Romy the Cat wrote:

I disagree with Mark that HDCD is beneficial for 88.2kHz/24bit...

manisandher wrote:

"A little information is dangerous. I find it incredible that people with absolutely no understanding about the technology and its application feel the need to speak with the authority of an expert. The HDCD processes "Peak Extension" and "Low Level Extension" are only available when creating 44.1kHz/16 bit material. These processes are user selectable and were not used in the creation of the BSO Classics masters. Further, we have not made a master in our history that utilized any  process that requires decoding to be played back correctly. Period. At 88.2 kHz the HDCD process alerts the PMI filter chip in the DA converter that the material was created  with an AD converter with a known anti-aliasing filter and noise shaping curve. The PMI filter chip can then utilize a reconstruction filter optimized for the specific AD converter used for the creation of the master. This process is invisible to all other DA converters and the HDCD dither/noise shaping is considered to be the dither of choice for high resolution material."

There was second factor that I found worth to note. Experimenting with Low Level Extension in past (all on 16 bit) I did found that Low Level Extension (or actually a mild Pacific’s proprietary compression) had a positive effect and was very much align with some of my critic of the BSO files. I am not exactly calling the BSO file as “intelligently uncompressed” but there is something in me feels that suggest me to do so. The BSO files are in a way as an un-groomed dog where dymick in a way obliterates the fine satellites of imaging. I thought that with “intelligent compression” that might be addressed. With FM files I have not an “intelligent compression” but over compression and dynamically the raw 88/24 files and not even close to FM result. However, from imaging position FM results are very much contestable and in some way more interesting. Do not forget that the 88/24 files and the FM files were most likely made from the very same feed.

I am writing it and it ridicules that I put myself in a position where it sounds like I am trying to solicit Mr.Donahue… to “properly” compress his files.  I certainly do not do it and I certainly do not imply that I know how they need to be compressed.  But I would also point out that from a certain perspective I do not consider the BSO files as “as good as they could be” and I did comment about some of my dissatisfaction in imagine department. A hint might be in the fact that Deutsch Requiem is better than other files), what different between them only Mark Donahue knows.

Finally, I am not specializing in mastering of recordings or about the evaluation of technologies to do so but rather in listed and discriminating the results. What I see in this exchange is that Mark unwillingly agrees with my observations. He said “If you can turn off the HDCD decoding you will find a significant decrease in the spatial imaging and overall imaging acuity” but it look like exactly how we, the people with no HDCD processors, hear the BSO files.

I think instead of blaming me in impersonisation of authority while have little knowledge Mr. Donahue might use his considerable knowledge and to make our BSO recordings to play via the HDCD-disabled processors without the “significant decrease in the spatial imaging and overall imaging acuity”. We are in away all rowing in a same boat and have the same objectives – to make BSO sound file as good as they could be. The relying on HDCD to engaged specific filters I think is not kosher way to do as the unfortunately there is not a lot of DACs with Pacific decoding chips out there and even those that can recognize HDCD and have on-board the complex interpolation HDCD decoding filter  are not necessary the worthy performers otherwise.  This conversation encourages me to rewire my Pacific and to try using it’s D/A for paying the BSO files. Nowadays one of my Melquiades out of service but when I will be back and when my installation will be up then most likely I  will try to do it.

Doing some further thinking by my mind that has “little understanding about the technology” I would like to share another concern that for my little mind does make since.

According to Mr. Donahue comments above playing the BSO files with a possessor that has HDCD decoding activated provide a “significant increase in the spatial imaging and overall imaging acuity” as the activation of HDCD decoding will inform the “DA converter that the material was created  with an AD converter with a known anti-aliasing filter and noise shaping curve”. OK, I can’t corroborate on it now, Mani keeps silence about the validity of this claim (most likely he just has no options tot shut down the HDCD decoding as most of the DACs has no such a manual switch). I do not believe that what Mark Donahue suggest is not true. Most likely it is what it is, but if it is so then… I do not like it. Let me to explain.

Let pretend that we have an objective to deliver to the end user the SoundMirror’s BSO file at the best possible condition in term sound quality. If the configuration and type of the filters that prevents frequency sampled at higher than one-half the sample rate (aliasing) so affective to sound (wish it true) then let see what filter might be the most optimum for us?

Mark Donahue went for the filters that were used by Pacific in their HDCD. I have no knowledge what kind filters they used in there BUT the HDCD was designed and was beneficial ONLY at 44.1KHz format. So, regardless what they did in there the signal had to be stopped at 22KHz with the brick-wall effect. Yes, the Pacific is very good 44KHz DAC…. BUT the SoundMirror’s BSO file is not 44KHz files and this set all bid off.

This is just a brainstorming of the concept, and if Mark or others read it then do not be damn to recognize it as criticism but rather as collaboration how to get best possible results. What I question: if the selection of the 44KHz-targeted brick-wall filter is rational for 88kHz files. People go for sharp and complex filters at 44KHz as the Nyquist point is right there, very close to the upper edge of signal. With the 88kHz we care less about it as the Nyquist point is very far and we can go away with much softer filter and with much more phase-friendly curve. This absolutely certainly would NOT be a filtration that was meant for strictly 44KHz operation.

Now, we all know that anti-aliasing filters in many ways have negative effect to sound. But remember: our abstract objective is to have the best sound on the user side. So, what SoundMirror do? They select the configuration where the best Sound will be under a condition when s user engages the anti-aliasing filters designed for 44KHz HDCD. Leaving aside the fact that there are not a lot of HDCD DACs out there let ask ourselves if it was the best solution for BSO 88kHz files? My answer is that it was not. I think the best would be if SoundMirror folks use not the 44KHz HDCD filters for their reference but rather the 88kHz with no filters at all.

Sound damn, isn’t it? Well not necessarily. I think that amount of users out there who would be able to take advantage of HDCD filters is no higher than the amount of people who would be able to play the files with no filter at all. The advantage of playing the 88kHz files with no post-conversion filtration of any kind enormous theoretically and practically. Why SoundMirror folks did not go there? Well, because SoundMirror is a part of industry establishment and they need to embrace all industry rules of the game, no matter how stupid the rules are. However, in the case the BSO files the SoundMirror own the rules, why wouldn’t they get for the best the best?

To get the best it would need very little: the SoundMirror need to use for this reference while they master the BSO files a R2R Multibit DAC with no post D/A filtration of ANY kind. The true Multibit has no wide- bandwidth noise, the signal is fully reconstructed and there is no need for noise-shaping. Averything is already there, just buffering the signal and feed it to amps - and you have the rawest possible, virgin signal not contaminated by phase anomalies of any filtration – it imposable to have anything better than this.  Talking about “imaging”, that is a manifestation of phase consistency…

Sure you will have remains of the 88kHz sampling rate right along with you signal, so what? When did you hear the 88kHz last time? There is however a reason why SoundMirror would not go there, the industry things… Take a look at the SoundMirror control mastering rooms:

• The monitors are B&W Matrix series powered by Threshold amplifiers.
• Dunlavy SC series speakers powered by Classe amplifiers
• B&W Diamond series monitors powered by B&W amplifiers.

Is any pattern that you recognize in there? Yes, there one: all amplification that SoundMirror uses is push-pull, solid-state, high power amplification that most likely uses feedback, a lot of feedback. Those amps would NOT be very friendlily with high content of HF power in input signal and they very much might go berserk, oscillating, burn tweeters and do other nasty things. However, a good no-feedback amp with reasonable amount of open-loop equalization as well as any no-fedback SET users have would do perfectly fine with unfiltered Multibit output and will not have ANY of negative effect from the absence of post-conversion filters. I wish SoundMirror use THAT as the reference not the filters that were constructed for 16 bit 44kHz. The HDCD was wonderful under the umbrella of 16bit Red Book CDs. Now we are way beyond it and I think the SoundMirror with their 88kHz files might move forward along with the opportunities the new format implies.

Romy the Cat wrote:

Now how much better would be if people abandon that idiotic SACD/DSD and Mark Donahue would record BSO right in PCM at 88.2 kHz/24bit or 172.2 kHz/24bit format? Then the files that we get will be as close to raw master files as it might be.The Cat

manisandher wrote:

Unlike you, I've never heard 4-bit DSD. But whatever DSD/SACD format Soundmirror is using, it's pretty damn good - it sounds nothing like the 'woolly' SACD I've experienced. Could the sound quality be improved? Sure, in the ways that you've outlined. However, the criticisms that you voiced in your earlier posts don't seem to me to be down to the original use of DSD/SACD recording...

I am not a big expert in SACD. I never had and SACD machine and the only reasons for it is that any single double layer SACD/CD that I have seen had absolutely horrifying quality of sound in CD layer. Most of those recordings were re-mastered and if you compared the regular CDs from beginning of 90s to the CD-layer on the SACD for 2001-2003 then it will be very clear that the idiots who produced them deserve to be castrated without anesthesia. It is not the format itself that turned me off initially but the extremely barbaric mastering that I heard on CD layer. In fact I trash quite a few SACD because of that.

I did not have any negative feeling about the SACD per say at that time but knowing that the VERY same people who vandalized the 16/44 layers were responsible for the SACD recordings I was very cold to SACD. In a few year I asked a friend of my who know very well the depth of digital design and mathematics to look into SACD and after his research he informed me that it looks very strange as it has not enough of “something” to do what format is need to do. I do not remember what it was but he has very deep analyses done. In another few years, sometime in 2005-2006 I learned that the DSD was originals designed not 1-bit but 4-bit processing and what I heard in 1999 was the Meitner’s own 4-bit non commercial prototype. Whwn I passed the information that the DSD meant to be 4-bit to a friend of my who did the math  he was laughing – as according to him ONLY then the whole DSD idea made engineering sense to him.

It would worth to mention that Sony killed the 4-bit DSD as it was “expensive” and embrace the 1-bit DSD – the format that has no theoretical capacity to become “properly implemented”. It is surely better then any 16/44, so what it is still a half-ass solution and I would prefer to stay with high-res (88K-172K) PCM as any editing in DSD anyhow is made by conversion to PCM. I wish the stupid SACD never took place as so many recording were in a way fucked up by SACD...

Romy the Cat wrote:

[Dima] feel that Lavry 924 is the only know to him properly implemented true Multibits DAC (and he knows them all) however he feel that there is much more “interesting” way to use this DAC. The DA924 is no oversampling Multibit at 88K and being Multibit is has no wide-bandwidth noise after the converter. The signal is fully shaped and juts care the residual sampling rate minus the sound frequency – means above 60K noise. So, Dima feel that in this case it makes sense for get rid the output stage and complicated and unnecessary output filter and feed mu load directly from the DAC deglitcher. In fact it is what he used on his Multibit for years and he feel that it is a way to go. Dima admit that it is not universal solution as the UHF nose would be bad or even damaging for some electronics, particularly if is it uses feedback , however in my case what Melquiades has no feedback and the “grid bias” acts as a mild low pass filter he feels that it would be problem-free. Probably if I try it I will put still a first order low pass filter after DAC somewhere around 18-10K (juts by hearing) and it will elevate the output impedance. I have 1 feet Dominus that I might use and I might position the DAC very close to load. What I am concerned if anyone who have playback capable for full-scale bass use such an approach. I am afraid that DA924 (that has a phenomenal lower bass now) might lose some of it driver and bass is the one that usually go first… So, it you use the Multibits-driver-through approach and have something to express on the subject then raise your voice…

Romy the Cat wrote:

I think that amount of users out there who would be able to take advantage of HDCD filters is no higher than the amount of people who would be able to play the files with no filter at all.

manisandher wrote:

Did you ever try this?

manisandher wrote:

But there is a BIG difference. If you don't have an HDCD filter, you just lose the added "spatial imaging and overall imaging acuity" that such a filter would provide. If you don't have a no-feedback amp, you lose your tweeters and who knows what else.

While I, waken up by Mani’s idea to us MP as the only DAC, was playing with new Pacific I have researched deeper the HDCD subject in the BSO recordings. Mark Donahu, from AudioMirror with his slightly patronizing attitude might consider himself as a person with “absolute understanding about the technology” but in his quest for being “special” he stepped away from Truth.

His sentiment that BSO files were HDCD encoded intentionally to in order to “utilize better reconstruction filters” was not accurate. I did find a suitable for myself configuration with my Pacific how I can read the HDCD and I can testify that it make absolutely no difference at 88/22. Further research of this subject led me to realize that the problem was not that there is no difference between “with HDCD” or “without HDCD” but that Pacific has non-defeatable HDCD decoding. Using different configurations it is possible to make Pacific to show off the presents of HDCD code or not but if Pacific does not indicate the HDCD presents but the HDCD exist in Digital_ IN then the HDCD filter is engaged automatically. So, the Mark warring that “If you can turn off the HDCD decoding you will find a significant decrease in the spatial imaging and overall imaging acuity” was absolutely incorrect. There is no ways to turn off the HDCD decoding PROPORLY in Pacific but Lavry Gold that do not understand HDCD does not indicate any “decrease in the spatial imaging and overall imaging acuity” what it runs against Pacific with HDCD EXPLICIT on.

Also, AudioMirror sentiment that the HDCD was used “strategically” during them “mastering” the files was not exactly accurate. Pacific during DD operation and rate or depth change automatically reinsert the LSB code. So, Mark Donahu most likely recorded in SACD as most of those folks unfortunately do nowadays (they believe that it is DSD), then they converted it into PCM, most likely 172kHz. Then they “master” it and down converted it to 88kHz. The down conversion they did with Pacific, which is unquestionably the best tool for the job. While Pacific run the 172kHz- 88kHz it inject the LSB code that makes people to feel the material has HDCD code.

It will be interesting if AudioMirror crate two parallel masters in 172kHz and SACD during recording, but it is hard to know truth. The negative moments that we hear at the BSO files are not from the technology but from “Mastering” and heavy use of multi-microphones…. but this is a whole another story....

"When encoding to 44.1/16, there are a couple of options for fitting the 24 bit source into the 16 bit delivery format. The user may use one or the other or neither, operator's choice. But at any setting of 88KHz or higher those options are not even available. You've got a Model Two, Mani, check it out for yourself. There are no options for turning off any so-called HDCD functionality. Unless you count turning off the flag sensing on the monitor input, which has no effect other than changing the blue light to yellow."

Romy, you record at 24/88.2 - is this the case with the Model Two in AD mode? Or is this just the case in DD mode?

(I won't be able to check for myself until early next week.)

Mani.

I hate when those industry people steps out really and from facts just to maintain the faces. What would be next: they would record at +5dB digital and suggest that this type of sound cure listeners shin cancer?

Yes, Paul Stubblebine is correct, in over 16-bit mode the HDCD is not available (or necessary) but the point that Mark Donahu made was that if the 88/24 material has HDCD code then it tricks Pacific to  use “better output filters”. I did not buy this idea as the “better output filters” would be still the filters that were designed for 44kHz. There is no such a thing as “reconstruction filters” as those people like to scare people. It is just a low low-pass filer, nothing more or less, though it might be quite complex in some cases. At 88K and higher the demand to the low-pass filer filter is MUCH less and I see why we need to search for “better 44kHz output filters”. Most likely pacific has different type of filters for 1X, 2X and 4X and it’s it.  I presume that Mr. Donahu comments that “you get improved audio if your DA converter properly decodes
the HDCD (Even at 88.2 kHz /24 bit!)” were made to justify that facts the when Pacific do DD down conversion it automatically re-injects the LSB code, that any DAC would recognize as HDCD encoding.

Anyhow, the point I was making is that when Pacific runs in AD mode, you feed Digital_IN (this is the only way how I use it) and your digital has LSB code (HDCD encoded) then Pacific most likely read and LSB code without you knowing it. If you switch you meter souse from Digital_IN to Ouput then the HDCD indictor will light up blue. It is kind of confusing but if I run DD mode with proper HDCD decoding and AD with the presumed “surreptitious” HDCD decoding and hear no difference then can I assume that that my presumption about the  surreptitiousness of HDCD decoding in AD mode was accurate?

Whatever the “right” answer might be I do not feel that it is important as I see a NEED to encode 2X at HDCD as an absolutely false idea.  Pacific just does it apparently in the way how Mark Donahu used it and Mark created a “justification”, substituting the specific fault of a given processor with a fable about the benefit of HDCD encoding at 88K.

Romy the Cat wrote:

Yes, Paul Stubblebine is correct, in over 16-bit mode the HDCD is not available (or necessary)...

Romy the Cat wrote:

Pacific just does it apparently

It would be nice if somewhere exist a community of Pacific users with some kind of knowledge base about different ways to used this processor, but I think this time has gone as the HDCD time has gone… In the world of 16.44 only the HDCD was good. Nowadays it is an anathema…

I do not have knowledge about the nuts and bolt of with process and I use just common sense and admit only what make send to me. If you wish a definitive answer then get in touch with Berkeley Audio guys, watch the agendas through…

Anyhow, Paul Stubblebine and Mark Donahue both right and both wrong. They just  used different definition of HDCD. Yes, Paul is right - there is no HDCD at 88K but what is the HDCD? The HDCD is application of pseudo 4-bit peak extension, custom dithering, low level custom compression and use of specific post-conversion filters. Mark was incorrect saying that the “peak extension” and “Low Level Extension“are user selectable. They are NOT. They are to a degree user-optimizeable but the can’t be defeat if the HDCD is implied in 16 bit. However, Mark does not talk about the HDCD but about the LSB code that Pacific write even at 88K. Think about LSB code as 19 kHz pilot signal for FM – it is an identifier that FM signal has also stereo code on side-band apt to 53kHz. What DAC with HDCD chip reads LSB code it engage all HDCD options as it thinks that the source was HDCD pre-processed (peak extension, dithering, low level compression, custom filters…) but it happens with only at 16/44. What happened what DAC runs at 2X or 4X is unknown to me. Mark Donahue feels that nothing happens but the custom filters get engaged. I do not know if it is true and if it is then I would like to have not the 44kHz HDCD filers but the 88kHz as the Nyquist frequency is too high and there is a lot advantages at 88kHz do not use any stiff break-wall filtration.

Do not forget that Pacific was not made to use as you do – it meant to be mastering processor – run DD operation, to do A/D operation and the most important to prepare data for 16/44  HDCD. That is why Pacific has automated LSB code reinsertion in DD operation even if you run 88 to 88 bit conversion and the 16/44 HDCD is presumed to be final product anywhere.

If Mark Donahue in his initial replay do not behave as Napoleon Bonaparte and instead of suggesting that “it will improve audio if your DA converter properly decodes the HDCD (Even at 88.2 kHz /24 bit!)” he would say “because my damn Pacific injected the LSB code  during my DD mastering and you DAC thinks that it is HDCD encoded” then all this conversation would not happen.

manisandher wrote:

So, all the files you've recorded @ 24/88.2 have the HDCD flag? And you've never noticed? How can this be? I'm sure I've read your previous posts Romy where you state that you don't agree with HDCD for >16/44.1 and that you don't use it...

 I have commented before that with all my admiration of the fact the BSO made the 88/24 available I was not completely thrilled with the job SoundMirror did. Those zillion microphones that John Newton’s team stack in each empty space of Symphony Hall was in a way beneficial as it helped to keep the audiences noise down but it also created problem with “interesting” imaging of BSO. There was one ugly nuance that derived from the poly-microphone techniques. The BSO files are properly recorded at -2 db most of the time with very gimmicks but of you look at the last movement of the Mahler 6 then you will see that in few instances the signal enters very hard digital clipping. The most interesting is not the clipping itself but how the orchestral dB power rises. It is almost DC type of rise with characters of sound moving from full orchestral crescendo to simple high dB flooding of sound with a single note.  It never happens in normal music.

What I think happened in there was the SoundMirror’s poly-microphone bitten the BSO by its tail. Most likely  SoundMirror has microphones right among the trumpets or trombones sections and the microphones were calibrated for the section’s normal outputs. However, those horns are highly directional and it looks like some of the player in Mahler ecstasy was pointing the horns toward to the microphones, completely flooding BSO balance. It does not happen in all movements and I do not think that SoundMirror channels the mix between the movements. They might use the movements from other days but in the one that they have chosen they missed that very unpleasant change of orchestral sound at high dynamic level. It would never happen if they records with two microphones and dummy head. I also have absolutely no problem with standard 3 hanging BSO microphones from which FM is broadsided.

Mozart: Symphonies: 14, 18, 20, 39, and 41

http://www.bso.org/bso/shop/audioDetail.jsp?pid=prod3690050&area=&id=bcat13360032

Celebrating Carter's Century: Highlights from the 2008 Festival of Contemporary Music at Tanglewood

http://www.bso.org/bso/shop/audioDetail.jsp?pid=prod3720206&area=&id=bcat13360032

Mozart Chamber Music for Winds and Strings

http://www.bso.org/bso/shop/audioDetail.jsp?pid=prod2910346&area=&id=bcat13360032

The Levine’s Mozart is very fine (I was there), if they use the right concert, which I think they do. The tow other I did not heard. My only concern is that those CD were able foe long time and only now thy become available in 24 bit. So, there is a remote chance that they are not the doubles of the raw 24 recordings but juts upsampled CD. I still am going to but then to see how it goes…

I recently interact a lot at BMI site and there were some “interesting” articles in there about BSO recordings.

http://classical-scene.com/2011/10/02/bso-recording/

http://classical-scene.com/2011/10/11/9243/