Jump to content


  • Posts

  • Joined

  • Last visited

  • Days Won


Everything posted by kgallen

  1. kgallen


    Indeed. DAT like CD is 16 bits. Forever. MD may well have started as 16 bit (because it was restricted by the processing power you could put on a consumer chip in the 1990's), but if you can get the ADC/DAC and have the number crunching power in your DSP to do the FFTs/IFFTs to convert the samples to/from the frequency domain then 18 or 20 bit resolution is achievable with the format. Note: 16 bits to 20 bit is absolutely massive in dynamic range/noise terms. Don't let those little 2- or 4-bit increments fool you - it's not a change of "2" from 16 to 18, it's 4-times and 16 to 20 is 16-times - repeat 16 times the dynamic range. This can be seen with the increasing dynamic range specs as ATRAC evolved[*] (expressed in dB - a logarithmic, not a linear scale). Beyond 20 bit, you're not designing for the human ear, because even 16 bit dynamic range is pretty much at the practical ear/brain limit anyway. That's not to say you wouldn't want more "bits" in the audio chain that's being used to generate the content, i.e. the recording studio, because as you combine and mix the tracks you want to be able to deliver a final master "tape" that can achieve 16 bit dynamic range (i.e. the noise floor is still at least 16 bits, not degraded below). I will (try and) stop now, because I admit I will start getting out of my depth with the maths... [*] Note within product families the improved dynamic range specs are due to better (more expensive) ADCs, DACs, better quality (electrically less noisy) power supplies and better filter components in the higher spec machines - the 9xx compared to the 5xx etc.
  2. kgallen


    This is a beautiful interview, if you can drag your way through the translation. You can feel the Sony engineers battling to get the concepts across to the interviewer who is hell-bent on reducing the discussion to "my (digital) bits are better than your (digital) bits". ATRAC is far cleverer than that. It's Adaptive, it's in the name. Nothing is fixed. It can "mould" itself to the music to capture what is needed to reproduce it most accurately. The ATRAC encoder is a thing of mathematical beauty and that Sony were able to implement such mathematical complexity in electronics for a consumer product in <1992 is frankly astonishing. http://minidisc.org/mj_ja3es.html Also recently I've been reading on here some mention of "fixed" and "floating" in the context of "fixed point" and "floating point" (terms we would traditionally associate with IEEE 754). "Floating" in the MD world is somewhat different (although you could loosely relate SF to exponent). This is covered quite nicely here in Figure 5, then related to Scale Factor in Figure 4. As I whined about the reddit thread above I'll also note that in respect of the 24-bit ADCs, the ADC generates the samples in the <time domain> whereas the "throwing away" that ATRAC does for compression is in the <frequency domain> (after the Fast Fourier Transforms the CXD DSP does). This key point is thoroughly lost on the contributors on that reddit thread. They were going on as if the "throwing away" was discarding samples and/or ADC bits in the time domain. Wrong... One other point, as I've just been looking at related websites that talk about digital sampling in an ADC. Some of their diagrams are very misleading. Can I state here and now, that as long as an ADC samples a bandwidth-limited signal at at least twice the highest required frequency, then the DAC will reproduce the analogue signal perfectly. Even if that ADC was just a few bits. The number of bits of an ADC does not affect the quality of the waveshape once reproduced (number of bit affects the dynamic range and noise floor). Just because you sample and quantise a real-world analogue signal at discrete time intervals to discrete quantisation steps (i.e. integers), you do not get back out a steppy rough analogue signal (like some of the diagrams I have seen suggest). If you put a perfect 20kHz sine wave into an ADC sampling at greater than 40kHz, this sine wave will be reproduced perfectly and smoothly once you have passed that discrete digital sample set back out through the DAC and LPF. OK I'll stop ranting now (maybe).
  3. kgallen


    I'll add myself to the list. I've just clearly demonstrated I'm no wiser. Or at least I'm incapable of navigating a website. I'll say that's a software problem not a hardware one. I'll get my coat.
  4. kgallen


    If the stated intentions can be achieved - i.e. a robust long-term archive of the valuable technical and machine documents then I think it's a worthy exercise. If it's a Twitter for the airing of random misnomers about the technical details then it's just a gossip site. We currently have here, Tapeheads plus .org. Whilst we're all on a voyage of discovery and none of us were actually on the Sony design teams, I don't think as yet either existing forum has degraded into one of arbitrary clueless contribution. At least on here I think we own up when we get to the limits of our knowledge. That's not to say we don't push the boundaries of our learning, but I don't think we make a habit of claiming knowledge where we have none.
  5. kgallen


    I just had a look. Bumped on some wild-ass discussion about "wide bit stream" and 24-bit DACs where most contributors are just shooting in the dark. Not the wrong end of the stick, the wrong forest.
  6. Not for you guys, I’m still here! :-D From recent perusing I notice you’ve all got quite a bit of history here!
  7. It’s only because I was faffing about with joining the CD-450, which only has coax SPDIF, to the MD-350 which is only TOSLINK. So I ended up having a play. I think the MD-350 gets quite favourable reviews - along the lines of ‘Choose Sony or the Tascam MD-350 is the only other that comes close’. I bought it (£120) when the 530 didn’t work - the 530 being the first machine I bought in 2019 in my ‘second wave’ of MD whose fault brought me to this Forum.
  8. I was playing about the other day with my Tascam MD-350 (Type-R MDLP, CXD2662) [I think I noted that somewhere here the other day, as yet another distraction...], recording in SP, LP2, LP4 using both some pop and classical pieces (recorded over SPDIF from a Tascam CD-450). LP4 was better than I expected but I could definitely hear a "flutter" in trumpet solos (although not so much on flutes). LP2 sounded as good as SP (Type-R) to me. With pop I think you could easily get away with LP4 quite frankly! Playing back the LP4 on the 480 (Type-S) it didn't sound any better to me than it did on a non-Type-S machine (MD-350 or MDS-E12). I did find out that the MDS-E10 sounds a whole load better than the Tascam MD-CD1 though... the Tascam was as rough as a badgers rear bits (actually the CD section is terrible - flat as a pancake, the MD section isn’t too bad).
  9. Yep the 520 has the Sony in-house CXD8607 DAC/ADC [probably 3rd-party devices were not up to scratch at that time] with a Sanyo LA9615 front-end amp, whereas the 530 moved to the AK4524 24-bit ADC/DAC, which became the standard for a fair few generations of machine, with a front-end amp based around Mitsubishi M5218 op-amps. uC is slightly different - newer version of the M30 series. The pinout is almost identical between machines with the 530 having a couple of uC pins assigned to keyboard - which in the 530 are not used.
  10. Honestly I have no clue! (As in they both sound great to me!) My assumption is that the 520 and 530 are essentially identical inside, aside the MDM5A to MDM5D (and hence CXD2654 [ATRAC1 v4.5] to CXD2656 [ATRAC1 Type-R]) drive change. They have the same case, buttons and display. Same menus. Same I/O essentially (530 adds Control-A1). I must admit I haven't compared schematics so they may as well be completely different, but as they are one product iteration apart I doubt it [now I'm going to have to compare aren't I to see how wrong I am again...]. I did read a review recently that concluded the 520 sounded better than the 530 (530 more "metallic"), but I couldn't corroborate that. I use the 530 because it's got the Control-A1 ports, it's wired to a Sony CDP-XE530 via optical and Control-A1 so can do Type-R CD copies. (This is why I would like another 530 or similar so I can do MD-MD dubs with title copy). Although in most cases I use a Tascam MD-CD1 for CD dubs in Type-R SP or LP2... I must admit that for listening (on cans) I tend to use my MDS-E10 (or Tascam MD-CD1, but the E10 audio quality is superior) which is next to me now as I work from home. Both are MDLP and a lot of the discs I've recorded recently are in LP2. When in the office it's my MZ-R700 portable (also MDLP). The 520 tends to get more "abused" these days as a DAC and headphone amp since it has an optical feed from my desktop PC, analogue feed from a 480 [no headphone socket] and coax feed from a Denon TU-1800DAB receiver. Useful for recording radio programs - although recently the 480 has been doing that in LP using its optical in...
  11. No, you carry on with pleasure! This thread had nowhere else to go, I just wanted to capture that 520 SOS review! :-) All the contributions here are well worth having, I love reading them.
  12. I wish we could update the links (and correct other errors, and update specs/data) on minidisc.org - but I don't believe we can. So second best I'll just post here. Here is the Sound On Sound Oct 1998 review of the MDS-JE520. It's back on the web as SOS rebuild their site from their archives (great one chaps - I was a mag subscriber for some time back in the day). Of course though the URL is different from the link on minidisc.org. So here goes. And I never realised "my beloved" was held in as much esteem as it was (is). Better still I paid £139.95 for mine rather than the £230 discussed. Although admittedly I was nearly a year late to the party so I guess you could argue it was "last year's model". But I don't care, I wouldn't have been able to afford the new 530. https://web.archive.org/web/20150607080326/http://www.soundonsound.com/sos/oct98/articles/sonyminidisc.html ps This is @Richard's fault, he's meticulously replacing all of the lost images in this part of the Forum - a part I never really knew existed. So now you'll have to put up with my ramblings here now too. pps SOS by Issue and by Article: https://www.soundonsound.com/more-sos-past-articles-coming
  13. -My apologies, on which schematic is this Q5 transistor you speak of.- <Hey, where has "strikethrough" gone when doing Edit???> Ah, on the 640 schematic, I was on 770 where it's Q34. This can be used in I2C for flow control - the slave holds SCL low to tell the master to extend the I2C cycle (I2C spec $3.1.9). But also you are right, this can also be used in multi-master I2C topologies as a means of bus arbitration (I2C spec $3.1.7). My second apologies, I am still commenting and disturbing your peace despite my assertions to the contrary :-(
  14. I believe I made (at least!) one incorrect assumption or statement earlier in my voyage of discovery. In reviewing/perusing my pdf attached above for 770, I realise that for a keyboard, the deck is an I2C slave (SCL is input to the deck, albeit with flow control from the deck using transistor Q34[770]/Q5[640], SDA seems input to the deck only). For PC-Link the deck is I2C master (or indeed could be master or slave) (SCL is output or bidirectional, SDA is bidirectional). This means that to support PC-Link, the "keyboard pcb" cannot have any active circuitry that would prohibit driving out either SDA or SCL. Active circuitry could of course allow this, but if the deck was intended keyboard-only then the active circuitry may be unidirectional as detailed above.
  15. @NGYAre you able to share (without major document preparation on your behalf!) the next step that I feel you know the answer to? I realise you want to fully validate your findings but of course this may depend on having a suitable deck and "vandalising" one's main machine is never to be taken lightly, so there could be a significant timeframe involved here. You are teasing us with the answer ;-) Is there anything next you need help with to bring this to a conclusion in our minds? I cannot help with the soldering iron, since the lead on mine will not reach to your workshop :-D Maybe you just want some space in your mind for this, without the chatter in this thread, in which case I abate!
  16. But on starting this, I did not know you were doing this so embarked on my own investigations per the initial request from @sfbp. I realise you want to present the perfect solution, but even if you share some thoughts then we can collectively learn and collaborate on the way. My reading of your comments was you still had issues to solve. If you want to present the ultimate solution and have all of the answers but are not ready to share, then that "heads up" to us is useful. i.e. I can butt out if this is all old news and there is a known solution yet to be shared. That said, I don't think I've wasted my own time since I've learned more about an aspect of these machines I hadn't investigated before. However if I'm presenting old news at the frustration of others then I can stop. Above reads bitter, it is not meant to! Greatest respect to you all here, you are great friends if such can be stated given this is an open forum.
  17. OK you clever people here is my latest cut-and-shut based on the insight from @M1JWR and @BearBoy Per @NGY's comments, the PC-Link (raw I2C) is not buffered in the 770 (despite the block diagram suggesting it is, below). The key here is that if the unit has a keyboard interface, then the I2C (IIC DATA, IIC CLK) also needs to be available raw on the DIN connector. The Keyboard interface has this 5-transistor circuit between the uC and the DIN. Also the PC-Link has these two extra connections IIC BUSY and IIC POWER. I believe that @sfbp has all these connections in place on his 640, however I need to review the details (but need to pop off for now for a work meeting, so be back later...) Sorry this is a pdf - exporting as png/jpg/gif didn't have the resolution (however I'll also upload a png just so you can see what's in the pdf, so you don't get too excited...) Keyboard_PCLink.pdf
  18. Well there you go, you've been there, done that! No point in me interfering here any more, you've done it! Hardware is fixed. That's me done. Firmware then. Over to you :-P
  19. On the 640 it's next to the rear panel Control-A1(II) connector. This picture is flipped. For you with the lid off, looking from the rear and down it will be to the right just above the rear panel just down from the power supply section (big black capacitors)/just across from the S-shaped heatsink.
  20. Zero-ohm link. Commonly used to option in and out circuit elements within a product range. See them all over the place. Could be a wire link but more usually these (and those) days a chip "resistor", 0R.
  21. Sorry for the cut-and-shut picture: I wonder if you chaps have a 4-pin or 6-pin connector for CN820 on your 640 main board. Then if R826 and R822 are fitted on your KB board (probably). If so, then I think you could rightly label your KB connector as "Keyboard (PC-Link)" with no changes...
  22. I2C spec - used to be quite hard to come by, but easily found on't web these days here. There are three traditional modes, standard-mode @100kbps, fast-mode @400kpbs and Hs-mode @3.4Mbps (not really ever termed baud on I2C).
  23. Well I guess we could argue details, but I2C is "Inter-IC" so not really designed/intended for off-board driving over a traditional cable - you'd use a spec like RS232 and it's brethren for that.
  24. It probably would work but not be robust. The uC I2C ports are not designed to drive a cable. Having a read of the datasheet for that I2C buffer gives the background of why that chip is required in such I2C topologies wrt capacitive load and current drive etc. If the USB interface inside the MN10 thing were mounted inside the 640 and wired directly in, I suspect the buffer chip would not be necessary.
  • Create New...