Apollo x16claims -129dB thd+N

[helloworld]

Now I guess I know how they get that high number THD+N. They use ess9016 in X6 X8 and X8p based on a review of Chinese website and use ess9028pro in X16 models based on a russian forum. I guess they were trying to narrow down the bandwidth as low as possible which leads to the THD+N very close to DNR.
ESS9016 has a DNR of 124dB, since they use two chips and use A-weighted, they get 129 dB of DNR is possible.
ESS9028pro has a DNR of 129dB, so they get 133dB of DNR is also possible.
For the THD+N prat, I guess it's just a advertisement trick.

 

[ryanmh1]

I think I found the same Chinese website the other day, completely by accident. They run all the pro audio gear they sell through an RMAA loopback. Most of the numbers are realistic, and within the bound of other third party tests. They have an Apollo X6 as measuring .00005% THD. That's -126dB. Incredible. They have the THD+Noise as -110.6dB, which is also pretty incredible considering that this is the composite performance of the DAC and ADC on a loopback, and that this is RMAA which tends to overstate the noise. I suspect RMAA either botched the noise measurement, or the ADC ran into its limits. Test the DAC alone on an APx555 and I wouldn't be surprised to see it beat everything else. My recollection is that Universal Audio is a lot of the former E-MU engineers, who were top notch.

 

[helloworld]

I think I found the same Chinese website the other day, completely by accident. They run all the pro audio gear they sell through an RMAA loopback. Most of the numbers are realistic, and within the bound of other third party tests. They have an Apollo X6 as measuring .00005% THD. That's -126dB. Incredible. They have the THD+Noise as -110.6dB, which is also pretty incredible considering that this is the composite performance of the DAC and ADC on a loopback, and that this is RMAA which tends to overstate the noise. I suspect RMAA either botched the noise measurement, or the ADC ran into its limits. Test the DAC alone on an APx555 and I wouldn't be surprised to see it beat everything else. My recollection is that Universal Audio is a lot of the former E-MU engineers, who were top notch.

Yes, I am not questioning their engineering, I am very confident that they can utilize the best potential of ess chips like RME did in adi2dac.

 

[amirm]

I was pleasantly surprised to see that they now support Windows! If you all can pester them, maybe they offer a loaned unit for testing.

 

[bennetng]

I think I found the same Chinese website the other day, completely by accident. They run all the pro audio gear they sell through an RMAA loopback. Most of the numbers are realistic, and within the bound of other third party tests. They have an Apollo X6 as measuring .00005% THD. That's -126dB. Incredible. They have the THD+Noise as -110.6dB, which is also pretty incredible considering that this is the composite performance of the DAC and ADC on a loopback, and that this is RMAA which tends to overstate the noise. I suspect RMAA either botched the noise measurement, or the ADC ran into its limits. Test the DAC alone on an APx555 and I wouldn't be surprised to see it beat everything else. My recollection is that Universal Audio is a lot of the former E-MU engineers, who were top notch.

I guess I found the same website as well.
https://www.exound.com/articles/f7d33f56-208a-4f0d-8317-9652a2d49b4e

RMAA's results are both A-weighted and 20kHz limited by default. I took the reference 32-bit float signal, added a lot of ultrasonic noise and demonstrated the difference between using vs not using the 20kHz filter, yet the published results from that Chinese website are still slightly worse than the bottleneck (analog input) so I am not very optimistic about their claimed -115dB input and -129dB output THD+N. DNR seems reliable though.
[edit] Didn't notice it is an x6 rather than x16. Their manual said analog line-in has 123dBA DNR and -113dB THD+N so the RMAA result is pretty close to the specs.

On the other hand ADI-2 Pro's RMAA test results closely matched RME's stated THD+N.
http://prosound.ixbt.com/interfaces/rme-adi2pro.shtml
http://prosound.ixbt.com/interfaces/rme/adi2pro/+24.shtml
http://www.rme-audio.de/en/products/adi_2-pro.php

PS: RMAA 6.4.3 and later show one more significant digit, for example a value of 0.03456 in recent version will appear as 0.0346 in older versions.

 

[chriskilbourn]

Sorry, I’m lost trying to keep track of these figures. How would the Apollo x6 DACs compare to the RME ADI 2 DAC? Not sure if there’d be much of an audible difference between the two units.

 

[helloworld]

Sorry, I’m lost trying to keep track of these figures. How would the Apollo x6 DACs compare to the RME ADI 2 DAC? Not sure if there’d be much of an audible difference between the two units.

Apollo x6 has much more features than adi 2 pro, such as mic preamps, 5.1 surrounding system, better dsp, a lot of internal channels for recording. To me it’s the best karaoke interface.

 

[JohnYang1997]

Can we just stop a little bit and get the bullshit out of the way? It seems that many of you don't even know how thd+n can be measured and how noise related to thd+n and how thd works.
Let's first get noise out, 0.000037% of thd alone is very impressive. means each harmonics is optimistically under 0.00002%. No matter how narrow band they used for noise the thd is the limit. If they count only 2nd harmonic it's still great, if they count 2nd and 3rd, it's impressive. Unless they don't count distortion at all but that is completely bullshit. But this has nothing to do with dnr.
Then you can use very cheap equipment to measure this performance. You just need a good notch filter and a noise amplifier.
If you go to okto research's website. You can see 0.000032% of thd (not thd+n). So at least it's an understandable number.

 

[Blumlein 88]

Can we just stop a little bit and get the bullshit out of the way? It seems that many of you don't even know how thd+n can be measured and how noise related to thd+n and how thd works.
Let's first get noise out, 0.000037% of thd alone is very impressive. means each harmonics is optimistically under 0.00002%. No matter how narrow band they used for noise the thd is the limit. If they count only 2nd harmonic it's still great, if they count 2nd and 3rd, it's impressive. Unless they don't count distortion at all but that is completely bullshit. But this has nothing to do with dnr.
Then you can use very cheap equipment to measure this performance. You just need a good notch filter and a noise amplifier.
If you go to okto research's website. You can see 0.000032% of thd (not thd+n). So at least it's an understandable number.

Well the bs is we have to guess how they measured. They should use AES 17 guidelines or specify in detail what they did. When they won't do that then you can be sure they are gaming the numbers a little bit.

 

[JohnYang1997]

I know this is an old thread. And after all this time. After gaining the ability to design.
9038pro has the ability to vastly surpass the output performance of AP. And it's better than AK4499.(less importantly).
In mono configuration, someone has achieved 143dB DNR with A weighting(better than datasheet). And I just made it 140dB(the datasheet performance) with 18Vrms output voltage(to minimize the contribution of AP's noise floor). What's more to be known is yet to find out.
And these professional companies should be capable of deliver however good performance they want just that they prefer to use as cheap components as possible that work and are good enough. This is probably the missing piece I was talking about early this year.
So this performance is very possible.

 

[bennetng]

I know this is an old thread. And after all this time. After gaining the ability to design.
9038pro has the ability to vastly surpass the output performance of AP. And it's better than AK4499.(less importantly).
In mono configuration, someone has achieved 143dB DNR with A weighting(better than datasheet). And I just made it 140dB(the datasheet performance) with 18Vrms output voltage(to minimize the contribution of AP's noise floor). What's more to be known is yet to find out.
And these professional companies should be capable of deliver however good performance they want just that they prefer to use as cheap components as possible that work and are good enough. This is probably the missing piece I was talking about early this year.
So this performance is very possible.

I guess being Thunderbolt is a death penalty on ASR already as Amir has never reviewed a TB interface, and if they are being reviewed, 4Vrms results will be put into the chart. Also, many people don't seem to care about the DSP capabilities.

 

[maverickronin]

Also, many people don't seem to care about the DSP capabilities.

How many are there that are even useful from a playback (rather than a recording) standpoint?

The idea of integrating all these DSP plugins into the hardware is pretty cool but I've tried looking through their webstie before and I don't remember finding anything that would be useful for playback besides equalizers which are already common enough.

 

[bennetng]

How many are there that are even useful from a playback (rather than a recording) standpoint?

The idea of integrating all these DSP plugins into the hardware is pretty cool but I've tried looking through their webstie before and I don't remember finding anything that would be useful for playback besides equalizers which are already common enough.

Some basic DSP capabilities can be useful in playback in a situation like this:
https://hydrogenaud.io/index.php?topic=104051.msg854152#msg854152
https://hydrogenaud.io/index.php?topic=104051.msg854399#msg854399
Hardware-based multiclient operation ensure glitch-free and low latency operation in multiple applications. Which can be useful in games and other interactive applications, not necessarily audio production. So some DSP power is good enough for these scenarios.

However DSP power in these UA interfaces could be overkill for typical consumers, with large amount of processors.

 

[Matias]

Their x6 model uses Sabre ES9016 DAC chip, so likely the x16 uses the same or ES9018.

Here are independent measurements which confirm the very low distortion of the smaller x6, so their claims for the x16 are very likely to be true indeed. Impressive.

https://www.soundandrecording.de/eq...iversal-audio-stellt-neue-apollo-x-serie-vor/

Also note that the impressive THD+N is given at -1dBFS, and its full scale is a very high 24 dBu or 12V RMS. So at normal residential 4V balanced the ratio would be a lot lower.

 

[maverickronin]

Hardware-based multiclient operation ensure glitch-free and low latency operation in multiple applications. Which can be useful in games and other interactive applications, not necessarily audio production. So some DSP power is good enough for these scenarios.

I'm totally on board with that - I have an RME ADI-2 DAC. Besides a 5 band PEQ it also has dynamic loudness equalization and crossfeed. No latency, no dropout, and I can get all that with other sources besides my PC.

I'm just asking about the kinds of DSP plugins available on the platform. I'm not too well versed in the pro audio world, but almost all of these seem to be effects processors for recording. Most of them seem to be emulating specific hardware. Someone might like the idea of using it to simulate a tube amp and there are a few equalizers. Unless I'm misunderstanding what some of these do that's about all I can see that would be useful for playback .

I don't see anything which would let me replicate the functionality of my RME.

 

[bennetng]

I'm totally on board with that - I have an RME ADI-2 DAC. Besides a 5 band PEQ it also has dynamic loudness equalization and crossfeed. No latency, no dropout, and I can get all that with other sources besides my PC.

I'm just asking about the kinds of DSP plugins available on the platform. I'm not too well versed in the pro audio world, but almost all of these seem to be effects processors for recording. Most of them seem to be emulating specific hardware. Someone might like the idea of using it to simulate a tube amp and there are a few equalizers. Unless I'm misunderstanding what some of these do that's about all I can see that would be useful for playback .

I don't see anything which would let me replicate the functionality of my RME.

A specific thing about Universal Audio is that they don't make audio interfaces at the very beginning, they made hardware accelerators in the past:
https://reverb.com/item/1708856-universal-audio-uad-1-pci-used

Which is an old product so not on the official website. They have a (somewhat) newer card on the website:
https://www.uaudio.com/uad-accelerators/uad-2-pcie.html

So incorporating the exact (or very similar) processors and codes in their audio interfaces is more like a sales technique to attract their old and loyal customers, to ensure predictable sound quality when using the plugins.

Emulation is never an easy thing, even in digital domain, as in different game console emulators. You may get higher resolution, more framerate and better texture filtering than the original hardware, as well as some annoying bugs and glitches ruining the experience. The same goes for audio plugins as well. Just for example Roland released some virtual instruments aimed at emulating some of their old hardware synthesizers, but the level of emulation is pretty unsatisfactory. Here are some demo audio files showing the bugs if you are interested:
https://forums.cockos.com/showpost.php?p=1861656&postcount=11

 

[Dave Tremblay]

Disclosure: I am the Chief Architect for Universal Audio. That said, I'm not in sales, and am not trying to sell you on Apollo; just trying to answer some of these questions. Hopefully that is OK with the forum rules.

I've read through this thread and there are two main themes here:

1) How does UA measure their devices and is it just marketing?

2) What is the reason to integrate DSP into an audio interface?

If I could answer briefly...

1) We have a very talented team of scientists and engineers that are very proud of the audio performance, both analog and digital, we've been able to squeeze out of the parts we're using. That measurement data is coming from Engineers and is certainly not intended to be misleading. I will ask about whether we are strictly measuring to AES-17. I know we have to tweak out our AP units to be able to measure our performance as we exceed the auto-ranging performance of the AP.

2) In addition to just wanting to make a great analog/digital interface, we do have a long history in digitally modeling classic studio gear. There are a few advantages of integrating that into an interface. The first is that all of our modeling plugins have the ability to be truly voltage calibrated. If you use a typical audio plugin in a DAW, you don't know what voltage a particular sample value is because every audio interface has a different 0dBFS to Volts reference point. Since we're integrated, our effects know this calibration and can behave exactly like the real hardware. The second is audio latency through the effects themselves. In a professional recording environment, a vocalist can often detect latency in their headphones greater than 3mSec. That is difficult to do through the host computer processing. Doing it in DSP, on device, we can deliver effect processed sound to headphones of sub 2mSec. The last one I wanted to mention has to do with a unique feature of Apollo. Our unison mic preamps have analog circuitry on the front end that allows us to alter the input impedance of the channel to match that of a physical piece of gear. Some microphones are very sensitive to the input impedance of the preamp that they are plugged into. When we model a mic preamp in our devices, we physically change the analog into to match the modeled gear, thus interacting with the microphone in the same way as the real mic preamp would have.

One more thing to consider when evaluating these devices is the analog circuitry on the outputs. We take this very seriously. A good example is headphone amps. You can have a great DAC, but if you listen on headphones, and your interface can't drive them well, it's not going to sound good.

Dave

 

[dfuller]

Disclosure: I am the Chief Architect for Universal Audio. That said, I'm not in sales, and am not trying to sell you on Apollo; just trying to answer some of these questions. Hopefully that is OK with the forum rules.

I've read through this thread and there are two main themes here:

1) How does UA measure their devices and is it just marketing?

2) What is the reason to integrate DSP into an audio interface?

If I could answer briefly...

1) We have a very talented team of scientists and engineers that are very proud of the audio performance, both analog and digital, we've been able to squeeze out of the parts we're using. That measurement data is coming from Engineers and is certainly not intended to be misleading. I will ask about whether we are strictly measuring to AES-17. I know we have to tweak out our AP units to be able to measure our performance as we exceed the auto-ranging performance of the AP.

2) In addition to just wanting to make a great analog/digital interface, we do have a long history in digitally modeling classic studio gear. There are a few advantages of integrating that into an interface. The first is that all of our modeling plugins have the ability to be truly voltage calibrated. If you use a typical audio plugin in a DAW, you don't know what voltage a particular sample value is because every audio interface has a different 0dBFS to Volts reference point. Since we're integrated, our effects know this calibration and can behave exactly like the real hardware. The second is audio latency through the effects themselves. In a professional recording environment, a vocalist can often detect latency in their headphones greater than 3mSec. That is difficult to do through the host computer processing. Doing it in DSP, on device, we can deliver effect processed sound to headphones of sub 2mSec. The last one I wanted to mention has to do with a unique feature of Apollo. Our unison mic preamps have analog circuitry on the front end that allows us to alter the input impedance of the channel to match that of a physical piece of gear. Some microphones are very sensitive to the input impedance of the preamp that they are plugged into. When we model a mic preamp in our devices, we physically change the analog into to match the modeled gear, thus interacting with the microphone in the same way as the real mic preamp would have.

One more thing to consider when evaluating these devices is the analog circuitry on the outputs. We take this very seriously. A good example is headphone amps. You can have a great DAC, but if you listen on headphones, and your interface can't drive them well, it's not going to sound good.

Dave

That's the kind of info I like to see! IIRC, you're using the TI/Burr Brown PGA2500 programmable mic preamp on the front end?

 

[Dave Tremblay]

That's the kind of info I like to see! IIRC, you're using the TI/Burr Brown PGA2500 programmable mic preamp on the front end?

That is correct. We have some tricks in there, but the PGA2500 is the basis of our programmable gain.

By the way, I did confirm that we have AES-17 settings turned on with our AP units to spit out those specs, but I wasn't able to confirm quickly that we are following a full AES-17 procedure.

 

[Matias]

By the way, I did confirm that we have AES-17 settings turned on with our AP units to spit out those specs, but I wasn't able to confirm quickly that we are following a full AES-17 procedure.

Could you please post the graphs here using the same settings that Amir usually uses in his DAC and ADC reviews? Consider this free publicity.