over sampling or upsampling

Tell me their difference?

meridian 588 use upsampling or oversampling?
Ayre CX-7 use upsampling or oversampling?
Thanks

 

Someone will me correct me soon if I'm wrong I'm sure, but I believe there is no difference whatsoever; just two different words for the same thing.

Edited to add: Ah, I see I was indeed not quite right...

 

Generally, there's no difference. Upsampling was invented by the marketing guys at some companies in an attempt to distinguish them from the masses when they weren't actually doing anything different at all .

They are both a way to increase the sample frequency before the D/A conversion so that the "brick wall" filter doesn't have to be so steep and hence has a less detrimental effect on the sound.

Michael.

 

the only 'difference' afaikt is that upsamping is increasing the sample frequency by a non integer amount i.e. 4.3537.....x for 192 upsampling. and that oversampling is by an integer amount i.e. 8x, 16x, 128x. the only reason for this odd increase factor is that home cinema amps and dvd players have pushed the prices of dacs that accept these frequencies down below good oversampling dac/filters. also a chip to deal with lower frequencies is often cheaper (if you work it out 8x oversampling is 352.8khz against 96 or 192khz - and that's only 8x).
there is however a lot more going on than the up/oversampling rate. there's the interpolation scheme, filter types, word size (16bit, 18 bit, 24 bit ...) etc. which are just as, if not more important to the final sound of the player / dac.
cheers


julian

 

Does 'oversampling' imply a whole number multiplier and 'upsampling' conversion to a particular rate? IOW if you 'upsample' CD to 192kHz it's a whole different can of worms to a straightforward oversample to 176kHz.

In general sample rate conversion, as in 44.1kHz to 192kHz (times 4.3537...) is to be avoided except where absolutely necessary, so 'upsampling' involving such conversion is a bad thing. A needless complication.

Paul

 

Upsampling, can be either a direct or multiplier or part of, IE 96/192Khz.
The SRC (usually a bit rate enhancer also) takes a 'good guess' at the interpolation curve and 'fills' in the dots in you like from say 88.2Khz (2 times red book) to 96Khz or 174.6<>192khz, IT cannot ADD what is not there, it just fills in the missing pieces of the curve from the information it has recieved.
Over sampling taking the 44100hz(red book) and directly multipling by the frequency by a factor of 4/8/16/32/ or 64 times so if your dac/cdp is 32 times O/s thats 1.411200Mhz theroy being by spreading the digital signal over a higher and wider bandwidth the 'digital noise' is lost the signal is then returned to red book land by either a brick wall filter, or spline interpolation algorythym, but its still in essence 44.1khz.
generally I have found that upsampling players and dac's (one exception so far the MSB PLatinium series 3 time u/s 133.3Khz), the effect is to' flesh out' the sound, to 'stretch the sound stage' and increase the detail and sound less digital like, those of you with switchable sample rates will have proberbly noticed this effect
IMHO O/Sampling gives a more 'real feel' to the sound, although Upsampling can and does reduce the digital glare effect in stock form.
The upsampling buzz word has been banded about for 5 or so years now, done correctly it can produce good results, although the twats that devised cd in the first place, should have had this method already implemented before it was released on the unsuspected world 20 odd years ago, that way the initial disappointed wouldn't have been so great

 

Here's a question.

How exactly does it "fill the gaps" ??

I mean if you take a sample of 20khz using redbook - you'll get a maximum of 3 sample points. When it comes to recreate this 20khz signal out of the dac, redbook relies on the fact that it is bandwidth limited to 20khz to get a 20khz sine wave out for the input signal. It has to be a 20khz sinewave linking these three sample points as that is the highest frequency the system can reproduce due to the brick wall filter.

Now consider this. IF you were to just increase the brickwall filter to 40khz but keep the sampling frequency the same, then the resulting output of a 20khz input would be closer to a triangle wave, as it can now more quickly swing the output voltage to get to the next sampling point in a straighter line.

So... does the upsampling/oversampling recreate a series of sampling points following the 20khz curve? or does it basically draw a straight line between the three "true" sampling points. IF it does the latter it would seem to me that the resulting output will be even less accurate than the 20hkz bandwidth system would allow in the first place.


Just a thought. I'd be interested to know exactly how these systems improve anything over the standard redbook spec.


GTM

 

there are a number of ways of filling in the gaps.
1 ) just replicate the preceding data.
2) replicate the data and add some noise or dither
3 ) join the dots with a straight line.
4) use a spline to create a more natural curve.
5) use a spline but fudge it using proprietory algorythms such as wadia's digimaster or legato link to try to mirror the music more accurately.

at the end of the day though they are all just guesses although soem can be surprisingly effective.
i would think that 192k src would have to at least use a straight line interpolation algorythm or things could sound funny.

cheers


julian

 

From CD it can never be more than the essence of 44.1, however it gets converted.

I'm extremely sceptical of 'upsampling', I don't see how it can better 'over sampling', and I can see many ways of screwing up the process. As usual though anything can happen with real world systems.

Paul

 

Ive heard 2upsampling cd players, the denon 1s at work and Roz's S.A.T fix. Must say, with rozzars cdp in my system, switching from normal to upsampling there was barely an audible difference. Id say the sound gained abit of colour to make the presentatin abit more full bodied but only by a tiny margin.

Ultra

 

FWIW I used a Monarchy 24/96 upsampler in between my transport (Teac T1) and DAC (Chord DAC64) for a while and I got rid of it. It could upsample to 48khz or 96khz and also increase the bit depth to 24 bits but however I had it configured it seemd to take some of the life out of the sound and make it sound a bit dull. It was a pretty subtle difference but it was a downgrade IMO so I sold it. Several DAC64 owners have found that it works best with a straight 16/44.1 input - not even the bells and whilstles of the ludicrously priced Chord Blu transport (176kHz 24 bits etc) could improve on good old 16/44.1

I'm with Paul on this one, upsampling (meaning doing a non-integer sample rate conversion) is unnecessarily messing with the data and usually detrimental.

Michael.

 

Of course, if they'd not invariably used vinyl-acetate masters to create their AAD releases, the harsh, and bass-light sound wouldn't have been so much of an issue either. Witness - exhibit A, the first 2 Howard Jones albums on CD. Treble so harsh you can cut glass with it.

I know that by 1985 (when the 2nd HJ CD came out) the hardware itself wasn't so bad in the harshness stakes - that old 1985 Technics SLP2 I have can even "do" Panterra without going brillo pad - something my 1991 Sony CD couldn't

As for early CDs all sounding bad - I'll race you my 1983 RCA copy of David Bowie's "Hunky dory" any day against the crap EMI remaster from 2001 - a music-free zone, that one!

Philips have been using oversampling since the CD101 in 1983

 

So what do you guys prefer, 44.1, 88.2, 96, 192?

I dunno, I've tried them all, 88.2 seems a fairly happy medium, I guess.

 

please reply it

 

From the Meridian web site (http://www.meridian-audio.com/p_588.htm) :

"High-quality 2-channel output using 192kHz 24-bit DACs"

That implies upsampling (to 192kHz 24bits).

Michael.

 

I'm a good ol' 44.1 man me, with a few :meister: enhancements though I have to say the Burmester 001 isn't bad at all thats 96Khz

 

Amir, I'd have thought you'd have got the idea from this thread that it doesn't really matter whether the Meridian uses Upsampling or Oversampling really. At least not in any way that would make me decide to either buy one or not based on this fact; just listen to one!