Linn Records upsampling

[Jim Collinson Linn Staff]

Hi Folks,

Writing to keep you all up to speed on the situation and let you know the latest developments.

We have been working closely with ARTS on this and I can confirm that the label’s titles are genuine 24/96 recordings and none have been up-sampled from CD.

We talked in some detail with their engineers before penning the original deal with the label, and understood their recording engineer’s techniques and workflow to be of the highest standard - these are indeed great recordings.

It appears, however, that when batch exporting their catalogue for delivery to our servers there has been some form of heavy low-pass filtering, or that the audio has met some piece of outboard that was only 44.1khz capable, and the high frequencies have been lost. This process was completed post-master by a member of staff at ARTS that is no longer with the company, so sadly we can get no more detail than that!

All their audio was delivered to us around two years ago and our systems, some of them automated, did not catch this discrepancy. These systems have since seen a significant overhaul and every album delivered, from whatever source, now goes through a full manual check in addition to our automated quality controls.

We are in the process of getting the entire ARTS catalogue delivered again. This will be done meticulously, and for each title the label are getting the original recording engineer on board to revisit the original sessions to ensure that the workflow, and purity of the audio is preserved.

This isn’t a speedy process but starting today, and over the coming months, we will be making the titles available once more, and contacting customers individually to make them aware and allow them to download again. These ARTS 24-bit recordings are great and sound even better with the full frequency range restored.

Many thanks for your patience and valuable feedback,

Jim - Linn Records

[rowlandhills]

Jim - What a great response! It's nice to know that this much attention is getting paid to this type of issue. Thanks.

[spoon]

(2010-11-11 00:31)linnrd Wrote:  

(2010-11-10 10:14)DonEscoba Wrote:  OK, I know now how to check the frequency spectrum (with Audacity or alike) and how to look if there is a sharp cutoff at 22khz.

But how do I check if the sample size of a music file has been 16 bit originally and has merely been padded to 24 bit with zero or random bits? Is there a reliable procedure to measure the effective sample size?

I seem to recall some converters that can be set up with an option to decrease the bit depth IF AND ONLY IF the 24-bit file is padded with zeros. I cannot off-hand recall which one(s) but if anyone else is familiar with one or more, it would be interesting to work with. Also, if it is possible to incorporate a converter that does this in dBpowerAmp then Spoon would be the person to chat with.

There is no easy way of checking bit depth. Frequency is this easy one, but bit depth beyond the simple check for zero padding is much more difficult, the reason being:

1) Take a 16 bit file of a certain frequency.
2) Convert to 24 bit file (zero padded)
3) Convert Frequency >> another Frequency

Now you have a file where the 24 bits is not zero padded and not filled with noise (because a frequency transform uses FFT). The audio would not have a dynamic range of 24 bit, but that is not to say a 24 bit recording would have a full 24 bit dynamic range which could be checked in any case.

[linnrd]

(2010-11-26 17:20)spoon Wrote:  

(2010-11-11 00:31)linnrd Wrote:  

(2010-11-10 10:14)DonEscoba Wrote:  OK, I know now how to check the frequency spectrum (with Audacity or alike) and how to look if there is a sharp cutoff at 22khz.

But how do I check if the sample size of a music file has been 16 bit originally and has merely been padded to 24 bit with zero or random bits? Is there a reliable procedure to measure the effective sample size?

I seem to recall some converters that can be set up with an option to decrease the bit depth IF AND ONLY IF the 24-bit file is padded with zeros. I cannot off-hand recall which one(s) but if anyone else is familiar with one or more, it would be interesting to work with. Also, if it is possible to incorporate a converter that does this in dBpowerAmp then Spoon would be the person to chat with.

There is no easy way of checking bit depth. Frequency is this easy one, but bit depth beyond the simple check for zero padding is much more difficult, the reason being:

1) Take a 16 bit file of a certain frequency.
2) Convert to 24 bit file (zero padded)
3) Convert Frequency >> another Frequency

Now you have a file where the 24 bits is not zero padded and not filled with noise (because a frequency transform uses FFT). The audio would not have a dynamic range of 24 bit, but that is not to say a 24 bit recording would have a full 24 bit dynamic range which could be checked in any case.

AAUUUGH!!! I have forgotten all my DSP from my undergrad days. I should've caught that one, Spoon. Thanks. I hang my head in shame.

[DonEscoba]

(2010-11-26 17:20)spoon Wrote:  There is no easy way of checking bit depth. Frequency is this easy one, but bit depth beyond the simple check for zero padding is much more difficult, the reason being:

1) Take a 16 bit file of a certain frequency.
2) Convert to 24 bit file (zero padded)
3) Convert Frequency >> another Frequency

Now you have a file where the 24 bits is not zero padded and not filled with noise (because a frequency transform uses FFT). The audio would not have a dynamic range of 24 bit, but that is not to say a 24 bit recording would have a full 24 bit dynamic range which could be checked in any case.

I understand. Thanks for the clarification.

[Robert_C]

(2010-11-26 17:20)spoon Wrote:  

(2010-11-11 00:31)linnrd Wrote:  

(2010-11-10 10:14)DonEscoba Wrote:  OK, I know now how to check the frequency spectrum (with Audacity or alike) and how to look if there is a sharp cutoff at 22khz.

But how do I check if the sample size of a music file has been 16 bit originally and has merely been padded to 24 bit with zero or random bits? Is there a reliable procedure to measure the effective sample size?

I seem to recall some converters that can be set up with an option to decrease the bit depth IF AND ONLY IF the 24-bit file is padded with zeros. I cannot off-hand recall which one(s) but if anyone else is familiar with one or more, it would be interesting to work with. Also, if it is possible to incorporate a converter that does this in dBpowerAmp then Spoon would be the person to chat with.

There is no easy way of checking bit depth. Frequency is this easy one, but bit depth beyond the simple check for zero padding is much more difficult, the reason being:

1) Take a 16 bit file of a certain frequency.
2) Convert to 24 bit file (zero padded)
3) Convert Frequency >> another Frequency

Now you have a file where the 24 bits is not zero padded and not filled with noise (because a frequency transform uses FFT). The audio would not have a dynamic range of 24 bit, but that is not to say a 24 bit recording would have a full 24 bit dynamic range which could be checked in any case.

You can end up with an vitrually indistinguishable 24 bit file in point 2 above, simply by interpolating samples rather than padding with zeros.

Very difficult to check that the dynamic range of a recording goes beyond the 16 bit limit of 96dB anyway - studios and concert halls just are not this quiet! Recording engineers probably gain ride the final result anyway - certainly on non classical recordings. The main benefit of 24 bit is probably lower distortion. 24 bit recordings seem to lose that granular, glassy sound that 16 bit cd can exhibit.