Jump to content

robbo1802

  • Posts

    16
  • Joined

  • Last visited

  • Country

    Australia

Retained

  • Member Title
    Newbie
  1. I had thought that the current sources used in the output stage would only be configured as a FIR when the BB chip is in DSD mode. Are you saying that they are also configured as a FIR when the BB chip is used in PCM mode? I hadn't gotten round to thinking about some of the implications of the sigma-delta modulator schemes as regards sound quality. As you have been at this a while, perhaps you can help me out. Can you describe how the modulators influence the sound quality? Do they create deterministic uncorrelated errors (artifacts?), deterministic correlated errors (distortion and/or artifacts), stochastic correlated errors (modulated noise), stochastic uncorrelated errors (raised noise floor and/or colour it), alter the tonal balance, some other SQ issue, or a combination of these? This would be most helpful. I will answer your other posts when I get some time and after some thought on how best to discuss the Information Theory issue (can't spend all my time on the computer, we are on the move and mobile internet is not guaranteed - bloody telecoms companies!). One question that may help me, do you believe in the Shannon-Hartley equation? If not, I will not be able to discuss the Information Theory issue so please let me know. The equation is shown below for other readers of this thread who may not be familiar with it, of course Google is also available. >:-) C = B * log2(1+ S/N) where C is the theoretical maximum channel capacity in bits/sec, B is the bandwidth of the channel in Hz, S is the average signal power and N is the average noise power (the equation applies to a Gaussian channel so the PDF of the noise must be Gaussian, fortunately this is usually the case) . Beware the S/N is NOT in dB, if you wish to use dB you will have to convert it to the equivalent linear number, e.g., for 20dB signal noise enter 100 for S/N. Entering the theoretical maximum figures for Redbook CD in the equation will give a rather obvious answer. As an aside, when I started in engineering no one thought we would ever get to 30% of the Shannon limit and it was at 25% or so for many years until a couple of breakthroughs in mathematics and then overnight (or so it seemed) everyone has a modem on their desk that is operating within spitting distance of the limit. Okay, a little bit of exaggeration but that is what it felt like. Regards, Bob
  2. Really! The only bit that provided good anecdotes was the very early years with all the electro-mechanical gear and it certainly didn't feel all that interesting at the time - I was banished to that area for being a bad boy. I am retired, they haven't done a thing right since I left! :-/ One of the problems that many companies face is too many choices. One of the low speed networks I spent a lot of time on was engineered to have a lot of features that were never marketed to customers as the features made the product too difficult to market (and bill!). This does have parallels with DACs where the capabilities of the chipsets are not realised in the end design. There are always some bad products that make money as the few alternatives are just as bad. Ahh, Motorola, used to be one of my favourite companies, I used to code the MC680x family in hex on the ferry to work (knew the instruction set off my heart) as we could not initially get a cross compiler for our Altair clone. Worked with a lot of other Motorola stuff and then our company took up Six Sigma - never forgiven! Music is mostly classical recorded in any era, in contemporary music no one has written anything good since 1986. :-) Unfortunately, even when remastered, many of the my favourite albums have poor sound due to a combination of bad recording practices and subsequent deterioration of the multi-track masters (should they have survived). Sadly, I still can't listen with other people's ears. Regards, Bob
  3. In this case I was referring to the actual conversion stages and to a lesser extent the analog FIR filter. I appear to be relatively insensitive to changes in the digital filter responses - I suppose I am just lucky. >:-/ Regards, Bob
  4. Hi, I am not sure what you mean but I will make a guess, I will assume that you mean that using HQplayer means that you can have a DAC without a chip as HQPlayer is doing most of the conversion. I think your first assumption is that HQPlayer will do everything that I want on all hardware platforms that I want to use. Let us assume that this is so. The next assumption appears to be that I believe that it is a good idea to convert everything to DSD128 or higher before it gets played back. Let us assume that this is so. The DAC still needs something to turn the electrons on and off. This could be a simple discrete switching device followed by some filters (or not if simplicity is the goal and that the rest of the audio chain can handle the resultant high frequency hash without deficit or failure). It could also be a more sophisticated discrete implementation with cascaded transversal and traditional low pass filters such as the DSC1. It could also be a BB DSD/PCM179x chipset which is purposely designed to turn those electrons on and off and provide the first stage of the cascaded filter. Now back to the first two assumptions. I will concede that I have not yet trialed HQPlayer so maybe it could do all that I want on all the platforms I may wish to use. However, I currently do not have a DSD enabled DAC. The second assumption is not so easy for me to parse. One of reasons I want to have a DAC capable of PCM and DSD is to see if I am comfortable with DSD. Earlier in this thread I stated I am an ex-telecommunications engineer and like so many telecommunications engineers Information Theory rules supreme - you can't beat entropy. Unfortunately a simple view of DSD and classic information theory are not a good fit. You have look at DSD (and delta-sigma itself for that matter) at just the right angle to think it might be a good idea. This is not to say that delta-sigma is wrong, it just means that you have to make some serious assumptions about the nature of the signal you are encoding (and your own hearing -important) in order to get past limitations determined by information theory. I have not accepted these assumptions at this point and may need convincing (or not) via abx. Even should I accept that DSD is a suitable format, there are still psychological issues that may factor into my being comfortable with the overall audio chain. Regards, Bob
  5. Hi Jay, No I do not have the capability for DIY at the moment. The wife and I travel full time around Australia in a motorhome and have done so for about 9 years. In any case I do longer possess the eyesight and hand skills to do surface mount. At the beginning, I did state that I was after a commercial product that was portable (not super portable) and could be battery powered and had a headphone amp (although that is not a deal breaker). I was starting the survey to get some ideas about narrowing the search. I could probably go to a unit that used an external 12V plug pack for power - I have no problems using an VRLA battery to power the thing. I do not discount the possibility of moving various functions to different areas of the replay chain. I also don't discount that it gives you more flexibility in determining the characteristics of the filters, conversion algorithms, etc. I consider them distant because both can be used with other components. I may have a need to use the DAC with other sources. I don't think the filter provided by the manufacturers are ambiguous at all, quite the contrary, they are often fixed in stone (literally). It is really a question of whether you believe that you need the flexibility presented by HQPlayer. My listening is quite adaptable, I have no difficulty at all listening to electrically recorded 78s if the performance interests me (I draw the line at acoustically recorded in most circumstances). All I require is that the sound I listen to is the best possible for the source, this is a psychological thing which I can only describe as a comfort level, that is, I am comfortable with the sound, if I feel the sound is 'wrong' I get uncomfortable and consider the sound unconvincing. This has some quite strange effects on how I built audio equipment and felt about recordings. For instance tonal balance was not an issue, it could vary widely as long as I had no definitive reference to compare it to such as the instrument from which the recording was made, however if I did have the reference in my head then there were serious issues as I "knew' it was 'wrong'. Distortion on the other hand was a big problem especially the type of distortion common with analog tape recorders and phono cartridges/vinyl records. A similar problem occurs with poorly controlled resonances throughout the audio band but I have no sensitivity to reasonable phase shifts (really time delays, so sensitive in bass but not mid/treble). Strange isn't it, but there is a saying, I can't listen with your ears and you can't listen with mine. I am not sure I quite understand what you are saying here. I will address one thing that is common in audiophile circles with which I have a great deal of difficulty. If I have misinterpreted you, please accept my apologies - and a blanket apology to anyone who finds this offensive. [Rant on] It is the lone hero myth. The lone hero rides into town and with nothing more than a soldering iron and an idea of what constitutes good sound and defeats all the corporate engineer lackeys with their multi-million dollar development platforms and test equipment. He can do his because he listens to music and everyone knows that corporate lackey engineers don't listen to music, have tin ears, are not passionate about their work, are only interested in money and stock options (as if!), and live only to saddle consumers with awful and evil products. This is typified in the following from this link: iDSD micro Crowd-Designed! iCLUB members - V4.08 BETA firmware (page 136) - Page 57 " Many ADC/DAC Chip's that deliver outstanding performance seem to result from lucky accidents, where someone in the big corporate hierarchy got it right. They got it right either because they knew exactly what they are doing and design for sound quality (something that is not easily, if at all "measurable"), or by sheer accident (the million monkeys, million typewriters and Shakespeare gig). In the history of digital audio there have been a number of ADC and DAC Chip's (or discrete solutions) that were outstanding. " and to a lesser extent:iDSD micro Crowd-Designed! iCLUB members - V4.08 BETA firmware (page 136) - Page 5 this is the IFI designer talking about the DSD1793 "It was the result of excellent work done at Burr-Brown Japan before it became TI and was meant to be used in High End universal disk players with excellent performance for CD replay, Audio-DVD and SACD. Unlike most newer Chips from TI and many others, it retains a true multibit core for PCM replay, based on the visionary (and under-appreciated) PCM69 AND it retains Burr-Brown’s unique and excellent approach to DSD replay, first introduced with the DSD1700 Chip." Clearly the inference here is that BB got it right and TI stuffed it up afterwards, after all TI must have fired all the BB designers (American corporate lackeys are more evil than Japanese corporate lackeys) and of course TI engineers apart from all the other evils have no experience in DSP or designing audio silicon. If you have been following this thread you will know that I believe there is only one later chip and in the areas that affect SQ it is not much different to the DSD1793 - I believe Miska disagrees with me here. I am not saying that lone developers can't design good audio and I do not mean any insult to HQPlayer. I just don't believe in this myth that the engineers always get it wrong. Even worse, most lone (hardware) developers just don't have the resources to do it right or even know when they get it right. Since they can't measure it, they say measurements are immaterial and when someone else measures it and reports how badly it measures they say it sounds good because it measures bad - it is all part of magic. [Rant off] If you haven't found it on CA already, at least one member has built a DSC1 so you should be able to search it. BTW, what's your background? Mine is in discrete devices and the development and commercialization of a direct excitation mechanical vibration / resonant frequency instrument, which is my technical interest in DA and AD conversion. For music applications, I'm a chump at this stuff! Jay I am retired but I was an evil corporate lackey engineer in the telecommunications sector. Started when data transmission meant telegraphy (110baud max but mostly 50baud - baud is a measure of asynchronous speed defined as 1/minimum period between transitions so basically max of 50 bits/sec) and sub 300bits/sec propriety modems. I worked right up to Plesiochronous Digital Hierarchy of 565Mbits/sec but escaped in 2002 before getting sucked into SDH (the ITU's version of Sonet) implementation, there was a pile of stuff (not all data transmission) in the middle. Most everything I worked on is gone, even some of the buildings! Being in telecommunications does colour how I think about electronics, especially regarding cables, connectors, and data integrity. It also means that some of the loose terminology used in audio forums winds me up, one of the reasons I don't often visit forums. Audio was always a hobby, got serious in the mid 70s, other things have periodically got in the way since the mid to late 80s. On the other hand I have never stopped being serious about music. I am especially passionate about piano - can't play but have owned a couple. Did amateur recording of pianos, small ensembles, etc in the mid to late seventies and into the 80s using a number of "semi-professional" analog tape decks. One thing you don't want ever to do is work out how much the audio gear of the 70s and 80s cost in todays dollars. Especially, don't work out what those Berkshire Hathaway shares you could have bought with that money would be worth today. :-) Regards, Bob BTW I apologize for the length of this reply, I lost it a bit there.
  6. I haven't looked at transversal filters for a long time and don't really see a need to strain the brain about something that I will not put to good use so it will take your word for it. >>;-) I expressed myself poorly there. I was mostly referring to the crosstalk between the purely digital stages of the chip and the final conversion stages. Reduction of the currents in the digital stages will reduce the radiated magnetic fields thus decreasing the crosstalk. I was aware of the relationship between effective impedance and noise - memories of building moving coil pre-amps a long long time ago. One thing that the chips have in their favour is a very controlled environment, this is not so easy in the discrete world. Discrete ladder DACs in particular occupy fairly large tracts of PCB space and thus present significant design issues. I even think there are a number of potential benefits from the chip manufacturers putting the I/V convertor on the die as long as the op-amp function is of the best performance and the power supplies are separate from the rest of the die. I wasn't aware of the ram-hammer effect - sort of reminds me of the problems with vinyl records when the tracks were cut too close together, you got an effect in some ways similar to tape print through. I will watch with interest how the manufacturers deal with quantum tunneling at 10nm and especially 7nm - have you ever looked at how large a 7nm feature is at silicon matrix spacing level. Ooops, off topic Yes I saw that, I was going to ask you about the hash in the FFT of the IFI compared to the TEAC UD501. I have been plowing through this tread: iDSD micro Crowd-Designed! iCLUB members - V4.08 BETA firmware (page 136) I am now suffering from hype overload, so many people in audio seem to have gone to the Bob Carver School of Audio Component Naming (don't get me wrong, I thought Bob Carver was a great designer but his naming!). Anyhow, in amongst all the hype, exaggeration, and fawning there were some facts. One fact was the frequency at which the internal switching power supply operates - 640KHz - on first inspection a lot of this output hash appears to relate to this power supply frequency and related frequencies. I can't see anything much that appears to relate to switching hash which would be heavily correlated to Fs. However, I am out of practice with this sort of thing and somewhat archaic in my views, I would (if I ever went back to engineering) much prefer to use a good wideband analog CRO to look for switching glitches. These FFTs and the square wave results for the IFI micro do raise questions about the design of circuitry surrounding the DSD1793. In order to compare it with the PCM1795 (in my view the only latter design in the BB range), I would have to find some measurements where it is implemented in a similar way, ie. same clocking and as far as possible output stage (optimized for overall filter performance - I/V, of course, can't be the same). I have not found this information anywhere. Okay, I shall wait for future postings on development and related performance measurements. So a DSC2? I shall apologise to many people in advance but I just don't see the justification for the huge prices of much of modern high end audio. Sure it costs money to do design, development, marketing (Gods! so much marketing) but in my opinion the value for money is often sorely lacking. This is all hidden in marketing terminology which often describes minor variations on old themes as breakthroughs of breathtaking proportions. This is especially galling when the manufacturers tout the benefits of very expensive passive components while using, as you said, a $2.50 chip as a major component. Apart from cost cutting and a reduction in power usage, I can see find no reason to use the DSD1793 over the PCM/DSD1792A, the PCM/DSD1796 or even the PCM1795. Wow, the IDSD Nano IMD doesn't look too good, I will have a look around and see if I can find some similar results. I greatly appreciate you posting your measurements but I sometimes have difficulty comparing apples with apples as you don't always show the same measurements for the various DACs. I understand you are often looking at aspects of the DACs that particularly interest you but to be fair to the DAC manufacturers any comparison should be between identical measurements at (as close as possible) identical test conditions. Then again, finding any test results at all in the mess that is the internet is good thing. Again, no complaints, just comments. Regards, Bob
  7. I have had a quick look. Although under some circumstances (probably more common with HQPlayer) the performance of the software player and DAC can be inexorably linked, to me they are quite separate. I would like to keep as close as possible to the topic of DAC chips or discrete implementations. I would agree that the design of the chip can greatly affect what needs to be done in the rest of the design, e.g. does it include the I/V stage thus fixing the performance of this part of the design, or can the digital filters can be bypassed thus increasing flexibility, etc.. How a software player interacts with the DAC chips is far more distantly linked and it think it is off-topic. In contrast, Miska's DSC1 (and the still unreleased DSC2?) is very much on topic, both as a design and as a demonstration of principles. I will probably be lambasted for this but I treat commercial products differently from DIY projects. I don't see the necessity to constantly praise people for the effort they put into commercial products, after all they are getting paid for that effort (whether it is commensurate with the effort expended is another issue). I will praise the product if it works well and represents good value for money. Regards, Bob
  8. Hi, Thanks for the suggestion, I have never asked about borrowing demo units from a retailer. I am limited in options as I am currently traveling around Australia in a motorhome with my wife and have been doing so for just under 10 years. This is why the Stax are in storage, I doubt they would survive the environment, I have had 2 pair of dynamics headphones fail already. I know what you mean about relaxing, I always find that no matter how good the sound, if headphones are not comfortable I can't relax so they are a fail. I like to get out in the open away from distractions and just flow into the chair thus the requirement for a battery powered DAC/headphone amp. Regards, Bob
  9. I take it you are referring to your DSC1. Yes, a strangely appealing design. >>:-)>> Although the BB DSD capable chips appear to use a similar design philosophy to the DSC1, they do not have the classic comb filter response. I have not got much further with my investigation of the BB chips but I suspect that they use some form of 'multi-phase' FIR output stage to improve the performance over the more straight forward approach and it is this which results in the lack of regularity in the filter response. As to the DSC1, well you are facing the daunting prospect of getting the last 5% of performance out of it without making it too elaborate of over-engineering it (read unnecessary cost). I will be interested (if you go ahead with DSC2) to see which way you go, especially with the PCB layout. I don't know what PCB tools you have access to but working out the absolute best layout regards, jitter, digital noise/crosstalk, etc. is going to be no small feat. Of course, all those chip designers had all those incredibly powerful design and test tools to work with. They also had the great advantage of miniaturisation on their side. I know most people get concerned with the close spacing in chips but of course this is accompanied by smaller feature size and lower currents (and potentially voltages) - one has only to look to the incredibly shrinking microprocessor feature size and its related shrinking power consumption to see the advantages of reduced feature size. I have had a look at the DSC1 measurements, they look very good. I don't know much about your Pico test rig regarding it residuals and resolution but could the J test and IMD results stand to be improved? This is not criticism as I applaud you for producing the design in the first place and putting it in the public domain in the second. Unfortunately, circumstances preclude me from contributing in any meaningful way. These same circumstances mandate that I buy all my audio gear these days - such is life. Regards, Bob
  10. Hi Jud, A designer could provide their own digital filter (or not) and bypass the filter of the PCM/DSD179x but the chip is still there. Alternatively, a designer could convert everything to DSD256 and bypass 90% of the PCM/DSD179x functionality but the chip is still there. Alternatively, the designer could convert everything to DSD and use a simple 1 bit switch to do the D/A and finally get rid of the chip. Another alternative is to do all the digital processing in the FPGA and implement a ladder DAC in discrete components. However, there are both advantages and disadvantages of using ICs for DACs, especially ladder DACs. The designers of all those ladder DAC chips spent a lot of time optimizing their circuit designs, I find it difficult to believe that individuals or even small companies could devote the resources required to emulate the work done by the major chip designers over quite a period to time - of course, I could also be quite wrong. Regards, Bob
  11. Hi, The documentation dates of the BB chips suggest that all the chips except the PCM1795 were designed at the same time. The PCM1795 is possibly just a later redesign of the PCM1796 to provide 32bit precision for the digital front end. The thing I don't understand about the unit pricing is the difference between the DSD1792A and the PCM1792A as they appear to be, from the datasheet performance, the same chip (apart from the digital interface). Indeed the PCM1792A appears to offer more functionality that the DSD1792A. I really think that all these chip design are based around a small number of functional blocks cut and pasted together to create the different versions. For instance the higher current versions could be easily fabricated using two paralleled output stages. The different digital filter responses are also quite easily implemented. It would be interesting to see if TI/BB are still using the same foundry process to manufacture these chips today as they did at the beginning of production. We could tell a great deal from just a visual inspection of the die of the different chips. I will give your measurements of the IDSD (DSD1793) and the Teac (PCM1795) a look and try to work out what they imply. Of course the PCM1795 has a much better PCM digital filter and doesn't have the compromise of internal I/V convertors. I will see what I can find about the ESS chips but ESS appears to be quite secretive. Thanks for the reply. Regards, Bob
  12. I just pulled those slew rate figures out of the air to illustrate a point, I didn't mean them to apply to DACs. I am aware of the relatively low slew rates found in analog audio. Nevertheless, when it comes to D/A conversion, issues relating to switching transients (much faster than the related analog signal transients) can become a problem. I doubt that modern DAC chips still suffer from these switching transients in the current switches of the DAC chip output stages but I don't know about discrete implementations. Regards, Bob
  13. Auditions are the best way to make good decisions but I find it difficult to make careful assessments in retail premises for all the usual reasons. Most retailers do not have the facilities to allow good testing against a known reference not even with my own headphones. Unfortunately, I have limited access to other situations where I could properly audition DACs. So I am pretty stuck with regards the listening test. I will have a look at the Chord Hugo. Regards, Bob
  14. While it is true that that execution is important, I would think it more likely that poor execution can make a good chip bad rather than good execution making a bad chip good. The best possible solution is good execution and a good chip. For example, if an application requires a slew rate of 20V/ms but the op-amp only has an internal slew rate of 18V/ms then no amount of execution will ensure that the application operates properly (short of something like adding an extra voltage gain output stage). The op-amp could be replaced by one with a much higher slew rate but then the application may also fail because of execution problems (compensation and layout problems could cause high frequency oscillation, etc.) So the best result is to use a faster op-amp and then get the execution right. The slew rate of the op-amp can be found from the datasheets. In DACs the chipset puts limitations on how the DAC will perform and no matter how good the execution, the limitations will remain - whether these limitations are important or not is another issue. So I first look at the datasheets to get an idea of what limitations the chipset puts on the resulting DACs. But, as I said, looking at the specs for the chips was just the start. Regards, Bob
  15. I am starting to look at D/A convertors (finished units) after being away from audio for some time. To get an idea of what to buy, I have started to do a survey of the available D/A chips. I am aware that there are many other factors apart from the D/A chipset to consider but this is where I thought I would start. I am looking at chipsets that support at least 24bit 192K PCM and native DSD, or as close to it as possible. At this time, I have ruled out discrete implementations although this may change. Does anyone know of an up to date survey of all the chips currently in production? I have started off my own survey with the seemingly large TI/BB catalogue. I have downloaded and looked at the data sheets to try and get a handle on what is going on in the audio D/A range. Fortunately it appears that the rather large number of chips can be sorted into distinct groups (based solely on the datasheets). I have grouped the chips together based around their performance specifications, filter implementations (both digital and analog), and analog outputs but independent of their digital interfaces. I have also listed the 1k prices and datasheet dates. Apart from the PCM1795, all the chip datasheets were released in 2003 (anyone know when the chips were designed/released?}. I have come up with the following groups: group1 DSD1791 $2.36 Mar03, DSD1793 $2.36 Mar03, PCM1791A $2.22 Mar03, PCM1793 $2.18 Mar03 - no DSD. low cost, Built in I/V convertors, PCM stop band rejection 82DB, DSD filters are same as next group, DR 113DB at 2.1VRMS, THD+N 0.001 group2 DSD1792A $11.18 Feb04 (DSD1792 Mar03), DSD1794A $11.18 Aug04 (DSD1794 Mar03), PCM1792A $7.08 Feb04 (PCM1792 Mar03), PCM1794A $7.08 Aug04 (PCM1794 Mar03) - no DSD. high cost, current output 7.8mA P-P, PCM stop band rejection 130DB, DSD filters are same as previous group, DR 127DB at 2.1VRMS out DR 132DB 9VRMS Mono, THD+N 0.0004, datasheet specifies different analog output filters for PCM and DSD to realize maximum performance group3 DSD1796 $3.10 Dec03, PCM1796 $2.70 Dec03, PCM1795 $2.90 Mar09, (3.9mA P-P current) 32bit? low cost, current output 4.0mA P-P, PCM stop band rejection 98DB, DSD FIR filters appear to be simpler with lower rejection ratio compared to the previous groups, DR 123DB at 2.1VRMS, THD+N 0.0005, datasheet specifies different analog output filters for PCM and DSD to realize maximum performance. From the datasheets, to me all the chips appear to be pretty much the same design with relatively small variations. From the above I think that it is a reasonable assumption that all the chips in each group will have the same potential SQ and that all the chips will conform to a family SQ. So a start, but, I have really only had a preliminary look at the datasheets, plenty of reading left to do. One thing I have not mentioned is that my requirements are for a battery powered (trans)portable DAC/headphone amplifier. It does not have to be super portable, I don't listen to music on trains etc. but I do like to listen to music outside with open backed headphones. I mostly use a Win8 tablet/laptop convertible as my music source. Obviously, one DAC I am looking at is the IDSD Micro. So perhaps with this in mind I had better look at the current requirements for the DAC chips, in this case at fs=192K DSD1793 etc Idd 28mA Icc 16mA - no extra power required for I/V DSD1794 etc Idd 45mA Icc 37mA - extra power required for I/V PCM1796 etc Idd 21mA Icc 20mA - extra power required for I/V Hmmm, the DSD1794 group have considerably more power requirements compared to the other 2 groups with the DSD1793 being the most economical (taking into account the internal I/V convertors). Indeed, given the low analog current requirement of the DSD1793, I wonder what the compromises were. So that leaves, AD, AKM, Cirrus Logic/Wolfson, ESS to look at (am I missing anyone?). Again, has this all been done before? Bob
×
×
  • Create New...