Jump to content

John Dyson

  • Content Count

    90
  • Joined

  • Last visited

About John Dyson

  • Rank
    Freshman Member

Recent Profile Visitors

The recent visitors block is disabled and is not being shown to other users.

  1. John Dyson

    Lies about vinyl vs digital

    A significant component of that warmth is IMD (intermodulation distortion.) IMD can either sound 'harsh' or 'mellow' depending on certain mathematical attributes. Tape compression tends to have the 'mellow' attribute. The tape compression tends not to be too awful until the IMD starts becoming more apparent as a kind of distortion (e.g. birdies are an extreme case.) Analog tape requires some expertise to use to its maximum quality -- recording at a moderately high level to end up well above the noise, but not so hot as to cause significant distortion that sounds like distortion :-). It isn't a really difficult problem unless using tape professionally and trying to get the absolute maximum quality when dealing with multiple generations. Using tape at speeds of 15ips and 30ips help to allow using hotter signal levels, get above the hiss, and if possible -- try to avoid noise reduction systems. 30ips has other troubles, but if you really, really want to avoid NR -- then 30ips will be your best bet. Every NR that I know something about (I'll be learning a LOT more about C4 soon -- so I don't know much about that) produces significant distortion or has problems with transients.
  2. John Dyson

    Still better SQ from CD.... Help..

    I agree... Something similar happens when streaming from the internet. If this kind of thing (proper buffering) wasn't true, it would be impossible to listen to material directly over the internet without jitter so bad that it would be horridly unlistenable. All of the timing issues from the 'disk' are handled as you said, and the only real 'jitter' issue is that of the final clock and final D/A that resolves the digital into analog. There are numerous mechanisms for measurements that imply jitter from the final clock and D/A but not all of those measurements really measure jitter. Much of the apparent transmission from other parts of the CD player result from ground bounce, capacitive coupling (not really much) and radiation (probably measurable, but not much.) Ground bounce (that is, not everything in the circuit grounded to RF standards, and the various resistances and inductances interacting with varying currents can cause noise in the output.) Some of the noise might also affect the internal oscillator, and even the 'improved' results from using an external timebase can stil manifest jitter because of the sampling point and ground issues as I noted above. Again -- I agree with you 100%, and there are all kinds of reasons for you statement to be true. However, it is also true that better designed devices can produce a better audio signal if internal clocking, grounding and adequate shielding is done. Having SOME RF background would probably avoid the seduction of the relatively low 'frequencies' in audio equipement. Gotta remember that the clocks have hard edges, lots of HF content, and can cause serious grounding issues if not carefully dealt with. John
  3. John Dyson

    Still better SQ from CD.... Help..

    Contrary to common audiophile knowledge, any common and recent CD player has the rotation timing decoupled from the output clocking. All of the necessary queueing and resyncing happens such that the exact bits that are on the CD (modulo proper decoding of the data on the CD -- which is pristine if the CD has few errors), so the only thing that you need to really worry about is proper queueing without missing data, and the last clock in the chain (converting to analog) not having noise (including ground conducted noise.) If this wasnt' true, how could you POSSIBLY listen to streamed data over the internet with the sometimes seconds of data delays? The answer is effectively data queueing and resynch -- similar as what happens on a CD/DVD/etc. Guraranteed that the clock noise (jicoming from the originating site is fully decoupled, and the only noise that you need to worry about is from local sources due to various kinds of electronic/radiation/ coupling. The 'jitter' is more importantly caused by ground bouncing, and not any kind of conduction due to D-flip flop windows. Given this -- ONLY THE LAST CLOCK IN THE CHAIN CAUSES THE REAL JITTER. So, the measurements that talk about all kinds of clocking noise reduction -- need to look deeper and more fundamentally than the clock edge on D flip flops -- even though SOME noise can pass through, but it diminishes very quickly and the window is so small that the noise is attenuated. The noise really comes from radiation (from strong, high current surges causing radiation, and MORE IMPORTANTLY, ground bounces from the current flow directly with an current/inductance thing.) The biggest source of the noise is ground conduction and to some extent radiation to/from the local cicuitry. THAT is what is measured when looking for clocking problems. (I know -- I design all kinds of things in that arena.) PLEASE WATCH FOR SNAKE OIL as there is a lot of money to be made from dyseducating people... The acutal jitter comes from that last clock -- other sources are actually (more than likely) ground noise conduction. Remember the effective 100's of msec of 'jitter' over the internet -- of course, if you cannot listen to the internet stuff because of that 100s msec of jitter due to propagation issues -- there is something really wrong. John
  4. John Dyson

    Compress FLAC flies for mobile

    I think that going to flac is a good thing -- but unless I use a set of esoteric options on flac (to maximize compression), the compression doesn't seem to take very long. The main reason for staying in the 96k realm is for pro purposes. There is a tendancy to stay in multiple-of-96k relm (sometimes 48k, but that isn't optimal for signal processing.) -- and that is where I like to stay. Once there is ONE conversion, then the there is little other conversion loss. I have zero religion to stay at the standard rates because my audio processing program actually runs optimally at about 64k-72k :-). However, for interchange (again for pro applications, often the source material is 96k/192k), I tend to convert everything to the 96k multiple/submultible realm. For listening reasons, there is little/no reason to convert TO 88.2k, but in a pinch, there is nothing wrong with using 88.2k for processing, but it seems to be relatively less common in the profesisonal circles that I have been dealing with. John
  5. John Dyson

    Compress FLAC flies for mobile

    If the player supports it, I dont' accept anything but flac or opus -- there is a significant issue with time resolution where material with overlaid/delayed vocals are smeared badly (to the point of being of multiple/overlaid vocals/instrucments being indistinguishable) using lame/mp3 -- even at 320k. At 256k, opus retains some of the seperation, isn't quite as smeary, but isn't quite perfect either for my hearing - generally better than mp3. Flac (or other non-lossy format -- I am not religious about lossless compressed formats) is the gold standard, but in a portable/non-professinoal situation - the very best lossy compressed format is okay with me. For listening only, even 16bit flac is okay & superior to mp3 or opus, and 48kHz i(instead of 44.1kHz) is more important than 24bits. The cost of 96k vs 48k isn't usually very severe using flac. 24bits instead 16bits tends to be more costly, and for listening only makes a subaudible difference in noise. I don't have a negative feeling about AAC, even though I havent played with it much -- opus is good enough for my hearing (again -- in portable sitautions, or in the worst case demo files -- the loss using opus isn't quite as serious as on mp3.) For serious professional work, I need flac/wav/whatever uncompressed at 24bits/at least 96k, and for professoinal listening only I need 24bits/48k, but can get by with 16bits/48k. I know that this subject is about portable, but if I am going to process a recording, I really need 24bits (or FP) at 96k or better. Low sample rates are troublesome for certain kinds of processing (up conversion for processing is undesirable, but sometimes necessary for best quality.) But, I think that the apparent consensus for listening, stayling with at least 16bit lossless is a good thing. 48kHz is nicer instead of 44.1kHz, mostly for the micro sized possibility needing subsequent processing (loss of original.) Additionally, my own superstition is that 48kHz is better than 44.1k, because conversion to 96k (I know about 88.2k, but I'd rather have 96k) potentially has less troubles. All this is my opinion -- I am not interested in imposing my thoughts on anyone else.
  6. Nowadays, you often don't need to go through the fft mechanism for normal sized filters -- CPUS are pretty fast now, and some kinds of errors are less of a problem. (There are other kidns of errors that are really unpleasant using the direct approach.) On my project, using the FFT is probably more of a problem than it is worth, even though I have written FFT based NR programs in the past. I need to be incredibly careful about any errors. I guess if one is using a 'stupid' interpreter that doesn't use AVX/SSE2 (or NEON), then there is a real speed issue, but using AVX can often give an 8X speed increase when doing FIR over not using the instruction (almost) if one is very careful in the implementation. FFTs seem to get less of an advantage from SIMD instructions -- but I'd expect that some advanced FFT routines (probably GPLed like FFTW probably is) might be able to take advantage of the super useful SIMD instructions. (The SIMD instructions can often give nearly the same speed of perhaps 8 multiplies as using 1 on normal multiply instr.) There are even multiply-add instructions that can give an additional boost. Also, some of the more advanced Intel machines can even do more SIMD operations. So, for shorter filters (perhaps 512 or smaller), the idea that a 'null' filter (just passes the data) being a multiply by 1 is true. BTW -- recent compilers are sometimes pretty good at finding ways of using SIMD, but the code still needs to consider the best way to write the code for the compiler to take maximum advantage. John
  7. Yes -- a filter which does NOTHING is effectively a single multiplication. If someone can hear a *1.0 operation, then something is wrong somewhere. It sure could be buggy software. (There are definitely issues that can happen with precision truncation, but as long as someone is sanely using 24 bits or floating point, then there are seldom problems -- but there CAN be.) When manipulating low frequencies, one might use decimation & interpolation, and bad algorithms for that can be problematical. In my project, I have avoided using decimation & interpoloation except for a singular rate conversion -- things can become complicated, and my project is complicated enough.
  8. I hate to bring it up -- but I have found that HDtracks unknowingly sometimes sells DolbyA encoded material. The symptoms that you describe (terrible sound stage, overemphasized (actually compressed) high frequencies, and weak midrange) -- that sounds like a description of undecoded DolbyA. My HDtracks Carpenters' album is also DolbyA encoded. John
  9. John Dyson

    Lies about vinyl vs digital

    Gawd -- I cannot believe that myth about 'stair steps' hasn't gone away yet... It isn't just irritating, it is really sad. John
  10. John Dyson

    Lies about vinyl vs digital

    I will c rtainly strongly claim with direct personal experience and NOT just from technical facts -- mp3 is okay to bits-compress audio, and produces plausible results, but at my level of listening (even though I have a 14k limit), mp3 sucks. For casual listening -- not-so-bad, but for critical listening, it looses too much information. Attributes other than audio quality -- I don't care. Making money, do it any way you want. Social ventures -- be happy and make friends. however, for audio quality -- give me accuracy, and that will win every time -- assuming everything else is equal. Of course, sometimes material has deficiencies, and distracting those problems with the material with other (more beloved deficiencies) might be beneficial. :-). John
  11. John Dyson

    The flaws of blind listening tests

    The DolbyA distortion doesn't really sound 'bad', but tends to soften the sound -- and that is one reason why it hasn't been very aggressively fixed. DolbyA encoded/decode CAN cause a lost 'precision' in the sound (gawd -- hate to use terms like that :-)). Wrt simple anti-distortion mechanisms... R Dolby acutally used a rather ingenius and simple one on the DolbyA (but it still becomes too fast), and that is that crazy diode network that also doubles as a rectifier. (If you look at the 361/cat22 schematic and see the 4 JFET compressors -- the fancy diode network is on the right side of the compressor circuit.) That circuit should be looked at SUPER carefully, because he actually made the attack slow for small increases in signal level, then speed up rapidly for fast increases. It was a cool design with his very careful way that he finessed the log/exponential characteristics of the diodes, when the know-how was a little less common in 1965. * One manifestation of DolbyA distortion is when there might be a chorus of two, three or four people, and somehow their distinct voices are softened into a chorus (it really isn't 3-4 clearly distinguishable voices anymore.) That is one commonly manifested kind of distortion occuring on DolbyA. The effect is similar to tape compression (which is a form of distortion), where it softens the sound in a mysterious way. A good group to check for the distortion might be Simon & Garfunkel when they are singing clearly without much accompanyment. WRT Orban -- I think that he is the one with the phasing patent. I use a similar technology (but super different design) in my decoder to cancel out practically all of the distortion (it really doesn't get every last bit, but it does amazing things.) I always avoid infringing on patents, however I really didn't consider his approach at all for what I have been doing. (His thing is really cool also.) Using his phasing concept can be helpful in hiding some of the IMD from fast limiters (and it actually implements part of the fast limiter.) Doing what Orban (I do hope he is the person with that patent) did back in the 1970's was pretty good forward thinking also. I am working very hard right now, trying to get a semi-final release ready. Needed a break, so did some forum trolling (hopefully not in a negative way!!!!) John
  12. John Dyson

    Lies about vinyl vs digital

    Yep -- I did a programming change incorrectly -- there are two streams of the audio signal level measurment -- really hard to explain without context, but there were two versions of signal level measurement the almost same value (approximately) that were erroneously mixed together (by using the same IIR filter twice, rather than using the individual filters that exist for each stream), and produced a gain control signal with seemingly almost random noise. . So, the audio was being modulated by a noisy gain control signal (in fact, the signal levels were different in time also.) It created a veil in the sound -- that I could not hear because of severe hearing fatigue until this morning. 62yrs old and extreme hearing fatigue made it difficult for me to hear the problem until this morning. I have been exhaustively tsting for the last few days -- otherwise simly adjusting a nearly perfect emulation. I botched a change near the last minute (gonna check back through my source control to find out when I created the botch.) -- I did find I did the botch on Jan18 between 14:22 and 20:34... I keep checkpoints fairly often. I cannot fix bugs purely by listening most of the time, but I can tell when there are errors, or hear things that aren't quite right (sometimes even visualizing things like signal envelopes because of experience of knowing what various things sound like.) Considering the software task, it is impossible to do what I have done without being able to hear. There are NO written specifications for the process other than two kinds of ancient hardware (EACH WITH LOTS OF SELECTED COMPONENTS.) Basically, my project emulates 4 FET gain control element feedback compressors which are in a feedback loop with a band splitter. The dynamics are horrendous to match, the frequency response vs signal level is horrendous to match. If it wasn't for the multiply nested feedback loops with a band splitter in the loop, the compressors alone (4 in the bigger loop) wouldn't be too difficult to emulate. But, the whole thing hasn't really been sucessfully done before now. Any minor changes or 'code cleanups' entail risk -- cannot be lazy at all when making changes. (Straightforward transformation of certain kinds of feedback controllers sometimes connot be done like on a DolbyA -- so various very difficult and tedious techniques need to be done. Even mathematically creating a model is so complex that I think that it borders on totally impossible. Simpler feedback ssytesm wouldnt' be so bad. There is one other way to emulate a DolbyA -- and it is patented, and I don't think even completely implemented. I might have used the patented technique if it wasn't patented -- but also there are some severe limitations for distortion reduction in the Sony patent design. My technique affords the ability to do the distortion reduction, but is incredibly difficult to make correct. My design is NOT tweaky, but is very tedious to make sure that all of the effects of every little thign that affects the gain or timing is modeled correctly. I have done complicated things in my life, but this thing is the MOST complicated by far, and it is pretty small -- maybe 20k lines of code. I become so severely fatigued that I make time wasting mistakes once in a while. VERY FRUSTRATING FOR EVERYONE ON THE PROJECT. John
  13. John Dyson

    Lies about vinyl vs digital

    To make those with audiophile-speak happy, I just fixed a bit of 'fuzz' in the sound from my software. Can we agree on that word? (IN fact, I wouldn't know how to describe the problem in the sense of signal damage, because it was a total botch.) Is 'botch' okay also? :-). John
  14. John Dyson

    Lies about vinyl vs digital

    No, I don't try to teach 'this stuff'. I have sometimes given classes to people who already know the basics, and that is much easier. I don't think that anyone can represent some of the concepts that I need to deal with in a few paragraphs and/or without diagrams. The problme is this -- there are lots of shortcuts taken, creating a kind of 'creole', and it really isn't needed for the level of complexity normally talked about here. THere are really technically meaningful words and descriptions for almost everything said. However, I am not suggesting that everyone be technically correct, but simply do not dismiss those of us who are do try to help, and do live in the world of closer-to technical correctness. Just because I might write something fairly close to being accurate in ways that can be reproduced by technical means -- doesn't make it mumbo-jumbo. It is really the normal language -- day-to-day used by engineers, and I have already simplified it. John
  15. John Dyson

    Lies about vinyl vs digital

    I understand that needlessly complicating things isn't a good thing -- it frustrates me also. Here is a little anecdote (true and recent example.) I am working on an incredibly complex piece of software -- seems like it is impossible to make it work perfectly in the way that I want. Here it is: I have an input signal (recorded music), where I need to synchronize (make the timing the same) the signal (recorded music) with the timing critical dynamic gain control within a few 100usec (volume control at the exact correct time.) The other thing that is because the way that the signal level (volume, but with implication of precise timing) is measured, there are undesired variations in the measured signal level (measured volume, but with the implication of precise timing.) I need to listen careful to the shape of the signal envelope (the measured loudness with great precision) by using my well trained sense of hearing to detect the shape of that envelope (measure the hardness of the signal with some timing precision -- basically the shape of the gain control/measured level curve.) Right now, I think that I got the timing correct (got the recording and the volume control matched up) and got the waveshpe corrrect (made sure that the volume control changes in the way that makes the resulting recording sound correct.) However, now I have a phasing problem, where the phase of the signal & sidebands resultign from the fast changing gain control is producing the wrong set of sidebands.... (Quick explanation --there are numerous kinds of distortion, we all know about harmonic, intermodulation, transient intermodulate, etc... The kind of distortion that I am dealing with exists only when there is fast gain control, and it is like modulating an AM signal and getting sidebands. If you create the upper sideband in a gain control device, the sound is 'softend', if you crate the lower sideband, the sound gets 'harder.') * the above is the world and thinking -- in relatively simple terms instead of the very complex things that the programming really does* Okay, we could strain to interpret in more common language what I am writing about... Until the subject becomes very very technical (and important for making the project work correctly.) Can you see how very difficult some of the technical work really is (I mean, not speaking of the 'art' of development, but the deep nuts and bolts to actually make a complex device work correctly?) The details that I approximated above aren't nearly at the depth that I am trying to deal with... Now, I have a group of people speaking of 'air' -- a new term in the context (relatively so), but I do know -- similar to my hearing being able to roughly determine the waveshape of the gain control attack -- the term 'air' should be defined and/or refined to actually mean something concrete and coherent. If you want more 'air', then work with the REAL engineers who can actually give you 'air' (people like me ), rather than make me feel like my really hard/accurate thinking and language are both 'mumbo-jumbo'. John
×