna salad with the cucumber and tomatoes and it was sublime! The fresh and great scent of the cucumbers took me back to wonderful days of summer!

Appreciate any donations using: https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/

Thanks for this-- always exciting to see professional studio gear test this well! Someone posted brief test results for the model I have (UAD Apollo Twin). Is it possible to tell from this data how the Twin compares to the Apollo 16? (I don't know how to read any of the data tables-- I am only here for the summaries and the home-grown food photos!

 

[Dave Tremblay]

Do they measure like the X16 too? That would be interesting.

Our lower cost products have slightly lower specs than our flagship x16, but still very good.

 

[Labjr]

I wonder how Metric Halo compares?

 

[FreaK]

What's the SINAD of the chili peppers and tomatos?

 

[Frank Dernie]

I wonder how Metric Halo compares?

Indeed.
I have a ULN2 and LIO8 with 4 microphone preamps.
Used as DACs I can say they sound the same as the other DACS I own level matched and blind.
In fact it was this that stopped me bothering about DACs 10 years ago since I discovered it is it is in styling, ease of use and functional requirements that they differ, not SQ.

Having written that, for recording the ADC performance, ease of use and software stability are actually super important.

 

[Hector Achilleus]

I wonder how Metric Halo compares?

Not really competitive I guess.
I don't see any AP measurement on web, but for example MH LIO/ULN8 use first generation AKM 192khz DA (AK 4395) by early 2000's : competitive noise performance (in the -120dB realm) but absolutely not SINAD performance (-100db or worse).

Anyway...

Used as DACs I can say they sound the same as the other DACS I own level matched and blind.
In fact it was this that stopped me bothering about DACs 10 years ago since I discovered it is it is in styling, ease of use and functional requirements that they differ, not SQ.

...perfect point.
At these levels of performance -100 vs -120 SINAD is pure gymnic IMHO

 

[JohnBooty]

What's the SINAD of the chili peppers and tomatos?

Their SINAD (Seasonal INgestion of Aromatic Delicacies) rating is around 150dB (delicious Bites)

 

[TNT]

I still think you cherish filter attenuation a little to high in general. For the conversion to work perfectly, everything above fs/2 should be attenuated. I think this means say -120 dB down or more. Here only 6,8 dB down. One need to ponder on a stimuli that really challenge the possible problems with low attenuation just above fs/2.

 

[DSJR]

Despite perhaps not really being entirely suitable for domestic use, I'm glad it's been tested to help show the way for domestic dersigners if nothing else (I do apologise for saying that, as simpler and cheaper products are catching up fast with this model if it's used simply and without the pro extras).

 

[PeteL]

The high output of X16 should allow you to drive amplifiers such as Purifi and Hypex without their input buffers and get better measured performance than you can with high-fi DACs. B

It may, but generally speaking I'd be cautious with this recommendation. NCore/Purifi module are designed to be used with A Hi Zin LowZ out buffer, regardless of the needed input gain. In this case we could be OK. The X16 Zout at 100 ohm is relatively low for balanced line out.. The NC500 module, for example is 1800 ohms (purifi 2200) so we get a ratio of 18 in this case, it's ok but a small number we normally aim for much more, and it's not rare to see balanced line outs with 600 ohms Zout on other products, so we should not suggest that enough output lead to better performance with those modules, altough in this particular case it may hold true, by fluke .

 

[Francis Vaughan]

Something that this device exemplifies is the manner in which AVRs are such terrible value. This thing tears the head off even the most top of the line AV Pre-pros in terms of both analog and digital performance. The difference is that it isn't paying licensing costs to that whole gamut of money grubbing IP vendors whose badges adorn the front of AVRs.

 

[jaykay77]

Just curious...in an audiophile or home theater setting how would one utilize all of those channels? What’s the implementation?

 

[AudioSceptic]

This is a review and detailed measurements of the Universal Audio Apollo X16 Interface (ADC, DAC and DSP). It was kindly sent to me by the company for testing and costs US $3,499. The X16 is the longest requested interface for us to measure going back to when I had barely started measuring audio electronics. Interest was driven by seemingly impossible distortion and noise specifications. So it is very fortuitous that we finally get to test it.

The X16 is designed for rack mounting:

View attachment 94384

The only computer interface is Thunderbolt ()TB). It was hell getting TB capability on my PC motherboard. Once there, I had to buy a cable which at $40 (from Apple) for a 2.5 foot version, was outrageous. Then again the interface is running at ultra high-speed with very low latency so the passive cables are short by definition. This made it hard for me to see the display on the unit so I have no feedback on it. However, I did use the extensive software package/control that came with the unit, to the tune of 600 megabytes!

Yes, Apple cables are a rip-off, but I see that even the cheapest TB cables at Amazon UK are around £20 for 2 metres. Perhaps they just can't be made below a certain price?

Was it really hell getting a TB connection on your PC? I thought that would be trivial by now. This article is from 2015! <https://uk.pcmag.com/laptops/45633/what-thunderbolt-3-means-for-pc-connectivity-an-explainer>

 

[AudioSceptic]

Just curious...in an audiophile or home theater setting how would one utilize all of those channels? What’s the implementation?

I can't imagine this being used anywhere outside a studio.

 

[ElNino]

It is a shame Thunderbolt gets painted as a Apple/Mac thing. The one sentence summary is that that its is a couple of lanes of PCI with all the neat smarts invented for Firewire for connectivity (hot reconnection, reconfiguration, clock management, isochronous guarantees etc). There is nothing all that is specifically Apple about it. It isn't hard to find PCs that have Thunderbolt, but it is impossible to find a Mac without.
In terms of utility in building anything slightly complex it beats USB totally. It is just that most consumer level users will never need its capabilities.
The level of backward compatibility possible is remarkable as well. I have an ancient Apogee Duet - the original 400Mb/s Firewire version I use now as a desktop DAC. It lives on the end of a Thunderbolt-2 to Firewire converter, which itself lives on the end of a Thunderbolt-3 to Thunderbolt-2 converter, and despite deprecation of support by Apogee, the entire thing works perfectly in Catalina.
Providing access via a PCI lane does raise interesting security issues, so it is perhaps not all roses. USB devices can't initiate transfers into your machine.

I agree. The absolute nicest feature of Firewire IMHO was the ability to synchronize clocks on multiple devices over the bus without needing to use a word clock. So, for example, you can have a Topping D90 driving your L/R speakers and synch a MOTU 828mk3 to the D90's clock for a bunch of sub and surround channels just through the transport alone. (I'm not talking about creating an aggregate device, which requires ASRC.)

You can even do this at different sample rates, for example synch a device running at 44.1kHz to a device running at 88.2kHz without needing two word clocks and something like a Big Ben that supports multiple word clocks that are multiples of each other. So you can do your DSP at 44.1kHz and have an upsampling app do 44.1->88.2 with no risk of clock drift.

Unfortunately, this feature requires low-level driver support and was only supported by a few manufacturers in Firewire's heyday. CoreAudio still supports it, but implementing modern driver support is next to impossible because Apple's documentation has gotten poorer and you really need to find developers with tribal knowledge. It still works on older Mojave-and-earlier computers though.

 

[AudioSceptic]

I agree. The absolute nicest feature of Firewire IMHO was the ability to synchronize clocks on multiple devices over the bus without needing to use a word clock. So, for example, you can have a Topping D90 driving your L/R speakers and synch a MOTU 828mk3 to the D90's clock for a bunch of sub and surround channels just through the transport alone. (I'm not talking about creating an aggregate device, which requires ASRC.)

You can even do this at different sample rates, for example synch a device running at 44.1kHz to a device running at 88.2kHz without needing two word clocks and something like a Big Ben that supports multiple word clocks that are multiples of each other. So you can do your DSP at 44.1kHz and have an upsampling app do 44.1->88.2 with no risk of clock drift.

Unfortunately, this feature requires low-level driver support and was only supported by a few manufacturers in Firewire's heyday. CoreAudio still supports it, but implementing modern driver support is next to impossible because Apple's documentation has gotten poorer and you really need to find developers with tribal knowledge. It still works on older Mojave-and-earlier computers though.

It's a shame FW always cost more than USB as it was better in so many ways. I fear that TB might also remain a niche interface, despite being an Intel (not Apple) innovation. The price of TB peripherals is so high that few bother with it, and USB 3 is fast enough for most. Does anyone know if licensing alone is responsible for the high prices?