1000 Albums

Today I hit a minor bragging point milestone, I have now copied 1000 of my audio albums (that I have purchased) to my hard drive for playback with Roon.

 I wasn't thinking this was a lot, but some might.  I know audiophiles with 4000 or more albums on their hard drives.  And lots of audiophiles, me included, have over 1000 commercial albums in some combination of formats such as LP's, CD's, tapes, MP3 files, etc.  But it takes some work even to get CD's copied to the hard drive correctly   (though I now have a fairly streamlined and decent process using XLD), with the metadata correct, in the best folder, and checked for errors and avoiding duplication.  (I'm figuring 20 minutes median average time when there are no issues, but sometimes it can literally take days of cleaning and stuff to get a good copy.  This is especially true of digital copies of vinyl albums.  But even plain old CD's can be a lot of work.  Recently I tried to get around the strong buzz made when I was reading a particular CD.  So I copied at 4x onto a CD-R using my SmartAndFriendly CD copier.  Bu when I read that copy onto my computer with XLD, it had lots of Accurip errors, only one track was copied correctly.  So I went back to the original CD and copied it using my computer DVD+R drive.  Even with the loud buzz it copied correctly.  I'm now a bit worried about the SmartAndFriendly copier which I thought was perfect.  This isn't the first time I've made an "inter-copy" and in previous cases it was the best solution.  To make sure my backup disc wasn't harmed in this process, I removed the backup hard drive from on top of the DVD+R drive, after turning off backup and shutting it down.  Etc.  But usually it's the metadata which causes the most grief with CD's.)

Large numbers of audiophiles have access to millions of albums online through services such as Tidal, QOBuz, Apple, Google, Youtube, and others.

But somehow "my" albums are much more important.  If I go randomly searching through catalogues such as Tidal, I may have to listen to quite a few albums before I find one I like, or even tolerate, even in categories I think are good.

Sadly, this might even be true of a few "my" albums.

And this got me thinking of back in the early days of my audio hobby.

It has seemed to me like other people always knew exactly what albums to buy, and which to play, but this was never true for me.  I struggled at record stores.  How did anyone ever know what to buy?  More than once I had a friend pick out good albums for me at a record store.

Now my tastes were broad, but also very peculiar.  Never much listening to AM radio (which was BIG when I was a kid) and avoiding commercial rock and pop stations on FM, I was not tuned-in to any stream of recommendations like cool kids were.

I mostly liked classical music, always touted by my mother.  Though it seemed like most of our records were show tune albums, beloved by my older sister.  I only started listening to the radio in 7th grade (after buying a clock radio, I quickly progressed audiophile fashion to having my growing "hifi" turned on by a timer) but didn't do much until 1969 or so, when I bought a reel to reel tape recorder from a friend who was leaving state.

THAT was when I really got into listening to FM radio, at least for awhile.  It always helps to have "something to do."  But the radio I listened to was almost entirely non-commercial stations like KPFK and KCRW.  So I still wasn't getting tuned into ongoing popular music (or even "classical" music).  The tape recorder had a foot pedal to turn pause on and off.  So I'd have it set up in "record pause" and press the footpedal, by the side of my bed, when I heard something really good, and I created tapes of the best "songs" and "episodes."  But as would be true with VCR's and DVR's decades later, it's very likely I spent more time lovingly creating these tapes, than actually listening to them.

I only learned of the Beatles because that same friend who sold me the reel to reel tape recorder, and then he was gone, and my next friend was more into Crosby, Stills, Nash, and Young, which I didn't much care for.

So I really didn't get connected to popular music in high school.  Only classical music, which I was performing as a student cellist in the High School Orchestra.  And whatever I happened to hear otherwise casually.

But I had two different sets of friends introduce me to Pink Floyd when I was in College.  I heard Dark Side of the Moon in December 1973 while getting high on marijuana for the first time in a friend's dorm room.  Later some really cool girl invited me to a party when Wish You Were Here was being rolled out.

In the late 70's one friend of mine really loved "The Wall."  And yet another got me into Steely Dan.  A very nice (but way out) guy I only saw a few times got me into Eno.  Also some were into Progressive and Electronic Rock (King Crimson, ELO).

So that's about it, about all I knew, in popular music, by the time I was 24 (1980).  You couldn't just look up endless articles about any band online like you can today.  You either kept up with radio, relevant magazines, or had lots of cool friends.

I met an elderly Jazz Musician in the 1980's, and he helped me pick out some albums.  And some girl I danced with did likewise.  Finally I was liking New Age Jazz generally, which featured prominently on a San Diego radio station.  I might have bought a Windham Hill record or two, but somehow it wasn't as catchy as the radio.  (Only about 10 years ago, I rediscovered what I'd been missing all along.  Shadowfax and Pat Metheny.)

Now I know a lot more, obviously.  But still it's hard to find new things, and I simply hate all the recommendations given to me by streaming services.  They don't know my taste at all, or actually it seems like they know exactly what I like and then insist on giving me the exact opposite.

Firstly, for more of the time, I'm looking for music without words, and that's virtually impossible to find on streaming services.  There is No Such Category, since any kind of music can come with words, especially when you venture beyond known classical composers.  I myself create Electronic Music (including my latest album, Mythic Rocks, now on all streaming services) with no words, but if I try the category of Electronic Music on any particular streaming service, nearly every song recommended will come with words...  I've even come up with several conspiracy theories to explain this.  They need the words for subliminal brainwashing of some kind.

Anyway, that's one reason why it's good to have your own personal collection of music, though nowadays you could just have use the "playlist" or "favorites" or "My Album" features of whatever streaming service(s) or music manager(s) you are using, and accomplish the vital "pre-selection" of music without actually "owning" anything.

But then there may be things that are hard or impossible to do.  Right now I mostly listen using my mplay playlist generator (which now features splay and tplay for even better play management) and it's easy to make playlists for music on my hard drive, but not so easy to make textfile playlists of albums on streaming services...if it's even possible I haven't yet figured out how to do it and they might never make it possible).  So right now I need lots of "owned" music to keep my all-waking-hours automatic jukebox working.

And there are other things.  Often there are individual tracks which are duds and I really want to remove them.  That's much easier to do (or even uniquely possible again) when you own the music files.

HDCD recordings need to be properly decoded, which is not possible on my system (or most nowadays) without doing it in advance on the files.

Finally, I have even gone to the point of editing tracks of music when they have peculiar misfeatures but are otherwise indispensible.

But the "automated" (scripted) playlist generation is the big thing for me now.  I'd get much more use from the streaming services I use if they let me create and use text playlists.  As it is, I'm thinking of cancelling Tidal and keeping just one paid service, QOBuz.

Anyway, back in the day when I was struggling all by myself to pick out new records at the record store I'd end up with a variety of conundrums.

In generally they'd be pushing "the latest thing."  But "the latest" album of some band you knew little about might well be their worst.  What you'd really like would be the "best" album by that band.  To find that, you'd have to get away from the end-caps and other displays and into the long aisles of recordings.  Even then you'd have a hard time figuring out which album of some band would be the best one to buy.  If you asked the sales person, likely it would be "the latest" which would be recommended.

I loved reading the recommended lists in The Absolute Sound and Stereophile.  But I could never remember what had been recommended when I got to the record store.  (And at most record stores such things would be impossible to find anyway...).

Somehow there was always a huge mental gap between the "independent" recommendation of albums and the purchasing process.  And even in our hypercommercialized online world, it seems there still is.

Then back in the Vinyl days, whatever I bought, after some long protracted selection process, would invariable be warped or have some skip or something.  And I'd never get back to the store on time to exchange.  Or if I did, the next copy (preselected by the store staff) would have an even worse problem.

So it's not surprising that when I get down to the never-before-copied parts of my collection, I could be running into a lot of "clunkers" I really don't want to listen to that much.

OTOH, it may just as well be, and actually is most of the time, more like re-discovery.  Things that sounded awful in days back when they were put at the bottom of the pile are now sounding amazingly good.  At various times in the past, my system of the day could make some things, and perhaps a lot of things, sound awful.

I didn't myself get a decent handle on this until I started doing acoustical measurements, around 1998.

But even then, there were endless issues, including amplifier bias problems, and even the discovery of the requirement for a down tilted response on electrostatic systems which was only my discovery last year, when I applied a slight boost at 1kHz to make everything fit this curve better (and sound much nicer).  I've made the most progress in the actual "sound" since I've been retired, now nearly 4 years.

Anyway, gems are turning up as I get to the bottom of my bedroom rack.  Such as Laura Marie, a musician I heard at a San Antonio Luminaria.

There's still much work to do on this pile, and there are many other piles.

One thing that this copying work does is create a new opportunity to listen to things as if they were new, again.

And I also benefit from my mplay playlists which ensure I ultimately hear everything.

In addition to decoding (or transcribing...which might be slightly better in some ways) HDCD's, I try to make transcriptions of all DVD-Audios and SACD's  too.  Since my system runs on 24/96 and 24/48 anyway, there is no benefit from preserving DSD from SACD's or other sources.  (I personally don't believe there is any benefit anyway, all the important characteristics of DSD are captured by 24/96, nothing better than 24/96 is needed for anything except possibly production.)  And vinyl too.  I prefer to transcribe vinyl whenever I'm playing it at all.  And those transcriptions become part of my collection too.  Perhaps even multiple transcriptions, comparing different players, turntables, or techniques.  (I put some redundant copies in a different folder so they don't get auto-selected for playlists.)  Actually despite thinking about it a lot, I don't have that many LP transcriptions (and many I do have are ready to be upgraded with better recordings on better systems).

Trick for when Album Leveling doesn't work

 Level adjusting is very serious business.  Back when I first got Sonos in 2006 and tried "random play" the difference in levels made many things unlistenable, often because too loud.  (And yet, I'm a person that generally likes "realistic loudness."  But in a playlist of music, something that is too loud sticks out like a sore thumb.)  I believe that virtually all the inexplicable differences that audiophiles claim to hear due to tweaks that have no explanation that is reasonable to audio engineering are actually level mismatches in disguise, since we can hear differences as small as 0.1dB (though it's not easy).

Now I rely on Roon's "Album" Volume Leveling, which works great in most cases.

Sometimes, however, it doesn't.  Typically when I am playing back programs recorded from FM radio, for example, which are highly compressed.  Or my own compositions.  Roon fails to dial back the maximum level to make up for the high degree of compression or high level of deep bass/extreme highs.  It does no good if I make the recordings at -6dB, Room simply adds +6dB gain.  Depending on various factors, Roon either amplifies up to peak level of 0dB or it reduces below that level if it thinks it's going to sound too loud.  But sometimes if fails to reduce, or even increases the level when it's not needed.

Two commercial albums that exhibit the same problem are Beautiful Vol. VII by Andean Fusion and Respighi Ancient Airs and Dances Suites Nos. 1-3 by Rico Saccani.  The latter is actually what inspired the present experiment.  Roon album leveling increases the level by 2.2dB.  It sounds much better without that boost.  Beautiful is already recorded at maximum peak level and Roon reduces it by a whopping 4.4dB but that's still not enough.  It doesn't help when in addition to being highly compressed, albums are very dry (little reverb or ambience).  This can make recordings sound painful even in the absence of distortion or clipping.  "In Your Face."  A slight distance is more comfortable.

I discovered I could stop this by appending a "needle drop" click (excerpted from recording vinyl) amplified to maximum level.  But needle drops at maximum level are a bit annoying.

A single cycle at 1kHz was just as bad or worse.  I didn't even want to try a single peak cycle at 20kHz, it would exercise the maximum peak level on a 150W amplifier that is powering my supertweeters.

But my subwoofers have internal digital EQ which steeply rolls off bass below the tolerable limits of the woofer.

So I tried a single peak cycle at 1 Hz (1 sec).  It works well, and I hear hardly any sound at all (just a very soft click).  This can be easily added to the end of the last track of an album, or any digital recording, using Audacity which can generate any tone, even 1 Hz, of any duration down to 1msec.  While playing a single peak cycle at 1 Hz, I see the digital meters on my EQ going to 0dB slowly and back.  With a peak level signal, Roon will not add any gain.  When the 1 Hz peaks at only 0.999 in the Audacity generator level control, Roon adds +0.1dB gain, so I decided might as well go all the way to 1.0 then Roon won't mess with it (Roon can't, other than lower the level).

(The first time I did this, I was scared out of my chair by a temporary shadow caused by a plane flying overhead just at that moment.  I didn't intend to invent the black hole speaker.)

I'm not sure if this would work on all systems.  Most do reject deep subsonics quite well however.

Reducing hum from -94dB to -100dB

It seems like I can improve the S/N from the Denon DVD-9000 dramatically, over 6db, with a simple tweak which might apply whenever unbalanced interconnection is used.  (Or it might be peculiar to this set of components, etc.)

I connected an extra set of stereo cables from my Denon DVD-9000 to the Emotiva XSP-1 Preamp.  I deliberately connected the unused surround outputs to an extra of the preamp.  As the outputs are not carrying any signal during this test, they may act as a better ground connection than the initial stereo pair of cables connected to the main output.  It also turned out that just adding one unused cable made nearly all the difference.  Adding a second unused cable gave little additional improvement, so I suspect more similar cables would only be for the very obsessed.

The results are shocking.  I got over 6dB improvement in peak unweighted S/N ratio from having the extra cables.  Most is clearly hum but some is noises that have a 60 Hz period.  This is similar to improvements I see with balanced interconnects over single ended.

For the test shown I had the Denon DVD-9000 playing a regular CD (not HDCD which might be tricky) but in pause for the entire duration of the test.   The Emotiva XSP-1 gain was set at 0.0dB and not changed.

In the starting condition, I had the two extra cables connected from the surround outputs on the Denon to another input on the Emotiva.  At two minutes, I removed the first extra cable on the Emotiva side.  At 4 minutes, I removed the second extra cable.  At 6 minutes I removed the one of the cable connections on the Denon Side,   Around 8 minutes and 40 seconds I removed the second cable.  Then about 10 minutes I reconnected one of the extra cables.  Around 11 minutes I reconnected the second extra cable, so it was back to the starting condition (to be sure nothing else had changed...and it hadn't changed appreciably, compared to the main effect, which is huge).

DVD-9000 with and without (center) extra interconnects

There is visibly far more hum and noise (it's mostly hum but with some strange high frequency bursts that repeat every 1/60 sec) in the center condition which is just using the two stereo interconnects.  As soon as just one unused extra connection is made, the hum and noise drops 6dB.  Then when the second cable is added it improves a few tenths of a dB more.  What is remaining with the extra cable(s) seems to be almost entirely random high frequency noise and the high frequency bursts at 1/60 sec are eliminated.  It seems hardly worth trying more than two (I have no more good cables to try anyway...I had to buy this pair last week)...but just adding one unused cable is huge.

(Possibly it has something to do with the grounding on the Denon DVD-9000...but everything about the DVD-9000 seems impeccable.  BTW, I have Direct Mode set to 2 which shuts off video output.  Also it could have to do with the grounding on the XSP-1.  The extra cables were connected to Input 3, which is in a different bank.)

Another curious thing is that even connecting the cables on the Denon side alone seems to help.  I also saw that kind of behavior with the Video output--having a stub video cable connected (but not connected to anything else) makes the DVD-9000  quieter, even in Direct Mode 2.

Something that might be even better is connecting a very heavy wire, possibly mostly buss bar, to the grounds of two plugs and use that as the "extra" interconnection.  I think I've seen things like that sold as tweaks before.

Both my original interconnects and the extra interconnects are both Blue Jeans Cable LC-1, which has the best single ended shield configuration and plugs for hum and noise reduction.  The originals, unabused, are just a few years older.  These are very solid and thick cables custom made by Beldon with a double braided shield, and terminated by Blue Jeans with very solid Canare RCA jacks.  It's hard to imagine they could be broken or even slightly out of spec.

(Somehow, leading up to these tests, I had intended to upgrade the previous cables to LC-1 to see if that helped.  From above, it looked like I had been using standard grade Radio Shack cables from the mid 2000's.  Possibly intended only as a temporary fix until I got better cables.  But when I actually got close enough to change the cables I could see they were actually LC-1's.  It's funny how when the picture is unclear my mind can instead give me a clear picture of something wrong.)

Previous measurements shows unweighted peak S/N ratio around 93dB for the Denon DVD-9000, and 95dB for the Oppo BDP-205.  That 2dB difference was entirely hum and showed the benefits of balanced interconnection.  Both players actually have "120dB S/N" specification, the limiting factor is my Tascam DA-3000, which can only sample with 113dB S/N, which like all factor specs is A-Weighted, meaning the unweighted S/N is likely in the 95-105dB range.

With these changes, it looks like the Denon will be either identical to or maybe even better than the Oppo, at least for practical purposes when sampling with the Tascam DA-3000 for transcription or playback through my DSP based system.

If the test seems strange...it's partly because it followed an earlier series of tests which were inconclusive because I could never be sure that either I had reset the starting conditions in an identical way, or that the Denon hadn't started muting in some extra way.  Leaving the player running the whole time and going back to the starting condition at the end were the ways to remove these uncertainties that arose in previous testing despite attempts to control the starting conditions rigorously.  The Audacity screenshot was made after 80dB of Audacity digital amplification.  The noise seems higher than some previous tests because the Emotiva volume control was set to 0dB instead of -4.5dB (which I had been using for HDCD since the Denon actually expands peaks at least that high above the normal CD-level 2V output, though I discovered on album Poems of Thunder that in some cases HDCD peaks reach 5dB).  So add 5dB to the wideband unweighted peak S/N levels for comparison relevant to HDCD which is mainly what I use the Denon for.

Here are the peak levels from various parts of the test (correcting for the 80dB gain but not the additional 5dB available output on HDCD).

First 2 minutes (extra cable pair connected)

-91.05dB   (would be -96.05dB including 5dB extended HDCD peaks...this is better than previous data for Oppo which was -95.3dB)

Second 2 minutes (only one extra cable connected to Emotiva but both on Denon)

-90.3dB

Third 2 minutes (two extra cables connected to Denon but nothing else)

-87dB

Fourth segment (only one extra cable connected to Denon)

-85.5dB

Fifth segment (no extra cables connected to Denon)

-84.4dB (not including some early and late peaks which might be due to connection in progress)

-84.3dB (including suspicious early and late peaks)

Sixth segment (one cable reattached between Denon and Emotiva)

-90.0dB

Seventh segment (both extra cables connected between Denon and Emotiva)

-90.6dB

I could immediately add 5dB to all these numbers for HDCD because HDCD peaks are 5dB higher on the Denon, so I have to turn volume down 5dB on the preamp.  (Although, then the noise of the Emotiva itself would be a larger factor than it is, so results are not completely predictable.)  Then these are all unweighted peak measurements, which would improve by 10-25 dB with RMS and A Weighting.  And I don't know how much of these measurements are proportionally hum and ultrasonics.   It would be complicated to do weighted and RMS measurements on these data (I tried but ran into a snag) and only useful for bragging rights anyway since I don't see how I'm make these improvements significantly better without engineering a fake "grounding bus" interconnect as I suggested above.  This present improvement is great and I'll look for it in future connections.  (Though perhaps I should do more investigations to be sure this is a real effect and not a fault of the old cables, for example.)

My guess is that these numbers are showing at least 108dB properly A weighted RMS S/N, but with a mean guess of around 116dB S/N RMS A-Weighted based on the idea that RMS and A-Weighting would improve things about 20dB...but that can't be true since it exceeds the specs of all equipment I'm using (except the Denon itself, at 120dB S/N).  In reality all the noise levels are somewhat additive, so say if all 3 units (Denon, XSP-1, and DA-3000) had 120dB S/N, the result would be about 110dB S/N.  But the Tascam DA-3000 (balanced in) is at 113dB and the Emotiva XSP-1 is at 115dB (unbalanced in to balanced out).

Though the second additional cable gave only a fraction of a dB additional benefit, the 60 Hz component is still not entirely gone, so there might be some further benefit from grounding bus, etc.

*****

I have decided not to do any further investigations on this for the time being.  It could be, for example, that the initial cables were faulty (though they're fairly new and very rugged Blue Jeans LC-1), so if I swapped the brand new "grounding" cables for the initial ones, I might then have the benefit of low hum with only one pair.

Or it could be how the "grounding cables" are routed differently, perhaps out of EMI fields.

Adding cables might not always help.  I'm not claiming that at all.  I think only that if you see a hum issue, such a thing might be worth investigating.

In this cause there was no audible "hum" issue that I ever noticed.  It's deep down in the noise floor, deeply below the ambient noise in the room.  At -90 to -100dB or so in the electrical signal, at max output around 100dB, it would be 0-10dB, or -24 to -4dB "A Weighted".  If the two initial cables were somehow defective, and they would both have to be defective for there to be hum, one would expect much higher hum level than that.

There could be a straightforward explanation such as that each cable lowers the ground resistance.  Except it doesn't seem to work so incrementally.  Instead, the first extra cable made nearly all the difference.

It really seems like there were additional ground currents between the player and preamp, and routing these additional currents lowered the hum induction that occurred in the interconnect.

But why are these ground currents choosing to go through the additional cable(s) ?

Adding one more cable would in theory not be adding that much more conductance.

Instead, it looks like there is something peculiar going on.  Perhaps the grounding of the surround jacks on the denon is closer to the power supply ground.  And because of that, the extra currents DO choose to go through the new cable(s) selectively, giving a reduction in induced hum.

The Denon 9000 is grounded and so is the Emotiva XSP-1 (IIRC...or it's connected to grounded gear).  Ground loops may be to some degree inevitable, and perhaps it's a wonder they were already so low.

I think therefore that this finding is almost certainly some kind of fluke as I've described, not a general rule.

But it certainly looks like, I have already stumbled upon a lucky optimum.  Any change is likely to make it worse.  Then, assuming micro degrees in cable routing are also a contributing factor, I may find it hard to make it this good again.

And to do this test in the fairly incontrovertible way I have (as opposed to separate tests, in which some extra variable might have changed) means more wear and tear on my equipment, including my beloved Denon DVD-9000, which is very long in the tooth.  (In fact it already started acting up, with the drawer not pulling all the way in once, perhaps in protest.  Now I see it may do that so it looks OK but doesn't register the incoming disc, showing only 0's on the display.)  

To do a proper test proving the initial cables were the problem, I'd have to do lots of cable plugging and re-plugging while it's in pause with a spinning disc.  Almost every time, with the very tight LC-1 Canare connectors, I have to push so hard it jostles the Denon slightly.

To do this test, to make the Denon about as good as the Oppo (which uses balanced connections!), I thought it was worth this, but just once, not for a series of follow up tests to PROVE the causal mechanism, which may not even be possible.

To reduce such wear on my equipment, I'm ending this test here, at what appears to be some kind of lucky optimum in hum reduction.  "If it ain't broke, don't fix it."  At  least for now.  Next time the Denon is removed I'll know to check the cables again.

Stuff in progress

I have some suspicious evidence that HDCD software conversion isn't doing quite as well as my Denon DVD-9000.  The software converted version is indistinguishable visually from DVD-9000 output using Audacity at high signal levels, where you can clearly see the HDCD system working (and I now think it is in fact essential for high dynamic range recording and 16 bits...because real high dynamic range recordings can have 17dB above average peaks).  But at medium levels where the HDCD is barely working, what seems to be happening is there are tiny super-peaks in the 24/96 data from the Denon, so the Denon output looks ever so slightly peakier.

What I think this means is that HDCD is being applied at the 2x oversampling level (or higher), not to the original data stream, and in fact I've always believed that was how it was done (and why no simple software conversion tool would suffice...unless it was doing a compatible form of oversampling as well...and the digital filter changes I always believed were being done).

It's not at present evidence of variable filtering because the recording in question, Poems of Thunder, does not use the variable filter feature!  At least in the output, perhaps variable filter was used in recording (or perhaps not).

Notably Reference Recordings don't show the Transient Filter flag in their HDCD's either.

I was meaning to investigate that, then I noticed my recording had clipped, so I tried lowering the preamp level to -5dB (-4.5dB has been my standard for HDCD for years, to account for the fact the Denon actually puts up to 4.5dB more output with HDCD's than anything else...which it appears nobody else does, or perhaps other like California Audio Labs did too back in the early days of HDCD, I don't know.

So I re-recorded, and then I noticed something I should have been thinking about from the moment I noticed -93dB hum (60Hz) in recordings from the Denon: I'm using a standard grade Radio Shack 3 foot interconnect for the Denon and my Emotiva Preamp.  I've ordered a cable of my new standard (especially where shielding is important): Blue Jeans Cable LC-1.  It's thick and has the best shielding system and low capacitance too.

So I'm holding off on more tests until I get that cable upgraded, and I might even be able to demonstrate a measurable cable difference here, if the LC-1 lowers the hum a few dB.  And what will it do elsewhere?  It might even raise the apparent noise level because low capacitance.  We'll see.

It can't affect the Oppo which is connected with teflon balanced cables.  Recordings from the Oppo show no visual evidence of 60 Hz in their output.  But they show about the same level of high frequency noise, which is apparently therefore the fault of the sampler I'm using now, the DA-3000.

I think the Lavry AD10 is a tad better, the specs suggest 4dB better.   I need to get that fixed too.

Even then, the AD10 will still have quite a bit more noise than the Oppo....even a bit more than the Denon!

It hardly makes sense to concentrate on making reference transcriptions of anything, including vinyl, until I get the Lavry fixed.

Investigating the Denon DVD-9000 noise

Denon 9000 noise tests

To get a better understanding of the noise produced by the Denon DVD-9000 with Emotiva XSP-1 and Tascam DA-3000, I recorded a series of tests shown above unedited and unamplified.  The Emotiva gain was at -4.5dB required for HDCD playback (which produces 4.5dB higher peak levels than anything else with the Denon DVD-9000) however peak output from the player will still reach -0.2dB from clipping the DA-3000 (at -9dB reference level).   

(Strange as it sounds, the Denon 9000 has greater S/N with HDCD than anything else, including DVD-Audio, because the HDCD peak output levels are higher than anything else while the noise from the player is pretty well constant.)

At the start I played bits from the last track of the Mephisto & Co.  HDCD, showing the noise before the track and after pausing.  Finally I just let the track fade out into silence for 5 minutes as the player kept running.  Then I shut the player down.  Then I shut the preamp down.  Then I connected a video cable from the Denon to my TV (through a video isolator).  Then I turned everything back on and tried the Pure Direct switch.

Without any added gain, the noise level is so low it's hard to see what's going on.  Using the audio editing program Audacity I edited out the major peaks shown above and then was able to amplify the rest by 66dB, giving the following result.  Now the noise is a thick line of varying height which can be studied.

Here is a blow by blow account of the edited version of the graph.

The first 5 minutes are just after the HDCD has stopped playing (in Pure Direct mode 2).

At 5 minutes the player is shut down.

At 15 minutes the preamp is shut down.

At 20 minutes a coax 75 ohm video cable is connected from the Denon to a video isolator that connects to a nearby TV.

At 25 minutes the TV is turned on (on a different circuit).

At 30 minutes the preamp is turned back on.

At 35 minutes the Denon is turned back on.

At 39 minutes, Play is pressed on the Denon followed by Pause as quickly as possible.

At 44 minutes, Pure Direct is turned off.

At 49 minutes, cable to TV is removed.

At 54 minutes, a stub cable is attached.

The noise falls noticeably when the player is turned off, and still more when the preamp is turned off, and comes back when the preamp and player are turned back on.  Other than that, there's not much change!

When I turn the Pure Direct mode off with a video cable connected to a TV, the noise rises about 0.3dB to the highest level in this recording (44 minutes).  The noise falls again when the cable to the TV is removed and reaches the lowest fully operating level when a sort stub video cable is connected to the player only.

*****

I can typically amplify noise portions from the Denon 9000 not containing special peaks by as much as 93dB in Audacity.

It appears this "noise" is about 70% 60 Hz hum, which gets knocked down 27dB by the traditional A weighting.  So if were all hum, that would mean a 120dB S/N ratio.

I'm not claiming it's exactly that good, perhaps more like 108dB for my entire recording chain here.  Consider:

Denon DVD-9000 S/N: 120dB

Emotiva XSP-2:  (unbalanced input to balanced output)  112dB

(balanced in to balanced output is specd at 115dB and is typically 117dB)

Tascam DA-3000 (balanced input): 113dB

The specs (confirmed elsewhere) indicate it's not going to be as good as 112dB, perhaps more like 108dB S/N.  And that seems plausible with my measurements and a guestimate in a previous post.

I can't fully isolate the parts for testing...may just have to live with that hum.

I investigated the noise under a series of conditions:

1.  Prior to playback

2.  After playback

3.  Player turned off

4.  Preamp turned off also

5.  Video cable connected to TV (using isolator)

6.  With and without the highest "Pure Direct Mode"

The Denon DVD-9000 does not cheat with the visible noise level.  There does not appear to be any circuit that suppresses noise during pause or stop.  The noise level before and after the studio noise on any album (which I can clearly see when amplified by 70dB in Audacity) kicks in is identical to the pause noise.  Not much difference when the Denon is even turned off (that's how quiet the Denon is!).  When the Emotiva XSP-1 is turned off, the noise does go down a few dB.  Then I'm looking at just the noise of the DA-3000.  

It does appear the Emotiva is the limiting factor, just as the specs suggest, but not much different than the DA-3000 itself so the noise only falls an additional 2dB or so when the XSP-1 is turned off.

If and only if a video cable is connected to my TV does the "Pure Direct" mode make any difference.  With the video cable connected to the TV, the noise goes up about 0.3dB when "Pure Direct" is turned off.

I measured less noise with the video cable connected and Pure Direct than with the video cable not connected and Pure Direct, but that was probably an experimental fluke.

The least of all noise was measures with short stub video cable connected, and Pure Direct mode.  So that's my setup for the most critical transcriptions.  The stub video cable makes it easy to connect or disconnect the video cable to the TV without getting behind the Denon DVD-9000, which is usually impossible.

Testing an Oppo BDP-205 connected to the Emotiva through balanced inputs instead, I could amplify the output by just over 95dB, a bit more than 2dB higher.

The Emotiva alone is at least that much better with balanced inputs, let alone how good the Oppo is--119.8dB S/N in fast linear phase mode according to Archimago, which on Archimago's recommendation I've just tonight started using...and now that I am using the fast linear phase mode...it's finally sounding as good or better then the various Denons.  (I'm now thinking that Denon's "AL 24 Plus" as used in the DVD-9000 is a linear phase type system too, implemented with FIR probably.  Denon even claims it's "adaptive" suggesting multiple filters...)

So I'm shocked but with either of these two players, the player as such is not the limiting factor but the preamp and the DA-3000, with balanced output players getting some benefit from the lower Emotiva noise with banaced inputs.

Actually this is not fully a test of the preamp because shorted inputs were not used.

The noise from the Denon DVD-9000 is what you could call interface noise, cased by the single ended connections and output resistance mostly.  The balanced "interface" of the Oppo is a mere 2dB better.  Their respective DA chips are hardly even part of the equation with -120dB and -130dB noise respectively.

No Evidence for HDCD Transient Flag playback operation

Ignore the other channel on bottom, simply compare the top (played on Denon DVD-9000, (attenuated by Emotiva XSP-1, recorded by Tascam DA-3000) with the bottom (original AIF file copied from CD by iTunes). 

Here is visual comparison of a transient with HDCD decoding, and the raw "no-effect HDCD" digital data (having no other feathers than Transient Flag enabled).  The decoding is done to 24/96 from the one player which I thought there was still hope might implement the Transient Flag, the rare and expensive Denon DVD-9000 (which claimed next generation HDCD decoding, one of the first not to use PMD100).  To see the effect of the transient flag, one would have to use a higher sampling rate (I used 24/96 in recording the transient on top...that should be sufficient to show a variable playback filter if it existed).

Nevertheless, they look identical.  Comparing the two graphs I sometimes see a new detail on one, but then when I go and check the other, it has that same identical detail too, as often happens in subjective audio comparisons.  (To "hear differences" may mean not bothering to go back and check when something allegedly "new" is hear.)

Now actually, I don't know that Transient flag is operative on this stretch of audio, it was just the most visibly clear transient in the first track of Fresh Aire III "on HDCD" which has Transient Flags enabled but uses no other HDCD features (peak extension or gain adjustment).

I always thought this Fresh Aire box set sounded especially magical on HDCD players.  However the best evidence now is that in every case it was just lighting the HDCD light, and doing nothing more (on the user side...it did indicate you had an authentically unaltered HDCD recording made with Pacific Microsonics Analog to Digital Converters (either the One or Two).

And those converters were said to be among the best sounding ever...  Many used them because of that AND because of the convenient Peak and Gain Adjustment features, which meant setting setting the exact levels was less important and you didn't have to worry as much about clipping since far out peaks were automatically compressed (and then re-expanded by an HDCD decoder).

When the HDCD Peak and Gain features ARE used, as much as 1 bit or 6dB of effective dynamic range is added (about 4.5dB on the album Mephisto & Co. according to my measurements).  This is not always insignificant.  Inexpensive audio equipment including preamps and amplifiers can have 10-20dB more dynamic range than conventional CD's...120dB.  We can experience at least 130dB (but I recommend not going above 110dB for sake of your hearing).

The argument that the Objectophiles make is that acoustic noise makes this irrelevant.  I don't accept that...acoustic noise is something fundamentally different.

It makes sense to use 24/96 (or 24/88) and nothing higher, as those are essentially beyond the capabilities of even the best audio electronics anyway and can't in any way be felt to be inadequate.

And most of the benefit comes from the first extra bit and bump in sampling rate.

In theory, HDCD did both, with the second coming from a variable playback filter that could be bumped up and down to make transients more transient.  It also made it possible to distribute "media" and "audiophile" versions of recordings at the same time, with audiophiles (who aren't seeking maximum loudness) getting the benefit of the wider (closer to the limits of perception) dynamic range.

The standard story is the variable playback filter was never implemented because the patent for such variable filters was owned by Ed Meitner of Museatex.  HDCD still uses a variable filter prior to digitization, where it's most important anyway, and that's what triggers the flag.  Many experts say that even if implemented the effects of the variable playback filter would be very marginal.

That appears to be the case, and my doubts are mostly evaporated by this test.

But HDCD still delivers nice sounding encoder/decoders with almost 1 bit of dynamic range added.  It's only irrelevant now because true high definition like 24/96 is so easily accessed (if it's even made available).

HDCD files should be expanded if they can't easily be played on HDCD dacs (which are becoming harder to find).  Expanders include HDCD.exe, Foobar2000, and ffmpeg.  I use ffmpeg, which is said to use good code from Foobar2000 if you see the HDCD statistics being reported.

Since these decoders were based on reverse engineering, it's possible that licensed decoders in players are slightly more accurate.  But it hardly matters for the peaks we are talking about.  And it's likely the software decoders are in fact quite accurate as many Objectophiles poured considerable effort into getting them correct.

There is no need to do anything when only the Transient Flag is enabled.

AFAIK and according to these data, anyway.

Software Converting HDCD

I'm finally converting the HDCD albums on my harddrive to 24-bit, using ffmpeg.

I should have done this years ago.  

Instead I depended either on playing the physical discs themselves in my Denon DVD-9000 player (and resampling back to digital for my crossovers and EQ's), or running the SPDIF single into the DVD-9000 from my computer through Sonos.

I spent many hours carefully setting up a HDCD-resampling path to my system.  Every time I changed the system, I had to move things around to keep it still working.  I was doing more of that setup stuff than actually playing HDCD's it seemed.

But there really wasn't any reason not to use a software converter, either foobar2000 or ffmpeg.

A few years ago I installed ffmpeg on my computer mainly so I could just TEST what kind of HDCD features discs had.  I didn't pay attention to the fact that the script was actually doing full conversions of HDCD files so it could make reports about them, and then just trashing the conversions.

Here's the simple script hdcd-convert that will convert all the files that are wildcard specified:

#!/bin/bash

rm junk.flac 2>/dev/null

for name in "$@"

do

    ffmpeg -i "$name" -af hdcd "$name".hdcd.flac

done

I run this script in a terminal with a command like this:

hdcd-convert *.aif

This does both the peak extension and the gain adjustment.  AFAIK nobody does the variable filters in playback, and I'm finally coming to believe that (though I still have a bunch of test resamples of my favorite HDCD discs played on the best HDCD player I have...the Denon DVD-9000).

Checking the actual Denon 9000 manual, no mention is made of Museatex (as I previously reported incorrectly) nor their patent on switched playback filters, which btw is:

5,388,221

The Denon manual does list a bunch of patents used for HDCD, but not that one.  In fact the listed patents for HDCD all have higher numbers than that one too.

5,479,168

5,683,074 

5,640,161 

5,808,574 

5,838,274 

5,854,600 

5,864,311 

5,872,531

The Denon DVD-9000 provides no way to turn HDCD decoding off, so I cant test both ways through the same device.

The peak extension is a VERY BIG DEAL on the discs that have it, which include Mephisto & Co., one of my all time favorites, and most other Reference Recordings CD's (though strangely about half of mine either don't have HDCD or don't use it except for flipping the useless Transient Filter flag).

I'm searching through my other discs for those that use HDCD.  Neil Young, Grateful Dead, and many others used it.

I'm pretty much in line with Charlie Hansen's thinking about HDCD.  It had some good aspects and some scam aspects.  I do not consider it entirely a scam.  In many ways it makes sense, and it could have been a solution to the "loudness wars" (where newbies use the effectively compressed undecoded discs, and Audiophiles get the fuller dynamic range).  But for various reasons, we got the unabated loudness wars instead.

Charlie dismisses the potential usefulness of the transient filter flag (TF) in a weird way.  He says that when the sampler uses a filter with higher leakage of aliases, it makes no sense that the playback decoder would also use a filter with higher leakage of aliases.

Funny because Charlie had his own digital filter design that had very nice impulses but far more alias leakage then the usual digital filters.  He should have known why someone would do that: for better transients!

Actually, in my mind it makes no sense to do what Charlie did, use a "transient correct" but high leakage filter All The Time.  That would contribute grunge to dense passages.  The variable filter option sounds like it might be a better idea than his to me.  The TF should only be used for simple transients, not thick textures.

Arnie Kruger however had a very simple but (mostly) devastating critique of HDCD:

The core problem is that the higher bit depth is not realized in the actual signal, not even in the master recording.

It is easy to make 24 bit recordings where the bottom 9 or more bits are noise. As a matter of fact, that's always what happens when you make real world recordings.

At playback time it doesn't matter whether the bottom 9 or more bits as represented in the analog domain were noise sourced in the original recording environment, or whether they were sourced in the dither of a 16 bit medium. Its all noise, not music.

In this same thread Arnie puts HDCD in the same category as "true high resolution" formats like DVD-Audio, SACD, though it only provides about a bit or so of effectively increased resolution.

(There's no actual increase in resolution at all, except that it's shifted over a slightly wider range, making it "effectively" increased where it's more needed.)

But Arnie simply believes none of that is necessary.

While I completely agree that electronic noise in the recording studio is just as bad as electronic noise in the playback system, acoustic "noise" in my opinion isn't really noise at all...it's part of the product...the ambience.

Early indications are that the noise level from my Denon DVD-9000 transcriptions is around 92dB or so without any kind of weighting.  With proper A Weighting that might improve to around 105dB or so I think, but I'm wondering why I'm not getting closer to the spec for my DA-3000 sampler which is 114dB or the Denon DVD-9000 spec of 118dB.

But I argue now and still that "noise" and "lack of resolution" are not the same thing.

I still believe it's important to capture the "resolution" even if a tad bit of noise is being added.

So that's my second disagreement with Arnie.  Though I feel a bit less strongly about it (the difference between noise and lack of resolution might not be very big in many cases).

Anyway, with my transcription system having unweighted noise around -92dB, when I amplify it way up (by 90dB) I can see when the actual noise from the recording starts a fraction of a second before the music itself.  The "noise" (which includes ambience of course) is easily much higher than -92dB on the recordings I've looked at.

The hardware conversions I'm doing using the Denon 9000 will always have more noise than the software conversions.  There probably isn't a real reason to do the hardware conversions, but I'm doing them to satisfy myself that they aren't necessary.