Jump to content

DH

  • Posts

    9
  • Joined

  • Last visited

  • Country

    country-ZZ

Retained

  • Member Title
    Newbie
  1. DH

    Amarra 2.3

    OK, so I just did some more testing, and it appears I can load media into Amarra Playlist from across the network no problem, just not through iTunes. They sound fine, at least on first inspection. So, Amarra's internal buffering routines to the side, that does not appear to be a problem, and I could dump iTunes and use the upstairs server. And that in turn raises the question whether Amarra sounds any better than the Logitech Touch on their own terms. Can't answer that since I don't have one. Next I will try to see if I can remotely mount other iTunes libraries. So far that has not worked, but perhaps with some tinkering...
  2. DH

    Amarra 2.3

    Sounds like it is definitely not an average network! I would be interested in understanding better how it is configured. One of my intermediate ideas was that I would be able to mount the server's music library on the MacMini--I was hoping to cash in on what you say, namely the fact that the network ought to be as transparent as the SUN network (or close to, since I don't see that I can have a unified HFS across the network in the Appleverse, but perhaps I am just too out of date). I totally agree with what you say about disk fragmentation. Perhaps I will have to periodically wipe the Mini and restore it. Plus I am not quite the purist I pretend to be, since I have some other stuff running on it.
  3. DH

    Amarra 2.3

    (No update is out, except 2.3.3 from Dec?) Paul, I take your point if by streaming you mean serial transfer. But I was merely making a distinction between streaming as it is used today to describe audio and video transfer in contradistinction to eg TCP, where you have guaranteed transfer but, as a result, time lags. I may be mistaken in this, but I believe that most of the audio and video network streaming protocols that I have occasion to use (Netflix, Apple's various implementations) involve some sort of error correction to avoid time lags. This usage seems to agree with Wikipedia's, though that is perhaps not a point in my favour! The development project I mentioned in my first posting was SUN/UNIX based, in the 80s, so I worked with a diskless workstation on the development network, on the trading floor network, and on the road with stand alone workstations. You always got the "smoothest" operation when the source code and the data were local. The Apple network I am now configuring essentially replicates that network wirelessly, and my experience (=the crappy Apple streaming) seems to indicating that this latency issue is at work here as well, for the same reasons. I simply don't know how to otherwise explain what I've observed, and the more I read about the intricacy of clocking, the more plausible that explanation seems to me.
  4. DH

    Amarra 2.3

    Steve, I agree with most of what you say (eliminate as much as possible from the path), but once that is done, what is simple? Is, e.g. floating point processing simpler that integer-processing? Part of my irritation with Benchmark is that they have fooled people (i.e. me!) into thinking that "bit-perfect" input is all you need. So for instance, according to Benchmark, my iPad will transfer and indeed stream music to my DAC "without losing sonic quality... setup has no sonic degradation or tradeoffs." Now, I haven't tried this yet ($20 dongle required), but I will bet my bottom dollar (or 20 of them) that this is a load of crock, and that on the basis of all the other configurations I have tried. As is pointed out in one of the links I mentioned above, the mere fact that eg Unix will not under normal circumstances distinguish between memory calls for data or code, you can see right away that there is in fact a lot of complexity that has to be dealt with, but which the DAC manufacturers obfuscate for marketing reasons. So yes, minimise extra components, but don't infer that a focus on simplicity will solve all problems. I would indeed go so far as to say that the problem of transferring these mathematical objects into temporal objects (if I may be Aristotle to Plato) is fundamentally insoluble--time lags will always be present--and that there is still a long way to go before we even get close. (http://www.benchmarkmedia.com/discuss/feedback/newsletter/2011/06/1/ipad-streams-high-resolution-audio-dac1
  5. DH

    Amarra 2.3

    Hi Paul, Please accept the below in the spirit of cooperation and dialogue. First a technical point, namely that one must distinguish between streaming and genuine data transfer. So it is incorrect to say that “unless you have the music file memory resident, you are *always* streaming it from somewhere or another.” Data transfer must be error-free. Audio data *should* be error-free and correctly timed—the bits must arrive at the DAC more or less at the right time. DAC manufacturers such as Benchmark (my DAC) obfuscate this issue by claiming that their reclocking eliminates the latter problem, but this is misleading, because unless you have an infinite buffer, timing mismatches between arriving data and the data being grabbed by the DAC will eventually manifest themselves. In streamed data, error-correction is sacrificed for time-accuracy. In a standard data sourcing (off disc, network), time accuracy is sacrificed for bit-perfection, but it doesn’t matter since everything downstream is asynchronous. The relentless focus on bit-perfection sweeps the critical timing issues under the rug. The problem for us freaks is that we need both. This cannot be achieved unless the data are local. Best is memory resident, second best internal drive resident, worst across a network with high latency. You have a wired network, which is better than WiFi (which I gave up on), but the latency is still extremely high. As discussion here makes clear (http://www.extron.com/company/article.aspx?id=digisigproc_ts), standard computer architectures are asynchronous, thus the problem is that your PC (mac, whatever) has to be turned into a DSP, which can only be achieved with varying degrees of success. James Moorer basically developed all the systems that underly modern DSPs. Interesting read here (http://musicthing.blogspot.com/2005/05/tiny-music-makers-pt-3-thx-sound.html). What Amarra is doing is turning your computer into a DSP, which means turning it into a synchronous device sourcing data off the local drive, with a large buffer (cache), and then sending it in the correct time sequence to the DAC, error-free. There are also the filter functionalities, but I haven’t started looking into that yet. My impression is that Amarra uses double-precision integer-processing in preference to floating point (as does Pure Music and most other progs out there), but here again I have more reading to do. The architecture I gave up on is close to what you describe. I have a server bank upstairs, various Macs and PCs, iPads and iPhones. All can stream source data to each other, and they can also stream it onwards to other devices for “display” i.e. rendering on whatever DAC that machine drives. But I began to notice very early on that the quality of these streams varied depending on which machines were in the path, and that made me realize that the streams were very far from error-free. Thus my first step was to drive the Benchmark DAC with a directly connected USB source (eg a MacBook Pro with data locally resident). This led to dramatic improvements, however there was still a lot of high-frequency noise (a lot = detectable to a freak). This is when I started reading up on Pure Music, Amarra etc. In short, I think that given your system, you could squeeze a lot better sound out of it if you drove your primary DAC from a local computer that can feed it with synchronous error-free data. This does not mean giving up the capacity to stream (I still have that functionality across all devices using the standard OSX and Unix components). But it does mean that when you want the highest quality, you can have it.
  6. DH

    Amarra 2.3

    I don't know whether my system is better. It's pretty old. On the other hand, every time I go to a store, the speakers sound worse than the KEFs. I know there are newer better speakers out there (love to get Linkwitz Labs stuff), but I've never heard them. In any case, I am not here to compare the size of my...speakers. ;-) But regarding the Touch--it appears not to have any local storage. I no longer believe that any network streamed sources are acceptable, though I was originally convinced of the opposite. Pure Music also does not accept streamed sources, and it's the only thing I've heard that comes close to Amarra (and I hope we are always talking about the Full version here--my impression on testing was that the Full is the actual product).
  7. DH

    Amarra 2.3

    I think this passage may give some indication of what is going on, but this is just speculation at this point. From: 48-BIT INTEGER PROCESSING BEATS 32-BIT FLOATING POINT FOR PROFESSIONAL AUDIO APPLICATIONS (1999) - James A. Moorer Sonic Solutions "With recent improvements in DSP cost/speed performance, it is now reasonable to use integer DSP processing of two 24-bit words in parallel. We can perform all the calculations in a signal path at 48-bit precision. Of course, only words up to 24 bits can be exchanged with DA/AD converters, although any size sample from 24 through 48 bits (or more) could hypothetically be stored on a hard disk (with corresponding reduction in the audio capacity of the disk). It is worthwhile to compare 48-bit integer processing with 32- bit floating point and characterize the differences. It is well-known that for linear processing, the error in floating-point computation is equivalent to error in the coefficients that multiply the signal. Analysis of error in linear computation then becomes an analysis of coefficient sensitivity. In integer computation, there is generally a tradeoff between roundoff error and dynamic range. One of the complications in integer computing is that this tradeoff must be considered at every stage of the process, thus placing additional burden on the designer. In floating-point, this tradeoff is made automatically, which can have unanticipated side effects as well. [...] In 48-bit integer processing, we can divide the word up into three portions: some number of high-order bits will be used for dynamic-range, or “headroom.” Some number of low-order bits will be used to keep track of a the fractional portion of the sample. These are sometimes known as “guard bits.” The remainder of the bits form the audio sample. One reasonable allocation of a 48-bit word would be a 32-bit audio sample with 8-bits of headroom and 8 guard bits. This would allow summing of up to 256 channels of audio without overflow."
  8. DH

    Amarra 2.3

    Obviously I have not tried every device out there, but I have tried many permutations of hardware and software up to this point, and can say with some confidence that a Logitech Touch is most unlikely to match up with Amarra. I recommend that anyone interested in these issues visit James Moorer's website (http://www.jamminpower.com/jam.html). The issues dealt with there are most certainly not incorporated into any Linux or other OS standard software. Read up on his work at Lucasfilm and the design of the ASP processor there (also "Deep Note"). The last six months have forced me to completely revise my comp sci biases, i.e. all DACs sound the same, the issues are minor. It seemed to me that with sufficient buffering, WiFi ought not be an issue. But I now have to concede that all that was a mistake. There is a lot more going on here, and the problem is that very few people are expert in digital, analog, and the musical aspects. My suspicion is that the guts of Amarra, particularly in Playlist, are completely proprietary, independent of most of the OS. If I see another company out there with the track record and product depth of SonicStudio, I will certainly audition it. And then there is just the evidence of my ears, which all my friends confirm.
  9. DH

    Amarra 2.3

    I’d like to thank all the posters in this thread, whose opinions I’ve made much use of over the past six months as I have configured my home network, trying to integrate all audio and video without compromising on quality in any way. Here are the results of my experience, from which yours may of course deviate. At least as far as audio is concerned, Amarra has been the capstone to this project, which involved dumping my first solution--streaming audio to an AppleTV then into a Benchmark DAC--in favour of a dedicated MacMini feeding video into a SONY TV, with audio over USB into the DAC. I did this once I tried out the various players discussed on this thread and others, and realised that WiFi latency was incompatible with good sound, and that the Apple audio software (and hardware, i.e. AppleTV) had to be taken out of the out of the path. I still have the AppleTV connected through optical to the DAC for streaming photos, videos off local mobile devices. Frankly, I am puzzled by many posters’ apparent indifference to the sound of PureMusic and the lower end versions of Amarra in comparison with Amarra Full. On my system (KEF105s, Perreaux amp, BM DAC1) the difference between Amarra Full, particularly in Playlist mode, and all the other options I have tested over the past 3 months can scarcely be overstated, and that was also the immediate reaction of all of my non-audiophile friends. For me, Amarra sounds like nothing else I’ve heard. I have had this (by now vintage) system for 26 years, and during that period I have had a variety of high-end CD players, turntables, etc. Listening to Amarra is like listening to an entirely new system. I feel as if I have been regiven my 700 odd CDs, since each one sounds new. Again, I know most of this music inside out, so to hear entirely new instrumental parts is quite a surprise. This was not the case with PureMusic, nor with the other Amarra versions, and here again all listeners agreed immediately on comparing. My wife liked it so much she offered to buy Amarra when the trial ran out, and was unimpressed by my observations that Pure Music was perhaps better wrt price/performance. This has never happened to me with a tech product of any sort, including audio hardware. ;-) I also have had no installation problems, and no hiccups since buying (first-time) the new release. 2.3 (?), which I was using as a demo, did have hiccups, but these have disappeared with the new release. I am at present listening to gapless Bach cantatas in Playlist mode with no difficulties. ->However
×
×
  • Create New...