• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

RME ADI-2 DAC FS - AKM Versus ESS Measurements (DAC, Preamp & Headamp)

It's practically the same as expected in most things.

Sadly the main difference is possibly ESS IMD hump but can't be measured with Cosmos ADC since itself has ESS IMD hump.

I think the RME has the hump but we won't know until we see an AP-555 measurement

Hope @amirm could do it.

Or @MC_RME can share his measurement
 
I cannot run reliable SMPTE IMD Versus Level tests (in intend to find any trace of ESS Hump) with my Cosmos. Main reason being that IMD humps appears with any DAC I tested anyway.
The E1DA ADC itself introduces some (very low level) IMD hump.
It hardly shows in stereo mode, but in mono mode, noise being lowered, it's apparent in measurements
 
AKM's THD plot suggests it has a simpler analog circuit design like a single ended analog gain stage. It's a reasonable sign of subjectively musical sounding converter, because it resembles acoustics instruments' behavior. ESS's random spikes may not be ideal even if THD is slightly lower, and it suggests ESS sounds more digital, less natural subjectively.
 
Last edited:
AKM's THD plot suggests it has a simpler analog circuit design like a single ended analog gain stage. It's a reasonable sign of subjectively musical sounding converter, because it resembles acoustics instruments' behavior. ESS's random spikes may not be ideal even if THD is slightly lower, and it suggests ESS sounds more digital, less natural subjectively.
There may be some truth in this (I have no clue) if the THD levels weren't that low.
At this kind of level, you defenitely can't hear any of them anyway.
You'd need to have the THD components 60dB higher or so to -maybe- start playing a role.

This looks more like trying to justify a legend, to my eyes.
 
Last edited:
RME ADI-2 DAC FS - AKM Versus ESS Measurements

View attachment 215776

Hey Folks,

Today's review is slightly different from usual, with not only one, but two products tested. Well, it is not exactly two "different" products either, strictly speaking... You don't get it? Alright. What about two different units, but with the exact same commercial designation, and looking 99% the same ? I am not sure if the RME ADI-2 DAC FS still needs to be introduced at this point, but for the very few who are not yet familiar with the name: It is a DAC, preamp and headphones amp, packed with tons of features and DSP settings. I mean: tons. Basically, more than any competitor at the same price point. Surprisingly, while RME initially is a company known to be part of the "pro" world, the ADI-2 DAC has been a tremendous success in the audiophile community these last few years.

What you are about to see is a direct comparison with measurements between the (discontinued) "Version 2", that came with AKM AK4493 DAC chip, and the revision that is actually available, now equipped with the ES9028Q2M from ESS. No need to mention why this change happened at this specific period. AKM. Fire. You got it... Having both units available for loan at the same time has not been an easy task at all. Fortunately, two French members volunteered... so, kudos to @Oldschool10 and @Starus for making this review possible! By the way, the ADI-2 DAC FS has been launched at 999€ initially, but MSRP recently went up to 1099€.

Since its launch back in fall 2017, there have been three iterations on the market, without the product's name has even known any modification. In fairness, this brought a bit of confusion for potential customers, even if RME has always been very transparent about reasons and revisions available on the market. For those looking especially for the used market, let me clear that up, once and for all:

- ADI-2 DAC FS "V1". From fall 2017 to fall 2019. Came at first with AK4490 chip and a small remote. Initial review by @amirm here.
- ADI-2 DAC FS "V2". From fall 2019 to mid-2021. Updated with AK4493 and the new remote, which by the way works much better. It also has been reviewed by @amirm and is the one of the two units compared today.
- ADI-2 DAC FS "V3". From mid-2021 to nowadays. Is now implemented with ES9028Q2M and is our second contender.
- "non-FS" ADI-2 DAC is not a thing! Please stop asking about it!

To avoid any (more) confusion, both units tested below will be simply called either "the AKM" or "the ESS".​

View attachment 215777

As you can see, there is absolutely no way to distinguish one over the other upfront. The different meter's color can simply be tuned in display settings, so I chose both cyan and orange on purpose (did not want to mess with my data). What you do not see on pictures is minor contrast variations between the two screens, which I believe to be IPS. The ADI-2 DAC is overall a very well-made piece of gear, build to last with no assembly flaw. The main knob is pleasant to use, with some quite responsive volume acceleration. On the other hand, the controls with front buttons are not that intuitive, to be perfectly honest. That is the kind of gear that really takes times to get familiar with. Hopefully, it comes with probably one the best manual I have ever seen for any audio product. It (still) is, by the way, Made in Germany.​

View attachment 215778

Now, you may point out minor differences. In fact, there are only two. The hole above the power plug and a letter added after the serial number (the ESS being the one you see above). Speaking of the power plug, one of the most common mistake ADI-2's owners are doing, is to forgot to lock it in, causing power clicks and sudden shut-downs. The rest is very common: S/PDIF and USB-B inputs, RCA and XLR outputs, and that's it. However, there are hidden features that you would not guess. The USB-B input can act as a digital recorder from S/PDIF. It may also be used to upgrade firmware (RME still releases some, with new features and customization). Last but not least, the three digital inputs may be switched automatically (the last one to get a signal), which is a very comfortable feature I would like to see more often.​

View attachment 215598
The "C" letter next to the serial number is one of the few way to confirm if you have an ESS ADI-2 DAC.
I could talk about functionalities and features of the ADI-2 DAC for days, but this is a measurements-focused review. So, it is now time to explore raw performance:

Measurements

Disclaimer: Measurements you are about to see are not intended to be as precise or extensive than what you get from a 30k€ AP. There is obviously both hardware and software limitations here, so not quite apples to apples comparison with Amir's testing. Still, this data is enough to have a pretty good idea if the gear is bad or not, stellar, broken, or sub-par...

- ADC : E1DA Cosmos (Grade B). Minimum phase filter. Cosmos APU as preamp for lowest levels.
- Software : RMAA 6.4.5 PRO, Multitone Loopback Analyzer 1.0.15 and REW V5.20.7.
- Method : 8 runs for each test, then I choose the closest to the average. All regular tests are running 24bits / 44.1Khz except for Jitter and Multitone ones.
- RME ADI-2 DAC: Set to Auto Ref Level - On, with multiple levels used, see details. Both units are running the same firmware (0.50). Most tests are done with USB-in and XLR outputs, unless stated otherwise. All DSP/EQ settings are off.​
- Prior to this publication, results have been showed to Matthias Carstens (@MC_RME), I added a few notes based on his remarks.

Let's start with the main testing. I chose to measure at 15dBu (≃4.36V) through the balanced outputs. This appeared to be a fair compromise: it was close enough to output usually measured from balanced DACs (often 4 to 4.3V) and yet gave me near-0dBFS level when using REW, despite the Cosmos fixed input sensitivity . This was acheived by lowering the volume at -4dBr, with the ADI-2 being set to Auto Ref Level. In details: 13dBu (+ 6dB for balanced), then -4dB Volume.​

Summary (4.3V - 15dBu)

Test
AKM
ESS
Frequency response (from 40 Hz to 15 kHz), dB
+0.03, -0.02​
+0.07, -0.02​
Noise level, dB(A)
-120.7​
-120.5​
Dynamic range, dB(A)
120.4​
120.5​
THD, % (RMAA)
0.00008​
0.00012​
THD + Noise, dB (REW)
-114.8​
-112.9​
IMD + Noise, %
0.00047​
0.00050​
Stereo crosstalk, dB
-120.4​
-121.7​
IMD at 10Khz, %
0.00036​
0.00044​

There is no question that both FFTs showed very specific shapes of their own. There is, however, absolutely nothing that I could see to be any close to be audible. All variations measured are landing way down below -120dB. Besides, it is not that easy to proclaim any winner whatsoever. While ESS overall THD+N has been calculated as "worse", it also has lower noise floor at sub frequencies, with only one dominant third harmonic. What will need further investigation, tho, is the 0.29 dBFS difference between both units, while they are supposed to run the exact same dBu level... Looking at raw RMAA numbers, the two devices performed very similar to each others. Repeatable, but small variations are seen in THD % and Crosstalk. But let's have a look at graphs for deeper observations:​

Frequency Response

AKM Left
AKM Right
ESS Left
ESS Right
From 20 Hz to 20 kHz, dB​
-0.03, +0.03
-0.02, +0.04
-0.03, +0.09
-0.03, +0.10
From 40 Hz to 15 kHz, dB​
-0.02, +0.03
-0.01, +0.04
-0.02, +0.07
-0.01, +0.07

Noise Level

AKM Left
AKM Right
ESS Left
ESS Right
RMS power, dB​
-119.2
-119.4
-119.3
-119.7
RMS power (A-weighted), dB​
-120.5
-120.8
-120.3
-120.7
Peak level, dB FS​
-86.6
-86.7
-101.1
-101.8
DC offset, %​
-0.0​
+0.0​
-0.0​
+0.0​

Dynamic RangeView attachment 215783

AKM Left
AKM Right
ESS Left
ESS Right
Dynamic range, dB
+119.2
+119.5
+119.4
+119.7
Dynamic range (A-weighted), dB
+120.3
+120.5
+120.4
+120.7
DC offset, %
-0.00
+0.00
-0.00
+0.00

THD + Noise (at -3 dB FS)

AKM Left
AKM Right
ESS Left
ESS Right
THD, %​
0.00008
0.00008
0.00014
0.00010
THD + Noise, %​
0.00036
0.00035
0.00036
0.00033
THD + Noise (A-weighted), %​
0.00035
0.00035
0.00034
0.00031

Intermodulation Distortion

AKM Left
AKM Right
ESS Left
ESS Right
IMD + Noise, %​
0.00047
0.00046
0.00051
0.00049
IMD + Noise (A-weighted), %​
0.00032
0.00031
0.00041
0.00041

Stereo Crosstalk

AKM Left
AKM Right
ESS Left
ESS Right
Crosstalk at 100 Hz, dB​
-118
-118
-122
-122
Crosstalk at 1000 Hz, dB​
-120
-119
-121
-120
Crosstalk at 10000 Hz, dB​
-111
-108
-111
-108

Intermodulation Distortion (swept tones)

AKM Left
AKM Right
ESS Left
ESS Right
IMD + Noise at 5000 Hz,​
0.00036
0.00037
0.00044
0.00043
IMD + Noise at 10000 Hz,​
0.00038
0.00035
0.00043
0.00042
IMD + Noise at 15000 Hz,​
0.00036
0.00036
0.00047
0.00044

Alright. Now you got the whole picture. As for the regular THD+N FFT, we see that each chip implementations do show a specific behavior depending on the test. It is, again, absolutely not obvious to proclaim one better than the other. Ultimately, I would say that the ESS would be kind of "cleaner", thanks to its lower Noise Floor. Before someone asks: no, I cannot run reliable SMPTE IMD Versus Level tests (in intend to find any trace of ESS Hump) with my Cosmos. Main reason being that IMD humps appears with any DAC I tested anyway.

What happens when pushing both ADI-2 DAC to 0dB (19dBu - 6.9V)?​

0dBr output - 19dBu​

Test
AKM
ESS
Frequency response (from 40 Hz to 15 kHz), dB
+0.03, -0.02​
+0.07, -0.02​
Noise level, dB(A)
-122.3​
-122.3​
Dynamic range, dB(A)
122.1​
122.2​
THD, % (RMAA)
0.00009​
0.00010​
THD + Noise, dB (REW)
-114.7​
-116.6​
IMD + Noise, %
0.00058​
0.00041​
Stereo crosstalk, dB
-122.3​
-123.0​


Well, that is unexpected. ESS took back the lead when we reach 0dBr. Matthias explained later that ES9028Q2M uses its own THD compensation, that is supposed to work best at higher levels. You also see the pattern of harmonics to be slightly different than at -4dBr. By the way, the 0.29dBFS higher output for the ESS remains.​

A little word about Auto Ref Level. The ADI-2 DAC has this (unique?) feature, that allows you to work with four analog steps when you are using its volume control. Common digital attenuation in most DACs would degrade SNR (and so, SINAD) theoretically by the same amount of dB that are reduced from the max volume. ADI-2 DAC attenuation is relative to either analog Ref Levels: -5dBu, +1dBu, +7dBu and +13dBu (add 6dB for XLR) and it can switch automatically between each. That makes the ADI-2 DAC a very precise and competent preamp. As a little downside, you will hear a little click when switching between Ref Levels. Auto Ref Level is way better explained by Matthias himself here.

Here are quick SINAD captures at the same dBr with the two ADI-2 DAC. Note that the smallest levels (-71 and -45dB) appeared to be too low to be properly measured by the Cosmos ADC (minimum input sensitivity: 1.7Vrms). So I use the Cosmos APU as a preamp to compensate:
THD+N Versus Level
Level
AKM THD+N
ESS THD+N
-71.5 dBr​
-64.9 dB​
-65.6 dB​
-45.5 dBr​
-89.6 dB​
-88.7 dB​
-30.0 dBr​
-99.4 dB​
-99.9 dB​
-20.0 dBr​
-108.0 dB​
-107.6 dB​
-12.5 dBr​
-113.1 dB​
-113.6 dB​
-4.0 dBr​
-114.9 dB​
-112.9 dB​
-1.5 dBr​
-116.3 dB​
-114.0 dB​
0.0 dBr​
-114.7 dB​
-116.6 dB​

Note: Both units may in fact have a touch more headroom, allowing them to go up to +2.5dB. I didn't capture results at these levels, but noticed, looking at RTA, that AKM THD+N went down to -110dB THD+N when pushed at that level. @MC_RME told that ESS does a better job in that regard, as the 0dB results would suggest.

As pointed above (0.29 dBFS difference at 4V and above), there are indeed minor to noticable variations that I observed in output voltage. Measuring output with a true-RMS DMM gave me these numbers:​

Voltage variations @1kHz
Level
Predicted Volt/dBu
AKM
ESS
-20.0 dBr​
0.69 Vrms​
0.68 Vrms​
0.70 Vrms​
-12.5 dBr​
1.64 Vrms​
1.62 Vrms​
1.67 Vrms​
-4.5 dBr​
4.11 Vrms​
4.09 Vrms​
4.22 Vrms​
0.0 dBr​
6.91 Vrms​
6.91 Vrms​
7.13 Vrms​

Assuming the output stage is supposed to be the same between both devices, I assume that it could be relative to the own digital attenuation behavior of each chip (?) @MC_RME, if you have any thoughts...


Filters
View attachment 215206
The last filter available is specific to each version. If you see Brickwall filter appears in settings, you've got an ESS revision. If SD LD is available, it has to be an AKM AK4493... If you see neither, then you have an original ADI-2 DAC (AK4490).

AKM
View attachment 215790

ESS
View attachment 215791

There is not a lot to talk about. What you see is typical of filters from either AKM or ESS with most recent DACs. In fact, AKM filters do have the exact same frequency responses than what I measured from the AK4499 in the Monolith Liquid Platinum.


Back to my usual benchmarks, I wondered if S/PDIF could be affected by the chip revision:​

USB Versus S/PDIF

Test
AKM USB
AKM Coaxial
AKM Toslink
ESS USB
ESS Coaxial
ESS Toslink
Frequency response (from 40 Hz to 15 kHz), dB
+0.03, -0.02​
+0.03, -0.03​
+0.03, -0.03​
+0.07, -0.02​
+0.07, -0.03​
+0.07, -0.03​
Noise level, dB(A)
-120.7​
-120.6​
-120.6​
-120.5​
-120.6​
-120.4​
Dynamic range, dB(A)
120.4​
120.4​
120.5​
120.5​
120.5​
120.5​
THD, %
0.00008​
0.00008​
0.00008​
0.00012​
0.00013​
0.00013​
IMD + Noise, %
0.00047​
0.00056​
0.00052​
0.00050​
0.00056​
0.00053​
Stereo crosstalk, dB
-120.4​
-120.3​
-120.4​
-121.7​
-121.0​
-120.5​

Nope! This should be predictable results for any DAC out there, but it is not always the case. Competent design = no variation between optical, coaxial or USB. Period. I expected nothing less from RME.


At this point, I feel to have enough data to draw a conclusion. Yet, just to be sure, I wanted to check other outputs. First, unbalanced line outs:​

XLR Versus RCA

Test
XLR AKM
RCA AKM
XLR ESS
RCA ESS
Frequency response (from 40 Hz to 15 kHz), dB
+0.03, -0.02​
+0.03, -0.02​
+0.05, -0.14​
+0.05, -0.14​
Noise level, dB(A)
-120.7​
-119.8​
-120.5​
-119.3​
Dynamic range, dB(A)
120.4​
119.6​
120.5​
119.4​
THD, % (RMAA)
0.00008​
0.00009​
0.00012​
0.00013​
THD + Noise, dB (REW)
-114.8​
-112.5​
-112.9​
-110.4​
IMD + Noise, %
0.00047​
0.00053​
0.00050​
0.00060​
Stereo crosstalk, dB
-120.4​
-120.1​
-121.7​
-119.5​


Definetly the same story.

At last, a quick test out of the headphones amp. Note that it is only a run at fixed revel (which basically is @2.8Vrms) with no load.

Headphones out (High Power, -7dB Level)
Test
AKM HF High -7dB
ESS HF High -7dB
Frequency response (from 40 Hz to 15 kHz), dB
+0.04, -0.02​
+0.08, -0.02​
Noise level, dB(A)
-103.6​
-102.5​
Dynamic range, dB(A)
103.6​
102.6​
THD, % (RMAA)
0.00018​
0.00014​
THD + Noise, dB (REW)
-101.6​
-100.7​
IMD + Noise, %
0.00256​
0.00279​
Stereo crosstalk, dB
-92.1​
-91.6​

Note: @MC_RME had remarks about these plots. They do not match his results at all (same setting, High Power, -7dBr). It should reach 109dB SINAD with a proper rig, and under 32, 300, or 600 Ohms loads. One of the issues in my data is obviously the lack of load board, IMHO, and the still-not-optimal use of the Cosmos when single ended is in question. Anyway, these are only intended to catch "differences" between the two DUTs, certainly not to evaluate pure performance of the (amazing) ADI-2 DAC headamp.

Jitter
(48Khz, 1-24Khz bandwidth, 16 averages, 262K FFT)

AKM USB
View attachment 215795

AKM S/PDIFView attachment 215796

I computed one S/PDIF graph, since both coax and toslink gave me the same results runs after runs (either AKM or ESS, for that matter). Honestly, what you look at is how it should be with any properly designed DAC: Almost no sidebands, very few, if no variations between all digital inputs.

Hum, now there is unfortunately a little degradation when measuring S/PDIF with the ESS... If you look closer at the levels that we are talking about (more than140dB below the signal), it is absolutely not a serious issue. But if seeking for absolute technical perfection, then the AKM won this round.


Multitone 64
(192Khz, 20Hz-22Khz bandwidth, 10 averages, 262K FFT)

AKM
View attachment 215799

ESS
View attachment 215800

Both ESS and AKM showed absolutely great performance in Multitone. Looking closer at the plots, though, you see that the AKM tends to have a very small rise in harmonics after 10kHz.


Conclusions
After all these boring graphs and numbers, please allow me to turn a bit more personal... I think I forgot to mention something during this entire review. The fact that I was already very familiar with the ADI-2 DAC FS. In fact, I happen to personally own three of them, that I unfortunately had to sold since. I needed a more "lifestyle", all-in-one unit, that could also act as a streamer. Putting my hands back on it, for the purpose of this review, reminds me how accomplished an audio product can be. If you ask me: it is simply the best DAC there is. It sounds like nothing (or like anything you want with DSP tweaks), it is reliable, it does absolutely everything flawlessly with second to none support from the manufacturer. These "subjective" appreciations, added to my data or any measurements that have been published until then, leave no doubts about my recommendation: Of course, it is absolutely recommended!

Now, back to much more objective conclusions, and to the very goal of this review: Is the ESS revision a downgraded ADI-2 DAC FS? Well, there is nothing in my data showing that it could be the case. We saw that AKM may tend to have better Jitter rejection, or negligibly lower THD+N at some volumes, while the ESS shown to be more competent dealing with noise floor, crosstalk, or when pushed around maximum output levels. Not that anybody should care, considering the degree of transparency that is in question. In short, it has been an accomplished design from the start, and it still is by today. So, here are my two cents: if you order a brand new ADI-2 DAC FS, you don't have to feel fooled. If you buy a used one with AK4493, you will be perfectly fine either. The end.

Flanker rating: Draw Game
I make my living in data acquisition and analysis of electronic devices albeit not in the audio world. So I'm wondering how much your measurement variations are statistical noise due to sample to sample variations, temperature effects, grounding, cabling and connector effects, etc. as opposed to meaningful differences between the implementations When physicists and engineers make precise measurements they usually give a ranges - if would be more credible to say for example THD is 112.1db +- 0.1db if that really is the case.
 
Bravo VintageFlanker !
 
if would be more credible to say for example THD is 112.1db +- 0.1db if that really is the case
Ideally, yes. But not easy to achieve, is it ?
And it's less of an issue if you're comparing two results in the same context, I guess, rather than to try to make absolute measurements.
 
I make my living in data acquisition and analysis of electronic devices albeit not in the audio world. So I'm wondering how much your measurement variations are statistical noise due to sample to sample variations, temperature effects, grounding, cabling and connector effects, etc. as opposed to meaningful differences between the implementations When physicists and engineers make precise measurements they usually give a ranges - if would be more credible to say for example THD is 112.1db +- 0.1db if that really is the case.
He described his method: "8 runs for each test, then I choose the closest to the average".
 
Agree with rfb99. Specially when you are comparing two samples, the correct, and probably only meaningful way, is to report value and deviation, so that the reader can tell if the values are actually different or not.
But well, we all know that is not the purpose here... until someone comes and say chip x is better because is has 0.5 db more sinad :D

Thanks Flanker and all other members/reviewers for the refreshing different articles and hard work. Keep it going!
 
Congrats to VintageFlanker ! That's indeed an extremely indepth comparison. The differences measured are of such an infinitesimal magnitude they need very precise instruments to be detected. 10/15 years ago, such a test would have bean a pure nonsense, blame it on the unsufficient precision of instruments at that time.

This brings me to a simple question : I suppose that make definitely no difference for listening ? Well, we all know the answer I guess, but I would appreciate a confirmation by VF.
 
Until the very end, I thought this was Amir’s review. And then I read “Flanker rating: Draw Game” ??? What does this new terminology mean? Then I scrolled back up top and there it was @VintageFlanker !!! Bravo Zulu Sir. Extremely impressed by your work. This question has been haunting many folks about what the chip change really means. And now we know. Much adu about nothing as you have so eloquently expressed and backed up with a plethora of measurements.

Congratulations Sir!
 
My first thought upon seeing the thread title was, "this is about the most useless thing to measure as I bet RME has made sure they are essentially the same". Of course we need somebody to confirm it and you did a fine job.

You could try the two different Baby Face Pro FS units next. I have one with the AKM chip. New ones are ESS.
I bought the Baby Face Pro FS for use with M1 Macbook Pro this past March (ESS DAC) and subjectively, it's awesome. Objectively - it's for YouTube compression so nobody will notice the difference!

I'm a huge RME fan (loved the ADI2 FS AKM before replacing it with something that also had room correction) and have been using the RME HDSPe AIO Pro installed in my PC as my dedicated sound card/DAC. Regardless of the minor differences between these DAC chips, the overall user experience of RME products and their accompanying software is unmatched in this space.
 
Back
Top Bottom