Issues with Large Files

[Budge]

On 23/01/17 17:24, RS wrote:
>> From: Budge
>
>> Sent: Monday, January 23, 2017 11:45
>
>> You will have to ask somebody who knows about these things but when I
>> chased Linn this is relevant part of the reply:-
>
>> "The issue is heavily influenced by but not directly related to track
>> duration. (Possibly too much technical detail - it's caused by the
>> number of entries in the 'stsz' table. Problem files have huge numbers
>> of entries, often with repeated small values; the mpeg spec allows but
>> doesn't encourage encoders to combine these blocks together.)"
>
> This page describes stsz.
> https://wiki.multimedia.cx/index.php/QuickTime_container#stsz
> There are two modes, uniform sample size and different sample size.  I
> assume the sample size is the number of bits to encode the level, namely
> 16. If there is a large table it must be because the different sample
> size flag is set.  For a 9200s file the number of sample sizes would be
> 441,600,000. Unless the sampling frequency deviates from 48kHz, the
> table size is directly related to track  duration.  I do not know why
> the different sample size flag should be set; maybe it is the presence
> of the thumbnail.  You could try running it through ffmpeg with -vn
> -acodec copy.
>
> Interestingly the Linn amplifiers support 24 bit encoding and a sampling
> frequency of 192 kHz, which would make the stsz table 4 times as large.
> (As the article you linked to said, 192kHz sampling is pointless or even
> counter-productive, and there little to be gained by 24 bit encoding for
> a production version.)
>
> I did ask whether the number of samples per frame was always 1024.  It
> seems it can also be 960 but 1024 is more common.
>
> I thought you had given the name of the opera your Linn player would not
> play, but going back through the thread all I could find was that it was
> 9200s, so it is fairly short.  Der Rosenkavalier is just under 5 hours.
> I downloaded it from dashhigh1 and played it on my Triax satellite
> receiver, which cost a tiny fraction of the cheapest Linn, and it played
> all through. I skipped through it, but it played to the end without
> problem, as have several shorter operas.
>

Again see my earlier posts.  I am convinced I am experiencing two 
different problems, corruption with profile which is occurring on short 
files and the long file Linn problem.  I did say they all play on my RPi 
without any problems.

The article to which I linked identified some technical benefits from 24 
bit during record production but explains clearly why not relevant at 
the listening end. Linn, however, from the days of vinyl lps, made and 
continue to make their money from the Golden Eared brigade who tend to 
believe that more must be better, whether bits or cash.  I have four 
Linn devices purchased because they worked reasonably reliably at a time 
when all similar kit used different proprietary standards but not as 
effectively or reliably. Unfortunately Linn is also all proprietary but 
the hardware works and is OK using apps like BubbleDS Next and 
BubbleUPnP for control points.  (Linn apps are inferior) I doubt, 
however, I shall buy any more Linn devices.