|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
147.10.167.184
In Reply to: This is just silly posted by Charles Hansen on December 22, 2005 at 09:08:41:
Regardless of how you choose to interpret one statement in an FAQ, the facts are these: an invention ("HDCD") is legally described by the patent, not by the implementation. The patent clearly spells out the mechanism for playback filter switching. Whether or not this is implemented is immaterial, and you have absolutely no proof (other than a distrust of Pacific Microsonics) that what was described in the patent has not been implemented.*** Regarding the signal level issue, you are simply wrong. If you understood the compansion process used in HDCD a little better, you would realize your error. I will try to explain this to you. ***
I think it is you that doesn't understand. The Microsoft implementation of HDCD automatically attenuates the digital signal by 1 bit (or approximately 6dB). If Peak Extend is engaged, the top 3 dB in the digital signal is expanded to 9dB. If Peak Extend is not engaged, no expansion occurs, therefore the result does not exceed -6dB.
Therefore, if the digital samples stored on the CD has signals between 0 and -3dB, and if on the output of WMP these signals are reproduced as -6 to -9 dB, then clearly that particular album/track does NOT have Peak Extend. Otherwise these signals will be expanded to 0 to -9 dB. Understand?
*** A DSP chip that decodes HDCD is also not a "low cost" item. ***
Not to you, perhaps. But to a large manufacturer, a DSP *is* a low cost item. Last time I checked, the pricing of these things in bulk is quite reasonable.
No one is forcing you to support HDCD. But your comments on what it is and isn't are wrong. Please don't mislead potential customers from thinking they are getting optimal performance listening to HDCDs on non-HDCD players, because they are not.
Follow Ups:
Christine wrote, "You have absolutely no proof that what was described in the patent has not been implemented." Just as you have no proof that it was implemented. Without further information, we will have to agree to disagree.On your second point, you stubbornly persist in being flat-out wrong. You wrote, "The Microsoft implementation of HDCD automatically attenuates the digital signal by 1 bit (or approximately 6dB)."
This is incorrect. Here is the text from the Microsoft document explaining gain scaling:
A key feature of the HDCD process is Peak Extend. Peak Extend increases the dynamic range of
Redbook CDs by 6dB. The Peak Extend feature is selectable on the Model 1 (or Model 2) HDCD
encoder - the recording or mastering engineer has the ability to set Peak Extend ON or OFF on
the Model 1 HDCD encoder. So it is possible to have HDCD recordings with, or without Peak Extend.
Because Peak Extend adds 6dB of dynamic range to the top end, the “average” decoded
signal level will be 6dB lower than an HDCD recording without Peak Extend, or a non-HDCD
recording. Unless the decoded level of Peak Extended and non-Peak Extended recordings are
matched using Gain Scaling, Peak Extended Recordings will be 6dB quieter than non-Peak
Extended recordings and this is not acceptable. All HDCD decoders must either 1) lower the gain
of non-HDCD recordings and HDCD recordings without Peak Extend by 6dB, or 2) raise the gain
of the Peak Extended recording 6dB. This is a requirement of the HDCD license agreement.So contrary to your unfounded assertion, the signal is *only* attenuated during playback if it was recorded with "peak extend" enabled. Therefore you *cannot* determine if a recording was recorded with "peak extend" enabled simply by examining the peak signal level of the decoded output.
However, I just realized that there actually *is* a relatively simple way to tell if a recording was made with "peak extend". Pin 5 on the PMD-100 goes "high" when the recording employs "peak extend". (At least this is the case when analog gain scaling is employed in the DAC. Tying pin 19 low on the PMD-100 enables internal digital gain scaling, but the datasheet doesn't say if this disables the output of pin 5 or not.)
So if one had a DAC with the PMD-100 (it may need to use analog gain scaling for this to work), then it would be easy to tell what percentage of HDCD discs use "peak extend".
*** Christine wrote, "You have absolutely no proof that what was described in the patent has not been implemented." ***My point was I don't need to prove it, since an invention is described by the patent, not the implementation. You on the other hand is clearly wrong, since your interpretation of an admittedly badly written sentence does not correlate with what's described in the patent. What is or isn't implemented is completely irrelevant, since when we are talking about HDCD we are talking about the technology, NOT a specific implementation.
*** On your second point, you stubbornly persist in being flat-out wrong. You wrote, "The Microsoft implementation of HDCD automatically attenuates the digital signal by 1 bit (or approximately 6dB)." ***
Charles, I am describing what WMP does when playing HDCD. If you don't believe me, check it out yourself. Would you like me to send you the actual WAVE files as output by WMP vs what's stored on the CD so you can clearly see this 6dB attenuation for yourself? I can send you an example of a disc showing Peak Extend in action, and other without Peak Extend.
Again, before you call someone wrong, it helps if you actually know what you are talking about.
PS - The MediaTek all in one "DVD player on a chip" does HDCD decoding. This is used in $50 made in China DVD players.
On the first point, either you are being taken in by deliberately misleading marketing fluff, or I am being cynical about their lack of forthrightness. I can provide *lots* of evidence for my position, but none of it is out-and-out proof.The only additional thing I've found since yesterday is the block diagram from the datasheet for the PMD-200. The section in question is labeled "2xFs Interpolation Filter" (note the use of the singular once again). Like every other reference to the decoding filter, it leans toward the idea that there is one single filter. I suppose it is possible that this filter has variable coefficients, but that seems quite a stretch to me. Again, we won't know until we have some additional facts.
On your second point, I'm not sure I'm following you. Your idea of comparing undecoded data with decoded data is a good one. This will clearly show if the level is being changed by -6 dB. According to the HDCD specification, this will be done only to those HDCD discs that employ "peak extend".
If I understand you correctly, you are saying that WMP lowers the level of *all* HDCD discs by -6 dB. If this is correct, I have two questions for you:
a) Why would WMP not follow the HDCD specification?
b) How do you then tell the difference between an HDCD disc that uses "peak extend" and one that does not?
On your third point, I'm afraid you don't understand the situation. Just because MediaTek (a Taiwanese semiconductor foundry) sells MPEG decoder chips for a low price to high-volume Chinese manufacturers doesn't mean that it is even *available* to Western-based low-volume manufacturers. If you don't believe me, try to get a lowly datasheet or even a sample chip. And even if you could somehow get a hold of the chip in small production quantities, it would be nearly impossible to use one in a CD player. MPEG decoder chips represent a monumental programming task. For example, Madrigal spent five man-years developing the Proceed PMDT DVD player. When they released it (a year behind schedule) it was so bug-ridden that they wouldn't even submit it for review.
So we at Ayre certainly wouldn't go down the path of using a MediaTek MPEG decoder as a way to provide HDCD playback on a CD player. But there is certainly more than one way to skin a cat, and you are more than welcome to design and sell your own CD player.
*** On the first point, either you are being taken in by deliberately misleading marketing fluff, or I am being cynical about their lack of forthrightness. I can provide *lots* of evidence for my position, but none of it is out-and-out proof. ***No Charles, you can download the patent from the US Patents Office web site. It is a *legal* document that describes what HDCD is, and not "deliberately misleading marketing fluff".
As for your evidence, you haven't actually given any, apart from you misreading a sentence, and irrelevant material/guesses concerning one particular *implementation* of HDCD.
*** If I understand you correctly, you are saying that WMP lowers the level of *all* HDCD discs by -6 dB. ***
Actually, it lowers the level of ALL CDs (HDCD encoded or not) by 6dB (when you select the 24-bit output mode, which is required for HDCD decoding). More precisely, it maps the 16-bit digital signal into a 24-bit word just below the most significant bit. This is to allow for Peak Extend encoded discs to expand into the most significant bit.
I'll attach four graphs to illustrate, taken from two HDCDs (one encoded with Peak Extend, one without). As you can see from the graphs, the presence or absence of Peak Extend is fairly obvious.
The non-decoded ones are digital rips from the CD, done using EAC in secure mode.
The decoded ones are the output of WMP playing these CDs, exactly as WMP would send to a WDM driver for an audio card (these waveforms were captured by the WDM rerouting the output into a WAV recording). No editing have been done on the output - ie. they are exactly what comes out of WMP.
*** Just because MediaTek (a Taiwanese semiconductor foundry) sells MPEG decoder chips for a low price to high-volume Chinese manufacturers doesn't mean that it is even *available* to Western-based low-volume manufacturers. ***
The MediaTek decoder is used on the Oppo DVD player, which is designed by a (smallish) US company. It is also used in many DVD players from boutique manufacturers such as NAD (Canada/UK) and Arcam (UK). I think all three companies qualify as "Western-based low-volume manufacturers."
It's not a bad decoder, and can certainly be used as the basis for a good DVD player (eg. the Oppo).
PS - I don't understand why you think it would be difficult for the original PMD chip to support multiple filters. Again, if you had actually read the patent, you would realise the filters only differ by three parameters. It would be very easy to build a chip that essentially has one filter, but allow the parameters to be dynamically reloaded.
This is an HDCD disc without Peak Extend, by the same artist (Joni Mitchell), taken from her album "Both Sides Now". This is track 3 (Comes Love).Not the extreme peak limiting on this track. It looks like it's clipping, but when you zoom in, you realise it's not - all the peaks have been limited to just under 0dBFS.
As you can clearly see, on a disc that has no Peak Extend, WMP does not expand the peaks, so the entire track does not have any peaks exceeding -6dBFS.
This is what Adobe Audition reports for the digitally rip (non-decoded) wave file:
Left Right
Min Sample Value: -32425 -32418
Max Sample Value: 32399 32416
Peak Amplitude: -.09 dB -.09 dB
Possibly Clipped: 0 0
DC Offset: 0 0
Minimum RMS Power: -89.95 dB -89.84 dB
Maximum RMS Power: -4.54 dB -5.32 dB
Average RMS Power: -17 dB -17.27 dB
Total RMS Power: -15.63 dB -15.85 dB
Actual Bit Depth: 16 Bits 16 BitsUsing RMS Window of 50 ms
By contrast, this is the statistics for the WMP decoded output:
Left Right
Min Sample Value: -16212.5 -16209
Max Sample Value: 16199.5 16208
Peak Amplitude: -6.11 dB -6.11 dB
Possibly Clipped: 0 0
DC Offset: 0 0
Minimum RMS Power: -95.57 dB -95.55 dB
Maximum RMS Power: -10.56 dB -11.34 dB
Average RMS Power: -23.03 dB -23.3 dB
Total RMS Power: -21.65 dB -21.88 dB
Actual Bit Depth: 24 Bits 24 BitsUsing RMS Window of 50 ms
As you can see, even for an HDCD disc that is NOT encoded with Peak Extend, the output of WMP is not identical to the non-decoded data.
In particular, if you zoom in on both files, you will find the low level information have been subtly altered, which is consistent with what the patent describes as "low level gain manipulation".
I think this is sufficient evidence to completely destroy your theory that there is no benefit or material difference to HDCD decoding on non-Peak Extend discs.
So, bottom line is HDCD decoding does make a difference, regardless of whether the discs are flagged with Peak Extend or not.
So let's review what we've learned here:1) Microsoft's implementation of HDCD decoding is different in WMP than it is for stand-alone CD players. I had no idea that this was the case, and I have no idea why they would do this.
2) By comparing the non-decoded output of an HDCD disc versus the decoded output, you can tell if "peak extend" was used in the recording. I hadn't thought of this method before, and this is a useful tool. (This method cannot be used with stand-alone CD players.)
On these points, the facts are self-evident and there is no room for disagreement.
Next you claim that by comparing the non-decoded output of an HDCD disc versus the decoded output, you can *probably* tell if "low level extension" was used in the recording. In this case the evidence is not quite as clear as with "peak extend", so I will reserve judgement on this. The statistical analysis you posted clearly shows a 6 dB shift in levels (as noted in point #1 above). Once this is compensated for, the biggest difference between the files is less than 0.5 dB (for "minimum RMS power"). Without further information, I don't think we can say one way or the other. In theory this method should work, but it is likely to be tedious, and I'm not clear that you've correctly identified a disc employing "low level extension".
And finally we return to the reading comprehension issue. You claim to "destroy [my] theory", but the problem is that you are setting up a straw man. I never said what you claim.
The HDCD encoder has two compansion schemes that are optionally engaged by the mastering engineer -- "peak extend" and "low level extension". (Furthermore, there are two varieties of "low level extension".) However if BOTH of these are *not* engaged, then there is no need for HDCD decoding.
I will recap one more time, in hopes that this will sink in:
1) When HDCD was introduced, it represented a true sonic breakthrough for the CD format. This was because the HDCD encoder was head and shoulders superior to the only other widely available choice at that time -- the horrible sounding Sony 1630. However, most of the gains found with HDCD were due to the encoder's features that DO NOT require decoding:
a) Dithering.
b) Switchable on-the-fly anti-alias filters.
c) Keith Johnson's discrete analog circuitry, excellent power supplies, and low-jitter clocks.
Nowadays there are many great sounding A/D converters available (none of them with HDCD encoding), but 10 years ago it was a completely different story. And while there are still many HDCD encoders in use, there are none available for sale today.
2) Unless an HDCD disc was recorded with "peak extend" and/or "low level extension", there is nothing for an HDCD playback machine to decode. Hence there is no advantage to using an HDCD-capable player with these discs.
You and I have both described methods to distinguish HDCD discs using "peak extend" from HDCD discs that do not use "peak extend". It may be possible that your method is also able to distinguish discs that use "low level extension". What is yet to be determined is what percentage of HDCD discs use "peak extend" and/or "low level extension" and would therefore benefit from HDCD decoding.
Please note that I never said that these type of discs *wouldn't* benefit from HDCD decoding. Instead, what I said was that HDCD decoding didn't provide the full benefits commonly ascribed to it for two reasons:
a) Much of the sonic benefit from HDCD discs does not derive from the HDCD process per se.
b) Many (we don't know how many) HDCD discs are not encoded in such a way as to require decoding.
*** On these points, the facts are self-evident and there is no room for disagreement. ***I think you will now agree that previously when you were asserting that I was not able to distinguish between discs that have peak extend or not, you were wrong. Also, when you persisted in calling me wrong in my description of what WMP does, you were also wrong.
Again, perhaps an apology would be in order?
*** Once this is compensated for, the biggest difference between the files is less than 0.5 dB (for "minimum RMS power"). Without further information, I don't think we can say one way or the other. ***
I've already said: further information is there, by inspecting and comparing the actual shape of the waveforms. You are more than welcome to repeat what I've done, and evaluate the differences yourself. What you will find is that these differences are on every HDCD disc (or at least in my case every single HDCD I own).
*** The HDCD encoder has two compansion schemes that are optionally engaged by the mastering engineer -- "peak extend" and "low level extension". (Furthermore, there are two varieties of "low level extension".) However if BOTH of these are *not* engaged, then there is no need for HDCD decoding. ***
No, you are still wrong. Unlike Peak Extend, the low level manipulations are NOT optional. Again, read the patent.
*** I will recap one more time, in hopes that this will sink in ***
Your recap is still wrong, and is not consistent with what the patent says.
*** Much of the sonic benefit from HDCD discs does not derive from the HDCD process per se. ***
This is debatable, since it is merely your opinion, and you have not provided any substantiation for this statement. Any it is clearly wrong for discs with Peak Extend, but we can argue about whether the benefits of dynamic filter switching and low level gain manipulation is "substantial."
*** Many (we don't know how many) HDCD discs are not encoded in such a way as to require decoding. ***
You are wrong, because the low level gain manipulation is NOT optional. It's related to the way HDCD does dithering.
You are amazing in your persistence in being wrong. You must love to be wrong. Repeat after me:THE PATENT IS NOT THE PRODUCT.
THE PATENT IS NOT THE PRODUCT.
THE PATENT IS NOT THE PRODUCT.
Please refer to the operating manual for the Pacific Microsonics Model Two HDCD encoder:
==========
Low Level Extension is an average signal level based low level compression / expansion
system used on HDCD 16-bit amplitude encoded recordings which very gradually raises
gain a preset amount when the average signal level drops below a preset threshold. During
HDCD 16-bit decoded playback the compression curve is expanded back to linear gain by
the HDCD decoder using a precisely mapped inverse of the compression curve controlled by
a hidden code, producing a dynamic range and resolution floor beyond 16-bit. During
undecoded playback low level information normally lost by standard 16-bit players is preserved,
providing more accurate timbral and spatial reproduction.
There are two modes of Low Level Extension, “Normal” and “Special”. Normal mode begins
to affect the input signal 45 dB below peak level, gradually raising the gain 4 dB as the
level drops over an 18 dB range. Special mode begins to affect the input signal 39 dB
below peak level, and gradually raises the gain 7.5 dB over a 26 dB range.Use of Low Level Extension is optional in the HDCD 16-bit encoding process.
==========Since you apparently have a problem with reading compehension, I will repeat this last point:
USE OF LOW LEVEL EXTENSION IS OPTIONAL IN THE HDCD 16-BIT ENCODING PROCESS.
You seem to have a real aversion to reading the patent, but prefer to quote from a specific implementation.If you did read the patent, you may realise what the patent refers to as "low level gain manipulation" refers a broad number of things, including low level extension. It is not a synonym for low level extension.
Anyway, the proof is in the empirical results, which you have ignored. Again, based on my experience, the WMP HDCD decoder seems to manipulate all HDCD content, regardless of whether specific features are engaged or disengaged.
If you can find a specific HDCD disc that passes through the WMP HDCD decoder unscathed, then perhaps you may have a point. But I'm willing to bet you can't.
Christine wrote, "HDCD is defined by the patent, not the implementation."Yep, I'm pretty sure that's the way it works in real life.
There they are in the mastering studio, getting ready to create the master for their new album. The band is there, the producer is there, and the mastering engineer there, all working hard to make the best product they can. The studio is equipped with a Pacific Microsonics Model Two A/D converter (which along with the Model One is the only way there is to make an HDCD disc).
They are debating which settings to use, and then someone shouts, "STOP! We have to check with the patent!"
So they go on-line and download the patent. They discuss it for a while and then agree, that even though there are menu options for turning "low level extension" on and off, and even though these options are explained in the manual, that they MUST NOT turn off "low level extension". After all, that's what it says in the patent.
Tell me something, Charles, have you ever been in a recording studio? Have you been involved in mastering any recordings? Do you actually have any idea at all what constitutes "real life" in a studio?
One of my best friends (whom I have known for over 30 years) is a well-established recording engineer. His credits include Alanis Morissette, Ringo Starr, and Dada. I have sat in on several recording sessions with him.I have also sat in on mastering sessions with George Marino (Sterling Sound), David Glasser (Airshow Mastering), and Gus Skinas (Super Audio Center). If you haven't heard of these guys, do a Google search and see what you find out.
So yes, I do have an idea of what constitutes "real life" in a studio. And I must say that not once in all my times in a studio has someone been concerned about looking up a patent, even though in some cases these people were actually using an HDCD encoder. Funny, that.
This is "Black Crow", from Joni Mitchell's "Shadows and Light" (Disc 2, Track 1), ripped via EAC in secure mode.This is what a non-HDCD player would play. As you can see, the waveform is peak-limited (all the peaks are suspiciously the same height) - this is due to Peak Extend in action.
This is how WMP has decoded the track. Note the attenuation of the entire waveform by -6dB, followed by expansion of peaks, which are now clearly of different heights (louder than -6dB).
This post is made possible by the generous support of people like you and our sponsors: