The Ultimate Converter DA/AD Loopback Shootout Thread!

[nms]

Hey guys, welcome to the largest audio interface DA/AD converter test thread on the web. Glad you could join us!

This is an ongoing test process that we've been expanding on an evolving along the way. You'll quickly notice that everything relevant to the test process gets challenged, and after 8 months of doing this we've covered quite a bit of ground. We're adding new units on an ongoing basis and will also be expanding into posting Right Mark Audio Analyzer evaluations soon (freq response, THD, IMD, dynamic range specs) so watch for that. I add all results to the front page for easy reference.

*Note: DA/AD loopback tests done with one computer using one master clock source tell most of the story (coloration of the analog front end, filter design, distortion, conversion quality, freq response, transient response etc) but are not a good way to test clock drift since the same clock will be used to send and receive. When you're doing first capture any clock instability is of course stamped on your audio. One way to reveal it is if the length of the recorded audio (in samples) is less or more than the original source (when playing the source from another DAW with both systems using a separate clock, thereby not concealing clock differences. I do currently use a Motu 828mk2 in my setup so I can have all my synths wired up, but I use a Mytek Stereo 192 for recording and use the excellent Mytek clock as master clock for my studio using a bnc wordclock cable out to the Motu.

As for where I'm coming from.. I didn't start this all, just came onboard part way through and refined & expanded it with the goal of trying to make it as accurate as possible. I had the same skepticism and recycled hearsay driven assumptions about this or that piece of gear that many of you have. At the time my plan was to eventually pick up a Prism Orpheus or SSL Alphalink, but it wasn't a pressing matter since I was using the fantastic (according to internet lore) Steinberg MR816x. I'm sure many of you heard the same tales of how the MR816 stood up to the sound of the Orpheus. Suffice to say, as of Feb 2012 my views and gear took a much different direction. Did it sound good? Does an Orpheus sound good? Yeah they do. Most options available today do sound good, so keep that in perspective despite how low some units ranked here. This test isn't about what sounds good. It's about what sounds BEST.. at least where accuracy and transparent conversion between digital and analog realms are concerned.

The thread is a community effort to get to the bottom of converter quality differences in a more reliable and quantifiable manner free of subjectivity and brand expectation. I started this new thread for organization purposes and for a fresh start expanding on the ideas people came up with along the way for increasing accuracy of the results. Particular credit goes first to Kraku for starting the original thread and collecting so many converter samples from people and to didier.best for getting the Audio Diffmaker analysis stats going which we used to rank them in order of transparency. I later went over them all again myself re-analyzing with Diffmaker using a bit more precise an approach. Audio Diffmaker is program for analyzing the exact differences between 2 audio recordings by performing a precise subsample time & gain alignment of the waveforms then inverting the phase and nulling against each other leaving a differences file containing only what was not cancelled out. If you want a program to test transparency in round trip interface passes, that's the one you want. Doing null tests without subsample phase alignment can lead to varied inaccurate results.

Although the segment we're basing the transparency rankings on is the same, the new test file we're using here is a 55 second file that also contains the following tones for purposes of further investigation:

- saw wave at 147hz taken from my Virus Ti2
- 998hz sine
- 60hz sine at -1 dbfs (to test low freq performance, an important one)
- 60hz sine at -0.2 dbfs (clipping test)
- 11khz sine at -1 dbfs
- white noise
- drum loop from a well produced house anthem by SHM
- clip of Aaron Neville's "It feels like rain" which was the test file from Kraku's thread where we're continuing on from.

This is a loopback test going directly out and back in (DAC to ADC) where the results are recorded. You'll find here the results for 1 pass, 4 passes, and 10 passes of the full file. The files you'll find posted are of identical sample length and RMS levels are all matched to the original within a hundredth of a db. For matching levels I referenced only the 998hz test tone section of the recordings. This means if one converter has a hyped top end or a low end rolloff, as some do, you will hear it. If something sounds louder on one than the other it'll be because of the unit, not mismatched recording levels. You'll hear whatever flaws each unit has, played back at you in what I feel is a true form.

To make things easier, along with providing the full combo files at 1, 4, and 10 passes.. you'll find in the folders seamless loops of the saw, drum loop, and 2 loops made from the Aaron Neville track. You could easily take for example the saw samples of 2 different converters and play them on seamless repeat in a player to listen for any difference in tone. This method can be much easier to spot differences than switching back and forth manually at random points in 2 tracks since it allows you to really focus in on the differences. A spectrum analyzer on hi res setting is also useful for comparing some of these clips.

As results are added they'll be run through Audio Diffmaker to get analysis figures of the transparency of the recordings vs the original and will be listed in the results section in order of rank as we go.


To Listen To The Audio Files: (usually behind on updating these, just pm me if you want the files for a particular unit)

Any units with the name hyperlinked have the level matched downloadable audio files for 1, 4, 10 passes plus the null difference file. Right click & save as to download.
Original Source File For Comparison Purposes :: includes extracted loops for seamless playback reference & null testing against original file to compare with any of the other converters :: Supplied by nms & Kraku


Null Test Difference Files for all converters (what's left over after the null test. Insight to what's lost or distorted in the 10 conversion passes)

[nms]

The Test Process:

-Play and record the full length test file then make a recording of that file and continue on in that manner til you have done 10 passes.
-Make sure any timestretching/warping is turned off on the initial play file.
-Route directly out and back through line in, bypass preamps where possible. No patchbays please.. or if you had to then mention it so it can be noted
-Use shortest/good cables you have preferrably
-24bit recordings at 44.1khz
-Use the included test tone to set the levels of your converter to unity gains. This usually means opening the output level to its 0db marker on the interface's software routing mixer and then adjusting the input levels so that the -1db test signal is coming back in at exactly -1dbfs using your DAW to reference. Take care to match the L&R levels as precisely as possible.
-Upload your RAW results for 1 pass, 4 passes, and 10 passes of the DA/AD loop and I will RMS match, chop & run them through the analysis and add the result stats here.

Download the source file and calibration test tone here



Transparency Rankings via Audio Diffmaker Analysis Stats

How To Read The Results: From top to bottom the list flows in order of transparency using the Correlation Depth figure provided by Audio Diffmaker. Sample rate error is the amount of clock drift. The next 2 figures (Correlation Depth & RMS) listed in dB are the ones to look at for converter transparency. Higher numbers = more transparent. The RMS value is the measured level of the difference file created by Audio Diffmaker after performing a phase and gain aligned null test against the original. Quieter is better here as a perfect copy would null leaving nothing behind. The more distorted the waveform is, the more audio will be left behind after the null test. The nature of this distortion or color is difficult to kinow to know. While a low ranking unit like the UFX may not look so good, remember this is a measurement of transparency. If you like the sound of it then go ahead and use it!

LYNX HILO // RMAA test report Best all around score. Great job Lynx!
sample rate error: 0.1145 ppm. Corr Depth: 27.1 dB (L), 27.4 dB (R) // RMS: L=-38.7 R=-38.8 // Single pass noise floor: -117.6, -117.6 (RMS)

APOGEE SYMPHONY
sample rate error: 0.5004 ppm. Corr Depth: 31.2 dB (L), 31.6 dB (R) // RMS: L=-31.9 R=-30.9 // Single pass noise floor: -113.4, -113.4 (RMS)

APOGEE DA16X - AD16X
sample rate error: 0.4992 ppm. Corr Depth: 24.1 dB (L), 26.2 dB (R) // RMS: L=-31.8 R=-30.9

APOGEE DA16X - AD16X (World Studios' unit) // RMAA test report
sample rate error: 0.4985 ppm. Corr Depth: 24.8 dB (L), 25.0 dB (R) // RMS: L=-31.8 R=-30.8 // Single pass noise floor: -108.1, -108.6 (RMS)

LAVRY DA10 (828mk2 ADC)
sample rate error: 0.0014 ppm. Corr Depth: 20.3 dB (L), 20.4 dB (R) // RMS: L=-37.2 R=-37.4 // Single pass noise floor: -102.5, -101.7 (RMS)

METRIC HALO ULN-8
sample rate error: 0.0025 ppm. Corr Depth: 20.3 dB (L), 20.4 dB (R) // RMS: L=-37.2 R=-37.2

METRIC HALO LIO-8 // RMAA test report 44khz - RMAA test report 88khz
sample rate error: 0.0031 ppm. Corr Depth: 20.2 dB (L), 20.4 dB (R)

MOTU 828 MK2 BLA MODDED
sample rate error: 0.0046 ppm. Corr Depth: 20.3 dB (L), 20.4 dB (R) // RMS: L=-37.7 R=-37 // Single pass noise floor: -102.1, -102 (RMS) (Still wondering how this can be a BLA modded unit.. where's the improved noise floor they sell you?)

MOTU 2408 MK3
sample rate error: 0.0081 ppm. Corr Depth: 20.3 dB (L), 20.4 dB (R) // RMS: L=-37.6 R=-38.3

MOTU 828 MK2 (ghetto3jon's unit)
sample rate error: 0.0021 ppm. Corr Depth: 20.2 dB (L), 20.4 dB (R) // RMS: L=-37.7 R=-38.1

MOTU 828 MK2 (nms' unit) // RMAA test report
sample rate error: 0.0003 ppm. Corr Depth: 20.2 dB (L), 20.4 dB (R) // RMS: L=-37.7 R=-38.1 // Single pass noise floor: -101.6, -101.8 (RMS)

UNIVERSAL AUDIO 2192
sample rate error: 0.0064 ppm. Corr Depth: 20.1 dB (L), 20.3 dB (R) // RMS: L=-36.4 R=-36.6

DIGIDESIGN 192iO
sample rate error: 0.0012 ppm. Corr Depth: 20.0 dB (L), 20.2 dB (R) // RMS: L=-37.1 R=-37.2

LAVRY DA10 to MYTEK STEREO 192 ADC
sample rate error: 0.0024 ppm. Corr Depth: 19.9 dB (L), 20.1 dB (R) // RMS: L=-37 R=-37 // Single pass noise floor: -104.7 (RMS)

MYTEK STEREO 192 ADC (828mk2 DA)
sample rate error: 0.0076 ppm. Corr Depth: 19.9 dB (L), 20.0 dB (R) // RMS: L=-36.7 R=-36.5 // Single pass noise floor: -104, -98.1 (RMS)

MYTEK 8X192
sample rate error: 0.0019 ppm. Corr Depth: 19.2 dB (L), 19.3 dB (R) // RMS: L=-36.4 R=-36.3

LYNX AURORA
sample rate error: 0.1103 ppm.Corr Depth: 19.0 dB (L), 19.1 dB (R) // RMS: L=-36.7 R=-36.5 // Single pass noise floor: -113.9, -113.9 (RMS)

ALESIS HD24XR
sample rate error: 0.0007 ppm. Corr Depth: 17.9 dB (L), 18.1 dB (R) // RMS: L=-35.2 R=-34.7 // Single pass noise floor: -112, -112.5 (RMS)

TC KONNEKT48
sample rate error: 0.0025 ppm. Corr Depth: 17.1 dB (L), 17.3 dB (R) // RMS: L=-34.7 R=-34.3 // Single pass noise floor: -98.5, -98.6 (RMS)

PRISM ORPHEUS (Undertow's unit)
sample rate error: 0.1144 ppm. Corr Depth: 15.3 dB (L), 15.4 dB (R) // RMS: L=-33.4 R=-33.1

PRISM ORPHEUS
sample rate error: 0.1094 ppm. Corr Depth: 15.3 dB (L), 15.4 dB (R) // RMS: L=-33.2 R=-32.9

MOTU 828mk3 // RMAA test report
sample rate error: 0.4926 ppm. Corr Depth: 14.1 dB (L), 14.4 dB (R) // RMS: L=-29.8 R=-29 // Single pass noise floor: -105.6, -105.2 (RMS)

Mytek Stereo96 ADC (828mk2 DA)
sample rate error: 0.0042 ppm. Corr Depth: 13.7 dB (L), 13.9 dB (R) // RMS: L=-31.6 R=-31 // Single pass noise floor: -106.4, -106.9 (RMS)

MOTU 896
sample rate error: 0.0114 ppm. Corr Depth: 12.9 dB (L), 13.0 dB (R) // RMS: L=-31.1 R=-30.6 // Single pass noise floor: -104.8, -105.2 (RMS)

TC IMPACT TWIN
sample rate error: 0.0155 ppm. Corr Depth: 12.8 dB (L), 12.9 dB (R) // RMS: L=-31 R=-30.3 // Single pass noise floor: -107.1, -104.5 (RMS)

ECHO AUDIOFIRE12
sample rate error: 0.0163 ppm. Corr Depth: 12.6 dB (L), 12.8 dB (R) // RMS: L=-30.7 R=-30.1 // Single pass noise floor: -109.2, -109.2 (RMS)

FOCUSRITE SAFFIRE PRO 40
sample rate error: 9.8984 ppm. Corr Depth: 12.5 dB (L), 13.0 dB (R) // RMS: L=-25.7 R=-26.1 // Single pass noise floor: -97.8, -99.4 (RMS)

KRK ERGO (The Ergo DAC resamples the digital signal to 96khz before each pass)
sample rate error: 6.3453 ppm. Corr Depth: 12.3 dB (L), 12.4 dB (R) // RMS: L=-20.7 R=-19.9 // Single pass noise floor: -94.7, -95.4 (RMS)

ACCESS VIRUS Ti2 AD fed by Steinberg MR816x DA. Virus clocking used. *For Virus users wondering about Virus line in ADC quality. Clearly the MR816' weak spot is ADC transparency while the DAC is cleaner
sample rate error: -0.4835 ppm. Corr Depth: 11.3 dB (L), 11.5 dB (R) // RMS: L=-28.1, R=-27.7

UNIVERSAL AUDIO APOLLO // RMAA test report
sample rate error: 0.129 ppm. Corr Depth: 10.3 dB (L), 10.4 dB (R) // RMS: L=-27.5, R=-26.9 // Single pass noise floor: -114.4, -114.4 (RMS)

APOGEE DUET
sample rate error: 0.441 ppm. Corr Depth: 10.3 dB (L), 10.6 dB (R) // RMS: L=-26.2, R=-25.7 // Single pass noise floor: -97.4, -98.4 (RMS)

LUCID 88192
sample rate error: 0.4668 ppm. Corr Depth: 10.2 dB (L), 10.4 dB (R) // RMS: L= -27.9 R= -27.3 // Single pass noise floor: -112.9, -112.9 (RMS)

AVID MBOX PRO 3
sample rate error: 0.5039 ppm. Corr Depth: 9.9 dB (L), 10.1 dB (R) // RMS: L=-26, R=-25.4 // Single pass noise floor: -101, -101 (RMS)

METRIC HALO ULN-2
sample rate error: 0.0316 ppm. Corr Depth: 9.3 dB (L), 9.4 dB (R) // RMS: L=-27.6 R=-27

RME FF400
sample rate error: 0.0321 ppm. Corr Depth: 9.2 dB (L), 9.3 dB (R) // RMS: L=-27.6 R=-27.1

BEHRINGER ADA8000 (preamps on all inputs of this unit give it a disadvantage for line in transparency)
sample rate error: 0.0208 ppm. Corr Depth: 8.6 dB (L), 8.8 dB (R) // RMS: L=-27.2 R=-26.6 // Single pass noise floor: -102.9, -102.9 (RMS)

RME Multiface 1
sample rate error: 0.047 ppm. Corr Depth: 8.3 dB (L), 8.2 dB (R) // RMS: L=-26.8 R=-26.1 // Single pass noise floor: -102.5, -101.6 (RMS)

MOTU HD192
sample rate error: 0.0358 ppm. Corr Depth: 7.9 dB (L), 7.9 dB (R) // RMS: L=-26.5 R=-25.8 // Single pass noise floor: -114.4, -115 (RMS)

FOCUSRITE SAFFIRE LE
sample rate error: 0.2543 ppm.Corr Depth: 7.7 dB (L), 7.9 dB (R) // RMS: L=-26.5 R=-26.1

ECHO LAYLA
sample rate error: 0.2464 ppm. Corr Depth: 7.2 dB (L), 7.4 dB (R) // RMS: L=-25.1 R=-24.6

M-AUDIO 2496
sample rate error: 0.0458 ppm. Corr Depth: 7.0 dB (L), 7.5 dB (R) // RMS: L=-25.8 R=-25.4

FOCUSRITE LIQUID SAFFIRE 56 (preamps on all inputs of this unit give it a disadvantage for line in transparency)
sample rate error: 0.12271 ppm. Corr Depth: 6.6 dB (L), 6.7 dB (R) // RMS: L=-24.5 R=-23.9 // Single pass noise floor: -100.1, -100.1 (RMS)

APOGEE DUET2
sample rate error: 0.4352 ppm. Corr Depth: 6.6 dB (L), 6.6 dB (R) // RMS: L= -25.4 R=-24.8 // Single pass noise floor: -102.4, -99 (RMS)

MOTU TRAVELER
sample rate error: 0.0683 ppm. Corr Depth: 6.4 dB (L), 6.6 dB (R) // RMS: L= -25.2 R=-24.7 // Single pass noise floor: -105.8, -105.4 (RMS)

ROLAND MMP2
sample rate error: 0.4746 ppm. Corr Depth: 6.2 dB (L), 6.3 dB (R) // RMS: L=-24.8 R=-24.3

RME FF800
sample rate error: 0.0793 ppm. Corr Depth: Corr Depth: 5.5 dB (L), 6.1 dB (R) // RMS: L=-24.4 R=-24.2 // Single pass noise floor: -110.9, -110.9 (RMS)

EMU 0404
sample rate error: 0.0898 ppm. Corr Depth: 5.2dB (L), 5.5 dB (R) // RMS: L=-24.2 R=-23.8 // Single pass noise floor: -110.2, -110.5 (RMS)

YAMAHA n12
sample rate error: 0.124 ppm. Corr Depth: 4.4 dB (L), 4.6 dB (R) // RMS: L=-23.4 R=-23 // Single pass noise floor: -103.7, -103.8 (RMS)

SSL Alpha Link
sample rate error: 0.1281 ppm. Corr Depth: 4.2 dB (L), 4.4 dB (R) // RMS: L=-23.2 R=-22.8 // Single pass noise floor: -109.2, -109.5 (RMS)

PRESONUS FIRESTUDIO PROJECT
sample rate error: 0.0459 ppm. Corr Depth: 4.2 dB (L), 4.5 dB (R) // Single pass noise floor: -84.1, -84.1 (RMS)

STEINBERG MR816X (preamps on all inputs of this unit give it a disadvantage for line in transparency) // RMAA test report
sample rate error: -0.1729 ppm. Corr Depth: 3.5 dB (L), 3.6 dB (R) // RMS: L=-22.6 R=-22.1 // Single pass noise floor: -91.7, -92.1

RME UFX
sample rate error: 0.3646 ppm. Corr Depth: 3.0 dB (L), 3.0 dB (R) // RMS: L=-22.2 R=-21.7 // Single pass noise floor: -108.4, -108.4 (RMS)



**Note: Sample rate correction is also enabled here so that clock drift wouldn't throw the results, but you can see the clock rate correction needed for each of them (rate adj).


Spectral Analysis Of Null Test Residuals: (This is an RMS measurement of what's leftover after the null test following 10 passes. This outlines where the converters are less transparent.)

.

[nms]

Now we take a look at some snapshots of a few key areas that are exposing performance differences here.

Do you like fast transient response and punchy tight low end? I certainly do.
This is how the 60hz sine wave and sawtooth look after 10 passes and matched levels. Of course these are things you can see and listen to yourself if you downloaded the result files for any of the converters.

Interesting that the 828mk2 is looking the best in both areas while the 828 mk3 appears nowhere near as clean.





Compared to the old tests we'd been doing I removed the first 10,000 samples and the last 12,000 samples from the difference files since those occasionally contained glitches due to the lack of pre/post-roll on the earlier test file.

[saciestudio]

IMO sine waves should be circa -20dB so the converters operate near nominal level. At such high levels as your files there will be probably more distortion than when correctly using the gear near nominal level.

I think everything should be at -20dB RMS except for the songs and drum loops which might really require louder levels in real world use.

Cool tests BTW

[nms]

I'm calling myth on that one! That's the exact reason those sines are at that level in the test file. A decently designed converter should be able to operate without suffering increased distortion with high operating levels so long as it is not being clipped. There's no amplification there.. just an intended neutral transfer of material taking place well within the available dynamic range. Since the MR816 is one of the lowest performers, that's a perfect unit to test a lower level though so I'll do that today.

I think if those 3 Motu can perform at the consistent level they do, right up next to the MH unit, that's a sign a decent converter should be able to handle it no prob.

EDIT: Just want to point out that the segment of music used for the null tests above is -19db RMS making the following few posts of discussion irrelevant

The -1dbfs sine waves are there purposely to see how the units handle it in the snapshots and none that have been tested show any improvement when recorded quieter. Could reduced performance happen when tracking 8+ channels at once peaking at -1db? Possible but untested and appears irrelevant here as we're not doing this.

[saciestudio]

Okay I disagree with you, I do track everything circa -20 RMS and it doesn´t add up more than the analog (and acoustic) noise floor, at least in my system.
Also, if one would use those tests as a reference about an unit for a DAD loopback, it´s probably for use with outboard gear that does operate at nominal levels. And last but not least when I worked with broadcast the standard level for aligning equipment was always nominal, usually -20dB.
I own both a firestudio 2626 and a profire 2626 so when I have the time I´ll run the tests through.

[nms]

There are diff standards for diff studios and broadcasting, but that's to do with uniformity, not because the converters can't do it cleanly. Any good modern converter should handle it fine. Hell as it is tonnes of ppl intentionally clip the AD.

I'm checking the MR again though at -10 just to see because at this point the only conclusion seems to be that the ppl on here praising it aren't very good at testing gear.

[saciestudio]

Yes, I know they are supposed to do it cleanly. They are also supposed to be used at reasonable levels, and I think testing with higher levels might make the difference between the best/worse units bigger than it is in real world use. Also I know there are different standards, but the nominal levels are used by engineers when designing and testing this stuff.

IME stuff sounds better recording at lower levels, especially small crest factor stuff such as classical guitar, but I´ve experienced this with other sounds, in DAWs and also in big digital desks, not just with preamps but with trim pots too. The high end sounds of the nails in classical guitar recordings get nasty much more easily at higher levels for example. YMMV

[nms]

No the MR816 is coming out at the bottom because that's truly where it ranks in the list for transparency. It suffers an audible low end loss, which for no good reason has never been mentioned on GS Previously but was immediately apparent when I began testing it. The 60hz sine loses .8db with the MR. The Motu 828 loses just .1db there after 10 passes. Like I said, if an old Motu can pull off that kind of result there's no good reason for any good converter not to be able to handle it as well.

The Diffmaker rankings use only the music clip at the end to compare and that piece is -19db RMS. Any piece that can't handle that well has no place in my studio. I'm buying a Mytek ADC this week to take over tracking duties and when it gets here I'll test the MR with that to see if it is at least acceptable as a DA. The results we've found from this testing have been far more informative than any of the pages and pages worth of crap I've ever read posted on converters here or elsewhere. A lot of it is expectation bias and people regurgitating what others say.

Here's why I included the saw wave in the test file. You can see the saw wave has been quite distorted by the Steinberg unit while the Motu 828 mk2 and original look identical still after 10 passes. For some reason the damage appears to be mostly happening in the positive phase of the wave form:

[saciestudio]

I may have not been clear on this: it´s not that the results will be bad because the converter cant handle, its because the gain setting at unity might be best achieved at nominal levels, and that´s about the analog front and not the conversion itself. That said, I´ve never stated anything about the MR... and I do find the results useful too, was just trying to give my view on the subject, which Im sure no expert in.

Ive aligned my IO the way you suggest a long time ago and I remember that although stuff matched at -1dB they didn´t at nominal levels. Nowadays when I use outboard gear I set the gain using nominal levels, it matches better through the dynamic range and the method works fine in any equipment.

-Use the included test tone to set the levels of your converter to unity gains. This usually means opening the output level to its 0db marker on the interface's software routing mixer and then adjusting the input levels so that the -1db test signal is coming back in at exactly -1db using your DAW to reference. Take care to match the L&R levels as precisely as possible.

[nms]

Really? I just double checked my MR and the tone comes in exactly on mark whether the tone is -1dbfs or -18dbfs. Anyways the only thing it will affect in this case is everyone using the same test conditions.
Truthfully, I haven't seen anthing to even indicate that this calibration makes a difference to the test results after Diffmaker matches the levels but it just removes an extra possibility for error.

Just added an interesting one. The MR816x feeding my Virus Ti2. As I suspected, clearly the place where the conversion transparency fails is in the ADC of the Steinberg. The DA is cleaner though below many of the other units still. It's really a shame the preamps in the MR816x can't be bypassed. Biggest flaw of the unit. The clocking appears better in the Virus as well.

[saciestudio]

Gonna double check that then maybe tomorrow... Dont remember the exact situation stuff didnt match but I do remember it happened and was solved by aligning with nominal levels maybe it was a specific scenario with some outboard and I extrapolated it in my mind.
Lets keep the tests coming!

[ssiicckkoo]

I was originally looking to purchase the RME FireFace400...but the MOTU UltraLite-mk3 Hybrid is looking pretty attractive right now. I'll be using it for transparent conversion alone. Why is there so much MOTU hate on the forums? Is there more to it?

[DaVogi]

your a doing a highly scientific converter test and you are using the preamp inputs of a unit instead of the direct a/d inputs accessible through the inserts?

maybe you should check your ULTIMATE test setup....

[nms]

Originally Posted by DaVogi

your a doing a highly scientific converter test and you are using the preamp inputs of a unit instead of the direct a/d inputs accessible through the inserts?
maybe you should check your ULTIMATE test setup....

You're saying a proper test of an 8 channel unit is to rig it through the one pair of 1/4" unbalanced insert jacks on the unit as opposed to using the intended 8 channels of balanced analog inputs that are intended for normal use?
I didn't test running the MR816 through the insert because I don't find it particularly useful if the only hope in getting a more transparent input is by means of the one unbalanced insert pair on the unit. I'd be willing to bet money though that it wouldn't be more transparent than the Motu's balanced i/o.

[prontold]

This is awesome. You guys really took this idea and ran with it. Unfortunately, I had to get rid of my MOTU because my new computer had no firewire. I'm curious to see how the Roland Octa-capture I've been using in its place rates. I'll have to do some passes with it.

On a side note, no idea if this makes sense, but maybe people like the sound of the MR816 because it has some sort of compression/low freq rolloff built in that you might use on most tracks anyway? Thats what it looked like on the saw wave image nms posted.

[nms]

I thought of that. If it was only a clean slight rolloff then it would not be that bad for tracking though disasterous for inserts in some cases. The stats are just so far off though that I can't trust that's the only destructive thing happening. The fundamental of that saw wave is 147 hz. If it's being affected to that degree at 147hz that's not a good sign for me. Also the manner in which the waveform distorts (lopsided taking place mostly on the positive phase) doesn't look good. I don't doubt a lot of people have the ability to regurgitate how one unit is better than another based on stuff they read or a flawed test as I've been guilty of that myself. I also don't doubt how easily we can be tricked into thinking one thing is better than another or wanting to reassure ourselves of our purchases. One thing I do doubt greatly is the quality and accuracy of testing done on units though by forum users.

I bought another high end piece that was hugely received on GS only to realize it had a worse noisefloor than any other piece of gear in my studio to the point I decided to sell it rather than deal with the hissing and wimpy output level. While there are certainly no shortage of people with opinions, the weight they carry can be pretty questionable!

[fredrikberg]

Hi!

This is my first post at gearslutz and I want to upload my testfiles for my HD24Xr.
I´m not shure how to do it. Testfiles are complete and in a zip-file.
I couldn´t match input/output since the hd24xr doesn´t have a trimknob.
I just put the original file on the hardrive on channel 1 and 2 and then recorded direct from out (1 & 2) to in (3 & 4) with shortest possible cable i could find.
Then copied channel 3&4 to 1&2 and repeated the process.
I included 1 pass, 5 passes and 10 passes.
Also, because i had to split the original file and join the resulting files in cubase I also
attached the original file splitted and the joined again (If cubase for some reason would alter the data).

Interesting thread. I also have a mr816csx and I always felt that my HD24xr had a more full and nice sound. Using my Mr816csx mainly for headphones and monitoring.

Regards

Fredrik Berg, Sweden

[Rupert Brown]

Wow thank you for this, seems very through. So would I be right in thinking your testing points to the UNIVERSAL AUDIO 2192 as being the most accurate of the lot?

[ZFire]

I always wonder if loopback testing of a converter is going to tell the whole story. Because if an ADC and a DAC are sharing the same low-level clock, then imperfection in the clock timing could potentially be masked (depending on the device architecture). Ie. If the DAC imparts distortion due to jitter, drift, etc., then the ADC could possibly un-do that distortion since it is using the same skewed clock. No idea if this is really a factor in real-world usage, but it makes me wonder about tests like this.

[ssiicckkoo]

Do you think you could post the .wav difference files? Perhaps with X amount of gain?

[nms]

@Rupert - No, the list is in order of transparency ranking so the Motus and MH are the top units.
@ssicckkoo - Here are all the null test difference files

Originally Posted by ZFire

I always wonder if loopback testing of a converter is going to tell the whole story. Because if an ADC and a DAC are sharing the same low-level clock, then imperfection in the clock timing could potentially be masked (depending on the device architecture). Ie. If the DAC imparts distortion due to jitter, drift, etc., then the ADC could possibly un-do that distortion since it is using the same skewed clock.

First, loopback testing doesn't tell the whole story. Comparing original captures from an identical source also tells some of it. It's just a lot easier to do loopback tests and unlike an original capture, a loopback also tests the DA.

Secondly, if a clock has bad jitter it will only be doubled by going out then back in.. it won't magically reverse itself since we're talking about error, not a calculated thing Even if the exact jitter was taking place on the DA & AD at the same time it still wouldn't reverse itself because of the latency in the signal.

Anyways, This is quite an eye opener which I just posted in the Motu vs Lynx shootout..
it can be difficult to focus in and catch the differences sometimes when switching back and forth at random points and a lot of different sounds happening. Analysis numbers are one thing. But snipping pieces in isolation can really expose stuff:

Motu 828mk2 BLA modded vs Lynx Aurora - guitar1

first half of the loop is Motu, second half Lynx. both RMS matched exactly. The Motu has a higher peak value which is evident in what you hear which is extended detail and a less flattened sound.

The Motu sounds a lot better there. Proof that converters sure do make a difference. This doesn't seem to be in line with those who say the Motus as an inferior converter and insist you have to spend far more for a "high end" result. So many recycled opinions passed around lol. The ongoing theme of all this converter testing so far is the Motus rock far beyond their price tag, and some companies are selling you products that aren't worth nearly what they sell for.

[nms]

@fredrikberg - I sent you your result files for the Alesis HDXR24. I didn't add this one to the list though because it's a recorder and not in current production.. while the main purpose of this is to test the commonly used current AD/DA converters. Respectable results though anyhow

sample rate adj=0.2187 ppm. Correlation Depth: 17.0 dB (L), 17.0 dB (R) // RMS: L=-31.60 R=-30.42

[fredrikberg]

Thanks!

Yeah, I think it´s a great machine. Also considering the pricetag.
You get 24 in/out at 44.1Khz or 48Khz (12 @ 96Khz).
Plus you can use it as a great standalone machine.

I paid like $2000 for a new unit 5 years ago. Seems like well spended money.

It got better results than Prism and slightly below Apogee.

/Fredrik Berg

[selmervi54]

Guys, I am interested in upgrading my Interface. From these test It looks like a Motu 828 MkII or III would be a great bang for your buck option. It looks to be transparent and hang with the big boys. It is way cheaper in cost then the others with similar specs. Originally I was going to buy the MR816 because of all the buzz on this forum, But it really comes in at the bottom of the list. Least transparent.

Am I crazy to think that Motu is the way to go?? I keep hearing comments about how Motu sounds like their are blankets over the speakers. This test seems to disprove this. What are your thoughts?

Also Motu seems to do a great job of keeping up to date drivers.

Thanks!

[nms]

The sad truth seems to be a lot of bad information gets passed around here. People often don't seem to be very good at testing gear properly. The tests here seem to be opposite to what people here say about Motu and the MR816. In all the gushing reviews of the MR not one person on here has see mentioned some of the flaws I found by testing mine. I say grab a Motu if what you want is transparent conversion. If you want preamps and don't mind a bit of coloration then an MR might make you happy. These tests only expose the transparency. It doesn't dictate what will sound best to you. I don't use onboard preamps so I plan to get rid of mine.

[6strings]

Thanks for doing this!

I did a blind shoot out with my 828mkII a while back (AD/DA conversion test against the raw audio file) when I was gearing up for an upgrade and came to the same conclusion with my own ears; the MOTU converters sound pretty transparent. Enough for me anyway. Again, thanks. Glad to hear my ears aren't totally shot.

I decided to put the money into other areas of my studio instead (mics!). I don't regret it one bit.

Don't worry about the grief you'll surely get about this. Slutz will be slutz you know!!


PS. Disclaimer: you gotta stay away from those on-board MOTU pres, however... (low gain and dull)

[jinksdingo]

I have witnessed the dislike posts on Motu here and I do not understand that all those who upgraded from a motu to others (now lower on the above lists) were convinced with their ears there was a vast audible difference.
Surely not all were telling porkies: there must be s'thing to it?

nms please elaborate; "I don't use onboard preamps so I plan to get rid of mine"

Thank you for the results by the way!
cheers

[nms]

Well from what I gather the preamps on the Motus may not be anything to write home about. The Steinbergs seem better. If I need a mic or instrument preamp though I use my 6176.. Not some onboard one. What I want from my converter is transparency and it appears the Motus have that in spades. Oh.. And people allow themselves to be fooled so easily. They let themselves easily believe there's some great difference so as to reinforce the extra money they spent on something that is supposed to sound better. Expectation bias based on the info they have been fed. That or they're going on the preamps which there is definite room for improvement. As for transparency though the test speaks for itself. I have some really surprising stuff to post soon about a few units.. Images included.. Stay tuned

[jinksdingo]

Good on you nms, proper evaluation by way of standard testing is definitely lacking for engineers compared to some audiophile forums and mags.

You are bridging that gap known as subjective reality to the real world objective reality.

From what you read or *hear described as a difference in quality *it could be as you have concluded; pre amp quality.

I know how many times I have EQed a channel to only later find it was the wrong channel EQ. But I could hear the changes! e

So if you are not using motu preamps knowing you have one of the best converters by test is reassuring.

Looking forward to your surprises

Well Done too for standing up against the general consensus of opinion and shinning a beacon of truth!