Upsampling?

[Flumpus]

Hey guys... I'm thinking about doing the upgrade thing for my MSB Link Dac 3... So I'm wondering, what exactly does the upsampling do? Is the difference big? I'd be really interested to see what some of yall who have heard it think... Thanks!

 

[DarkAngel]

To put it very simply when you upsample you add dither to original
CD signal to increase signal to 24/96 or more, this more easily allows noise/distortion to be removed from DAC process without doing harm to original signal.

There are many other factors to deal with to make a good DAC, upsampling is only one of these. BTW don't have a MSB LD3 so
can't comment on how effective their upsampling design is.

I do have Bel Canto Dac 1.1 which upsamples to 24/96, I think there is little doubt that the best sounding DACs do employ
upsampling in their design.

 

[Dusty Chalk]

Well, it's not really dither...at least, not in the way that I think of dither. Dither is noise deliberately introduced to mask artifacts of sampling. Upsampling is more like interpolation.

Upsampling is similar to oversampling. Oversampling is a way of artificially increasing the sampling rate so that they can put the filters higher and therefore have a smoother sounding DAC. But they don't do anything with the bit depth, so it's still 16 bit. Upsampling does both. It interpolates the existing signal to 24/96.

It is very smooth sounding, compared to un-upsampled CD's on the same DAC (I have a switch that turns it on or off because I also have the HDCD option, and the two are mutually exclusive, and I prefer to listen to HDCD encoded music through the HDCD decoder).

That said, getting the upsampling option was the second biggest leap in performance I got at the time, after getting the DAC in the first place. I would definitely highly recommend it. Have you ever run a coarse/grainy digital image through some smoothing software? It was like that. The HDCD option wasn't as big a leap in performance.

Don't get the wrong impression, it will not introduce information that is not there. It's nothing like getting a true 24/96 recording. I have several. Two in particular -- Alan Parson's Project, I, Robot and Schubert's Piano Trio No. 2 in E flat -- I have both, the CD and the 24/96 "DAD". (In the case of I, Robot, I compared against Alan Parson's own remastered collection, which has 4/5 of side one). There's just no comparison, 24/96 recording wins hands down. (Note: Classic Records has its own remastering process, and I have no idea if it is superior to Alan Parsons, but it is possible that that is the reason for the noticeable improvement.) In the case of the Schubert Trio (Arts is the label out of Germany, Trio Italiano are the players), I have the same exact recording and master (I didn't know they had it on DAD when I got the CD). The difference wasn't as noticeable, as it was glorious either way. Of course, I may have been biased, because I do get a bit carried away with this piece of music, it is far and away my favorite piece of chamber music (and, no, DarkAngel, not because of "The Hunger", I actually fell in love with it for the Fourth Movement), and this version is one of my favorites.

 

[Flumpus]

Interesting DustyChalk...

So would you suggest the HDCD and upsampling upgrades? I know you can't upsample and do HDCD decoding at the same time, so how does that work with only one option switch? If I turn the option switch off (I'm assuming that turns of upsampling), then play an HDCD, will the option light come on because it's using the HDCD decoder? Thanks for the info DustyChalk...

 

[thomas]

Upsampling increases the sampling rate of your audio by interpololation, basically generating new samples inbetween existing ones... Normal CD audio plays 44.1 thousand samples per second; when you upsample, inbetween each of those samples you generate a new sample, making the overal signal smoother when it leaves the DAC, before filtering.

Idealy, an upsampler should make CD's have the same resolution and detail as a DVD-A, DAD, or SACD, but of course in practice, it isn't perfect and adds a bit of distortion/jitter to the signal. Nevertheless, it still should be superior to regular redbook CD.

(oversampling also increases the sampling rate, but does so by adding "0"'s inbetween existing samples, instead of interpolating. It acutally decreases the quality of the signal, though an oversampled or upsampled signal is easier to filter. There is a BIG difference in sound between upsampling and oversampling!)

Quote:

I do have Bel Canto Dac 1.1 which upsamples to 24/96, I think there is little doubt that the best sounding DACs do employ

yep, many great DACs use this principal to increase the resolution of the audio without changing to SACD/DVD-A. But there is also another school of thought that believes all processing will add distortion to audio, and get rid of it all. MANY high end DACs today use this principal- no upsampling, no oversampling, no digital filter. THe audionote DAC, which sells for over $50 000, use this principal. Which is the ideal method? who knows...


HDCD is a totally different process, it starts off with a high quality recording (up to 192khz x 24 bit, i think) then uses an encoding process to store much of that info into CD audio (44.1x16). The HDCD decoder chip simply decodes the signal back into its original high-resolution form.

HDCD decoding and upsampling are definately seperate processes, and i'm pretty sure that you can't use both at the same time- a decoded HDCD track is already high resolution, can't increase it anymore with upsampling; an upsampled HDCD track loses all the hdcd encoding info, can't use the decoder

I'm not sure how the upgrade works, but if it goes before the DAC, then the HDCD decoding light should only come on if upsampling is turned off...then again, upsampling should be better than HDCD, and you might as well use it for everything, and forget about HDCD...

 

[Dusty Chalk]

Quote:

Originally posted by Flumpus So would you suggest the HDCD and upsampling upgrades?

I voted with my wallet, and my wallet said yes.

No, seriously. Get the upsampling option first. There's more of a return on your investment (it has an improving effect on more of your CD's). If/when you have US$160 to burn, then get the HDCD option. I don't know what it is, but I revisited my HDCD collection, and I noticed some sort of subconscious dislike of all of them. After I re-listened to them decoded, I liked them. This shouldn't be. I have no explanation. It could just be me. I am not saying that you will hear the same thing if/when you do a similar comparison (HDCD CD's decoded vs. undecoded). Quote:

I know you can't upsample and do HDCD decoding at the same time, so how does that work with only one option switch? If I turn the option switch off (I'm assuming that turns off upsampling), then play an HDCD, will the option light come on because it's using the HDCD decoder?

Yes, exactly.

Also, I should mention that this is not necessarily the case, only the case in the MSB Tech Link DAC III...and most every other DAC in the known universe. The thing is, one can decode entirely in the digital domain, but the only device that I know of that does this is Pacific Microsonics HDCD Processors I & II, and that costs thousands of dollars (I priced the old one once, the I, and I think it was like US$12K -- my wallet said "no" on that one). But it's just cheaper and easier to put the decoding right in the DAC chip, and that's where most manufacturers do this, including MSB Tech. I think this is an imposition by Pacific Microsonics, or maybe cost/technology/return/value/etc. in general.

I think Camelot is working on an upgrade that does the HDCD decoding before the upsampling. I think he said, "as long as I can do it all in one chip, and Pacific Microsonics just came out with such a chip", meaning he hadn't actually implemented it yet.

Also, there's an interesting little .pdf file on the HDCD site somewhere (somewhere under technical) about how they do it. It's basically similar to Dolby at a subtler level. It's a form of level dependent compression -- at low levels, it amplifies the signal, at high levels it attenuates. The HDCD decoder is constantly undoing this (in a HDCD encoded signal) in one of two ways, based on the LSB of the signal (the LSB spells out a pattern -- much like morse code -- in one of two ways, telling the decoder which filter to use).

So it doesn't make sense to me that I wouldn't like this. It sounds subtle and fairly transparent. Maybe I don't like the LSB, I don't know. It should be inaudible. I would not recommend you get it based on my experience. I would try to audition it for yourself.

Finally, the HDCD-capable DAC that comes with the HDCD upgrade (it basically is the HDCD upgrade) is a slightly better DAC, so that would be another reason to get it. (But we're getting really subtle now, even more so that HDCD.)

BTW, thomas, a couple minor niggles:

Oversampling doesn't interpolate zero's -- it interpolates the same value over and over again. Let's say you had 4x oversampling, and the incoming bitstream was like 1028, 1032, etc. then oversampling would deliver 1028, 1028, 1028, 1028, 1032, etc., whereas upsampling would interpolate different intermediate values for every value, after multiplying them by 2^8, of course. So something like: 1028*2^8, 1029*2^8, 1030*2^8, 1031*2^8, 1032*2^8 (actually, the intermediate values won't be exactly this, it depends on the surrounding signal, and what the upsampler interpolates those values to be; this was just an example). This is still wrong, though, and is therefore more incorrect than the 16/44.1 bitstream, as you correctly pointed out.

Also, I'm not sure it necessarily introduces distortion and jitter. It theoretically could remove jitter, depending on the superiority of its clock to that of the incoming digital stream. But you're right, theoretically it could go either way.

The rest of your post was pretty much right on.

PS You all can call me just "Dusty". Calling me "DustyChalk" is like me calling one of you "JohnSmith" or "MaryDoe". It sounds kind of formal and funny.

 

[Flumpus]

Thanks Dusty... cleared some stuff up for me

A couple more questions if ya don't mind...

How subtle is the improvement with upsampling? I've anything from "I can't hear a difference" to "it makes a world of difference"... This leads me to believe it's very system dependent, but I guess it can't hurt to try it out. Also, the P1000 power base, do you have this? Just wonderin' what kind of improvement it made.

 

[Vertigo-1]

Since HDCD's popped up in here, I wanna jump in with a question of my own...if I were to get one of those DACs that have HDCD capability, does this mean I could add HDCD decoding onto *any* player/transport that doesn't have it? Or does the transport have to stream out an HDCD signal first before the DAC will output one? I'm kinda guessing option one doesn't work really, otherwise lot's of people would have done it already to get HDCD decoding into their systems.

 

[Dusty Chalk]

Flumpus -- Upsampling provides subtle improvement to a system as good as yours. What I have found is that I no longer get fatigued. I used to be able to only listen to my system for an hour or two at a time, and I would have to take a break. Now...well, let's just say I haven't hit that limit yet, something else would usually come up first (food, sleep, other ... uh ... "nature calls" -type events). I've listened continuously (almost) for up to 10 hours straight.

Haven't gotten the P1000 power base yet (?!?!?) -- it's next on my list. Stereophile says that you should get that first. I wouldn't know. Though I am a believer in a good, clean power supply, so I will eventually get it. Thanks for reminding me. My wallet said something too, but it was kind of muffled, being where it is. I think it was whining...

Vertigo-1 -- actually, yes, that's exactly how it's done. HDCD encoding is done to a normal 16/44.1 bitstream, so there are no special requirements. As long as the transport doesn't adulterate the bitstream in any way, you can just add a DAC. As long as you have some sort of digital out that's compatible with the DAC, that's exactly one way to do it. I have yet to find a transport that adulterates the bitstream (in other words, I haven't had trouble with any).

I think the reason that few people have done this is that it's cheaper to get the Denon DCM-370 -- less than US$200! That is the cheapest way that I know of to get into HDCD. And I also know a few people that are quite happy with this upgrade.

 

[Vertigo-1]

Hmm...then I'm thinking...what if I were to get an SACD player, and add on HDCD to that with a DAC? The problem with that though I think is that SACDs don't pipe though the digital outs, and since the output end would mostly be connected to the DAC, that could get inconvient.

 

[thomas]

dusty...

don't wanna get into a long off-topic debate here (as i've known to in the past

but most of the sites i've seen indicate that oversampling inserts 0's inbetween samples:

Quote:

Oversampling consists of 2 parts, upsampling and digital filtering (see also Figure 2.4). During upsampling a signal x(n) at sampling frequency fs is changed to a signal y(n) with increased sample frequency k.fs by inserting k-1 zero-valued samples to the signal (in Figure 2.4, k=4). The insertion of k-1 zeros spreads the energy of each input signal sample over k output samples, effectively attenuating each sample by a factor k. Thus it is necessary to compensate for this by multiplying each sample of y(n) by k. After upsampling, digital filtering is accomplished by interpolating the samples of y(n), resulting in a signal z(n). The spectrum of z(n) is the same as the spectrum of the original continuous signal, but repeats around multiples of k.fs.

(b would be normal oversampling, once you add interpolation filters in c, it becomes upsampling)

Also, when you oversample, the jitter in the system (from clock instability, and other distortion) remains constant, but the sampling rate increases sharply. As a result the same jitter makes up a larger percentage of each sample, and has a larger overall effect. Since oversampling doesn't increase the resolution, and adds jitter, overall it degrades the signal...

from an article on my HD, by Ryohei Kusunoki, from MJ magazine: (can't remember the link)
Quote:

under an average jitter environment, the oversampling can not operate theoretically, and lowers the accuracy within the operating field. In short, just by oversampling the original data, 16bit accuracy can not be satisfied anymore.

Basically, oversampling in itself degrades the signal (contrary to manufacture claims), but it can improve the overall sound because the digital/analog filters can be much less intrusive if overampling is used...

anyways, back on topic

vertigo- yep, unfortunately i doubt there will be a standard digital out format for SACD, probably due to piracy fears. But several manufacturers are releasing players with propritary digital outs. most notibly is Sharp, which is planning a consumer lSACD ine that interfaces with 1-bit digtal class D amplifiers. while i doubt any audiophiles will want the amp, the cheap digital out may be great for hooking up a DAC.

But still, your idea is good for the time being, just use a switchbox to connect both your SACD analog out and DAC to a headphone amp. Then you will have high-res audio all the time. Still, i'd prefer an oversampler, it should work better than HDCD, and work for ALL recordings. I don't think i have ANY HDCD's in my collection...

dusty: one more thing i disagree with... i think

Quote:

That is the cheapest way that I know of to get into HDCD. And I also know a few people that are quite happy with this upgrade

isn't HDCD going to be incorporated into windows media player soon? the technology is owned by M$, and i see little reason for them to have it unless it will be incorporated into WMP.... unless they're trying to dominate the audiophile world next!

 

[Dusty Chalk]

Quote:

Originally posted by thomas ...most of the sites i've seen indicate that oversampling inserts 0's inbetween samples

Well that's just stupid! (Not you, implementing oversampling that way.) That's the explanation I heard (don't remember where I read it, I think it was in one of the DIY magazines that pre-dated AudioXpress). I just can't imagine engineers would implement it so badly...

PS Thanks for the link. Quote:

Also, when you oversample, the jitter in the system (from clock instability, and other distortion) remains constant, but the sampling rate increases sharply. As a result the same jitter makes up a larger percentage of each sample, and has a larger overall effect. Since oversampling doesn't increase the resolution, and adds jitter, overall it degrades the signal...

Well, no, I stand by my statement. It may locally multiply jitter, but in the end, you have the same amount of jitter you had before, since the signal does eventually come back to the 16/44.1 representation.

For example, let's say sample A is 16ps ahead of schedule, and sample A+1 is 3ps behind schedule, then the samples generated by the oversampling algorithm necessarily get more accurate to the theoretical "perfect" original signal, and then begin to diverge again. There may even be a sample in there somewhere that is right on. The net effect, overall, in terms of calculus, should be about the same, if not smaller. It's just been smoothed out. Quote:

Basically, oversampling in itself degrades the signal (contrary to manufacture claims), but it can improve the overall sound because the digital/analog filters can be much less intrusive if overampling is used...

Yeah, well, don't believe everything you read, that's all I can say to that. But also notice they said oversampling, and not upsampling, maybe this is before the days of upsampling as it is currently understood? And if it is as you say it is (with the zero's), then I have to agree! That would suck! Quote:

dusty: one more thing i disagree with... i think

Not sure what follows had to do with anything I said, so I don't think there is a disagreement. Quote:

isn't HDCD going to be incorporated into windows media player soon? the technology is owned by M$, and i see little reason for them to have it unless it will be incorporated into WMP.... unless they're trying to dominate the audiophile world next!

Gawd, I hope not (the latter that is)! That would suck! Please stay away from high-fi, Microsoft! As to the former, I have no idea (I am in no way related to Pacific Microsonics nor Microsoft), but other people have predicted the same thing.

But what's funny is that HDCD is just a smart compression technology that happens to be compatible with the uncompressed form. I can see why Microsoft would be interested -- they can use it with lots of multimedia -- pictures, movies, etc. Anything where what is represented is an analog signal. Imagine if IE decoded pictures with this subtly better look and sounded better to boot, and Netscape didn't -- you could see where most people would eventually take their business?