|
Post by Slinger on Apr 26, 2018 16:35:55 GMT
I've seen it mentioned and, as usual, there are two schools of thought #1 it sounds better and #2 it makes no difference. I wondered if anyone here had played around with whole number conversion ratios (44.1/88.2/176.4/etc) against the more often used 48/96/192 etc. rates.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Apr 26, 2018 16:54:04 GMT
You know what has always puzzled me. Why people mostly listen to old albums that were Recorded & Mastered in analogue find it important to have High resolution, or am i missing something?
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Apr 26, 2018 18:07:03 GMT
Direct multiples of red book deliver the best results imho
However IF the original file was recorded at a specific frequency then it is best played at its native recording rates
Using say 96/192/384 on red book or multipule variants of red book will incur the use of extra number crunching and the algorithm used to generate that particular frequency range will have an effect not always a beneficial one imho
We record at 24/96 / 24/176.4 and 32/352.8
Personally the bit rate has more of an effect than the sample rate does adding random bits of dither really work or is there more to it?
If you have a pass through function which will allow native recording rates to play naturally you may just find it is more positive
Also using an upsampling chipset / algorithm also introduces another reclocking section like a PLL which is usually incorporated in the chip set which maybe giving the results some have experienced
Perhaps we could run a digital session for those interested in making the most of the 1's and 0's at some point
|
|
|
Post by John on Apr 26, 2018 19:25:21 GMT
Perhaps we could run a digital session for those interested in making the most of the 1's and 0's at some point I like the thought
|
|
|
Post by MartinT on Apr 26, 2018 19:29:08 GMT
That would be interesting.
For me, it's not whole numbers but multiples. I find that CD rips (44.1kHz) upsample better to 24/88.2 while Spotify (48kHz) upsample better to 24/96.
|
|
|
Post by Slinger on Apr 26, 2018 19:34:53 GMT
You know what has always puzzled me. Why people mostly listen to old albums that were Recorded & Mastered in analogue find it important to have High resolution, or am i missing something? I think it partly touches on what Fluxed in Time is saying, in so much as it depends what quality they were re-recorded/remastered at rather than the original quality. If it was remastered at a decent quality then you want to be able to play it back at that quality otherwise you'll be degrading it. I know what I want to say, but it's difficult to simplify. Say something is recorded at (using an arbitrary number) what we would, today, class as a quality of 5, in 1963. It gets a straight-forward digital remaster at a modern-day quality of 9 in 2018. While the '9' doesn't actually make it sound any better, playing it back at less than 9 will make it sound worse. Does that make any sense at all? No? I'll get me coat.
|
|
|
Post by MartinT on Apr 26, 2018 19:38:56 GMT
Why people mostly listen to old albums that were Recorded & Mastered in analogue find it important to have High resolution, or am i missing something? Analogue is inherently high resolution.
|
|
|
Post by Slinger on Apr 26, 2018 19:44:22 GMT
Direct multiples of red book deliver the best results imho However IF the original file was recorded at a specific frequency then it is best played at its native recording rates Using say 96/192/384 on red book or multipule variants of red book will incur the use of extra number crunching and the algorithm used to generate that particular frequency range will have an effect not always a beneficial one imho We record at 24/96 / 24/176.4 and 32/352.8 Personally the bit rate has more of an effect than the sample rate does adding random bits of dither really work or is there more to it? If you have a pass through function which will allow native recording rates to play naturally you may just find it is more positive Also using an upsampling chipset / algorithm also introduces another reclocking section like a PLL which is usually incorporated in the chip set which maybe giving the results some have experienced Perhaps we could run a digital session for those interested in making the most of the 1's and 0's at some point Many thanks, that's just the sort of opinion I was looking for. I'm currently upsampling (or downsampling) everything to 24/96. Changing it means buggering about with the actual Roon Audio code so before I do start changing things (again) I need to make sure it's worth it. What I may try next is to leave the sample rate at 96 and set the bit depth to default to that of the original recording(s). Perhaps you could expand a little on what you mean by a "digital session" because it sounds interesting.
|
|
|
Post by MartinT on Apr 26, 2018 19:49:18 GMT
You'd be better off raising the bit depth to allow the interpolated point to be more accurately positioned and allow for dither to hide quantisation noise.
|
|
|
Post by Slinger on Apr 26, 2018 20:02:31 GMT
I thought you'd settled on 24/96, Martin. Have you gone up to 32 or am I misremembering?
|
|
|
Post by MartinT on Apr 26, 2018 21:13:44 GMT
No - I was thinking you meant staying with 16-bit. 24-bit is good
|
|
|
Post by Slinger on Apr 26, 2018 21:55:13 GMT
The idea I had was to keep the sample rate as I have it now, resampled up or down to 96khz, but allow the bitrate to be set by the audio file. Whichever way I do it will be a trade-off of some sort but that way 16 bit will play at 16 bit but 24 bit and above will play at whatever their native bitrate is. What is really needed is a way to set a minimum upsampling rate so, for instance, everything would be AT LEAST 24 bit. It needs some "IF THEN" statements of some sort.
IF bitrate <24 THEN bitrate = 24 AND samplerate=96 IF bitrate >=24 THEN bitrate == bitrate AND samplerate=96
Something like that, anyway.
|
|
|
Post by Stratmangler on Apr 26, 2018 22:02:52 GMT
It's not bitrate. The 16 or 24 refers to bit depth. The bitrate of an uncompressed PCM CD rip is 1411kbps.
|
|
|
Post by Slinger on Apr 26, 2018 22:08:31 GMT
It's not bitrate. The 16 or 24 refers to bit depth. The bitrate of an uncompressed PCM CD rip is 1411kbps. You're quite right, of course. DOH! Bit depth is what I meant. IF bitdepth <24 THEN bitdepth = 24 AND samplerate =96 IF bitdepth >=24 THEN bitdepth == bitdepth AND samplerate =96 ...makes a bit more sense. Of course, it would mean parsing every file before it was played... Sorry, still thinking out loud.
|
|
|
Post by MartinT on Apr 27, 2018 5:14:20 GMT
IF bitdepth <24 THEN bitdepth = 24 AND samplerate =96 Yes. However, I still think... IF bitdepth <24 THEN bitdepth = 24 AND samplerate =88.2...if the file is a CD rip at 44.1kHz
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Apr 27, 2018 8:10:26 GMT
The way to simplify matters here:- Red book standard is 16 bits 44.1Khz. The name Red Book was given to CD simply as the concept was written in red book nothing more! Let us look at the red book standard a bit rate/depth of 16 so liken it to a box 16 by 16 and inside that 1 second box you have 44100 samples in that time frame which equates to 1.35Mbit per second this gives you 1411 kbps in total which is CD (red book) standard (whether you play a CD or rip one at flac or wav) Now in the studio (remembering way back in time to the mid 80's) the 48Khz rate was used to record and a lot of studio's still use this format although they record at 24 bits. The 16/48Khz studio rate has 1.46Mbit per second which in turn generates 1536 kbps Now if we look at high resolution files or DVD Audio which has a bit rate/depth of 24 which translates into a box 24 by 24 although this time you are looking at 96000 samples in that 1 second time span this gives 4.39M bit per second which in turn produces 4608kpbs in total. To put this in to context comparing a red book file to a 24/96khz file (providing that all things being equal here!) the 24/96Khz files takes up over three times the amount of data required to make the same file in red book although it has 250 times the resolution of red book. 24/96 also has a theoretical S/N ratio of 144dB (human hearing between 128-132dB) My personal favorite recording rate is 24/176.4Khz (5.4Mbits per second) total rate 8467 kbps (direct multiple of red book) Now bit rates / depth there are a lot of DAW (digital audio workstations in studios) and some dac's (up-sample to 32bit rates) that use floating point this is to allow digital manipulation after the recording process has taken place to allow headroom and fine tuning of tracks (helps with any clipping issues, possible noise introduced during the recording process and sorting out any rounding errors made during the DSP) without loss of quality this is NOT true 32 bit recording. Genuine 32 bit depth has theoretical S/N ratio of 1680dB way, way beyond anything we could possibly interpret me feels If you use a algorithm to up-sample or an SRC (sample rate converter) you will induce more changes in the unit due to extra current being drawn to calculate the math and or use more function in the SRC chipset (as stated previously some SRC's have extra clocking facilities in the form of PLL etc) this may be the reason the quality is perceived to improve as the relock maybe the cause as opposed to the sample rate conversion imho. With regard to the Digital workshop we could organise an information day perhaps Martin or one of the other mod's could look into to this happy to set up a demonstration for you guy's and possible we could video it for those who could not attend?
|
|
|
Post by MartinT on Apr 27, 2018 9:39:05 GMT
Sounds a great idea, Tony, and happy to help organise it.
|
|
|
Post by Slinger on Apr 27, 2018 12:02:06 GMT
IF bitdepth <24 THEN bitdepth = 24 AND samplerate =96 Yes. However, I still think... IF bitdepth <24 THEN bitdepth = 24 AND samplerate =88.2...if the file is a CD rip at 44.1kHz Obviously, I was only generalising. Ideally, you'd need the ability to match different bit depths and sample rates.
|
|
|
Post by Slinger on Apr 27, 2018 14:08:21 GMT
Something along these lines perhaps...
IF (file-bitdepth <24 ) {file-bitdepth = 24 AND file-samplerate =file-samplerate * m }ELSE IF (file-bitdepth >=24 ) {file-bitdepth == file-bitdepth AND file-samplerate =file-samplerate * m }where "m" would be set by choosing a value from a numeric array of (for instance) 1 to 8. Anyway, I think that's quite enough of me trying to remember how to write code and mixing languages.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Apr 27, 2018 16:58:46 GMT
I had to go lay down after reading Tony post You know the best sounding CD to my ears were the HDCD CD's Im not sure of the Technical side but were they not like 20Bit info on a 16Bit? I once compared the Japanese issued King Crimson & Yes i had on both HDCD & later 24 Bit versions
|
|