Up sampling - good or evil?

[Jamtoast]

Hi folks

I've had a swim around the forum looking for answer to this but haven't found anything definitive. It may have been discussed ad nauseum and I can't find it. Or perhaps it's common knowledge and I'm showing my audiophillic naivety.

Anyway - to the question...

Like most DACs my Caiman (soon to be Bushmaster) can up sample to way above the 44.1kHz sampling rate of most of the (ripped CD) recordings I own.

Is the best approach to set the sample rate of the digital output from my Mac mini to match that of the recording (i.e. 44.1Khz) or does up sampling the stream to the DAC (say to 96kHz) add anything?

Intuitively a bit perfect rip sent at the same sampling rate as the original master is the 'purest' way to do it and up sampling won't add any sonic information that isn't in the original (but could add clock errors). This (http://www.stereophile.com/asweseeit/344/) (rather old) article seems to support that view.

Thoughts?

[Stratmangler]

I think the consensus is that you can't add information that isn't in the original.
However, if the material is upsampled then you may well be getting the DAC to use different filters automatically, instead of the brickwall filtering that CD players generally use.
You might prefer the sound of the different filters brought into play with upsampled material, but that is a purely personal and subjective decision.

[WAD62]

Does the bushmaster actually 'upsample' or is it capable of dealing with higher resolution files? :scratch:

I think (as Chris states) there's no single answer to this question, I think it probably depends on the DAC in question, and perhaps even the upsampling algorithm being used.

I have an old Beresford 7510 (the one before the cayman) in my study, and I upsample to 24/96 via FFSoX/ASIO (winamp plugins) and it seems to be marginally smoother at the top end, is it because I'm taking workload off the DAC, or as Chris states I'm activating different software inside the DAC...I don't know but it seems to like it.

However I've recently acquired an Audiolab M-Dac, and it hates upsampling/resampling of any kind...there is a marked difference when you move away from the native 16/44.1, even replaygain & crossfade compromise the quality (as they resample to a certain extent).

As we speak I'm messing around with my new toy, an Arcam rPAC, and am trying to decide which way to go...I can't decide as yet :scratch:

[Jamtoast]

I don't know but it seems to like it.

I had the same experience with my Caiman but then convinced myself that I must be making it up. Perhaps as you say - filtering.

That said it suits me to run 44.1 as some nasty things started happening occassionally at 96 - due I think to the Mac Mini not coping with sending to both SPDIF at 96 and HDMI.

[WAD62]

I had the same experience with my Caiman but then convinced myself that I must be making it up. Perhaps as you say - filtering.

That said it suits me to run 44.1 as some nasty things started happening occassionally at 96 - due I think to the Mac Mini not coping with sending to both SPDIF at 96 and HDMI.

I know what you mean Graham, sometimes the differences are so small, and it's not easy to do a quick enough A/B for a real comparison...

As I'm running the rPAC on my work computer I originally tried it without the upsampling, as it does add an extra CPU load, but for some reason the winamp ASIO out driver won't do gapless, unless I do upsample via FFSoX :scratch::scratch:

Perhaps a re-installation of the .dll may help

Edit; looks like it's a known problem, I must only have used ASIO in conjunction with FFSoX in the past:scratch:

However with my M-Dac it's clearer that it wants the native unaltered file, as it seems to trigger all of its sound processing gizmos (d3e engine stuff), and I'm pretty sure it's an upsampling DAC, so probably best to leave it to its own devices.

[sq225917]

Upsampling tends to mean slightly different things to different people. It should just mean sample interpolation, but it's used in place of ASRC and host of other things including oversampling.

There's no reason why upsampled data 'should' sound any better than lower res data. In reality though higher sample rate data is often processed with better filters that don't fold back into the audible.

I certainly don't think its something that should be avoided.

[StanleyB]

Like most DACs my Caiman (soon to be Bushmaster) can up sample to way above the 44.1kHz sampling rate of most of the (ripped CD) recordings I own.
Non of my DACs use upsampling. Having listened to upsampled music I have come to the conclusion that they make the sound softer sounding.
I personally don't recommend upsampling the audio in your PC before sending it to a DAC like mine. My DACs use oversampling, the frequency of which in itself depends on the incoming frequency of the original audio.

[Patrick Dixon]

I certainly don't think its something that should be avoided.

It isn't something that should be avoided but it isn't something that will necessarily be better either. As often, the detail of the implementation is more significant than the underlying technology.

And btw, that clock jitter thing in the article you linked to is a complete red-herring.

[Patrick Dixon]

Oh and Oversampling and Upsampling are really the same thing. Oversampling just seems to be the term used for Upsampling using integer multiples of the original signal.

[Reid Malenfant]

It isn't something that should be avoided but it isn't something that will necessarily be better either. As often, the detail of the implementation is more significant than the underlying technology.
Well said :)

I'm very happy indeed with my Monarchy Audio DIP 48/96 feeding 24 bit words at 96KHz to a dCS Purcell that converts that to DSD, which gets decoded via a dCS Delius.

The only way you'd get them off of me is from my cold, dead hands :eek:

[Welder]

Evil! Bad!
I mean good grief, aren't we always hearing bout interfering with the natural sound of the source an all that.:scratch:

As fer yer interpolation; well, it's not the original music at all, ffs.:rolleyes:

Hang on, my set up re-samples and up samples :scratch::D

[StanleyB]

What about giving your technical opinion John, instead of shouting monkey chants from the sideline.

[Welder]

Okay Stan, but not very technical I’m afraid.;)

I don’t think its worth worrying about.
Something tells me that up-sampling and over-sampling are to be avoided if possible, but I haven’t sat down and listened to the results of various sampling algorithms and it’s the algorithms used that imo probably make the difference.
I live quite happily with up-sampling but I am rather interested in at least the idea of NOS Dacs. It’s just the very few I’ve heard so far didn’t sound as good as the up-sampling variety.

[Jamtoast]

Non of my DACs use up sampling... ...My DACs use oversampling, the frequency of which in itself depends on the incoming frequency of the original audio.

My mistake Stan - you're quite right, I guess the up sampling (if that's the right term) of the 44.1 kHz signal is happening at the Mac end to be sent to the Caiman at 96. So any difference in the sound would be down to the processing at the Mac end?

A technical explanation of the difference between up- and over sampling would be great.

[Yiangos]

Hey guys

Not so "technical" myself but here's my findings regarding up-sampling.Back in the eighties and nineties,cds were recorded with something called "emphasis".I know because it shows on the screen of my DAC.Those cds don't like to be upsampled producing an overbright sound or sometimes a kinf of sound like the tweeters have holes in them.As for the rest,some like to be upsampled and some don't.If someone said to me "you can have just one cd player and chose between an upsample on and one that don't" i'd go for the non upsapling.

[Clive]

I've not exhastively auditioned the differernces but I suspect that upsampling whole multiples (eg 44.1 to 88.2) has a better chance of being transparent than say 44.1 to 96.

[Marco]

To my ears, evil - big time.

Whenever I've listened to up-sampling players, I dislike the false 'sheen' that they superimpose onto the music, and how the up-sampling effect makes music sound too 'nice', thus stuff which is meant to sound 'gritty', instead just sounds smoothed-off and rather 'polite'.

The effect just doesn't seem natural to me, compared with what I'm used to from my own digital source. For me, the K.I.S.S principle rules in hi-fi - and that means buggering about with the audio signal as little as possible! ;)

Marco.

[Macca]

Hey guys

Not so "technical" myself but here's my findings regarding up-sampling.Back in the eighties and nineties,cds were recorded with something called "emphasis".I know because it shows on the screen of my DAC.Those cds don't like to be upsampled producing an overbright sound or sometimes a kinf of sound like the tweeters have holes in them.As for the rest,some like to be upsampled and some don't.If someone said to me "you can have just one cd player and chose between an upsample on and one that don't" i'd go for the non upsapling.

I have an upsampling player (Cambridge 840) and lots of 80s & 90s CDs but have never heard this effect. I would agree with Stan that upsampling makes things a little 'softer', not necessarily a bad thing dependign on partnering equipment. Also agree with Marco about the 'sheen' although I have heard the same thing with non-upsampling players also. In my main sytem I prefer my Sony 761E over the 10 or so other players I have. That is a mid 90s Bitstream effort. No softeness or sheen., just the musical facts, ma'am.

[dave2010]

I think that the term upsampling applied to DACs may be used in different ways. Some simply raise the sampling frequency by a factor, say 4,8,16 and then apply digital filters to the resulting data stream. The advantage of this is claimed to be noise shaping. Some noise is moved out of the audio band.

Some use techniques such as delta-sigma modulation, and effectively recode the data as a series of pulses - maybe even just 0 or 1 logical values, which are then integrated using simple circuitry to give a voltage output. See, for example - http://en.wikipedia.org/wiki/Delta-sigma_modulation

I have vague recollections of these operating at least at 64 times the recorded sampling frquency.

In theory there is no point in upsampling the data stream as no additional information is made available, but there can be merits in terms of design, which has already been pointed out may be due to filter design. Some designs, such as the sigma-delta designs may also offer improved linearity. To consider a particular example, the (8 bit) data samples 1000 0000 and 0111 1111 represent analogue signals which are the minimum distance apart. Some systems with poor linearity may have problems when trying to render an alternating sequence of these. It's not a very interesting example, as the output levels should be low, but essentially the same problem can occur with more realistic data, and could well manifest itself as audible distortion.

Some class amps (e.g class D and T amps) also use techniques similar to delta-sigma DACs. It would seem to me to make sense for some units to have a combined DAC/amp if such high frequency modulation methods are to be used.

SACDs are particularly suitable for delta-sigma conversion, though I suspect that if a designer particularly wanted to convert the SACD data stream back to PCM internally using software, and then use a multi level DAC to ouput the signal it would be possible. Probably pretty pointless, though.

I don't know what problems/effects might arise by using a chain of such high frequency devices, such as a delta sigma DAC feeding a class D amp, though I suppose if the results were truly horrible we might have heard by now.

Upsampling is a particular case of sampling frequency conversion. Generally any distortion introduced is minimised by using powers of 2 between the incoming and outgoing sample frequency. Note that CD production may often involve a sample rate conversion (downsampling), which could be non optimal e.g 48,96 or 192 -> 44.1 kHz.

[Jamtoast]

Nice explanation Dave. Thank you.

[WAD62]

As I'm running the rPAC on my work computer I originally tried it without the upsampling, as it does add an extra CPU load, but for some reason the winamp ASIO out driver won't do gapless, unless I do upsample via FFSoX :scratch::scratch:

Perhaps a re-installation of the .dll may help


For other users of winamp/ASIO...

If you want to use 'unaltered' 16/44.1 with ASIO, and you need it to be gapless (doesn't work with the standard FLAC input plugin), then you can achieve this by using the FFSoX input plugin, and turning off 24 bit/replay gain/resampling/dither etc. And just use it to transcode the FLAC, this is then treated as WAV/PCM when it arrives into the winamp player, and gives gapless in the original resolution...:)

[Jamtoast]

So have now done some close comparisons with my set up and have consistently found that I like the sound when the Caiman is fed tracks at their native sample rate. There is a slight softening (or vagueness for want of a better term) that most tracks get when they are up sampled before being sent to the DAC. I also found the sound stage a bit compressed - less spatial separation in the mix especially when the music gets dense.

Most of my tracks are ripped from my CD collection so are at 44.1 kHz, but for the few hi res albums I have Audivana handles the sample rate switching so always sending at the native rate. Have also been playing with Bit Perfect and may switch to this as less intrusive interface and subjectively a marginally better sound.

It will be interesting to see if the Bushmaster when it arrives exhibits the same preference for native sample rate.

[dave2010]

Hey guys

Not so "technical" myself but here's my findings regarding up-sampling.Back in the eighties and nineties,cds were recorded with something called "emphasis".I know because it shows on the screen of my DAC.Those cds don't like to be upsampled producing an overbright sound or sometimes a kinf of sound like the tweeters have holes in them.As for the rest,some like to be upsampled and some don't.If someone said to me "you can have just one cd player and chose between an upsample on and one that don't" i'd go for the non upsapling.There may still be some CDs with emphasis. The original specification for CDs was worked out between Philips and Sony. The Europeans (Philips) were happy enough with PCM digital, but the Japanese engineers (Sony) suggested that they could get better noise results (lower noise floor) by using pre-emphasis with corresponding de-emphasis used in the CD player on playback. The pre-emphasis and de-emphasis were done using analogue circuitry (whether that's still needed I'm not sure - it might be possible to do it using digital filters now). The pre-emphasis/de-emphasis process was a similar approach to that used in radios and also for LPs and 78s, with their RIAA curves etc., though not necessarily using the same frequency curves.

I had a Philips CDP, and ran it for many years without noticing this. Then one day a light appeared on the front panel. Looking at the CD it was indeed a Japanese pressing, and the de-emphasis circuit had switched in. That was the only one I noticed using that CDP.

I think that there are still some CDs made with pre-emphasis. They should play properly on decent CDPs, as the standard requires that to be possible.

If ripping CDs with pre-emphasis they will tend to sound over bright unless the ripper takes note that pre-emphasis has been applied. I think there are some rippers which do this properly, but maybe not many.

[Yomanze]

Upsampling makes things sound more artificial to me, just like 'reclocking' DACs. My Theta DAC only uses the receiver's PLL, is sensitive as hell to transports, but when the transport is right it simply sounds more analogue than anything else I've had.

[StanleyB]

My Theta DAC only uses the receiver's PLL, is sensitive as hell to transports, but when the transport is right it simply sounds more analogue than anything else I've had.
Time to get a Bushmaster ;).

[Yomanze]

Time to get a Bushmaster ;).

My Theta DS Pro Progeny was a significant step up from my old Beresford TC-7510 in terms of dynamics, 'gravitas' and a liquid organic sound with way more seperation between instruments. Is this the case in the step up between the Caiman & Bushmaster?

[allsorts]

The sticky on the topic of osx upsampling led me to try upsampling cds into my 2009 macmini at 16/48. I went through audio midi and changed output to 24/96 to Audirvana Plus and then into Caiman. To my increasingly wooly ears, its noticeably the best sound quality i have achieved from my system. One other item to note was that any issues with getting the odd blip through the speakers disappeared when i upped the ram from 2 gb to 8gb.

[AlfaGTV]

upsampling cds into my 2009 macmini at 16/48.

Misprint? To my ears the very worst thing you can do to your computer audio "philes" is resampling from 44khz to 48khz. The music gets an unnatural "aura" to it, a little like an oversharpened digital photo using a visual analogy. I agree with the people hearing a smoother sound when upsampling to 88.2Khz/176.4Khz or even resampling to 96Khz or 192Khz. Not all music benefits from it and in most cases i prefer the original sample rate and bitdepth.
I have some DAC's to play around with the one i mostly prefer is an old Micromega T-DAC with a very full bodied sound, if a little grainy compared the Benchmark DAC-1.
But as always, YMMV.
Br Micke

[aBe]

I'd take vanilla over other toppings anyday.

I set every Latency and Buffer settings to zero or the lowest possible value, until just before it starts to sound 'clicky'.

On ASIO4ALL, I disabled everything except - Force WDM Driver To 16 Bit.
Enabling forced-16 Bit for some unknown reasons increases bass weight and opens up the sound a little bit :scratch:

[WAD62]

I'd take vanilla over other toppings anyday.

I set every Latency and Buffer settings to zero or the lowest possible value, until just before it starts to sound 'clicky'.

On ASIO4ALL, I disabled everything except - Force WDM Driver To 16 Bit.
Enabling forced-16 Bit for some unknown reasons increases bass weight and opens up the sound a little bit :scratch:

I can't detect any difference when setting the 'Force WDM Driver to 16 bit' on or off...so I leave it off.

There's quite a nice ASIO4ALL explanation here;

http://www.head-fi.org/t/221237/asio4all-explanation

Force WDM driver to 16 Bit
This option only has an effect if the supported bit depth of the WDM driver is larger than 16, but less than 24. Some AC97 devices report e.g. 20 Bits resolution but cannot actually be opened for more than 16 Bits resolution. Should this be the case on your system, this option provides a workaround. Originally, this was introduced as a workaround for an issue with the SigmaTel AC97 WDM driver.

...perhaps it's something in your setup.

As for buffers, unless you're playing/recording latency is not an issue, therefore the larger the buffer the better, they insulate you from potential dropouts caused by system contention, and have no detrimental effect on the sound...well not in my opinion anyway;)

Where's John(Welder), as I know he feels otherwise...:)