Thus far, there is no proven benefit for tinnitus, but I am monitoring it closely.
- Rich
-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Review and Measurements of PS Audio PerfectWave DirectStream DAC
- Thread starter amirm
- Start date
Thus far, there is no proven benefit for tinnitus, but I am monitoring it closely.
- Rich
At one time I owned a PS Audio Digital Link III DAC, a GCPH phono preamp, a GCHA headphone amplifier, several xStream Prelude SC power cables, several xStream Transcendent XLR interconnects and a pair of xStream Transcendent bi-wire speaker cables. I sold all the cables when I bought a bunch of Liberty Cable Z500 cables for pennies on the dollar. I replaced the power cables with Pangea Audio cables (which have since been replaced with $8 TrippLite cords). I sold the electronics after playing test tones from a test LP to setup my system. I heard the GCPH, a highly reviewed and well regarded phono preamp, clearly and loudly playing harmonics. I pulled out my iPhone and used my freebie dB Meter app and saw a whole string of harmonics. The second was only about 20dB below the 1kHz tone in loudness. The tones seemed to drop about 10dB per harmonic. I couldn’t believe what I was seeing and hearing. My new phono preamp plays the same 1kHz tone with no significant nor audible harmonics. It was on that day I discovered for myself that PS Audio produces garbage. The only good to come out of all this is that I was able to sell everything for what I spent. (I bought used.)
Martin
Martin
OP
- Thread Starter
- #1,603
No way is the output of the compiler different in each execution of the compiler. It is not like the compiler has a random number generator to spit out different logic every time you use it. I have managed development of ASICs using both Verilog and VHDL and not once did I hear the need to compile multiple times because each execution generated different results. The whole world of ASIC/FPGA development would turn upside down if your compiler generated different logic constructs every time you used it.
Run-time issues of course abound due to race conditions and ambiguity of the language. But that is not what we are talking about here. Their engineer clearly said Ted generated multiple "compiles" and only 3 sounded the best. These compiles would pass a binary comparison (sans headers and such) and so in no way could generate different outcomes. Indeed I think he said that they were the same yet "sounded different." It is like adding 1+2 and saying he is getting 3, 2 and 4. At some point they need to stop and say, "this doesn't make any sense to be so. Let's run a blind test to make sure we are not imagining this."
That's not what I said. What I've heard, and it seems plausible, is that the exact gate layout can change run to run without altering the functionality. Consider a software compiler that emits subtly different instructions sequences on different runs. They don't (usually) do that, but they wouldn't be useless even if they did.
In Canada? No way. In Canada, if you enter a store and do not buy anything (read plunging 1000s of dollars) they have a repertoire how to get even.
Where did you hear this from?
Someone far from anything audio related who was complaining about the Verilog compiler he was using being non-deterministic.
That's because all these sites, audio marts, audiogone etc. are sellers' sites, practically, financed by sellers altogether. Sellers gladly pay the fees and in return the site will ban any whistle-blower who even just mentions Audio Science Review, for example. The way they do it works like this.
There are on-duty fraternity dudes on their forums, such as audio mart forums and others. As soon as you publish some inconvenient truth, e.g. these measurements by Amir, these on-duty dudes start dissolving the discussion to the extreme. There is not just one dude, there are at least a dozen at any point in time. You engage in discussions with this fraternity sticking strictly to the audio engineering subject, but they fry you so much so that after a while a moderator shows up and points to some dozens and dozens of so called "rules" (even Stalin's birocracy had leaner rules and regulations not to mention USA constitution). And long story short - you are banned. Used audio trading is the most lucrative commerce activity in the world, second to none. Audio marts allow sellers to sell the same unit many times i.e. mark them as "Sold" when in reality they are not sold of course (they charge exorbitant insane prices e.g. $100 less than suggested MSRP, which by the way, is always 20% higher than the true MSRP ).
Many other tricks are used. So called "Blue Books" are essentially written (read forged) by those same sellers. Buyers only came to senses in the last few years since Hifishark portal was launched (with the help of portal and extensive googling one can get a pretty accurate picture of true Sold prices).
Other sites are no better than audio marts and audiogon. Audioshark is as bad as it gets. Audioaficionado is a prime example what audiofoolism really is. Audioaficionado was created by a dude who is a softer version of Charles Manson. He has a herd of followers who all call him Master Lu. What Master Lu says is like a Bible. For example, Master Lu pronounced once that $18,000 for Accuphase E650 is a very good deal. And the herd followed their master's epistle to the letter (btw, the Japanese MSRP of this mediocre noisy amplifier is $6000, which in itself is 2 to 3 times more expensive that it should be, especially having in mind how high general prices in Japan are).
KSTR
Major Contributor
Btw, thanks to revelations by Amir and many other forums, PS Audio DirectStream together with Bridge II and 12+ months of warranty can be bought used for $3000. The same without warranty, 3yr old goes close to $2000. But in the light of these measurements, even that price tag is too high.
OP
- Thread Starter
- #1,611
That's run time non-determinism, not compile time which is what PS Audio guy was talking about.
KSTR
Major Contributor
Yes, I understand. Then again, compile time is run time of the compiler so it could happen there as well, theoretically (and that would be a bug, in the end).
I have no expertise in compilers other than gcc and there I have never encountered non-determinism (non bit-identical final executable), provided the build system is properly sandboxed and you have managed to find all the places where someone used __DATE__ and __TIME__ macros in their code.
That is a completely separate issue. To illustrate what we're talking about here, consider a trivial example such a two-input NAND gate. If we compile the equivalent Verilog or VHDL for a specific FPGA, the compiler is free to use any of millions of logic elements to provide this function. Supposedly, some compilers do not necessarily pick the same one on every run. A software analogue would be a compiler that uses different registers to hold a temporary value if run multiple times. The machine instructions would be different, yet the program would function identically.
Since I have heard reports of such behaviour from others, I'm inclined to believe that PS Audio may have actually observed it too. The claimed differences in sound are, obviously, almost certain to be imagined.
KSTR
Major Contributor
With gcc, both for ARM and for x86 platform, I've never seen different binaries from different runs (after following the guidelines) but yes, theoretically it is possible.
What do you mean? What would they do lmao.
Great post. Thanks for sharing.
It is possible, that PS Audio has decided that regular firmware updates energizes the base.
Each release has to have a story to go with it.
The successor will be designed and it will start the cycle again.
- Rich
Each release has to have a story to go with it.
The successor will be designed and it will start the cycle again.
- Rich
I have never seen a non-deterministic software compiler either. That was just a hypothetical example. I do know that HDL compilers use parallel processing to speed up the builds, and it's conceivable that they do so in a non-deterministic way. Again, that does _not_ mean that claims of differences in sound should be believed. As is so often the case with audiophoolery, the purported cause is (possibly) real whereas the effect is not.
Plenty of examples. Off the top of my head, read reviews of a retailer called Planet of Sound
https://www.planetofsoundonline.com/
Google to see what customers say about these dudes. I visited some 7-8 different brick-and-mortar shops. They were all equally bad, so much so that it is difficult to pronounce a badarse winner. Remember that Canadian sellers (I call them Sufferers) are the only sellers in the world who create a separate section in their ads in which they pinpoint how much taxes they had paid for their products and how much you gonna save if you buy from them and save on taxes. I usually tell these dudes to start waving with their ears as that will certainly accelerate their (mockery of) sale. Dudes immediately complain to the Stalin's bureau of the site and the Gulag issues immediate ban order.
hmscott
Addicted to Fun and Learning
- Joined
- May 17, 2020
- Messages
- 604
- Likes
- 493
I was assuming they used different build-time command line switches, different compiler optimizations, perhaps even different libraries (versions) linked in - some houses have a history of using progressively newer releases of libraries over time and keeping them around for regression testing builds.
Anyway, there must have been binary differences in the builds output - they wouldn't be testing 3 copies of the exact same binary.
Similar threads
