To upsample or not to upsample? Any experiences?

[paradiziac]

I've been playing about with the Mac mini feeding my DAC via the MF V-Link.

I can hear a clear improvement if I upsample everything in the Mac mini to 24/96 before sending the Audio to the DAC, rather than outputting a "bit perfect" 16/44.1 signal (whatever that means!).

There's an improvement whether I'm listening to iPlayer, Spotify, or FLACs through Audirvana/Pure Music.

Also, all my FLACs now sound equally as good as the hi-res test material from HD tracks. :shifty:

I tried upsampling before on the PC using both Win 7's built-in sampler and the Sox resampler in Foobar. I couldn't hear much (any?) difference. Perhaps it was slightly better, but it was hard to tell because my various PCs were all less resolving than my new Mac mini (maybe noise issues from the USB ports on those machines? or the OS itself?).

If you're using a computer source, I'd urge you to give it a go--potentially a "free" upgrade!

Anyone else found upsampling on the computer to be better/worse/no different than letting the DAC take care of it all?

 

[CnoEvil]

I can't answer your question per se, but I've tried very different sampling philosophies.

Linn use the following approach in their DS range: http://docs.linn.co.uk/wiki/index.php/Up-sampling
and as I can testify, it sounds very well.

Audio Note on the other hand believe in the NOS approach here (under technology):
http://www.audionote.co.uk/downloads/dac2.1x_sig_2007.pdf

It sounds glorious in the context of their own system, but when I slotted it into mine, sounded incredibly detailed (probably due to no filtering,) but too forward and bright for my taste. I'm told the Black Gate Capacitors take an age to sound right, but I wasn't prepared to buy it on the off chance that the sound would come round in three months time.

It's not exactly what you were asking, but it is related (sort of). :shifty:

Cno

 

[altruistic.lemon]

You get automatic upsampling anyway, because that's what DACs do.

 

[paradiziac]

@Cno--yes, an interesting aside, I was aware of the 2 approaches (upsampling v non-up/oversampling with no/minimal filtering).

(Another aside, with no filtering you get a load of inaudible HF noise pouring into your amp...some amps/speakers might not like that--I wonder if that's why the AN didn't work out for you?).

When Linn speak about the need for "considerable processing power" it makes sense. I guess that's why a core i5 processor with 4 or 8GB of RAM does a very good job of upsampling (the main problem is the interface to get the data out cleanly). Perhaps hi-end dedicated streamers do a better job overall (?), but I have a computer in my living room which already has a very powerful processor inside--so I'm keen to try to get a good sound out of a piece of gear that I need anyway (regardless of audio).

In the Mac software (Audirvana +) you can actually play with a bunch of sliders that control all these settings that Linn talk about in the referred article. Linear-phase, minimum-phase, apodizing filters, Nyquist frequencies and all that jazz. So you can tune your source to your preferred taste rather than being stuck with the Linn design team's choice. (Though I haven't heard any obvious difference between any of the settings on my modest system...!)

I think the most interesting thing is that I found my computer does a better job at upsampling than my DAC. I just wanted to share that and was curious if anyone else has found the same--since the prevailing wisdom seems to be to aim for "bit perfect" output from your computer/streamer.

And...that "hi-res" 24/96 files do indeed sound better than 16/44.1, but probably not for the reasons that most people believe/are told.

 

[paradiziac]

You get automatic upsampling anyway, because that's what DACs do.

Yes, but my point was that my computer is doing a better job at upsampling than my V-Link+DAC--and I can hear it fairly easily even on my system.

 

[CnoEvil]

@Cno--yes, an interesting aside, I was aware of the 2 approaches (upsampling v non-up/oversampling with no/minimal filtering).

(Another aside, with no filtering you get a load of inaudible HF noise pouring into your amp...some amps/speakers might not like that--I wonder if that's why the AN didn't work out for you?).

My own (very uncomplicated) pet theory, is that the more forward presentation of the DAC, is nicely offset by their SET amps and musical sounding speakers ie it's a synergy thing; and then due to the simplistic approach, you can hear everything that's been recorded (ambient info etc), without any brick wall filtering.

 

[garyw77]

Anyone else found upsampling on the computer to be better/worse/no different than letting the DAC take care of it all?

To answer your last question, yes i have found upsampling to be very effective in my system.

Using Pure Music on the Mac Mini i have experimented with many settings. IMO The best seems to be to upsample to the maximum 384khz that the M2Tech Young Dac takes, this seems to open up everything.

 

[paradiziac]

@Cno--could be, but if AN really did remove even the gentle analogue filter as was reported somewhere, you have to wonder about all that inaudible HF--at the very least I expect there's going to be some freaked out cats and dogs in households with AN DACs going into systems with ribbon tweeters that go up to 40kHz!

You got me looking at Linn now...it's not easy to find any technical info of note on their website, but it seems that their top of the range Klimax DS has the same dual-mono WM8741 chips as my 300-odd quid DAC (and the rDac, which has a single WM8741)! Good to know as I think my DAC's power supply and output stage is decent as well, the main issue is the interface with the transport.

Guess I just need to figure out how to input my 384kHz upsampled data with "custom filter" into my DAC and bypass the DACs filters...

Even if it was only half as good as the Klimax, a 5K piece of gear for a few hundred notes is not to be sniffed at! I can see the attraction of DIY audio...

 

[tino]

I have tried a few upsampled files (upsampled off-line through Audacity) and on preliminary (albeit casual) hearing have not notice any improvement whatsoever on my Wadia 151 ... probably because the Wadia is performing upsampling anyway in real time anyway!

You may be interested to read their white paper (fairly basic non techie explanation) ... but if you peruse their web-site you may also find reference to their new upsampling DAC ... the 121 (24-bit 1.4 MHz upsampling!). Might give an M-Tech Young DAC a run for its money!

http://www.wadia.com/technology/digimaster/

http://www.wadia.com/products/decoding/121/

 

[tino]

... And as a counterpoint ... here is an interesting (techie) white paper on the virtues of NOS DACs ...

http://www.sakurasystems.com/articles/Kusunoki.html

 

[garyw77]

You may be interested to read their white paper (fairly basic non techie explanation) ... but if you peruse their web-site you may also find reference to their new upsampling DAC ... the 121 (24-bit 1.4 MHz upsampling!). Might give an M-Tech Young DAC a run for its money!

http://www.wadia.com/technology/digimaster/

http://www.wadia.com/products/decoding/121/

Thanks for the link, the 121 looks like a good piece of kit.

Unfortunately i wouldn't be able to upsample in Pure Music more than 192khz as that's the maximum input data rate the 121 takes and then the DAC upsamples.

The M2Tech doesn't upsample but accepts the 384khz from source.

 

[paradiziac]

Anyone else found upsampling on the computer to be better/worse/no different than letting the DAC take care of it all?

To answer your last question, yes i have found upsampling to be very effective in my system.

Using Pure Music on the Mac Mini i have experimented with many settings. IMO The best seems to be to upsample to the maximum 384khz that the M2Tech Young Dac takes, this seems to open up everything.

Cheers Gary, that's good to know, since the Young is one of the few DACs that will allow you to get a 384kHz sample rate out of a computer.

I'm also intrigued how non-oversampling works--there are some hi-end NOS solutions by companies reputed for making highly musical kit.

I think I'm less inclined to dash out and get an M-DAC, as amazing as it may be. Too many interesting developments in this area--it's becoming like computing: don't rush out and buy unless you have to, there will be something better/faster/cheaper next month.

Many may disagree if they haven't tried it, but my instinct is that we'll soon end up with a computer + reasonably cheap DAC being good enough for almost any amp/speaker combo.

All controlled via your phone/tablet computer of course

 

[CnoEvil]

It's good to have this subject rekindled, as I started a thread on NOS Dacs awhile back, which fizzled out quite quickly: http://www.whathifi.com/forum/hi-fi/nos-dacs

 

[paradiziac]

I have tried a few upsampled files (upsampled off-line through Audacity) and on preliminary (albeit casual) hearing have not notice any improvement whatsoever on my Wadia 151 ... probably because the Wadia is performing upsampling anyway in real time anyway!

I think it also depends on the quality of the software resampler used. Audacity isn't the best. Comparisons here:

http://src.infinitewave.ca/

 

[paradiziac]

It's good to have this subject rekindled, as I started a thread on NOS Dacs awhile back, which fizzled out quite quickly: http://www.whathifi.com/forum/hi-fi/nos-dacs

Well, I think it's fair to say that upsampling on a PC is much better than upsampling on a DAC. A DAC chip doesn't have enough computing power for a processor intensive operation and it's normally working in realtime. That's why Linn go on about processing power in their blah blah on the Klimax DS.

I've read that (one of?) the net effect of upsampling is to reduce jitter.

So maybe that explains the benefit I found. I know my DAC's relative weakness is sensitvity to jitter (a somewhat jittery SDPIF receiving chip and no re-clocking circuits etc).

The higher end NOS DACs by Audionote (and new Sugden CDPs) I would expect get a very good feed without much jitter and no doubt have additional de-jittering/re-clocking circuitry on the incoming stream.

I think ideally, I'd like to have a NOS DAC accepting crazy-high sample rates with gentle analogue (or no?) filtering and do all my own experiments/tuning on the computer side using a processor I already own and cheap/free industry standard software...

 

[acalex]

Really rookie question...what's the advantage of upsampling at pc's level? How does upsampling work? I mean...it is taking a 16bit /44KHz songo to 24/96 for example, right? This is what every DAC does normally?

Thanks

 

[paradiziac]

Really rookie question...what's the advantage of upsampling at pc's level? How does upsampling work? I mean...it is taking a 16bit /44KHz songo to 24/96 for example, right? This is what every DAC does normally?

Yes, I'm talking about upsampling on my PC a 16/44 mp3/FLAC/WAV to 24/88, 24/96 (or higher if your USB interface supports it--mine doesn't).

What the advantage? The simple answer: it sounds better on my system. I don't know why. I was curious if others have found this or not.

There are some logical reasons why upsampling on a PC is better than upsampling on a DAC chip. Upsampling needs a lot of processing power, doesn't have to be done in realtime, and digital processing is best done separated from the analogue output. The quality of the upsampling depends on the software algorithms used and the tuning/setting of digital filters, and if these are hardwired into a cheap chip that has to work in real time, compromises have to be made and the sound will suffer. That's maybe why you hear reports of people with older/cheaper upsampling CD players complaining that upsampling sucks the life out of the music. Even most software resamplers don't do a very good job, but some are very, very good and better than any hardware resampler. Two good ones I know are Sox (available as a (free) Foobar plugin) and iZotope (available in Audirvana + on the Mac). The default resampler in Mac OSX is pretty good as well, but not the Windows resampler (which is maybe why there is a lot of advice around saying that people should aim to get "bit perfect" files from their digital source). So basically, real "state of the art (but cheap/free)" software re-samplers are available now on our computers, so it seems natural to experiment.

Yes, most DACs upsample or oversample. But there are also good designs that do not, often from some of the more "musical" brands--very good (or even better?) sound is possible without it. I think it's difficult to separate the marketing blah blah from the reality, so it's better just to experiment and listen, rather than think upsampling/over-sampling is better overall than non up/oversampling. Like most things in hifi, the implementation is the key.

Since you're in Belgium, ever heard of the (Dutch) Metrum NOS DAC? It's a non-oversampling design using unspecified industrial chips. Not too many over here, but some (hype?) in the USA.

Some more reading:

http://www.soundstage.com/gettingtechnical/gettingtechnical200311.htm

http://www.digitalaudioblog.com/2011_01_01_archive.html

 

[paradiziac]

I would just add, that after testing the same tracks into my DAC with 2 different transports:

1: NAD 521i

2: Mac Mini -> V-Link (w/ Audirvana--iZoptope resampler at 24/88, 24/96 and 16/44 without software resampling)

The 2 upsampled files and CD transport was almost impossible to distinguish and 16/44 was inferior.

 

[acalex]

Thanks for explaining and for the links, I will enjoy the reading!
So that's the main difference in DACs, some will up/oversample in real time (hardware oversampling), other like the Metrum or the Audio Note ones took a different approach by removing this up/oversampling.

Still not sure why upsampling a track via a chip on the DAC would improve the sound quality...maybe it will become clear after I have read the 2 articles!

So what is a DAC doing after up/oversampling which I guess is the first "task" is performing once it gets the sound on its input?

 

[Frank Harvey]

Speaking with someone within the industry last week, he said their company had done testing to see whether it was worth upsampling 16bit material, and they found that it was better to play music into the DAC in its native form, and that upsampling had no benefit.

 

[acalex]

Speaking with someone within the industry last week, he said their company had done testing to see whether it was worth upsampling 16bit material, and they found that it was better to play music into the DAC in its native form, and that upsampling had no benefit.

Yes...but even in this case upsampling will be still performed by the DAC itself...correct?

 

[acalex]

Really rookie question...what's the advantage of upsampling at pc's level? How does upsampling work? I mean...it is taking a 16bit /44KHz songo to 24/96 for example, right? This is what every DAC does normally?

...

Two good ones I know are Sox (available as a (free) Foobar plugin) and iZotope (available in Audirvana + on the Mac). The default resampler in Mac OSX is pretty good as well, but not the Windows resampler (which is maybe why there is a lot of advice around saying that people should aim to get "bit perfect" files from their digital source).

....

So if I am using a windows system and playing a 16/44 FLAC ripped from my cd using for example Media Monkey with no plug-ins of any kind, this means that Windows will ANYWAY upsample the track automatically?

So obtaining a bit perfect would mean NOT allowing Windows to do this and letting the "native" digital track directly into the DAC which will take care of up-sampling before converting...correct?

 

[Frank Harvey]

Yes...but even in this case upsampling will be still performed by the DAC itself...correct?

I'm just passing on what he said

 

[paradiziac]

Yes...but even in this case upsampling will be still performed by the DAC itself...correct?

Yes, but it depends on the DAC design.

If you have an (up/)oversampling DAC, the DAC will oversample automatically. A non-oversampling DAC won't/can't. The reason some designs oversample and some don't depends on the chip architecture. Some DAC chips (delta sigma type e.g. WM8741) require oversampling, but other DAC chips (R2R e.g. TDA1541, PCM1704) can be used with no-oversampling. Some manufacturers and "audiophiles" think the NOS chips are better and gave a more natural sound (because of the negative effects of filtering when oversampling), but as they were more expensive to produce, the chip manufacturers gradually stopped making them and they are hard to source these days. So...as a consumer, it's not something to worry about, except to know that it's a fast-developing area. For now, we just have to choose the best sounding source! It's simply worth knowing as background knowledge that OS v NOS in DACs is a bit like SS v Tubes in amps.

Speaking with someone within the industry last week, he said their company had done testing to see whether it was worth upsampling 16bit material, and they found that it was better to play music into the DAC in its native form, and that upsampling had no benefit.

Hard to judge that without context, but it makes sense.

Maybe the benefit or otherwise depends on the quality/implementation of the resampler, the digital transport and the DAC. So whether you'd get an improvement would be down to each individual. If everything is optimized (as you'd expect on a manufacturer's test rig), I would say that there is less chance of there being a possibility to improve things. After all, increasing the sample rate doesn't retrieve information that wasn't there in the first place.

But in my flawed set up, upsampling is definitely beneficial and brings the computer up to the level of a CD player used as a transport.

 

[paradiziac]

Really rookie question...what's the advantage of upsampling at pc's level? How does upsampling work? I mean...it is taking a 16bit /44KHz songo to 24/96 for example, right? This is what every DAC does normally?

...

Two good ones I know are Sox (available as a (free) Foobar plugin) and iZotope (available in Audirvana + on the Mac). The default resampler in Mac OSX is pretty good as well, but not the Windows resampler (which is maybe why there is a lot of advice around saying that people should aim to get "bit perfect" files from their digital source).

....

So if I am using a windows system and playing a 16/44 FLAC ripped from my cd using for example Media Monkey with no plug-ins of any kind, this means that Windows will ANYWAY upsample the track automatically?

So obtaining a bit perfect would mean NOT allowing Windows to do this and letting the "native" digital track directly into the DAC which will take care of up-sampling before converting...correct?

In Windows Vista or 7, there's a setting in "sounds" (control panel or somewhere), if you check that it says 16/44, then Windows will leave it alone and you'll get the "native" digital track going directly to the DAC as you say.

Or you can use the same setting to set your output to 24/96 and let Windows do the resampling and see if it's better or worse.

When I tried it, it was hard to tell the difference. The upsampling maybe helps, but the Windows upsampler itself is not so great, so I think the two effects cancelled each other out.