The Classical Shop - FLAC clipping !!!

[johnb]

I was very impressed with last night's Petrenko's Shostakovich 10 and, this morning, bought the FLAC download of the Naxos recording from The Classical Shop (which is part of Chandos).

Out of curiosity I looked at the download in Sound Forge, thinking I might compare the dynamic range with the iPlayer HD. (Yes, I know I should get out more!)

I was startled to discover there was severe clipping on all the climaxes (and I do mean 'clipping' not limiting), probably caused by gain having been applied during the encoding process.

It would be interesting to learn whether this was just an anomaly or whether it is common with FLAC downloads from The Classical Shop.




And this shows where Sound Forge detected clipping (the orange verticals):

[PJPJ]

I have quite a few Chandos recordings as flac which I got from theclassicalshop and none has suffered from this (I checked them all) - I suspect they uploaded what Naxos supplied them. Not that that helps, and were this to happen to me, I'd ask for a refund.

[Gordon]

I have a few Classical Shop downloads of orchestral in FLAC and none of the movements that are of Chandos origin have any sign of clipping. Their plots look well behaved and the meter peaks just below 0dB on the software I use.

Now that doesn't mean much because we don't know how this 0 dB relates to the original A/D. Is it reasonable to assume that because it is software that it would be calibrated to be full scale of the bit range?? I would expect a healthy headroom allowance for A/D conversion and that means that subsequent FLAC coding [or anything else] should not be able to do anything bad but, as you suggest, one can change gain digitally of course if you accept a bit of round off [for small ratios - doubling is trivial to do but you may have to drop a bit] depending on your DAC. Your quiet bits don't look particularly enlarged to me.

Depends where they got the wav file [or AES or whatever] from in the first place? Can't see a master being clipped somehow.

Might be a bit of one off finger trouble somewhere. Can you hear it now that you know it's there?

[OldTechie]

I have some and they are variable. I have been trying to decide whether 24bit 96k is worth the extra money. I can't see that I should be able to hear the difference in the sample rate with my old ears with an 11 kHz cut-off. The difference between 24 bit and 16 bit should only be audible on quiet passages. So I started with a view that I should not benefit from the higher rate. Maybe I'm fooling myself, but the higher rate versions sound better on the ones I have and the difference is much more than I expected.

I have downloaded their CD quality versions and I think they come from the orginal publisher, because the levels do not match the 24/96 versions, and the relationship is not consistent. I only have three such to compare and I'm trying to devise a satisfactory test method. (If, having fixed the levels, I run an ABX test inside foobar2000, I can't reliably tell the difference after the first four or five comparisons. But if I listen to the music rather than the technical quality, I think I can. I bought a Mahler 4 and I'm stunned by the sound on the 24/96, and the 24/96 is slightly higher level than the 16/44.1 which just sounds very good.)

One is a Pentatone original. The CD quality version is 3.1db higher in level than the 24/96 version. It was the first one I tried. I started listening to the CD version whilst the 24/96 was downloading, and soon decided I was not too happy with it. It just seemed to be a slightly odd sound. However, I did not think it was clipping that disturbed me. I was so unhappy with it that I requested a WAV download instead of the WMV - but I could tell no difference between them. When I had the 24/96 I was surprised that it seemed very much better. The CD quality version is definitely clipped in places (not limited or compressed) but I don't like the sound in the parts where it is not clipped. It's some Russian violin concertos, and the violin solo sound quality displeases me to the extent that I don't really want to listen to it on the CD download, but somehow I can accept it on the 24/96 version. I have a suspicion that converting a surround sound original to stereo may result in a second-rate result. (Or maybe the slew-rate limit on a DSD original is audible - but again with my old ears it should not be.)

But here are some waveforms to show that it is the CD version that is clipped:

Top pair is the 24/96 download
Centre pair is the 16/44.1 download
Bottom pair is the 16/44.1 download attenuted 3dB and upsampled to 24/96 for comparison with top one on same scale.

The clipping instance is on the left upper track on a negatve going peak. This is one of 7 clipping instances in the first track of the download.




So, jb, was yours the CD quality download? If so, I suspect it is identical to the physical CD, and the 24/96 might be better.

But there again it may be the other way round. My Mahler has a short section where the waveform looks clipped even though it does not hit the digital peak. But it is the 24/96 that looks clipped, and the CD quality looks OK.

24/96:



16/44.1:



I wish commercial pressure would allow a few dB headroom on the final product, though I can't say I had noticed it on the Mahler until I checked the waveforms on a section where all the peaks were at an identical level. Maybe we all get accustomed to a bit of clipping on some peaks!

[Ariosto]

We should not accept any clipping in digital as it is a horrible sound, unlike the analogue days when a bit of waveform clipping was OK.

Good digital recordings should always have at least 6dB of headroom (-6dB), and many advise - 10dB, and some recommend -20dB. Even at minus 20dB you still have a signal to noise ratio better than CD and approching 120dB. (CD being 90dB if you are lucky - normally about 80dB).

It seems that the producers of classical CD's and downloads are getting the same disease as the pop world CD's where they push up the level to 0db and compress like crazy. It's known as the loudness wars.

[johnb]

On Friday I complained to The Classical Shop and they acted remarkably quickly. By Saturday morning they had asked Nexus to re-rip the CD, made the re-ripped FLAC available on their website and replied to me.

First of all please accept our apologies for the quality of this album. We do not encode the files here, only change their formats to offer Mp3, AIF, WAV and WMA. The files come to us in FLAC form from Naxos. As we have over 1.2 million files on offer you can understand that we cannot possibly audition every one!
We trust our suppliers to offer their best quality to us.
Anyway, I have asked Naxos for a new rip of this disc and I can now tell you that it is on site and ready for you to redownload. Please log back in and go to your order history to find the links to download again.

Very impressive.

The sad thing is that the re-ripped version seems identical to the original.

I'm tempted to buy the CD version out of sheer curiosity, to test whether the actual CD has the same clipping.

Gordon,
The clipping doesn't appear to be markedly noticeable to my ears. Though that might be because the climaxes are frequently dominated by the rasping brass.

[johnb]

Just an update to this thread:

Since downloading the flac files from The Classical Shop I've got hold of the actual CD.

The files ripped from the CD appear to be identical to the flac files - clipping and all. Make of that what you will.

[Gordon]

On Friday I complained to The Classical Shop and they acted remarkably quickly. By Saturday morning they had asked Nexus to re-rip the CD, made the re-ripped FLAC available on their website and replied to me.

Glad to see that someone is on the ball!! Good service even if it didn't solve the actual problem which looks as if it's with Naxos.

Gordon,
The clipping doesn't appear to be markedly noticeable to my ears. Though that might be because the climaxes are frequently dominated by the rasping brass.

Being pedantic, if you can't really hear anything nasty except congested sound at the peaks, what you have here is hard limiting so there won't be any severe harmonic distortion as you'd get with clipping which is where the channel cannot carry peaks so does what it can. In an analogue channel, particularly one with valves in it, it would soft limit progressively into hard limit and then clipping as the amplitude of the signal increases. Digital channels do not naturally have a progressive limit but a sudden and sharp clip where they run out of bits and this will certainly produce harsh distortion if done to excess.

Whilst I can undertsand a cautious engineer dealing with events installing a limiter why a CD producer would do it escapes me. Was it possibly a live recording from a concert taken from a radio station?