I can confirm that - upsampling to 384khz and sample-aligning yields a null-out to only -21.4 db - but I don't see anywhere on the page that implies that the files are meant for performance evaluation; only compatibility evaluation.

Where's the beef?

There's something even more disturbing about this. The website is for Naim's music label. It is entirely reasonable to assume that these test files to check compatibility are extracts from the actual files offered for purchase.

And that means that Naim have deliberately nobbled the 16/44.1 versions in order to make the "hi-res" files sound better, thereby encouraging customers to pay a 70% price premium to get the "best" versions. This is cynical and, quite frankly, immoral.

Have Naim stated that the signal is the same apart from the sample rate?

Without such a statement, surely a rational person is going to expect the files to be processed to sound a bit different and with the more expensive ones sounding a bit "better" at least to the ears of the intended consumers.

Naim is an "audiophile" rather than a "high fidelity" company/brand. The products are designed and marketed to appeal to people that highly weight audiophile attributes and are prepared to pay a premium to own them (the premium price almost certainly being part of the attraction). In this case it is music with higher sample rates but the other products are similar.

There is nothing wrong with producing expensive luxury goods that only appeal to the subset of the population that have freely chosen to adopt audiophile beliefs in preference to other competing beliefs like, for example, scientific beliefs.

There is nothing wrong with producing expensive luxury goods that only appeal to the subset of the population that have freely chosen to adopt audiophile beliefs in preference to other competing beliefs like, for example, scientific beliefs. In fact, since the audiophile thing became part of the mainstream of home audio in the 1970s, UK companies like Linn and Naim have done well whereas the established "high fidelity" home audio companies pretty much all disappeared although most of the brands are still around. I may have no interest in purchasing the products of companies like Linn and Naim because of their poor value for money given my beliefs but I would judge the presence of these luxury goods companies to have been a good thing for the UK.

Quote from: cliveb on 2010-04-09 08:42:15

And that means that Naim have deliberately nobbled the 16/44.1 versions in order to make the "hi-res" files sound better, thereby encouraging customers to pay a 70% price premium to get the "best" versions. This is cynical and, quite frankly, immoral.

Have Naim stated that the signal is the same apart from the sample rate? Without such a statement, surely a rational person is going to expect the files to be processed to sound a bit different and with the more expensive ones sounding a bit "better" at least to the ears of the intended consumers.

Quote from: honestguv on 2010-04-09 12:14:32

There is nothing wrong with producing expensive luxury goods that only appeal to the subset of the population that have freely chosen to adopt audiophile beliefs in preference to other competing beliefs like, for example, scientific beliefs.

Frankly, that isn´t the point here. I think that no one holds a grudge against companies like Linn / Naim. They appeal - as you say - only to several people, and these people represent a minority. Also, Linn / Naim over the years produced relatively no-nonsense components (if one ignores this ridiculous expensive turntable from Linn) and from what I hear they are also quite reliable. May that be as it is, it looks different with Naim´s musical releases.

The key to this thread is responsibility. They have to know that naive audiophiles will take these files to do comparisons such as "Why 24/96 is better". By using what I would call "improper practices" they ridicule themselves & the audiophile community. They may have done it in the first place to sell more high resolution downloads but by simply using altered mixes for the other format they present themselves not in the best light IMO.

Cavaille, your second paragraph appears to contradict your first paragraph. I don't think honestguv's point should be dismissed. The audiophile industry works the same way as the fashion industry. They both are selling exclusivity and style more than substance. Consumers and designers in both cases may not fully aware of the game. Enough hedge and mystery is put into the pitch by the businesspeople to keep them clear of misrepresentation charges. Looks like Naim may have cut it a bit close to the misrepresentation line on this one.

Just for comparison:


Again, the 24/96 served as reference. This looks like a frequency difference - but it isn´t. I compared the CD-layer of one of my SACDs with the SACD layer of the same disc. The CD-layer was grabbed with EAC, the SACD layer was recorded analogue with my ASUS soundcard while it was played back by my Pioneer.

1. The CD-layer is 1.06 dB louder than the SACD layer

There are a few free downloads (see right hand side of front page) of complete tracks from various albums which are also available in multiple formats (mp3/cd/hi-res) - has anyone checked those?

Like cliveb, I'm curious whether this is a one-off (i.e. a single intentionally rigged comparison), or standard on all their releases!

Quote from: Cavaille on 2010-04-09 11:58:34

1. The CD-layer is 1.06 dB louder than the SACD layer

I don't see this in your difference graph.  Or was it something you determined by other means?  Since the SACD was captured as A/D, how did you compare overall levels of it to the digital rip?  Peak normalization followed by crest factor calculation?

Quote from: krabapple on 2010-04-09 16:27:09

Quote from: Cavaille on 2010-04-09 11:58:34

1. The CD-layer is 1.06 dB louder than the SACD layer

I don't see this in your difference graph.  Or was it something you determined by other means?  Since the SACD was captured as A/D, how did you compare overall levels of it to the digital rip?  Peak normalization followed by crest factor calculation?

Crest factor calculation? Sorry, I don´t know what that means - don´t forget, I lack some basic knowledge    You can´t see level differences with iZotope Ozone. Therefore I determined level differences with ReplayGain; then I corrected the volume with WaveLab (I lowered the volume of the CD-layer).

I determined the input levels before I started recording with some tracks from CD with the help of their EAC grabbed counterparts. As it turned out the analogue combination of Pioneer/E-MU produced level differences of around 0.05 dB (compared to the EAC rip).

I did the same test with the ASUS: it records approximately 3 dB louder - so in that case here I had to lower the volume of the SACD rip for around -3 dB. Only that way the ASUS recordings matched their EAC CD-counterpart. Since I pre-determined that -3 dB correction I used it on all my SACD recordings. And in this particular case it turned out that the CD layer was louder.

If I read what you wrote correctly, you first calibrated your Pioneer/E-MU recording setup by comparing peak levels of an A-D capture of a CD, to a rip of the CD, and found that your chosen peak recording level happened to give you a very close level match to the corresponding rip (0.05dB difference calculated by comparing Replaygain values).  Then you used a *different* setup (Pioneer/ASUS) to do the same thing, and found that  ASUS required a 3dB cut of your initial chosen input level, to get A/D capture to match the rip (again by comparing Replaygain values)  Then you extrapolated that number (-3dB) to DSD A/D capture as a correction factor (i.e., you lowered your ASUS input level by 3dB compared to your original input level).  Then you compared the  Replaygain value of the A/D recorded DSD track to the Replaygain value of its ripped CD layer counterpart and found that they differed by 1.06 dB.  Did I get that right?

Quote from: krabapple on 2010-04-09 19:07:47

If I read what you wrote correctly, you first calibrated your Pioneer/E-MU recording setup by comparing peak levels of an A-D capture of a CD, to a rip of the CD, and found that your chosen peak recording level happened to give you a very close level match to the corresponding rip (0.05dB difference calculated by comparing Replaygain values).  Then you used a *different* setup (Pioneer/ASUS) to do the same thing, and found that  ASUS required a 3dB cut of your initial chosen input level, to get A/D capture to match the rip (again by comparing Replaygain values)  Then you extrapolated that number (-3dB) to DSD A/D capture as a correction factor (i.e., you lowered your ASUS input level by 3dB compared to your original input level).  Then you compared the  Replaygain value of the A/D recorded DSD track to the Replaygain value of its ripped CD layer counterpart and found that they differed by 1.06 dB.  Did I get that right?

Almost   

What you describe as "my chosen recording level" was (in case of the E-MU) the default input-level which would be 0 dBFS. I didn´t attentuate because I didn´t need to. The levels were correct right out of the box.

The inputs of the ASUS recorded 3 dB louder with the same configuration (i.e. 0 dBFS). An example: I ripped a track from a CD with EAC (being digital). Then I recorded the same track with the Pioneer trough the ASUS. The latter was configured using the default configuration (0 dBFS for the input). Then I compared the digital rip (serving as reference) to the analogue one and found that the analogue recording done with my ASUS was too loud. I replaygained both files and the difference was roughly 3 dB (to be precise: 2.8 dB).

I continued all my recordings with the ASUS using the default configuration (0 dBFS again or 100 % inside the mixer panel). That means of course that all my recording done with the ASUS were roughly 3 dB too loud. When I had finished recording I converted the recordings to 32 bit floating point and THEN I lowered the volume by the pre-determined level for every recording.

This is why a difference or ratio-based number is better than simple peak level, as a comparison metric.  It scales. The  ratio of peak to average level for a given signal remains basically the same across the range of useful and 'legal' A-D recording levels.  Your method may suggest that a DSD rip of your track would really be 1 dB lower in peak amplitude level than the CD layer rip, but you can't know -- because you can't do a DSD rip.  But you can be pretty sure that the difference in crest factors would be the same regardless of method (assuming no compressors/limiters in the A/D chain.)