Personal Observations on Resolution, Sample Rate and Data Compression

I think this sums it up very well. I also have heard real A-B comparisons and it's indeed different at higher qualities. I normally describe it as more air around the instruments, more tonal aspects, more natural.

Thanks for the analysis Huge, but why no 24/96? Based on the 24/48 conclusion it is likely that there will be little or no perceptible difference between 192 & 96 and thus paying extra for 192 material is probably rarely justified.

Hi SamS,

I wasn't trying to test all commercial distribution formats.

The reason for using both 16/48 and 16/44.1 was to test for the effect of the interpolation in the time domain - 16/48 is an integer ratio of 24/192, 16/44.1 isn't.

At some time I may try comparisons to 24/96, 24/44.1 and 16/96 to complete the picture.

I also don't know of anyone distributing variable rate MP3 as quite a few players can't handle that format.

Interesting post Huge - I do find that i can tolerate lower rate files on certain types of music. In the end out of the different file types i have, mastering is the key. I've got some Red Book, that you'd swear was Hi-Res, it's that good. 

Thanks for posting your observations

James

James,

You're absolutely right, no technical encoding of the data can do anything to make up for poor sound engineering.  One of the reasons for picking that recording was that it's very well engineered as well as being a dynamic performance.  Using a good master (technically and musically) was essential to getting a valid set of lower order encodings to test.

If it's a choice between: a good performance and well recorded/engineered with red book encoding or a less good performance and/or not such a good recording/engineering technique but with high res encoding; then in all cases I'll take the better performance and recording/engineering over the technical capability of the container.

Where is the raw data from the linear potentiometer regarding the magnitude of your "jump?"  Absent that . . .

The dog left it in the oven and it got cooked!

Hi Huge, sounds like you were having fun .

you might want to try with different down samplers and down sampling algorithms.. You will almost certainly get differing subjective assessments.

As far as MP3 encoding, the LAME encoder was always the master encoder in my expierience and in the mid 2000s I encoded much of my then collection with LAME, and today those encodings sit head and shoulders above anything I have since downloaded at identical data rates and a country mile away from iTune's in built encoder when ripping CDs.. LAME did allow a huge amount of tuning to optimise the encoding for the music type and the stereo field type.. Again you can increase accuracy for a particular master file.. So it's worth playing with the settings.. the variable rate is also interesting as you can change thresholds and how aggressively the bandwidth changes... Again all this can make the end effect more accurate at the expense of data reduction.

Simon

Very good post Huge, thanks. I'm doing also some test with DSD down sampling because I can not use it on my Uniti but some materials I would listen will probably be available in this format.

With my surprise I get the best result without following the integer ration in the decimation.

I hope in future someone share his experince also in this complex transcoding process.

DBpoweramp is apparently one if the best resamplers out there. On both frequency, phase, passband, transition and impulse behaviour.

Mr Spoon just pointed to a new test of his 15.1 resampler on src.infinitewave.ca 

It behaves even slightly better than Adobe Audition.

So not a bad choice at all by Huge and not one to better with another one.

Aleg, yes dbpoweramp is good, but there are many resamplers or SRCs out here and most vary in slightly different ways, as ultimately down sampling is a lossy process.

http://src.infinitewave.ca

It compares most of the SRCs out here going from 96 to 44.1kHz..  So it's another case of choosing your poison to a large extent.. It is about a subjective assessment and personal preference...

Simon

Given that the Naim streamers (and CD players) upsample red book data in their hardware during playback and we (presumably) like the Naim sound signature, I think it unlikely that we'd find an SRC to upsample the data in a way we'd find substantially preferable to the Naim firmware.

However, having said that, there may be something that shifts the acoustic signature subtly, adjusting it to personal taste, so there may still be reason to look at this approach.  Are you offering to run the tests and publish your results?