DISCUSSION: Upsampling audio devices

[Flint]

A recent post by Yesfan got me thinking of some new technologies we are seeing in the marketplace which we have not been discussing. In this case the technology is digital audio up-sampling - digitally converting lower resolution audio stream into a higher resolution format. Many audiophiles claim it is superior, others don't bother, and some (like me) are very skeptical.

Here's a great article specifically discussing Dolby's version of up-sampling used to "improve" the audio of movie soundtracks which were mastered at 48khz.

http://www.highdefdigest.com/blog/dolby-96k-upsampling/

I like the article because it is clear and honest and explains pretty well how the technology works and the inherent artifact (pre-ringing) which simple mathematical up-sampling causes.

What I found amazing, and many of you are aware of this psychological phenomena, is that a well respected company like Dolby Labs would use mind tricks to fool the audio writers they were demonstrating their fancy technology to. What did Dolby do? Well... first they fully explained what their technology does and how it is different from other technologies. Then they gave the writers a list of things to listen for while they auditioned content processed with their technology. BOTH of those activities will fool any listener into honestly believing they hear what they were told they would hear. That is the flaw in how we ALL hear things - our minds do so much analysis and interpretation of the raw electrical pulses from our ears that a simple idea can completely rewrite the mental algorithms temporarily (sometimes permanently).

So, of course some people heard what Dolby told them to. Those who didn't are likely like me who finds all these claims of greatness very unbelievable and skepticism holds back hearing what might really be there.

So...

My experience with up-sampling, which ranges from up-sampling audio using true professional grade audio production applications to auditioning professional, commercial, audiophile and consumer grade devices which perform up-sampling in real time, and listening to high resolution recordings which were up-sampled by the producers prior to manufacturing (but the process was not clear).

Devices:
In general I have found that the vast majority of the content where I can control the addition or removal of the process to sound exactly the same, though occasionally added noise and sense of sibilance was present.

Commercial content:
With content which was up-sampled prior to publishing of a disc or making available online it was nearly impossible to compare to the original since I didn't have access to the original (or the up-sampled version was re-mixed or edited which prevented a side by side comparison with the original CD) I cannot make a call on the results.

Professional Audio Software:
Doing the conversion in a non-linear fashion is typically superior since the algorithms can "look-ahead" in the audio data and make decisions based on what is coming (something a device cannot do with a stream of data) is in theory superior. However, I have not heard any clear audible difference between up-converted audio and the original source content.

So, I am not convinced.


What are your thoughts?

 

[Yesfan70]

So............


From the time you read and replied to my post in the other thread to starting this one (under 45 minutes) you did all that research and comparisons in that short amount of time to come to your conclusions? That's how I'm interpreting your post. :think:

 

[Flint]

No, I remembered quite a bit of stuff I hadn't thought about, then did some research and found that blog, then thought I would be fun to talk about since haven't had many new audio topics in quite a long time.

Since posting above I've spoken to some of my longtime friends who spend every day working in studios and mastering suites to start learning more from real world users.

 

[PaulyT]

Ok let me be blunt here (and Yesfan, this is not aimed at you, but at "audiophiles") : I don't understand why up-sampling would even THEORETICALLY be an improvement. You can't possibly use a computer post-hoc to add more information to a digital audio track that's "true" to the original sound waves in a recording studio. You can add OTHER information, maybe add some sort of digital effect that is subjectively pleasing. But you're not somehow re-capturing more of the original live sound, it's just not even theoretically possible.

I guess maybe some up-sampled stuff sounds better on certain gear if it's not able to natively handle a lower sampling rate or something very well, due to some software/hardware idiosyncracies. I dunno. Or again, it just adds something artificial that the listener prefers. It's the old (Soundhound) "fidelity" argument again. Up-sampling by itself may produce something you (a general listener) enjoy, but it's not making the digital replication of the original audio waves any more accurate w.r.t. the original.

This is not to say that a high sampling rate recording isn't superior, if the recording pathway is entirely run at a higher sampling rate, with appropriately capable gear (e.g. AIX studios). That's a different question. Whether the information >22kHz (the limit of a regular CD) is what makes for more realism... that I'm not convinced of, but not entirely convinced against, either.

 

[Lone Stranger]

I really don't know jack about this kind of stuff so I shouldn't say too much about this but I wonder if there is (theoretically) a difference in the quality in the D/A converters between an ordinary CD player and an upsampling device. Some 20 years ago I compared a Sony CD player with a 16 bit 4X oversampling D/A converter to a JVC CD player with a 1 bit 8X oversampling D/A converter. In my mind they both should have sounded the same using the same CD, amplifier and speakers since "bits is bits" but to my ears the JVC sounded very slightly clearer.
I used to own a Technics electronic keyboard that had a horrible D/A converter which was of poor quality and added a lot of its own noise in addition to the sounds.

 

[Botch]

I don't have a good feel for this vis-a-vis audio, but I've been doing some thinking about upsampling of 1080p to 4K, and I came up with this conundrum:
Let's say there's a screen shot where the left half is black, and the right half is white. The line runs straight down the middle, actually bisecting a "pixel" in the middle. On a 1080p TV, that pixel can't decide whether to display white, or black.
On a 4K TV, assuming the "upsampling/upconverting" algorithms are done well, the software realizes the situation, and that This Here Pixel is right in the middle of a divide between black and white. It stands to reason that it could assign the two left "quarter-pixels" to black, and the two right ones to white, and actually making the image sharper/clearer.
First, I don't know if 4K software can even do this, and secondly, I don't know if there's a similarity with "upconverting" audio. But, its fun to think about, if I'm bored, or in a staff meeting.

 

[Haywood]

Given that it has been proven that nobody can reliably detect the difference between 320k mp3 files and lossless files, I think it pretty much goes without saying that upsampling not going to yield any audible improvement.

 

[yromj]

Pauly stole my homework and turned it as his...I swear.

Botch, as for the upsampling video comparison goes, we had many discussions about this on the old forum. It boils down to the fact that you still can't add information to the image, regardless if the image is audible or visible. Take an image that's lower res than your monitors resolution and zoom in and see what happens.

Haywood's point is also extremely valid and can be extrapolated to point out that most people are listening to 192kbps .mp3 files on ear buds that came w/ their iPhones; so, what IS the point of this?

Oh yeah, marketing to sell something w/ a higher number. 24 has got to better than 16, right?

John

 

[Yesfan70]

Interesting posts I've read here. I myself don't know what to make of upconverting audio. I would think that once it becomes 16/44.1, then technically it's going to be as good as it gets as you all say. I do like the point Botch brings up. I agree you can't add any more info, but maybe the analogy he posted was what I experienced with my old setup. Some CDs did seem to sound "better". Maybe the typical audiophile would go on with the whole "like a veil lifted" speeches, but with my experiences, the differences were very subtle with the best recordings.

 

[Flint]

Back when HD audio was first being developed as a consumer application, the genius behind Meridian proved in a series of articles in Stereo Review (or maybe it was Audio) that the absolute highest resolution which could be audibly recognized and realized with the limitations of current electronic components is a resolution of 18bit, 88.1kHz. That was what he proposed as the highest HD audio format the industry should use because better wasn't audible and high resolution data rates were costly to process, store, and transmit.

 

[Haywood]

Flint mentioned the extremely cool Intel NUC computer line. I am seriously considering a Haswell i5 NUC with 8GB DDR3 and a 128GB mSATA card for my next media server (my content is stored on a NAS). They are not only incredibly small, but almost silent and have amazingly low power consumption.

 

[Deleted member 133]

Pauly stole my homework and turned it as his...I swear.

Botch, as for the upsampling video comparison goes, we had many discussions about this on the old forum. It boils down to the fact that you still can't add information to the image, regardless if the image is audible or visible. Take an image that's lower res than your monitors resolution and zoom in and see what happens.

Haywood's point is also extremely valid and can be extrapolated to point out that most people are listening to 192kbps .mp3 files on ear buds that came w/ their iPhones; so, what IS the point of this?

Oh yeah, marketing to sell something w/ a higher number. 24 has got to better than 16, right?

John

John,

I would agree in the case of straight upconversion, however expand the topic only just a bit to include frame interpolation and unfortunately we (not me - just the device/display) can actually add information that was not part of the original bitstream.

I don't think there's (yet) an audio equivalent, but I trust you get my gist. Like that violinist did not properly transition from one note to the next so we'll just add something to bridge the gap...)

Jeff

 

[malsackj]

I will go up on the limb and input the subject of processing in ambiophonics, to create a sound field from another room in the room you are listening from. Using computer data to generate surround fields for a room or space I want to perceive.

I can have the sound from a hall like the Kennedy Center, The Opera House,

or the echoes from being in the Grand Canyon. The computer data would be the interpretation of the room and calculated responses would be sent to the speakers based on the original signal.

So ambiosonics is trying to impart the room into the original signal by adding the calculated responses to additional speakers. Creating the surround output from the stereo source, based on the room chosen to be created in your listening room.

Up sampling can add to a 16 bit into 24 bit into 32 bit and so on. This move from 16 to 24 bit may not lower the noise floor or moving from 44 k samples to 96 k samples will not improve the signal because the add information will be an interpretation that adds one or two points between the two points on the sound wave form. They cannot predict or add a spike or other missing harmonic from a short term peak because it has no way to predict what the original sound is other than the beginning point and the end point that follows. They are in my opinion just taking to points on a graph and inserting one or more points between the two points to help make the wave more detailed and this does not improve the sound beyond the 16 bit CD sound.

With 5.1 or 7.1 in movies and music we don't need to have the ambiophonics to transport us to the room because the recorded material should be applied before we listen. With music do you want to choose what room you listen from ?

Kennedy Center,
Constitution Hall,
An Opera House,

MCI center no acoustics or bad acoustics.

The object of the listening to the recording is to enjoy the music, I will try to give you a perspective with my recording of the room, the experience, the performance. When people start having control of the room to enjoy the production, would you place the Orchestra in a small bar? How about the Beatles in an Opera House?

Does anyone take the stereo signal and tell the receiver to do a music with an church ambiance? Does anyone use the receivers other sound modes besides dolby digital for movies?


www.ambiophonics.org/

 

[Flint]

There are many problems with attempting to create acoustic environments in your home (or any location).

The processing necessary to recreate the acoustic of any other ambient location has been around since the 1980s. Yamaha was one of the most advanced leaders in this area. They visited all of the most recognized and popular concert halls and clubs in the world and created reverb / reflection algorithms which identically recreate those locations using a standard receiver at home. It was perfect in every way in making whatever was being reproduced by the speakers appear to be reproduced by those same speaker setup in the halls where they had created algorithms.

What was wrong with that?

Well, the recordings being reproduced by the speakers already had lots of ambient information in them. In other words, live recordings, like classical and jazz performances, already had all the reverb and ambience from the halls they were recorded in on the recording. Adding more ambience to recreate the experience in a different concert hall had a multiplication effect on the ambience and it became very overbearing and hindered enjoyment of the recording.

To prove the concept, Denon, and a few other companies, created recordings of acoustic performances with absolutely no ambient information. They basically created anechoic chambers for the performers to sit in as they played the music. In some sessions, to remove all ambient noise, the room was so hot huge blocks of dry ice were brought in and placed next to the performers to keep them and their instruments cools. I have a couple of the Denon CDs from those sessions and they are eerie to listen to over headphones or in well treated rooms. If you add a Hall setting to the processor, they start to sound more normal.

So, I am not a fan of the idea of trying to recreate the ambience of a space when the recording already has that in the recording.

 

[Rope]

Upsample this---------------------> bscene-buttred:

Rope

 

[Botch]

Does anyone take the stereo signal and tell the receiver to do a music with an church ambiance?

My Yamaha receiver has quite a collection of DSP programs. I played around with them for a bit after my system was first set up, but pretty much left them alone after that, agree with the above arguments that the "ambience" built into the recording was put there by the mixer/producer for a reason.
There's one exception, though. I have a couple stereo recordings of monks chanting in dank reverberant churches, and the church's ambience is well captured in the recordings. However, on these I tried using some of the Yammie's 5.1 church programs, and the effect is wonderful. I don't like hearing a "direct" signal of, say, a solo saxophone or a high-hat coming at me from several directions; it's not real. But, a bunch of monks singing in unison blend and no single source can be identified; having something like that come out of speakers all around you does sound real, and it provides a deep emotional impact. Which is, after all, what music is supposed to do!

The last DSP unit I bought for my keyboard stack, back when I was playing live, was a Yamaha SPX-900. It had one Reverb program in which you could design the room it would simulate, and it would provide a stereo effect of that space! You programmed the room's length, width, and height; surfaces (floors could be stone, wood, or carpeted; walls plaster or with curtains), even if there was wainscoting on the walls! It was a lot of fun to play with, but my last bands' sound man (who actually had a degree in it) proved to me that, in a club, the less effects the better, and I did find that both I and my bandmates could hear me better by getting rid of the smear; no sense in simulating a "club reverb" if you're already in a brick-n-mortar club!

 

[Huey]

Don't have much to add about the upsampling, but what if we looked at it from another angle, correcting the music for each individual person. Let's say we had some magic earphones or something like that, that would make the incoming signal in tune with our hearing deficits or strengths.

 

[Flint]

Don't have much to add about the upsampling, but what if we looked at it from another angle, correcting the music for each individual person. Let's say we had some magic earphones or something like that, that would make the incoming signal in tune with our hearing deficits or strengths.

There are two ways to look at that idea...

1) What any person hears when they listen to live music, say a jazz guitarist in the subway, will be filtered by the deficits in their hearing. So, when they get home and want to listen to a recording of a jazz guitarist the most ideal reproduction system will recreate what they heard in the subway. It will be lifelike and real, as if they were there. Why alter that, it is real!

2) Some people lose their hearing over time and miss the old days when things sounded different. Can we just apply a filter to compensate for what they've lost and reproduce the sound such that what gets to their brain is similar to how it used to sound before hearing damage occurred.


I tend to agree more with the first way of looking at it. One reason is that applying boosts to frequencies lost to a listener over time or from abuse & injury can actually make the damage worse. The reason most people lose hearing is from overexposure to those frequencies, and boosting those frequencies to compensate will accelerate the damage.

However, anything people do to enjoy music is good. I believe music is the best way to boost your soul and it is good for all aspects on one's emotional health. So, do whatever it takes to enjoy it.

 

[Zing]

However, anything people do to enjoy music is good. I believe music is the best way to boost your soul and it is good for all aspects on one's emotional health. So, do whatever it takes to enjoy it.

Amen!

 

[MatthewB]

I thought that when something was mastered say in this case at 48kHz how can you extrapolate more from what's not there without ruining the integrity of the original. That's like turning a turd into a diamond and selling it as a diamond.