Upsampling?

DTH · « **on:** 18 Apr 2003, 11:15 pm »

Can someone give me or point me to an idiot's guide to digital/analog conversion and the benefits of upsampling? I do have some knowledge here but I am curious to hear other folk's experience with this. I am in the process of researching D/A converters for use with my nOrh CD-1 and/or a PC.

TheChairGuy · « **Reply #1 on:** 19 Apr 2003, 12:17 am »

Like everything else audio, it's theoretical or real...depending on how you trust your 'ears' and know your level of gullibility.

I use an MSB Gold Link II with Power Base and 192 outputs (MSB does this for their DAC's) from Sony DVPS7700 transport. It sounds good, I'm happy with purchase, I have nothing to compare it to except the crap that passed for sound from the Sony without the upsampling DAC added.

Caveat Emptor.

witchdoctor · « **Reply #2 on:** 19 Apr 2003, 03:05 am »

Get a GW Labs DSP upsampler from audio advisor for a 30 day audition.
I like mine but get good cables.

Hantra · « **Reply #3 on:** 19 Apr 2003, 04:31 am »

Quote

I do have some knowledge here but I am curious to hear other folk's experience with this.

Upsampling allegedly allows you to theoretically obtain a higher bit rate than CD was designed for. Of course this is all theory, and to do it right requires amazing amounts of money. (See DCS DAC's)

The problem I have with upsampling is that the digital filtering is nasty. It adds so much garbage to the signal. Things all of a sudden appear in the music that aren't really there. In the past several years, we have learned to live with digital noise, and some have even learned to like it!?

Some folks will hear digital noise and say that it is merely extra "air", or "sparkle". When you live filterless, and non-oversampling for a bit, you begin to hear what digital noise really is, and how great life is without it. You begin to realize which sounds more like real music. (that is IF you know what real music sounds like)

I would strongly encourage anyone looking for a DAC to try any of the non-oversampling/filterless DAC's out there. I have listened to all I could get my hands on. From the 47 Labs Shigaraki, Audio Note's 3.1x balanced, Zanden, and the Tube DAC from Scott Nixon. There are subtle differences in these DAC's. I would say my favorite, and the best implementation of non-oversampling that I have heard is the AN 3.1x balanced. I will also say that it bettered the Nixon DAC only by an extremely thin margin. I'd say that the difference was like swapping interconnects. A bit of depth, and bass control was gained with the AN. But it was not what I would quantify as a $6000+ difference.

Try both! Try the modded out DI/O's and the Nixon DAC. That way, you're only out a few hundred, and you can see which you like better.

If you're not convinced yet, then ask me about JITTER. hehe

B

nathanm · « **Reply #4 on:** 21 Apr 2003, 03:32 am »

I figure it must be the same as upsampling a digital image. You are undoubtedly getting a bigger image, but there is not more detail. But the illusion is there that there is. It's kind of a grey area.

However, I can tell you easily if a photo has been upsampled, but I still can't tell the difference between upsampled and non-upsampled CD audio. Personally, I think at worst it could be called hype and at best a marginal change. If you want more bits record with more bits! Simple. Personally I think upsampling is somewhat of a marketing gimmick. It probably doesn't cost a fortune to put it in there and it's cheaper than using a higher resolution original. Hmmm. Sometimes I leave mine on because it makes me feel better.

Hantra · « **Reply #5 on:** 21 Apr 2003, 12:33 pm »

Nathan:

That's a very good point. The more you upsample a digital image, the more chance you have for visible errors in the image. Same with CD.

Not only do we have more opportunity for creating something that is not there (digital noise), but we also exponentially increase the chance of jitter having a very audible impact on the sound.

The level of acceptable jitter on a filtered digital conversion system is a mere fraction of the tolerance on a non-oversampled system.

Rob Babcock · « **Reply #6 on:** 7 May 2003, 07:24 am »

I think that the extra dither must be responsible for some of the improvements that people attribute to upsampling. The whole concept is a little nebulous, and not easily understandable. Even some manufacturers of upsamplers concede that they don't really know why it sounds better or why they work.

Do they work? I dunno, haven't had much chance to hear any of the better ones. On the one hand you wouldn't think that many people could be wrong, but of course the answer to that argument is Bose!

Still, I've heard enough positive from people I trust to be interested. I'd love to be able to compare the new Electrocompaniet, Perpetual Technologies, Musical Fidelity A3*24 & Bel Canto DAC2 to see how they stack up.

BTW, anyone ever use a GW Labs or Monarchy upsampler with the DI/O? I'm curious how that would work.

Sa-dono · « **Reply #7 on:** 7 May 2003, 07:10 pm »

Quote from: Rob Babcock

I think that the extra dither must be responsible for some of the improvements that people attribute to upsampling. The whole concept is a little nebulous, and not easily understandable. Even some manufacturers of upsamplers concede that they don't really know why it sounds better or why they work.

Do they work? I dunno, haven't had much chance to hear any of the better ones. On the one hand you wouldn't think that many people could be wrong, but of course the answer to that argument is Bose!

Still, I've heard enough positive from people I trust to be interested. I'd love to be able to compare the new Electrocompaniet, Perpetual Technologies, Musical Fidelity A3*24 & Bel Canto DAC2 to see how they stack up.

BTW, anyone ever use a GW Labs or Monarchy upsampler with the DI/O? I'm curious how that would work.

I guess chalk me up as one of the believers of upsampling. I haven't heard the Monarchy upsampler..but I do have the 18b..and they are supposed to be working on an upgrade where it will be have upsampling capability. It's supposed to be released in summer..and if it does happen, you can be sure I'll be getting my unit upgraded.

doug s. · « **Reply #8 on:** 7 May 2003, 07:49 pm »

rob,

i bought a gw-labs dsp from audio adwisor to use w/my modded di/o. no change at all w/the non-upsampled signal; extremely harsh, etched, & irritating in the upsampling mode. i am certainly glad a/a has a 30-day return policy!

i know of two other folks, (from other forums), that also had similar experiences.

w/o knowing any better, i'd hazard that if it upsampled to 88.2khz, it may have offered some improvements, as there wouldn't be any timing error correction that is needed to go from 44.1khz to 96khz...

doug s.

Carlman · « **Reply #9 on:** 7 May 2003, 08:11 pm »

Quote from: Hantra

Nathan:

That's a very good point. The more you upsample a digital image, the more chance you have for visible errors in the image. Same with CD.

Not only do we have more opportunity for creating something that is not there (digital noise), but we also exponentially increase the chance of jitter having a very audible impact on the sound.

The level of acceptable jitter on a filtered digital conversion system is a mere fraction of the tolerance on a non-oversampled system.

So, can you correct jitter with 'SteadyShot'?

I really appreciate Nathan's analogy with imaging. It's the best I've heard.... I guess it also helps that I've been doing photography for many years. I also sold camcorders and digital cameras at Circuit City. Coming up with good analogies to explain technical gizmos for the masses was one of the best parts of the job.

Back to subject... If upsampling creates/amplifies jitter, does filtering attempt to correct the jitter created by the upsampling?

I find CD's fascinating. I'm in awe of the technical processs of turning bits into music. I'm constantly amazed at how real something can sound from a CD. I played Pearl Jam's album 'No Code' back-to-back from Vinyl to CD and was astonished at how well both performed. The record still sounds a little more natural in the midrange. The CD player sounds almost as good in the midrange and is better at everything else. Maybe it's the upsampling?

(just kidding)...

Rob Babcock · « **Reply #10 on:** 7 May 2003, 09:59 pm »

I've heard that some units like to be upsampled at different rates. I've been curious about how the GW Labs would work with the DI/O.

For every person who's hated the GW I've heard on person who loves it. It may be a synergy thing or maybe just personal taste. I'd probably be more tempted to try the Monarch one, since it's a little cheaper.

Oxia · « **Reply #11 on:** 8 May 2003, 06:57 pm »

Quote from: nathanm

I figure it must be the same as upsampling a digital image. You are undoubtedly getting a bigger image, but there is not more detail. But the illusion is there that there is. It's kind of a grey area.

Nathan,

Your analogy of resizing a bitmap image isn't really the same thing as upsampling. Generally, when you talk about upsampling in digital audio, it's dealing with two concepts. First, there's the "upsampling" process, which inserts zero-valued samples between the original samples. Second, there's "interpolation", which is a process that generates the "in-between" samples based on the original samples. The end result is a mathematical approximation of the original samples, as though the source signal was sampled at a higher rate. So in fact, upsampling isn't trying to simply "blow up" the original into something larger. Rather, it is an attempt to fill in the gaps between the original samples to more closely replicate the shape of the original signal.

For a graphical analogy, take a look at figures #5 and #6 at this page:
http://www.stereophile.com/showarchives.cgi?540:6

Figure #5 is an undithered sinewave at 16-bits. It looks really ragged, unlike a "textbook" sinewave. Figure #6 is a sinewave sampled at 24-bits. Notice how much smoother it is, and how it more closely resembles a sinusoidal shape. Now the waveform in this graph was generated from actual 24-bit data, not the result of upsampling. However, it's useful as a graphical illustration of what upsampling and interpolation is trying to achieve.

doug s. · « **Reply #12 on:** 8 May 2003, 08:16 pm »

rob, ever hear anyone w/a di/o that likes the gw-labs upsampler? i haven't, but it doesn't mean they're not out there...

fwiw, it's my understanding that the gw-labs is the monarchy dip & monarchy upsampler combined in one box...

doug s.

Carlman · « **Reply #13 on:** 8 May 2003, 09:24 pm »

Quote from: Oxia

........
First, there's the "upsampling" process, which inserts zero-valued samples between the original samples. Second, there's "interpolation", which is a process that generates the "in-between" samples based on the original samples. The end result is a mathematical approximation of the original samples, as though the source signal was sampled at a higher rate. So in fact, upsampling isn't trying to simply "blow up" the original into something larger. Rather, it is an attempt to fill in the gaps between the original samples to more closely replicate the shape of the original signal.

A digital image is usually "upsampled" in that interpolation and rendering have occured while it was captured, to create the image. (In the analog to digital conversion) Interpolation is intended to render an image to be more 'true'.... the process is very similar to what you are describing for audio digital manipulation. Various devices (camcorders, dig. cameras, scanners) perform rendering, interpolation, etc. to different degrees using different methods, though. Some don't do it at all, some do it poorly, some use the same technology but one works better than the other... etc... just like CD players and DAC's.

When enlarging or otherwise manipulating an image, interpolation is occurring between bits to render smoother transitions to make a more believable image. This has to happen or there would be clear holes in the image as it was enlarged. Those holes get filled by the software which is written to interpret what is around them and fill the areas based on the existing image. Photoshop is a program that even allows you to control how much interpolation can occur.

So, there are 2 phases of upsampling in digital imaging as well... When the photo is first captured, by the CCD in a digital camera where it is processed and stored... and then again if any manipulation occurs... like fitting it to a certain size.

The analogy still works for me.

-Carl

Oxia · « **Reply #14 on:** 8 May 2003, 10:43 pm »

Hi Carlman,

Ok, I see where you and Nathan were going with that analogy now. I agree with you, and I think we are all basically on the same wavelength. I was just troubled a bit by this sentence from Nathan:

"You are undoubtedly getting a bigger image, but there is not more detail."

This made me think it was just a comparison between upsamping and "resizing", which would be wrong. My point simply was that the analogy breaks down if we ignore the impact of interpolation. In the case of both image processing and audio upsampling, the goal is to reduce (or eliminate) the aliasing artifacts caused by undersampling, and to create a final product that is a closer approximation to the original image (or signal). Therefore, in this way, interpolation is an attempt to recover detail that was lost due to undersampling, which is something that just rescaling could never do.

Rob Babcock · « **Reply #15 on:** 9 May 2003, 03:08 am »

You might be right, Doug, but I don't think so. As far as I know the Monarchy and GW Labs units are made by two different companies. But I wouldn't swear to that.

doug s. · « **Reply #16 on:** 9 May 2003, 11:19 am »

rob, when i read an article about the gw-labs upsampler a while back in s'phile, i tink sam tellig mentioned geoffrey wong as being the designer of both products, & that the gw-labs one was basically two monarchy's in one box... of course, *i* could be mistaken, too!

doug s.

Sa-dono · « **Reply #17 on:** 9 May 2003, 05:33 pm »

Quote from: doug s.

rob, when i read an article about the gw-labs upsampler a while back in s'phile, i tink sam tellig mentioned geoffrey wong as being the designer of both products, & that the gw-labs one was basically two monarchy's in one box... of course, *i* could be mistaken, too!

doug s.

I thought CC Poon did most of Monarchy's designs?

doug s. · « **Reply #18 on:** 11 May 2003, 05:25 pm »

rob & sa-dono,

ok, i did my homework...

cc poon *did* do the design on the monarchy products. but, i was correct that the gw-labs dsp *is* the upsampler & dip combined in one box. cc licensed godfrey wong (not geoffrey), to mfr the gw-labs dsp. sam tellig reviewed the gw-labs dsp in the feb '02 s'pile...

doug s.

Rob Babcock · « **Reply #19 on:** 11 May 2003, 06:39 pm »

Is CC Poon his real name or his "audio" name?!