There’s been some measure of controversy recently about upsampling with respect to hardware in general, as well as high resolution audio downloads and media, specifically.
In a recent article by the British high fi magazine, Hi Fi News and Record Reviews, editor Keith Howard expressed some reservations about the quality of what people are getting through high rez downloads and media services. His examples clearly show that some media we’re asked to pay a higher premium for aren’t as good as they claim to be and the message is: buyer beware.
The HiFi News article has been picked up by a few online sources, including iTrax.com a purveyor of downloadable high resolution audio. Here’s a link to a recent article written by Mark Waldrep. http://itrax.com/Pages/ArticleDetails.php?aID=32
It’s been suggested that upsampling (oversampling) is making something from nothing. Upsampling can be applied to many digital formats, from images to audio. Upsampling uses complex math algorithms to increase the sample rate and bit bit depth. For illustration, a standard CD has a sample rate (the number of “snapshots” of the moving audio signal) of 44.1kHz (44 thousand snapshots per second which defines the upper audio frequencies) and a bit depth of 16 (the numeric representation of the dynamic range). If you upsample this data, you can achieve much higher sample rates and bit depths; 192kHz and 32 bit as an example.
The “something from nothing” suggestion stems from the fact that this upsampled digital audio track has no more dynamic range and no more upper frequency content than its native state. So why upsample? The positive is three fold: because the sample rate is higher, the corresponding DAC filters used work at much higher frequencies removing possible sonic artifacts caused by these downstream filters. Higher bit depth allows designers to execute digital volume controls and dynamic expansion tricks without damage to the original dynamic range. Sample rate converters can digitally manipulate the incoming stream to eliminate jitter.
The negative is a hotly debated subject; the argument centering around the sample rate converters themselves and what possible damage is caused with all the manipulation, multiple clocks and so forth.
The controversy brought up in these articles are two fold: what actually is being claimed and sold by high resolution audio purveyors vs. what is being received, and questioning the benefits of creating high resolution digital copies of low resolution analog masters.
The first issue is clear and generally agreed upon by all concerned: upsampling an existing digital audio file and issuing it as a high resolution new release (at a premium price) is wrong and misleading.
The second issue revolves around the practice of high resolution digital mastering and reissuing of analog tapes; a medium whose dynamic range and frequency response are less than a standard CD.
Called out in both articles is HD Tracks, amongst others, for delivering “ham fisted upsampled” media that doesn’t perform as advertised and yet customers are asked to pony up a lot more money for this media.
This magazine’s experience with high resolution downloadable media is mixed, but generally very good. The idea (expressed below) that a high resolution digital copy of an analog tape is an “oxymoron” because analog tapes themselves are both amplitude and bandwidth limited to less than that of a standard CD is wrong in this newspaper’s opinion.
We contacted HD Tracks.com co-owner, David Chesky to get his side of the story and clear up some of the issues presented in this article (iTrax is a competitor of HD Tracks).
David responds. “Why do we sell hi res files at 192/24 when there is no musical energy up at 96khz? Lets look at it like this. In the current state of digital technology we do not have A/D chipsets that work at 44.1/16 that do not have pre and post ringing. This ringing destroys the music in the audio band. Audiophile labels do not record at 192/24 to try to capture what many think is air up over 50K. They do this as to keep the nasty anti-aliasing digital A/D filters as far away from the audio band as they can. This insures less corruption down where the music is. The sound becomes much more analog and relaxed. Think of pre and post ringing like this. You are riding down a street in a Ferrari at 180 MPH over a cobblestone road and someone gives you a camera and ask you to take some photos out of the windows. All these little cobblestones will cause bumps and the camera will not be steady and you will loose the focus. This is what happens in current A/D conversion as well at 44.1/16. Doubling or quadrupling that rate helps smooth the analogous ride.
How do we convert files from labels? All PCM recordings come to HDtracks at 88/24, 96/24, 176/24, or 192/24. All albums that come to use either DSD or SACD and are converted to 176/24 or 88/24. This is to keep the math even since DSD is based on multiples of Redbook CD ( 44.1/16).Since late last year, we began using independent mastering labs to check the files to make sure they are high res prior to their release. We began this policy because many of the non-audiophiles labels get confused about what are and what are not really hi res files in their delivery. We now check for brick wall limiting, and other signs of up sampling, and routinely reject products that are not true hi res. When we find that back catalog of albums are mislabeled, we pull those albums and ask the labels to replace them, or we have relabeled them to reflect what they actually are. Our customers are always right, and first and foremost we work to make them satisfied with their purchase.
The advantages of Hi res files on a disk server is that with these HD files you can find devices such as computers, or RAM drive players that can store the entire file in memory and play it back. This gives you a better reading of the file without the jitter and error correction caused by many disc players.
In the end, good music in hi res, delivered to your home, played back with a good device in memory can elevated the entire digital audio music experience to levels that were unattainable years ago. I believe it is not so much the digital aspect that people disliked years ago, but the execution of playback and its results. As an industry we are all striving for the same thing, that is, the pursuit of perfect playback in our homes.”
As Chesky mentions in his reply, most master tapes that are digital are recorded at higher sample and bit rates than their original 44.1/16 release and it’s an easy leap to understand what HD Tracks and other labels are doing when they remaster and then release a higher resolution master. That’s the easy part.
But when digital master are made from analog tape and that analog tape has technically lower resolution than even a standard 44.1/16 CD, does it do any good to master a higher resolution version if that resolution isn’t there in the first place? This, after all, is the central question posed by both authors of the original news.
The answer is yes and no. Yes if done properly, no if not. Let’s examine the no first.
As pointed out in the iTrax article, upsampling an digital master has little to no benefit and may actually be worse. It’s analogous to putting lipstick on a pig. It simply isn’t a good idea.
“My contention has always been that a standard definition recording from the past placed in a container that exceeds its fidelity standards remains a standard definition recording. We might be getting the best possible rendition of that older track but it is not the same thing as having a new recording done with live musicians at 96 kHz/24-bits. And it shouldn’t be marketed as such.” And herein lies the heart of the matter and where the misinformation is being generated.
No one is going to think that a high resolution track from the late John Coltrane is a new recording; but marketing it as a higher resolution rendition is, in our opinion, entirely accurate. One need audition any of the well mastered SACD and HD Tracks reissues to prove that to themselves.
Digital high resolution remasters of analog tapes are, if done properly, brilliant and should be supported. Great mastering engineers like Gus Skinas at the Super Audio Center in Boulder are quietly but steadily saving and preserving some of the best catalogs of music ever recorded. Much of this work is being reissued by Chad at Acoustic Sounds and the work is simply stunning.
Let’s not confuse upsampling and definitions on labels and misleading market speak as being the gospel. In fact, we would agree with the closing statement of the online article in spirit but would not want it to be construed as an implied condemnation or finger pointing exercise towards this fledgling industry of high resolution audio, an industry that is working hard at preserving and improving the state of digital audio for the better.
Forward to a friend and help us engage more readersYou must be logged in to post a comment.
PSTracks Staff
There’s more to this story on the PS forums http://www.psaudio.com/ps/forum/viewthread/2180/
“The article while posted by HD Tracks competitor was not written by iTrax. I too have downloaded some great high res stuff from HDT but have also downloaded some bombs. One of those companies that the article condemns is HD Tape transfers, which I find to be a great value. The company is using wide latitude recording of old master tapes from the 50s, 60s that is frankly fantastic. Maybe the same could have been achieved with lower sampling rates but they are not charging a premium for the higher sampled albums as others do. Two of the recordings I have downloaded are truly great.
iTrax is different in my opinion. Like other NEW recordings they are fortunate to be able to do their primary recordings at high sample rates and not converting old to new. So it is easy for them to stand behind the criticism since they are in a different market. “
Fred Bosick
HDTracks is lying. I get their emails offering new releases. I don’t mind and I enjoy reading about hi def releases of old music I like. But I was brought up short reading about a 96k/24bit “release” of Rush – Moving Pictures. It’s their first digitally recorded album – a great album! But it was never recorded at that sampling rate/resolution. All you’re getting is , maybe, a linear interpolation between original samples and bits 17-24 being solid zeros. I wrote their tech support about this and I got a generic reply. I don’t think the support person even understood my complaint. Most of our machines already oversample so there is absolutely no advantage in buying this “hi def” music product.
Staff
We appreciate your response but not sure we want to accuse anyone of lying. Let’s please keep it civil. David did say (in the article) that they have labeled some releases as high resolution when, in fact, they were upsampled as you imply. David also said they are getting them corrected as quickly as possible. Maybe they missed this one? I’ll ask David to respond. I can tell you the folks at HD Tracks are honorable people trying their best to make a good contribution to the high-end.
Michael
Of course, Chesky are not telling the whole story (to say the least)
http://www.whatsbestforum.com/showthread.php?1037-Upsampled-SACD-s
Bruce Brown is Chesky subcontractor that actually performs a transfer…
Werner
“The “something from nothing” suggestion stems from the fact that this upsampled digital audio track has no more dynamic range and no more upper frequency content than its native state. So why upsample? The positive is three fold: because the sample rate is higher, the corresponding DAC filters used work at much higher frequencies removing possible sonic artifacts caused by these downstream filters. ”
This is wrong. Yes, after up/oversampling the source file/data the filters in the DAC operate at a higher frequency. But during the act of upsampling another filter has to be used at the original data’s Nyquist rate, i.e. 22kHz for CD. And this filter may, or may not, be somewhat superior to the one in the DAC. But there are no certainties. Given sufficient resources one could implement any such filter as software or as hardware.
Paul McGowan
Not sure how that makes what was written “wrong”. In fact, going from 16 bits to 24 bits by upsampling gives you no more dynamic range than what was originally there – just as going from 44.1kHz to 192kHz does not extend the frequency response of what was placed on the disc in the first place. What it does do, as you suggest, is give you a different higher frequency filter set which is good – but at what cost sonically?
On the other hand, you can make a case that interpolating the data (filling in the missing data by “guessing” or predicting what would have been there can be effective in better sound.
I am just not personally convinced it’s a good way to go.
Vivek Kuvelker
I posted about my problem with HDTracks on my blog (http://bit.ly/HDTricks) & the full run-down is on the CyrusUnofficial forum at http://bit.ly/HDCrocks.
To summarise all of this, I bought the HDTracks’ Albert King & Stevie Ray Vaughan’s ‘In Session’ 2009 Stax Records re-master in April 2010. It’s described as a 24bit/96KHz HD album in FLAC format. As a result of problems other people were having with the company, I decided to check the album’s frequency plots using Audacity on the Mac. The results indicate that the downloaded music contains to musical information above 22KHz.
I asked HDTracks about this, but they didn’t seem at all bothered. After some external investigation & more moaning by me, they finally removed the album from sale on June 15th, but gave no explanation as to why they did this.
I’ve continued to ask them the following questions, but they continue to remain as quiet as a politician with a dodgy expenses claim:
“The evidence I have already provided indicates quite clearly that the Albert King & Stevie Ray Vaughan’s ‘In Session’ 24bit/96KHz album, which I purchased on 9th April 2010 from HDTracks, has either been upsampled from Red book (16bit/44.1KHz), or has passed through some form of HF cut-off from a higher-resolution source, perhaps unknowingly.
1 What specifically is the reason for the 22KHz high-frequency cut-off on each & every one of the tracks on the album?
2 What source material was used to produce the album download? Were analogue master tapes used for session re-mastering, or were the original recordings stored on 16bit/48KHz DAT?
3 Is the current ‘In Session’ download digitally identical to that which I purchased on 9th April 2010? ie Has any ‘new’ version of the album since been made available to your customers?
4 If no change in the download has taken place, do you intend removing the current version for sale to customers & offering a ‘true’ 24/96 version of the original recordings?
5 Do you have any intention of informing all previous purchasers of this album, of this potential upsampling issue?
6 If a ‘true’ 24/96 version of ‘In Session’ is currently available on your website, how do you intend offering those previous customers the opportunity for a free download of the ‘true’ HD album, along with some form of compensation?
7 If no ‘true’ 24/96 version of ‘In Session’ is currently available on your website, what do you intend doing for those previous customers who may feel that they have been deceived by HDTracks?
I still await HDTracks’ answers to these questions, but please bear in mind that I’m not prepared to wait for much longer.
Also, please do not ask again for methods by which you can provide me with a refund, as I will simply ignore this until I am satisfied with your company’s response to my questions.
Thanks again, Lisa.
Best wishes.
Mr Viv”
I think that the Canadian TV company that recorded the original recording weren’t keen on releasing the original analogue masters for re-mastering, or it was originally recorded on 16bit/48KHz DAT. I’m sure this was a problem between 1999 & 2003 with a lot of SACDs being produced from Red Book masters, instead of DSDs. Either way, I just don’t think the music was ever inherently recorded as a true ‘HD’ recording.
I wouldn’t mind if HDTracks were transparent-enough to provide enough information about their downloads to allow buyers to make informed decisions prior to making a purchase. This just isn’t happening, unfortunately.
They’ve quietly removed other albums in the past, without mentioning anything about doing so to existing buyers. That, I think, is ridiculous.
In their last response to me, David Chesky sent me this link:
https://www.hdtracks.com/files/DSD_to_LPCM.pdf
It’s a nice technical description of why they use ‘brick wall’ filtering in their recordings, but it still fails to answer any of my questions about why they pulled the album I bought.
Looking forward to hearing what you all think on PSTracks.
Viv