Jump to content


  • Content count

  • Joined

  • Last visited

  • Days Won


zenelectro last won the day on March 29 2012

zenelectro had the most liked content!

Community Reputation

207 Excellent

About zenelectro

  • Rank

Profile Fields

  • Location
  • Country
  • First Name

Recent Profile Visitors

3,630 profile views
  1. You need to sign NDA to get data sheet.
  2. Interesting - wonder what the difference is? TBH I'm kind of bummed out a bit with the Sabre chips, they should have kept the ASRC separate. It also has other challenges such as the increasingly low OP Z as you parallel DAC's for higher performance. Then there is the extreme sensitivity to PS due to close to zero PSRR It all adds up to Ferrari like performance that is often very difficult to extract in real life. OTOH, I've seen breadboarded 1794's with very little attention to PS and layout that exceeded data sheet distortion performance. Yes, very much agree.
  3. No, I don't own an operational 1794 DAC yet - currently working on it. However, that FFT is real I can assure you and I know others that have achieved similar performance. I have ESS DAC here, a 1541 0xOS and whatever comes through the shop for repairs / upgrades. The best most recent DAC was a hi level Accuphase with 2 x mono 9018's and it was very good but it still left me wanting more. There is something about the Sabre that I can't come to terms with. My measurement capabilities are EMU1212M which is actually very good but to measure these levels with EMU will require an IP notch filter to null the fundamental. IMHO, all these DAC's, ie; the Sabre, AK4497 and PCM1794 are capable of similar measured performance given the right analog stage. They all have advantages and disadvantages, it's really about where you want to compromise and what type of circuitry you prefer to work with. Terry
  4. There's a lot of talk about linearising of the R2R ladder but the measurements don't really show it and look typical R2R to me with a spray of upper harmonics. Also if the signal to noise is truly 120dB, then there is significant noise floor modulation going on. Notice the spray of high order harmonics. Also notice that this measurement at 0dBFS would have an RMS noise floor of about -105dB Now below is a really good DS (PCM1794) at 0dBFS Here the highest harmonic is the innocuous 2nd at -120, 3rd at -125 or so, everything after the 5th is below -145dB. Also bear in mind this was measured with Prism Dscope and a lot of this distortion (and noise) would be the Dscope ADC itself. The AP test set used to measure the R2R above has a better residual dist / noise spec than Dscope. Just sayin' Terry
  5. zenelectro

    PDX DAC project

    Ben, WRT doubling freq and jitter increase, that's not necessarily the case and freq doublers / multipliers are in fact often used to *lower jitter. Too complex to go into here but the short of it is XO's around 5 to 10MHz have lowest jitter inherently, especially close to carrier, IOW at low frequencies. So if say an 80MHz clock is required with very low close to carrier jitter, one option is to use 10MHz with a freq multiplier -> 80MHz. It's all about choosing compromises, as there will be higher jitter noise floor with this approach. So yes, you can make a very low jitter multiplier and it is used often for certain applications / requirements. But - to make a guaranteed low phase noise multiplier you will need very expensive test gear to verify that it actually works as desired. Terry
  6. George, WRT to this 'Hybrid' DAC topology you are describing, it is actually NOT an amalgamation of R2R and DS at all. From data sheet - Advanced Segment the upper 6 bits are processed separately to lower 18 bits. The lower 18 bits are Delta Sigma with 4 levels the upper 6 have 62 levels. Total = 66 levels. However this is all happening before the actual OP DAC itself in digital domain so to speak. The DAC's physical architecture is 66 x equal current sources which switch on / off at very high speed (say 64 x 8 x FS at 44.1), the x 8 being internal dig filter. For Sabre they are equally weighted resistors, from memory approx 50k each. So the original Sabre had 64 x 50k = 781 ohm OP Z per DAC channel. In stereo mode it is 4 x 64 = 256 equally weighted resistors = 195 ohms OP Z. -ALL- DAC architectures have their compromises including R2R. R2R have worse linearity and also generally a spray of low level upper order harmonics. They also have significant noise floor modulation. It really is a situation of choose your poison and in the case of R2R that poison happens to be subjectively sweeter tasting but still not necessarily actually truer to what was recorded - depending on implementation. Terry
  7. Yes that is right. The transformer is actually part of the low pass filter.
  8. zenelectro

    PDX DAC project

    The other option is to use a double speed clock and add divide by 2 with a simple flip flop circuit. What is the clock frequency? FWIW I've had quite a few PDX's across the bench for various types of work. Most didn't sound that good the biggest problem being clocking / jitter / etc. On one I improved the clocking / jitter and the sonic improvement was quite staggering. You can talk about RasPi / USB / Pink Faun / Ethernet etc etc etc etc etc etc etc etc but here is what counts: a/ Galvanic isolation between PC and I2S signal DAC side. I2S signal must go through high speed isolators. b/ Good quality Clocks on the DAC side of isolation. c/ Clocks must reclock I2S on DAC side of isolation. If you address all of the above the PDX will move into another level of sound quality. Good luck, Terry
  9. zenelectro

    DAC comparison

    The '63 against 1704 is not a cut and dry case. Yes plenty of people feel that more bits were better however AFAIK, the 63 was made with a different process and not entirely CMOS. The 1704 also had the step backward of no access to BP current offset. When this was disabled on 63 (and from memory) a good CCS substituted it was significantly improved. They are all pretty much unobtanium so.... life goes on
  10. zenelectro

    DAC comparison

    The data sheet performance of a DAC chip is absolutely achievable and in many cases exceeded. There are some DAC's that are particularly finicky, the ESS Sabres are an example, but there are others that I have seen breadboard achieving data sheet performance, the PCM1794 being one. The PCM1710 has significantly inferior performance in many ways - I won't go into them. If you understand DAC design, look closer at the data sheet and you will see. Having said above - Of course implementation is absolutely key to getting best performance - but the question was asked - so there's your answer... Terry
  11. Steven, From what you describe it sounds like you are really happy with the Pandora DAC - so I'm wondering why you are entertaining jumping ship, so to speak. The Pandora and in fact all Aesthetix DAC's for that matter are very different to the Bricasti M1. The Pandora is based on BB PCM1792 which is well known to be one of the most musical sounding DS DACs. They've added their own digital filtering and a tube OP stage. Let's look at the Bricasti - this is a pro based company and in fact the founders previously worked at Lexicon. Lexicon is a pro audio company and pretty much the father of high end Signal Processing AKA reverb, flanging, time modulation and space emulating effects. Looking at the Stereophile blurb, the Bricasti M1 is based on AD1955 DAC, high speed opamp I-V with discrete OP stage. IME this will get you a different sonic result than your Pandora DAC's approach and most likely move away from these subjective qualities that you are liking about the Pandora. Having said all the above, this is just my own speculative opinion take it for what it's worth - you will never really know until you listen to the DAC yourself. I'm just advising to proceed with caution, don't sell your Pandora until you can get to audition an M1. PS If you do audition the M1 I'd be interested in your subjective comparison. Good luck Terry
  12. zenelectro

    DAC comparison

    OK, I'll bight If you are just comparing DAC chips, which it appears you are, then I would give the nod to CS4382. It has better specs and Cirrus had been doing really good DS DAC's already for quite some time. BB never really got their DS act together until the Advanced Segment architecture AKA PCM179x. etc These are very high performance devices capable of competing with multibit if implemented correctly. cheers Terry
  13. Hi all, Just became aware of this thread (thanks RMP). My inbox says 47% full so I'm not sure what is going on with messages. Firstly, thanks LPG for the historical perspective, indeed your input and foresight WRT zero OS / 1541 DACs has been significant. WRT 3A/167M, holy ^&%$ I was aware they were getting expensive but those current prices are crazy. Hmmm from memory Steve still has a pair of mine WRT 6C45 Pi, as LPG states they are great tubes with a few caveats, they like to oscillate and they really need to be matched. The oscillation is no problem to sort but I've found they can vary all over the place so matching is important. Another option is WE437A which is almost exactly same tube as 3A/167M but I'm guessing by now it would be same price. I've contacted the owner of the Audiocentric DAC and will get this unit sorted. cheers Terry
  14. zenelectro

    ME850 mods designed by Art Vandelay

    Guys, I've been successful building BJT OP power amps that showed almost perfect 10kHz square wave into 1.5uF // 8R, (actually I got it up to 50kHz then cooked the OP stage due to too much level = bang ) but just a few things to think about: The edge rate of 10kHz can vary quite a bit depending on your generator, the amp front end BW and any low pass filtering in the path (cables / amp). This makes a significant difference to how the OP stage handles the leading edge and the massive curent demand if it's very fast. If it is very fast the OP stage layout (ground plane etc) starts to really come into play. It's also worth experimenting with the Base input resistors. WRT servo's... We did a huge amount of work on DC servos and found to make one completely dissapear sonically was a real challenge. Did find a solution but it was not easy by any stretch. People tend think servos are like a black transparent box and this is not the case at all. I think using higher speed more modern BJT's will generally sound better, even with similar OP stability into cap loads. WRT Emitter resistors, IME the smaller the better - if you can keep the whole shebang thermally stable. I have used NONE but that is a veeery tricky business...... 😃 😃 and usually end up with a small value. From memory, Technical Brain (love that name) Amps had none but did have a penchant for blowing up for a while. Good luck in your quest for ME pefection! T