Except those pressed into NOS services, of course.

To Shadders: please believe us. Whether called 'up' or 'over', it really is all the same,
and you always need a filter. When going up the filter has to be at half the source
rate, to avoid images being created, and when going down it has to be at half the
target rate to avoid aliases being created.

The reason that replay without oversampling ('NOS') works, despite violating the book,
is that images, in an ideal world, are innocuous, and when above 20kHz inaudible to humans.
Many amplifiers and loudspeakers are 'ideal' enough to make this a useful (minority) proposition
in domestic audio.

Aliases, on the other hand, are not to be tolerated.

Well, almost always ...


The distinction between integer and non-integer ratio sample rate conversion is not necessary:
again, both are mathematically the same stuff, except that implementing the latter is
a bit harder and thus more likely to be screwed up by a designer who is signal-theory-illiterate (and
trust me, there are many of those!).

It all depends on the point of view. Thinking as a mathemagician, or in the time domain, one may name it
'interpolation', whereas speaking as an signal theory expert, or in the frequency domain, one may name it
'filtering'.

After all, in any linear system any act of interpolation has a direct filtering action, and vice
versa, any act of low-pass filtering has an interpolating action. This is just one of the
many faces of the time-frequency duality in Nature.




Any distinction between oversampling and upsampling, in our context here (it differs in other contexts!), is entirely
artificial and was probably created by marketing forces.

In the 80s the CD player DACs that followed Philips' solution were called 'oversampling', because that term
really neatly described the process of running the DAC section at N x the 44.1kHz source rate while
using digital filtering for the signal reconstruction (ah 'reconstruction', another synonym to 'interpolation'
or 'low-pass filtering' here!).

(IRONICALLY the very oldest internal technical documentation of Philips describing
their dino-era combination of 14 bit DAC and filter named it ... 'upsampling'. Really.)

A while later hardware-based sample rate convertors appeared on the market, obviously intended
as translators between the most used standard rates in studios at the time: 44.1kHz and 48kHz.
These had to tackle non-integer ratio conversion on the fly, and as they were going up or
down, concomittant anti-alias or anti-imaging filtering. As this is pretty hairy to do (but not in
concept), quality was probable rather low.
A bit further down the line DVD appeared, along with rumours of DVD-A, and suddenly
production houses had a need for translation between any combination of 44.1, 48,
88.2, 96, 176.4, 192. So SRCs appeared that went up to 192kHz.

These then were adopted by the audiophool market, smacked in front
of a regular DAC (which until then happily oversampled CD to 176.4kHz or
352.8kHz) and configured for 192kHz output. Magic numbers appeared on
fascias and some bright spark called it 'upsampling' to distinguish it from
'oversampling'. A whole subculture emerged, bent on planting into
the global consciousness that 'up' and 'over' where totally different things,
and that the newest 'up' was sooo much betterish than the olde and crude
'over'.

Technically this was a major step back, of course, but another extremely sad
result was the massive pollution of related 'information' on the internet
and in audio publications, where ignorance of basic signal theory has been elevated to a religion.



Sad, as on a fundamental level all of this really quite simple.

Hi Richard,



The source in the example we discussed was an original 44.1k recording. As far as I know, the converters I use (ULN-8 from Metric Halo) do use filters. (I don't know that proper A-D conversion can occur without them.) Or did I misunderstand the question?

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

The crux of the sampling theorem is that subjecting the discrete samples to a specific kind of filter draws a line through
the samples (interpolates) in such a way that the original pre-sampled signal is recovered. This is the reconstruction process,
and it applies whenever the sample rate is increased (of which the conversion to the continuous time 'analogue'
domain is just one limit case).

Yes.

But that does not contradict the other.

Surely you can explain, or at least hypothesise why?

In the grand scheme of things,

not digital audio in particular.

Edited by user 09 March 2012 07:47:20(UTC)  | Reason: Not specified

I was talking to my friend Rich yesterday and he had a chance to make a comparison few of us have a chance to. Another friend is a mastering engineer with very good equipment and trained ears. He brought over 4 different versions of the same performance; 1. the high resolution original master 2. his remastering of it, also in high resolution, 3. the commercial CD. 4. the commercial LP. These were played back on his system, which consists of Quad 57s he got from me in the 70s [ rebuilt at least twice] , a Heathkit tube amp completely rebuilt, custom tube/hybrid line stage and an LP12 with Sumiko arm with good MC cartridge. CD was his Marantz 8000 series player driving a professional DAC brought over by the recording engineer. I have forgotten the name; it was one I was not familiar with. All I remember was that it lists for $2500 here and that both the engineer and my friend thought it was the best they had heard. I will find out the name. To shorten an already too long account all sounded different but the LP was the most unlike the original recording. This bothered Rich as he is a mostly vinyl man despite being into computer audio. It made him wonder if vinyl is doing something to make recordings more listenable at the expense of accuracy [ Ken will like that]. I wasn't there[ they are 800 miles away] so I didn't hear it but it does not really surprise me that a digital recording would sound more like a digital master than an analog recording made from that master. My favorite LPs are either direct cut or made from analog masters. Digital can sound very good and the Metrum has made some of my CDs sound a lot better; I just wish recordings in general were made to a higher standard. When you hear Soundkeeper or Yarlung CDs compared to the general run you can see what we are missing.

Why go through all that trouble. Just use a analogue master.

Absolutely, this is why it is so difficult to compare CD to Vinyl the chances of having like for like are minimal

Hi kengale,



While the preview function certainly makes life easier, the best cutting engineers would anticipate dynamic changes in the program and manually adjust the groove pitch to accommodate peaks, without manipulating (i.e. squeezing) the sonics.

George Piros was a master of this. I spent many an hour in his cutting room with him and very much enjoyed watching him ply his craft. And many an hour enjoying the Life he preserved in the records he cut.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

Yes.

Either a second head stack with a small tape bin between it
and the first stack> The arrangement required twice the
maintenance and alignment effort, and two racks of processing
gear between the tape machine and the cutting electronics.



But once digital delay lines became available you can
imagine how quickly these became popular with beancounters
and studio techs alike ...