• A Midsummer Night’s Dream - Compare Simultaneous DSD64 and DSD256 Session Recordings

    I was asked by the folks at NativeDSD.com to see if readers at ComputerAudiophile.com would be the right audience for testing the results of two different DSD AD converters, a unique experiment in recording music. My answer was a resounding yes; the DSD (and overall DA and AD converter) technology is nothing new to us here, and this would be a great and fun way to find out what our DSD-capable equipment (and our ears) are telling us about things like sweet spots within DACs, the importance of DSD bit rates, etc.

    The Question

    In the brave new world of High Resolution Music Downloads many music fans have asked a big question. Is there a difference in sound quality that comes from recordings made at different resolution levels and different Analog to Digital Converters (ADCs)?

    A Unique Opportunity

    The experts at Native DSD.com are giving you a unique opportunity to compare session files from a brand new performance of “A Midsummer Night’s Dream” by the Budapest Festival Orchestra under the direction of Ivan Fischer. The session was recorded at two different Direct Stream Digital (DSD) bit rates – Single Rate DSD (DSD 64fs) and Quad Rate DSD (DSD 256fs) with two of the industry’s best Analog to Digital Converters – the Grimm Audio AD1 (at DSD 64fs) and the Merging Technologies Horus (at DSD 256fs) Live in Budapest.

    The Engineers

    You can’t really do a comparison like this justice without top flight equipment and recording talent. On this recording, we have both.

    For this unique project, Jared Sacks, Founder of Native DSD and Producer and Recording Engineer of Channel Classics teamed up with Tom Caulfield, Mastering Engineer for Native DSD and a Veteran Recording Engineer and DSD Expert. The record producer was veteran Hein Dekker.

    The Recording

    Jared Sacks produced an analog mix of the performance with the Budapest Festival Orchestra. From that analog mix, Jared used his Grimm AD1 DSD 64fs converter and created a Stereo and Multichannel edition of the performance from the Grimm AD1 that was stored on the Merging Pyramix DAW system in Single Rate DSD (DSD 64fs).

    Tom Caulfield took the exact same signal and used the Merging Technologies Horus DSD 256fs converter to create a Stereo and Multichannel edition of the performance on the Horus that was stored on the Merging Pyramix DAW system in Quad Rate DSD (DSD 256fs).

    There was absolutely no post production involved in these files.

    Comments from the Engineers:

    Jared Sacks:

    “Setting up with the usual amount of equipment in the hollows of the MUPA Concert hall was more crowded than usual with Tom taking a side table to set up his computer to parallel record the Budapest Festival Orchestra. Since we always make an analogue mix during the sessions it was a simple task to split the feed close to the two converters without any loss of signal. Channel Classics has been using a custom made analogue mixer from the Dutch electronics master Rens Heijnis who also developed custom made battery powered microphone preamplifiers that we use in Channel Classics DSD recording sessions. All of this equipment is connected with 3T carbon cables from Van der Hul. For this DSD comparison, we selected a simple 4 minute Scherzo from the sessions.”

    Tom Caulfield:

    “Over time, Jared and I have discussed the merits of higher than 64fs DSD bit rate recording. He records all his projects with arguably the finest DSD A/D converter available today, the Grimm AD1. One of its characteristics however is it only operates at 2.82MHz, 64fs DSD – Single Rate DSD.

    All DSD encoding has as an artifact a modulation noise far exceeding the level of the incoming analog signal. DSD can shift this noise energy to above the useful audio frequency band, where it can be filtered. Using higher DSD bit rates (DSD 128fs, DSD 256fs) simply raise the noise envelope an octave for every doubling of the bit rate. The shifted noise envelope shape, and amount of noise energy remain the same. Just the frequency where the noise starts to become a measurable percentage of the lowest audio signal level doubles for every doubling of the bit rate. Also, at any DSD bit rate, the noise is uncorrelated to the signal, like tape hiss. That's very different than a correlated linear distortion or modulation.”

    The Expectation:

    So the discussion went; the actual encoding of the audio band should not be affected by the DSD bit rate, since regardless of the bit rate chosen, the audio band is well outside its frequency spectrum and influence…theoretically. The in-audio band conversion quality should be the sound quality determining factor, not the DSD sampling rate.

    Well, Let's Test That!

    Of course, there is only one way to really find out whether the theory and the expectation of DSD recording meet the reality. And that is to create a new recording on two of the top DSD converters and listen to the results. So they did!

    What resulted was a pair of exactly level matched Stereo and Multichannel files – at 64fs DSD from the Grimm AD1, and at 256fs DSD from the Horus. Both were recorded with identical Pyramix Digital Audio Workstations.

    Recognizing that there would be interest in comparisons at Double Rate DSD (DSD 128fs), they also converted the Quad Rate DSD recording from the Horus from Quad Rate DSD (DSD 256fs) to Double Rate DSD (DSD 128fs). That gives you, the listener, yet a third set of Stereo and Multichannel files to listen to and compare.

    The Story in Pictures

    Budapest Palace of Arts Hall: downstage microphone detail, including a stereo pair flanked by the three main ITU placed primary mics, and the two surround mics poking out of row 4.

    Upstage spot mic detail with producer Hein Dekker conversing with (not-pictured) stage personnel.

    It's ultimately all about the music! Follow along if you can.

    Ivan Fischer conducting, while vocalists Anna Lucia Richter soprano, Barbara Kozelj mezzo-soprano prepare to join in.

    Listening to a take: L to R, Jared Sacks recording engineer, Ivan Fischer conductor, Anna Lucia Richter soprano, Barbara Kozelj mezzo-soprano, with Hein Dekker producer looking on.

    Cables anyone? What it takes to mix in analog, record in DSD

    Merging Horus at lower left paralleling the Grimm A/D Converter at top. The red Cat-6 data cable contains the 256fs 5.0 channel DSD bit-stream being recorded.

    Now It’s Your Turn

    And now, it’s your turn to listen to the results of this historic recording session where you can test the quality and performance of different DSD bit rates and two top notch DSD converters.

    Below you will find links to the performance at multiple DSD rates. I invite you to download these files, free of charge from Native DSD. Once you have downloaded the files and compared the results, we’d like to invite you to report your results and comments as responses to this article below.

    I look forward to reading your comments and results. Make sure you tell us about your DAC, and the bit rates you listened to. And most of all, I hope you enjoy this fun exercise, and enjoy helping the folks at Native and Channel Classics make some DSD recording history!

    Note: Jared and I will be doing a seminar at Axpona Chicago on the 25th of this month. I hope to see some CA folks there, where more questions can be asked directly.

    The Files

    * DSD 256 from the original Merging Horus AD converter
    * DSD 128 downsampled from Horus 256
    * DSD 64 downsampled from Horus 256
    * DSD 64 from the original Grimm AD converter

    Click below*for the free DSD track downloads of the session files from Mendelssohn's A Midsummer Night's Dream recording sessions.


    Ted Brady

    Comments 96 Comments
    1. REShaman's Avatar
      REShaman -
      I have DL'd all four tracks even though my Wyred4Sound Dac2 DSDse is only capable of DSD64/128 and not 256. Intend to play through HQPlayer in Native DSD mode. Sorry I will not be able to compare the DSD256 track.

    1. Doak's Avatar
      Doak -
      Great! Thanks for this noble endeavor.
      Grabbed the DLs.
      Look forward to listening - and reporting.

    1. jhwalker's Avatar
      jhwalker -
      I love Jared's recordings with the Budapest Festival Orchestra / Fischer! I can't wait for this one to come out for purchase.

      I have downloaded all four of the stereo samples and will listen to them via HQPlayer > NAA in Native DSD mode, Direct SDM (i.e., no resampling / modulating). iFi Micro iDSD + Mr. Speakers Alpha Prime headphones.

      From what I've heard so far, I don't know that I'll be able to tell the difference :/ but I *think* I prefer each of the Horus versions to the Grimm - next to the Horus, the Grimm sounds a bit muted / soft. More later when I've had a few more listens.
    1. Bunpei's Avatar
      Bunpei -
      I'd like to make one thing clear.

      Which microphones were used for that recording session? My guess is "B & K". Is this correct?
    1. NiHo's Avatar
      NiHo -
      Ted, Thanks for making this available. I played three of the files from an SD card directly in my Resonessence Invicta → Bryston PowerPac 300s → Wilson Benesch Discoveries. Comments are:

      • Grimm DSD 64: Impression of a slightly blurred sound. Sounds more like a recording than a live concert.
      • Horus DSD 64: Clearer than the Grimm, and gives the impression of being more dynamic. Could this be due to a lower noise floor?
      • Horus DSD 128: The attacks and generally the shape of the notes are better defined than at 64. Starting to sound more like a live concert.

      As the Invicta doesn’t play DSD 256 I didn’t bother downloading that version.
    1. tailspn's Avatar
      tailspn -
      Hi Bunpei,

      It's primarily a mix of DPA 4003/6 (B&K) modified for 70 volt battery power, and Schoeps 2H and 2S.

      Equipment - Channel Classics Records

      The prinmary ITU and a and b mics are DPA omni's.
    1. ted_b's Avatar
      ted_b -
      Quote Originally Posted by Bunpei View Post
      I'd like to make one thing clear.

      Which microphones were used for that recording session? My guess is "B & K". Is this correct?
      Yes, Bruel & Kjaer 4006s. I've asked Tom to add more detail (oops, which he beat me to it.. Thx Tom)
    1. esldude's Avatar
      esldude -
      The Grimm file is .3 db quieter than the Horus. Just enough to make sure the Horus files sound better.

      Such a simple detail to get right.

      The Grimm clock also is around 5 ppm slower though this should be of no audible consequence.
    1. firedog's Avatar
      firedog -
      I also thought the Horus 64 sounded better than the Grimm, so I compared the Horus 64 to the Horus 128. Can't do 256 - yet!

      I thought the 128 sounded slightly more "open/airy", slightly more "live". Very slight difference, though. Not positive I would notice it every time.

      Would I pay for the higher res? Only if it was a recording I was especially interested in.

      Thanks to Native DSD and Channel for doing this.
    1. tailspn's Avatar
      tailspn -
      Quote Originally Posted by esldude View Post
      The Grimm file is .3 db quieter than the Horus. Just enough to make sure the Horus files sound better.

      Such a simple detail to get right.

      The Grimm clock also is around 5 ppm slower though this should be of no audible consequence.
      I'd be interested in how you determined the Horus is .3dB louder. I don't disagree, or agree, their tonal balance are different. We only had the Pyramix metering at the session to balance two independent systems, so in fact it's not so simple. Remember, these are raw DSD session files.
    1. EuroDriver's Avatar
      EuroDriver -
      I was very privileged to attend this session and I would like to make the additional points / anecdotes

      - Jarred Sacks and Hein Dekker were both at some point in their early careers orchestral musicians of the first rank. They hear stuff that the orchestra is doing not perfectly that even the conductor misses, and politely ask for a replay, often several times, until the performers get it just right. The working relationship between the J & H with the conductor and the soloists is a joy to witness, people with passion, working together, to create moments of sonic art and beauty

      - It was the first time that I have ever had the opportunity to change rows during a performance. The big take away for me was that there were sweet spots and bad spots between rows 20 and 5. A bad spot in my book is where the sound is congested. Microphone placement is key and that is where the experience of Jarred and Hein is priceless. The Hall in Budapest is tunable, there are many large side doors in the wings of the balcony levels which can be opened , slightly closed, or full closed which effects the reverberations. It was originally designed to be computer controlled, but the system does not work. Hein Dekker, after recording here more than a half dozen times, knows the acoustic character of the hall and is able to set the door openings manually

      - the performances were magical, a life experience I will never forget. Many thanks to tailspin aka Tom Caufield for arranging
    1. astrotoy's Avatar
      astrotoy -
      I just download the files. Thanks so much for making them available. I will play them through my Exa E028, which I believe can play all of the different versions. As of now, I plan to be in Geneva in June to hear the new Merging multichannel DAC. Don't know the specs yet.

    1. Superdad's Avatar
      Superdad -
      Quote Originally Posted by EuroDriver View Post
      I was very privileged to attend this session ….
      ...the performances were magical, a life experience I will never forget.
      Wow Edward, that must indeed have been a very rare treat.

      It is heartening to know that such care is still going into recording and preserving fine classical music performances. Our world's culture is much richer through the dedication and labor of those relent musicians and engineers/producers.
    1. bswietlik's Avatar
      bswietlik -
      Thanks for this enjoyable comparison! I just listened to three of the four recordings as my OPPO 105D/JRiverMC will not play DSD256, only DSD64 and 128. The Horus DSD128 recording was by far the superior sounding recording. Much more dynamic and open, in sum much more real. Next, was the Horus DSD64 and then the Grimm. The Grimm sounded veiled and stuffy, with weak dynamics and a lack of realism. I listen now exclusively to DSD music, and overall, I was quiet disappointed with the DSD64 sound from both Grimm and Horus --neither really seemed to have the sound quality I have gotten used to from DSD.
    1. realafrica's Avatar
      realafrica -
      I have my GEEK Pulse and foobar configured to play everything as DSD128. It was the only way I could play the double DSD and not have to keep changing the config..
      I first tried a comparison between the 2 recorders at DSD64. I could not tell any difference on my system with my ears. Then I played all the files, one at a time, skipping back and forth and choosing a few different short passages to use for comparison. I could not tell any difference on my system with my ears.
    1. realafrica's Avatar
      realafrica -
    1. esldude's Avatar
      esldude -
      Quote Originally Posted by tailspn View Post
      I'd be interested in how you determined the Horus is .3dB louder. I don't disagree, or agree, their tonal balance are different. We only had the Pyramix metering at the session to balance two independent systems, so in fact it's not so simple. Remember, these are raw DSD session files.
      Converted the files to PCM so such things could be easily compared.

      You are right, getting level matching with DSD isn't so simple.

      .3 db is enough to give an apparent tonal balance difference. Such small differences don't sound like loudness. Typically the quieter version might sound less firm and slightly less extended in the low end while less spacious in general. Whether that is the only difference in these files I don't know. But a fair comparison requires a .3 db bump somewhere in the playback chain for the Grimm files. I also noticed the 64fs Horus file has a good bit less ultrasonic noise above 27 khz than the Grimm. The downsampling probably used a steeper filter.
    1. tailspn's Avatar
      tailspn -
      Thanks for your feedback! I'll pass the Grimm 64fs through Jussi Lako's 7th order modulator, increasing the level by .3dB, and add to the Just Listen Mendelssohn folder. I'll also do the same with the Pyramix 64fs modulator so there's the opportunity to compare the same level change with two different DSD remodulators from the raw file.

      On the pyramix timeline, you can easily see the graphic waveform difference with an 4X expanded scale between the ultrasonic noise of the native 64fs and 256fs recordings. You're correct, converting the 256fs to 64fs does not yield the same noise signature as a native recorded 64fs of either the Grimm OR Horus. So yes, there's something in the filter shape that affects the result. I have not tried Jussi's DSD bit rate conversion to compare with the Pyramix yet, and that is an interesting variable.

      Thanks again for your input!
    1. tailspn's Avatar
      tailspn -
      Double post.
    1. analogsurviver's Avatar
      analogsurviver -
      This is my first post on the Computeraudiophile.

      Just downloaded the Mendelssohn tracks - and listened to them on iFi micro iDSD driving AKG K-1000s.
      Too bad it came to the 0.3 dB difference in levels between Grim and Horus - it is quite noticeable. Please note I do understand how difficult it is to set the levels exactly right - particularly for the DSD.

      Still, even if listening to the Grimm and Horus DSD64 in Korg Audiogate 2.x.y (latest version downloadable for 2. ) in 192/24 and high quality mode, Horus has much more believable dynamic range - even if a bit dry. As I have been recording to DSD128 for a couple of years on Korg MR-1000, I was naturally interested how DSD128 compares to DSD256. Much subtler difference than between Grimm and Horus, but is there. It is 02:12 here, I will do some more comparisons tomorrow. I am also looking forward to uploading of for 0.3 dB level corrected track for Grimm.

      Thank to everybody who made this possible - and I was particularly happy to see my countryman, mezzo Barbara Kozelj, participating in this recording. I had the pleasure of recording the dress rehearsal for Bach Johannes Passion, in which she also sang, two weeks ago at Cankarjev Dom, Ljubljana, Slovenia. I also attended at the second concert - she really performed splendidly. Which all makes me eagerly anticipating the possibility to purchase the DSD256 download of the entire piece when available.