maxmercy Posted September 22, 2019 Share Posted September 22, 2019 So by request, the comparison of DVD vs BD LOTR Extended Edition. I remember graphing them a few years ago, but with the LF content thread request, I dug a little deeper. Here is the comparo between DVD and BD Fellowship of the Ring: The DVD has the green Peak and red Avg graph, the BluRay the cyan Peak and the green Avg graph. Stats for the DVD mix: DC offset -0.000046 Min level -0.663155 Max level 0.593566 Pk lev dB -3.57 RMS lev dB -32.86 RMS Pk dB -10.99 RMS Tr dB -119.97 Crest factor 29.15 Flat factor 0.00 Pk count 2 Bit-depth 24/24 Num samples 657M Length s 13686.741 Scale max 1.000000 Window s 0.125 The DVD clips in 12 locations across all 7 channels, mainly in the Right Surround channel. Stats for the BD mix: DC offset -0.000048 Min level -0.629489 Max level 0.594391 Pk lev dB -4.02 RMS lev dB -33.07 RMS Pk dB -11.22 RMS Tr dB -234.33 Crest factor 28.36 Flat factor 0.00 Pk count 2 Bit-depth 24/24 Num samples 657M Length s 13697.685 Scale max 1.000000 Window s 0.125 The BD clips in only 2 locations across all 7 channels. The tracks differ only about 1/2dB all around, including only 1/4dB difference in dynamics. They appear to be very similar, likely the same track save for some minor differences when putting the whole thing together. Given SME's prior remarks, and the BD's lack of clipping, I think the BD is the track to get since it is lossless and may contain more HF content compression may take away. Looking at every channel's PvA, it is quite obvious why this track is held in high regard for LF content. ALL the LCRS channels extend to nearly 5Hz. This track may be amenable to BEQ. The Two Towers and Return of the King as I have time. JSS 3 Quote Link to comment Share on other sites More sharing options...
SME Posted September 22, 2019 Share Posted September 22, 2019 I don't know for certain, but I believe most likely the "extra" treble content on the BD is synthesized using a plug-in. I believe such plug-ins are applied in the post-processing step of all kinds of audio, both music and movie to "sweeten" the sound. I believe they work by adding content at higher harmonics of the existing content, mostly above 10-12 kHz. They may also identify and artificially extend transients. It's basically a way to boost the perceived spectral balance at the top end without boosting background hiss and nasty microphone resonances. It often sounds better than the "real thing" that it takes the place of. On a reference quality system, the added treble is likely to improve the perceived sound quality, but it could make sound quality worse on systems that do not reproduce the top-end with adequate accuracy. Lots of speakers have problematic resonances way up high that could be excited by the extra content and impart harshness, graininess, haziness, or excess brightness to the sound. I personally had this problem on my system up until the last 6-9 months or so after I optimized my HF response very tightly. (Alas, it's still not tight enough.) Also listener differences may apply here too, given that it mostly concerns content at the very top. Though I would not recommend counting on hearing tests using sine wave signals to decide whether top-end content applies to you or not. I hear sine waves to 17 kHz, but I'm fairly certain I hear contribution of frequencies up to at least 22 kHz in actual content. I can think of a few reasons this may be true. Probably most important is that we what we hear depends on the result of a lot of high-level perceptual processing that is highly adapted to the properties of real-life sound and real-life acoustics. For example, a tone at some fundamental frequency is almost always accompanied by higher harmonic overtones. A pure tone at say 19 kHz is unlikely to occur in isolation in real life, so our brains are likely to judge such a tone as spurious and filter it out from perception. However if that same tone occurs simultaneously with sine components at 3800, 7600, 11400, and 15200 Hz (all harmonically related to the 3800 Hz fundamental), then the brain might deem that provisional 19 kHz stimulus to be reliable and might incorporate it into perception. The highest frequencies affect apparent brightness, which relates very closely to perceived depth/distance because higher frequencies naturally decay more quickly as the sound passes through air. FWIW, I strongly suspect the same thing happens with ULF. I suspect hearing thresholds for ULF may be lower in the presence of higher harmonic content. However, there is a major variable here (really no different from the UHF) in that perception is likely to depend a lot on how accurately the higher frequency bass is being reproduced. Being that room effects both alters the innate bass response of speakers/subs and confounds efforts to fix these errors, very few systems reproduce the higher bass frequencies accurately enough to fully benefit from the ULF. Likewise, very few if any HF transducers (without DSP optimization) can reproduce the upper mids and treble accurately enough to allow the listener to benefit from the UHF content. Sorry this got a bit long, and yes, BEQ for LOTR would be quite exciting! I'm hoping to soon implement what ought to be a major accuracy improvement in my under 40 Hz response, so will hopefully be able to make even better judgments about bass quality soon. Quote Link to comment Share on other sites More sharing options...
maxmercy Posted September 23, 2019 Author Share Posted September 23, 2019 Even on a modest home system, I remember wondering what was 'wrong' with the sound of Batman Begins on Bluray. It turns out the BD would default to the lossy soundtrack. I couldn't put my finger on it, just that the lossless track sounded 'airier' and more 'crisp'. Looking at HF content in DVD vs BD I see the following: RED trace is the BD. There is a transition in the 14-15kHz region. Not sure I can hear that, and I'm not sure if this is not just an artifact of lossy compression. Has anyone done any experiments on what AC3/DTS compression does to a spectrogram/FFT? JSS Quote Link to comment Share on other sites More sharing options...
SME Posted September 23, 2019 Share Posted September 23, 2019 That sure looks weird! (Both curves.) I have no idea how they got like that, but this is my best guess: They applied a steep LPF around 13-14 kHz to everything (in the dub-stage?) but then later added synthetic top-end content (for home release?). And the extra content doesn't start until ~16-17 kHz, leaving that weird notch in place. As for the differences, the blue curve clearly as a very steep filter applied before 20 kHz, but the BD was allowed to extend beyond. It's certainly possible that the filter on the blue curve came from the encoder itself, but that would probably have been applied as a "pre-encoding" step rather than as an artifact of the lossy compression. One reason to brick-wall LPF before encoding is to try to reduce the entropy (i.e. information that cannot be easily predicted by identifying redundancy) in a range that's (presumably) less audibly important so that the encoder has more bits to use for encoding the rest. It probably only helps marginally though as there isn't a whole lot of room between 20 and 24 kHz even with "white" / linear frequency weighting. Along similar lines: I was a very early adopter of MP3. A major factor in sound quality with MP3 (and probably most other lossy codings) is the quality of the encoder itself. The decoder just has to follow a recipe to turn the compressed bits back into sound. The encoder has to decide what to discard in order to give up bits. When I migrated to LAME, arguably the best encoder available, I found a nice guide suggesting options to use to get the best performance (i.e. sound quality vs. bitrate). The guide recommended using the LPF option to brick-wall everything above 19.5 kHz. It seemed perfectly reasonable, and I didn't notice anything "wrong" at the time. But that was long before I had any kind of decent speakers or any of the knowledge I have now, and finally now on this system I hear a characteristic resonance common to all those old MP3 files. Thankfully it's fairly inoffensive most of the time but it's definitely degrading. OTOH, I also have a lot of MP3s I downloaded, streamed (back when "streaming" was basically pirate Internet radio), or obtained from other people which are encoded at lower bit rates than I used (e.g. 128 kbit fixed vs. 192 kbit ABR). Those sound much worse, but I still listen to them because I like the music and can't easily replace them with lossless stuff. My old MP3s also still sound better than whatever YouTube is using these days. Quote Link to comment Share on other sites More sharing options...
peniku8 Posted September 23, 2019 Share Posted September 23, 2019 @SME Check out iZotope's Ozone Exciter. It's basically the plugin that you're describing. It simulates different types of tube, tape and other distortion. It's just the nice type of distortion which is called saturation nowadays. It is very powerful but also very resource hungry, a few instances of it in a mix already max out my 4 core Xeon CPU. I feel the same about mp3. People with inferior knowledge talk down on mp3 because they just know the early stages of Fraunhofer or have only heard 128kbps files. In fact, when doing it properly (with VBR lame, slow encoding) you can make 128k mp3 almost transparent on most speakers/files. I listen to metal mostly, which is basically the worst case scenario for any compression algorithm (broad spectral content, low crest factor), and all of the files on my phone are converted to 140k VBR with my favourite audio converter dbPoweramp. If you convert from a very good source, that will be transparent on any usual pair of headphones with a given ambient noise around you. I was only able to identify 2/3 of the wav vs 320k files in a DBT correctly, but I'm eager to try this again when I built my new speakers with Ribbon tweeters. Quote Link to comment Share on other sites More sharing options...
3ll3d00d Posted September 23, 2019 Share Posted September 23, 2019 7 hours ago, maxmercy said: Has anyone done any experiments on what AC3/DTS compression does to a spectrogram/FFT? Only at the low end (you lose content at low bit rates) Quote Link to comment Share on other sites More sharing options...
maxmercy Posted September 23, 2019 Author Share Posted September 23, 2019 Prelim BEQ for Fellowship: Lots of under 3Hz noise, hence the extra highpass down low, now I just need the time to screen it properly. These are long films. JSS 2 Quote Link to comment Share on other sites More sharing options...
maxmercy Posted September 27, 2019 Author Share Posted September 27, 2019 The Two Towers Extended Edition DVD vs BD: Again, virtually identical graphs. Clipping in this film is more prominent, especially in the LCR channels, both on DVD and BD, with more clipping in the DVD version. Stats for DVD: DC offset -0.000001 Min level -0.461359 Max level 0.464815 Pk lev dB -6.65 RMS lev dB -37.72 RMS Pk dB -14.70 RMS Tr dB -1.#J Crest factor 35.74 Flat factor 0.00 Pk count 2 Bit-depth 23/24 Num samples 678M Length s 14125.920 Scale max 1.000000 Window s 0.125 Stats for BD: DC offset -0.000000 Min level -0.521735 Max level 0.440078 Pk lev dB -5.65 RMS lev dB -37.58 RMS Pk dB -14.72 RMS Tr dB -143.65 Crest factor 39.49 Flat factor 0.00 Pk count 2 Bit-depth 24/24 Num samples 678M Length s 14131.499 Scale max 1.000000 Window s 0.125 Similar stats with slightly more dynamics on BD, but essentially the same track. This film had some strange filters applied to the LFE channel, and full bandwidth LCR channels, extending even deeper than FOTR. It is BEQ-able, with a possible pre-post: JSS 2 Quote Link to comment Share on other sites More sharing options...
maxmercy Posted September 29, 2019 Author Share Posted September 29, 2019 Return of the King Extended Edition DVD vs BD: Very similar again, with the BD having slightly better dynamics and more clipping noted on the DVD version. DVD Stats: DC offset -0.000001 Min level -0.541110 Max level 0.589154 Pk lev dB -4.60 RMS lev dB -35.16 RMS Pk dB -10.58 RMS Tr dB -190.65 Crest factor 33.76 Flat factor 0.00 Pk count 2 Bit-depth 24/24 Num samples 758M Length s 15794.069 Scale max 1.000000 Window s 0.125 BD Stats: DC offset -0.000000 Min level -0.630937 Max level 0.592822 Pk lev dB -4.00 RMS lev dB -34.94 RMS Pk dB -10.14 RMS Tr dB -468.38 Crest factor 35.23 Flat factor 0.00 Pk count 2 Bit-depth 24/24 Num samples 758M Length s 15797.877 Scale max 1.000000 Window s 0.125 Crest Factors listed on all the stats are simple ratios of RMS vs peak amplitude, they are not in dB. This film has even more full bandwidth LCRS than the previous two films, with significant infrasonics in the back center channel (DVD and BD), and little roll-off in the LCRS save for under 10Hz. Basically, the DVD and BD presentations of these films are essentially the same. No tampering or shelving like we saw in 'Master and Commander'. JSS 2 Quote Link to comment Share on other sites More sharing options...
SME Posted December 16, 2019 Share Posted December 16, 2019 Hi @maxmercy, I'd like to request that you post the BEQs you did. You can post them in the private forum instead if you want. My wife has requested that we screen these films over her Xmas break, and I figure this would be a great time to evaluate your candidate BEQs. I can give honest feedback including sharing any tweaks I make myself. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.