Recording of February 2018: Chopin's Last Waltz

Chopin's Last Waltz
Chopin: Ballade 4 in f, Op.52; Fantasie in f, Op.49; Mazurka in c-sharp, Op.63 No.3; Mazurka in f, Op.68 No.4; Nocturne in E-flat, Op.62 No.2; Prelude in c-sharp, Op.45; Valse in A-flat, Op.64 No.3
Robert Silverman, piano
IsoMike 5606 (LP). 2017. Available from Acoustic Sounds and other on-line retailers; DSD files available from NativeDSD, Ellen Silverman, prod.; Ray Kimber, Aaron Hubbard, engs. DDA TT: 50:05
Performance *****
Sonics *****

At the 2017 Rocky Mountain Audio Fest, Ray Kimber was beginning to feel a little antsy. He'd just released his first LP, Chopin's Last Waltz, an all-Chopin program performed by pianist Robert Silverman, and while being part of the rush to re-embrace vinyl sounds great, pressing your first-ever long player can be a bit nerve-racking.

"No one had listened to it," Kimber told me from his home in Utah. "The LP pressing finishes the day before Rocky Mountain starts. They are driven over from Quality Record Pressing in Kansas in the QRP company truck. We had never used a turntable for any of our demos—we'd always just played back files out of the workstation. Oracle Audio is coming down with a turntable. Bob [Silverman] is going to be there. I was worried because we've never done a record [LP]. Once the records are set up and displayed, the night before the show starts, I kind of disappear. The next day, people listen in my room, and when I reappear, everyone says, 'It's amazing!'"

A concert and recording artist for more than five decades, Canadian pianist Robert Silverman's discography of more than 30 titles, for such labels as EMI Marquis Classics and CBC Records, includes complete cycles of the Mozart and Beethoven sonatas. His 1977 recording of Liszt's piano music received a Grand Prix du Disque from the Liszt Society of Budapest, and his 10-CD, 2001 set of all 32 Beethoven sonatas was shortlisted for the Juno Award. In 2010, he released a SACD of Mozart sonatas recorded with Ray Kimber's IsoMike process. He's also recorded several albums for Stereophile Records, the most recent being Beethoven's Diabelli Variations (2006) also recorded with the IsoMike technique.

Silverman's Chopin is an unqualified success. Although every composition presented here is a familiar selection from Chopin's oeuvre, Silverman's conceptions of them delve deeply into the composer's inherent passions for his music and his love of melody. The overall architecture of Silverman's playing is solid and sure. Taken at slow tempos, the well-known Fantasie in f, Op.49, particularly its placid Adagio, benefits from Silverman's deft, lingering touch. Perfectly projected, his statements are captured in ravishing, exquisitely balanced Direct Stream Digital (DSD) sound.

In a call from his home in Vancouver, British Columbia, Silverman explained why he decided to record Chopin this late in his career, and how he formulated his approach to the work of this troubled icon of the Romantic era: "I did not start off with any preconceived notions. I wanted to see where a deep and profound study, as well as my musicality—such as it is as I approach my 80th birthday—would take me. I thought, What if you just forget all these little Chopin-esque formulas that everybody automatically puts into their playing—including me, to a certain extent—and really study it as seriously as I studied Beethoven?

"So many people think of Chopin as a tunesmith, but he was as serious and as talented a composer as anybody, with a sense of counterpoint that's different than Bach, but better than anyone since Bach—and I include Beethoven in that, frankly. And he was an original thinker. He had a language that really was his own, and it probably did develop out of the way he played the piano. In some of the great pieces, Like the [Ballade 4, Op.52], or the Polonaise-Fantaisie [in A-flat, Op.61], he really did invent a different way of composing fairly large-scale pieces. I wish he was a nicer man, but I'm in awe."

The recording presents a beautiful, multidimensional sound image, with a wide dynamic range and gorgeously sharp, true tone colors. Ray Kimber's IsoMike recording technique is the result of much experimentation over the years. After reading about it, I thought it better to let Kimber himself explain it.

"I was tinkering with this very unusual crossover design and listening to the Fauré Requiem, and it struck me that there was something that was happening that wasn't being properly represented on playback. I tend to like gigantic, epic things with choir, organ, soloists, and maybe some brass thrown in. I found that those things were always highly disappointing on recordings, and glorious in person.

"I began to think, how do you hang a pair of microphones at an optimal distance for stereo reproduction—which I was thinking was five, six, seven, eight feet—and yet the microphones won't 'see' each other acoustically?

"In IsoMike, the two microphones see Bob's piano. The microphone on the right only sees the right-hand side of the acoustic of the hall because it has no line of sight to the other microphone—it's cut off by a baffle. That was a revelation to me, because it removed the interference. If the left microphone sees the right acoustic from the hall, all it does is interfere with it, muddle it, infuse it.

"I was philosophically opposed to using any processing like EQ, filtering, compression, or limiting, but if you don't use a limiter in the analog chain, then you stand a chance of clipping the DSD digital signal and the recording is ruined. But if you record it at a low enough level that you've got absolute confidence that you won't clip it, then all you're doing is throwing away bits of resolution or sacrificing signal-to-noise ratio.

"So we recorded in eight channels, and really didn't care if we clipped the main track by a dB or two, because we had a safety track running in sample-sync accuracy underneath it, a few dB [lower]. Say we clipped it by 2dB—we would simply go into the digital audio workstation, turn down the track by 2dB, turn up the safety track by the difference, and do a sub-millisecond splice.

"And then there was a medical issue that got into the record. After we had done the 4x, 256kHz DSD, Bob said we should make a record, and before I could stop myself, I said yes. Now, I had been meaning to have the impulse to say yes medically removed for years but it hasn't been done yet."—Robert Baird

Kal Rubinson's picture

No comment on the multichannel, Bob?

dalethorn's picture

So is there a non-DSD download, or would that be beneath the artistic level of this recording?

Ray Kimber's picture

There is a loss of information when a format with lots of bits is converted to a format with fewer bits. We had initially planned on only releasing this project in DSD256 and vinyl. This is the first IsoMike release in either format, so we are still figuring out the pros and cons of various decisions. We have made a decision to release the files down to a 96/24 resolution, it will take a couple of weeks for them to appear on the Acoustic Sounds site. We will do the conversion in-house, we have really good conversion tools. If this works out, we would probably follow a similar routine of first releasing vinyl along with the highest DSD format. Then approx 2 months later add some other formats. The IsoMike project is just for fun and to support the music program at Weber State University, so thank you for your interest.

dalethorn's picture

EDIT: I summarized my conversion issues in detail at the following. I believe that my additional tests in up-rez'ing MP3s to WAV and then reconversion to FLAC will dispel any notions of "low complexity" in the recordings.

Ray Kimber's picture

We will do the conversions in-house using either Saracon or Pyramix. The DSD files from nativeDSD do not have DRM, and DSD does not have built-in DRM, unless I am missing something here. It will take a couple of weeks to get the offerings on the Acoustic Sounds site, they indicated they are really busy and we are also tightly scheduled. I hope this helps.

dalethorn's picture

Edit: I summarized my findings in the post above.

dalethorn's picture

Edit: deleted in lieu of above summary.

Ray Kimber's picture

Hi Dale, I have been reading some of your posts over on computeraudio I have an idea. Is it possible that a recording without any compression or limiting be more compressable by a file condensing program? If you PM me on Facebook with your phone number I would be happy to visit with you. Or send it to I type kinda slow :)

dalethorn's picture

I sent the PM in Facebook. The whole idea of what's compressible and how much has been a fascinating exercise. One of the things that confuse the heck out of these processes is files that have a fixed size, such as WAV or MP3, depending on the sample rate. A number of folks who commented have very different experiences than me, probably because they use different software and techniques.

The thing that shocked me was having ripped lots of piano CDs to bit-perfect 16/44 WAV tracks, then converted those WAVs to FLAC in Foobar2000 and seeing the results, comparing those to the extreme compression from JRiver - just astounding. Anyway, I've ordered 3 new CDs of piano music per recommendations from the folks at CA, and we'll see how that goes when they arrive.

BTW, I posted a review on the NativeDSD site, and I don't believe I was exaggerating about the realistic tonality of this recording.

Kal Rubinson's picture

I don't know but conversion is easy.

dalethorn's picture

I tried working with support at the DSD site, but in the end there were too many roadblocks and it fizzled. I have a Macbook 12" that connects to the Internet and a PC that never connects to anything except USB drives that carry files from the Mac/Internet. If the conversion were truly simple (to 96k WAV) then I'd be on top of it.

Kal Rubinson's picture

I do not know what "the DSD site" is but you can download a trial copy of dbPowerAmp (or Mac equivalent) to perform the conversion. Also, many music player programs, like JRMC and foobar, will do it.

(0r just send me the file ;-))

dalethorn's picture

Edit: moving this text to the next comment.

dalethorn's picture

Edit: Summarizing my "struggle" in getting the DSDs converted to 96k and 48k files, and impressions of the sound....

My Macbook is able to play the DSDs in the Vox player, but I also bought JRiver Media Center for the conversions to the lower sample rates, for other devices. I'd recommend JRMC if you have an angel, otherwise who knows - it ain't simple.

The sound has a good sparkle, and the tones have a palpable quality, if that's an OK term. I have many other Chopin recordings as well as the 70 Silverman/Beethoven tracks, and in this one (Stereo format) the tones sound like I'm there. Really. I can't comment on soundstage or any of that - it's quite good, but with the several options listed (Binaural, Stereo, Multi-channel .....) I can't guess what others' expectations will be.

Kal Rubinson's picture

I tried foobar years ago and never succeeded in "getting it." I could often playback a file but I had trouble finding/organizing what I wanted to hear on a consistent basis. That said many others use, enjoy and recommend it.

Since then, I have used, to varying degrees of success and pleasure, about a dozen different music players on my PCs and, even, on a Mac or two. I now use JRMC mostly and Roon occasionally, rarely others. Foobar must be in my blindspot.

So, I would not impugn SACD/DSD for my incompetence with foobar. Sure, SACD/DSD tries very hard to prevent copying/reprocessing but that is moot for files legally purchased. Conversion to PCM (WAV/FLAC etc), on-the-fly or file-to-file, is easily accomplished by many programs. Surely you can find one to your liking.

P.S.: This is not the proper place for this discussion. If you can find one of my emails, contact me and let me help.

dalethorn's picture

I've been working with JRiver Media Center (and a Stereophile writer) for 2 days now doing conversions from these 352k DSD files to 16/44 WAV files, and I've found that those WAV files contain only about 300 KBPS of data, about the same as an MP3. The 16/44 CD rips I do at home contain the full data, and Level 5 FLACs from those files compress to about 65 percent of the WAV size. JRMCs WAV files compress to 23(!) percent as FLACs.

Note that the reason I'm posting this here is because I was told "conversion is easy", but not that the conversion would be a farce. So far only one answer was posted at the JRMC forum, and it ignored the issue.

If these DSDs cannot be converted to non-DRM files with full contents, then it needs to be stated here at this promotional point.

Kal Rubinson's picture

I am sympathetic but I do not agree with your last statement:
If these DSDs cannot be converted to non-DRM files with full contents, then it needs to be stated here at this promotional point.
It is nonsense to say that any physical or downloaded product is offered as anything other than what it is at point of sale. If a user wants to and is able to convert it to another format for his convenience, OK. If you can't for whatever reason, it is not the responsibility of the seller. Don't take this too personally but the conversions are easy.

P.S.: FLAC compressions are lossless regardless of the level of compression.

dalethorn's picture

You're being argumentative rather than offering a logical argument. Here are the relevant points:

I've been ripping CDs and buying high-res downloads for years. All of the digital audio files that I've acquired are open (not restricted) and convertible to equivalent or lower-res formats. In the case of converting to lower-res formats, none of the conversions I've done have lost information beyond the maximum that that lower-res container supports.

Point 2: A FLAC compression of a 16/44 WAV file (i.e. CD equivalent) has never, ever compressed to 23 percent of the WAV size at "Level 5" compression, and I have hundreds if not thousands of examples.

There are only a few possibilities: One, DSD is restricted, either inherently or by JRMC, to not be convertible to an open format at the full resolution of that format. Two, there is a bug in the JRMC converter. Three, I'm missing something in the conversion.

There may be other possibilities, but - I bought this album and got involved on your recommendation. I want to know where the problem lies, and there is indeed a problem. If you don't know, ask your contacts at JRMC. There's no logical reason to blame me - I want what I paid for, which is "easy to convert DSD".

Kal Rubinson's picture

Is there any evidence to suggest that the different sized FLAC files differ in data content/resolution?

dalethorn's picture

Two things I can offer:

1) Just rip a bunch of CDs to WAV and then convert the WAVs to FLAC. Try to achieve FLAC compression to 23 percent WAV size on anything except the MAC v23 JRMC. It's extremely unlikely you'll get anywhere near that. At Level 5 particularly.

2) Nobody wants to claim that JRMC is compressing FLACs to smaller sizes than MP3s - at least not so far. It would be interesting to get into this with the JRMC meisters, which I hope to get into eventually.

BTW, the reason I bought this album was the sound - amazing to me, and worth the extra effort if it can be established that the DSDs are down-res-able to a legitimate open format.

Edit: A properly ripped CD WAV file is 176.4 kb per second, or ~11 mb per minute.
Edit2: The FLACs from JRMC weren't just smaller, the play rate in Foobar was 290 KBPS.

cgh's picture

SIlverman's perspective is valuable. My level of appreciation for Bach's counterpoint really didn't mature until I started to seriously play Bach. For me it was classical guitar and playing the Lute and Violincello suites. Some of the suites I tried to play every day for years (inspired by the story of Casals). It was amazing how, even after years, layers of complexity were revealed to me through these daily meditations.