"Memory" player vs. Computer...

[hobieboy]

Not sure where this question should go hence post it in the General Forum...

I've been trying out different software players for Mac - Decibel, PureMusic, etc.

Everyone of them is definitely better than iTune. And turning on "memory play" is definitely better sounding too.

That got me thinking - Memory play sounds better probably because there's less jitter (no hard disk dependency). So, what about using something like the PS Audio PWT transport instead of a computer? The PWT is supposed to have way better clocks than the one using computer, or it should be as good as how software implements it.

Thoughts?

 

[Mr.C]

An asynchronous interface uses the clock from the external device (dac or spdif converter) rather than relying on the computer clock. Even with these devices (ie. improved clock) there are still differences heard with better software. The PWT defeats the purpose of computer audio because you need to spin a disk. The same technology with more convenience can be achieved with a computer.

 

[The Computer Audiophile]

Hi hobieboy - I think a few of your assumptions may be incorrect.

 

"Memory play sounds better probably because there's less jitter (no hard disk dependency)."

 

I don't believe memory play has anything to do with jitter. The data in memory doesn't have a clock signal associated with it yet.

 

 

 

"The PWT is supposed to have way better clocks than the one using computer, or it should be as good as how software implements it."

 

Not sure what is meant by this.

 

 

 

The PWT has 64MB of memory. Filling this memory is pretty easy with any music over 16 bit / 44.1 kHz. For example the smallest track on Diana Krall's The Girl in the Other Room 24/96 album is 123.5MB. Playing high resolution music on the PWT seems to defeat the memory player aspect unless one considers the fact that data goes through memory all that matters.

 

Comparing computers and the PWT is like apples and oranges. The PWT is really a transport that can read high resolution audio from a spinning disc. A computer or canned music server is so much more. PS Audio prides itself on the fact that the PWT has nothing to do with a computer.

 

One interesting part about the PWT specs is that PS Audio lists "440,000 gate FPGA" as a feature. Hopefully someone with more knowledge of component engineering can jump in here. I've been told that more gates is not a good thing.

 

 

 

 

 

[AudioDoctor]

if you use a PWT, you are no longer using a computer.

 

Are you advertising for PSAudio? Cuz of you are, they also make the PWD which can be used with a computer and network device...

 

[Audio_ELF]

I think I was thinking is something else...

 

[PeterSt]

One interesting part about the PWT specs is that PS Audio lists "440,000 gate FPGA" as a feature.

 

LOL!

 

[PeterSt]

There is so much more to (being) a memory player, that I wouldn't know where to start - were it about PC's/Mac's doing such a thing.

 

Anyway, this is nothing like making a buffer larger and next "have your memory player". IOW, it also is not about "audio plays from - or passes memory anyway".

 

I don't think the real thing exists on a Mac, but I am not as far as having examined them all.

 

Let me say in summerize that when not *all* is done to not defeat the benefit of "memory payback" by all sorts of other counterproductive stuff/activities, it just isn't working out. I hope it is allowed to give an example which is easily checkable : when Audirvana takes 1 minute for a 5 minute track to -in parallel- preprocess all which is needed in order to next don't need processing during playback (huh ?) ... what to say. Still Audirvana would be my choice, meaning that I am not negative about it per se. There is just a few more tads to it to let it really work (for a "memory player"). Not easy at all.

 

Peter

 

[goldsdad]

"example [...] when Audirvana takes 1 minute for a 5 minute track to -in parallel- preprocess all which is needed"

 

Peter, you very well know (but for some reason neglected to mention ), that that 1 minute is when 4x upsampling 44.1 kHz to 176.4 kHz.

 

About 1-2 seconds is the normal preprocessing time for that 5 minute track.

 

Anyway, what is your preferred technique for upsampling without lengthy delay before playback? A tentative guess is that you generate each upsampled sample just ahead of sending it to the audio device driver (or an intermediate buffer).

 

 

 

[barrows]

Indeed, the OP has a very good point. To respond to Chris:

 

"The PWT has 64MB of memory. Filling this memory is pretty easy with any music over 16 bit / 44.1 kHz. For example the smallest track on Diana Krall's The Girl in the Other Room 24/96 album is 123.5MB. Playing high resolution music on the PWT seems to defeat the memory player aspect unless one considers the fact that data goes through memory all that matters."

 

The PWT is a true "memory player" and always (and can only) plays back from memory, at all data rates up to 24/192. The FPGA mentioned is needed as somewhat complex processing must be done to manage the memory, and the reading of the disc. The PWT uses error correction when reading the disc, just like good ripping software, it will go back and do re-reads until it gets the read right, then the data is loaded into the memory. The PWT does not read full tracks into memory and then start playing-it loads the memory dynamically, this is done to avoid significant delays in playing a disc. The PWT is actually a computer, running a Linux based OS, which was custom produced specifically for the purpose of reading the disc, and managing the data in the memory-so the advantage it might have over traditional computer playback is that the "computer" is designed just for audio playback, and optomized for that purpose alone. It is important to understand that the PWT is a true "memory player" all music is asynchronously clocked out from the memory via fixed frequency oscillators (one for each base frequency) and these oscillators become the master clock for the DAC when used via the I2S connection with the PerfectWave DAC. The disadvantage vs. a computer based system, is only that the PWT relies on finding and placing a physical disc in the drawer, once the disc is in the drawer, the PWT shares all the sonic advantages of the best inplemented computer based systems.

 

Note that those who own the PerfectWave Transport and PerfectWave DAC generally report that the PWT via I2S into the PWD gives sonic performance at least equal to, if not better than, playback via the Bridge (ethernet) or via the best USB/Firewire-SPDIF interfaces.

 

I believe the Nova Physics Memory Player works similarly, but I do not know for sure.

 

 

 

[PeterSt]

Man, I am sorry ! Of course I know that ... we know ... :-)

And I even wanted to say it explicitly, but I guess I fell back to "preprocessing" only.

 

A tentative guess is that you generate each upsampled sample just ahead of sending it to the audio device driver (or an intermediate buffer).

 

Naahhh ... If *that* would be the case I would have made more errors in my previous post. So, my means (in XXHE) is a parallel pre-process just the same as in Audirvana, only it takes that elsewhere mentioned "spike" only (fraction of a second). That is for the whole track, or whatever fits in memory and 16x upsampling". When the track doesn't fit, it grabs more underway (tracks can be up to 9:59:59.99 long (9 hours+).

Notice the huge amounts of memory needed when files of 16/44.1 become 32/705.6 (32 times the original file size). Oh, cpu utilization is still a virtual zero during playback; the spikes use 100%, but of course (hehe) they occur when all is idle - in between buffer fills :-).

 

Regards,

Peter

 

[PeterSt]

I believe the Nova Physics Memory Player works similarly, but I do not know for sure.

 

As far as I recall, yes. A bit difficult to interpret the marketing blah though, at least as it was back at the time (which was around the same time XXHE emerged). Whether it was blah or not, at least it *can* work like how they described it (back then).

 

Peter

 

[Audio_ELF]

@barrows ... PS Audio specifically say that "Using an optical ROM reader , the PWT extracts the data off any CD or DVD in bit-perfect condition without using error correction and places that data into a special version of the PS Audio Digital Lens." -- http://www.psaudio.com/ps/products/description/perfectwave-transport?cat=

 

[goldsdad]

Peter, thanks for the reply.

 

You use a series of many very short bursts of high processor, etc. activity when the player is doing nothing else, instead of one lengthy period of high activity while the player and therefore the processor, etc. is busy with the simultaneous task of playing a track?

 

If all is idle (except the upsampling spikes) between buffer fills then the DAC or computer-DAC interface must have direct access to the computer's RAM. Correct?

 

 

[Paul R]

(grin)

 

I would suspect that Pure Music, Amarra, Fidelia, and Decibel all have the same kind of in memory processing.

 

To me "memory player" just means it preloads the data file, possibly converting it to a standard uncompressed format at the same time.

 

Macs, being Unix, do this as a matter of course. If you would get off that Windows kick....

 

-Paul

 

 

[PeterSt]

You use a series of many very short bursts of high processor, etc. activity when the player is doing nothing else, instead of one lengthy period of high activity

 

I don't think so. Or at least that's not what your symantics seem to imply (the "instead of").

You can consider what I do the same as Audirvana, but what Audirvana does (and/or how it does it) takes 150 times more cpu or something. I'm just finished more early.

What you implied, more looks like the opposite of memory playback. So, to be clear : I preprocess 5 minutes in 0.2 secs - Audirvana preprocesses 5 minutes in 60 secs.

 

Of course it needs some smartness to be able to win this "contest", like there are pre-pre processes (FLAC conversion amongst that), proxies/caches and, well, some stuff more which makes it a. unnoticeable and b. doing the real "SQ job" ?

I find the "unnoticeable" the most important and it is here where the development time goes into.

 

Paul, not sure what you were hinting at, but I would do it the same under OS/X. And if I understand you right and your hint is not a hint at all ... then you must be logically wrong, because that is now what I see (at all).

 

I was fighting with PM yesterday. Man, I almost got killed.

:-)

Peter

 

 

[Paul R]

LOL!

 

It was just a small joke - you see everything through "Windows" tinted glass. My first inclination is to see things through "Unix" tinted glass.

 

Pure Music is not known for being deadly, what kind of stuff did you run into?

 

-Paul

 

 

[PeterSt]

without a song.

Sort of.

 

It will be me - I'm just a user ! http://www.phasure.com/index.php?topic=1757.msg18477#msg18477

 

We know users ...

 

[goldsdad]

"...to be clear : I preprocess 5 minutes in 0.2 secs - Audirvana preprocesses 5 minutes in 60 secs."

 

OK, I get that now.

 

"...there are pre-pre processes (FLAC conversion amongst that)..."

 

Am I being dense again, or does that imply that your player requires files to have been specially prepared and stored ahead of supplying them to the player?

 

 

[PeterSt]

does that imply that your player requires files to have been specially prepared and stored ahead of supplying them to the player?

 

No.

Well, yes but you won't see that. Unless you find 8 FLACs comprising of a full "CD" converting within one second noticeable. i7 4 core that is.

 

Am I being dense again

 

Also No. :-)

 

But as I said. It takes some more to be a memory player. And as long as people think PM is not deadly it will take some more comprehension beyond, maybe;

For example "headless" as you Mac people call it. I'm not sure where that sits (so, tell me please). But headless is "xyz", and I don't think there's anything else anywhere like this. Not that this is important, but reading through the forums it seems easy to use fairly strong phenomena without knowing what it should really mean or imply, to next be happy with it and sit back. Like slowly being killed by ...

 

Ok ok, it can't be that bad. All is a matter of getting used to. Right ?

 

Peter (lacking a remote .. just saying ... :-)

 

 

[hobieboy]

Sorry that I wasn't clear enough in why I'm asking...

 

As Barrow pointed out, many of these memory players are in fact a Linux computer built to do one thing (ok - there are other brands/models that I know but I used PWT for convenience) which is to extract data off the disc, then (per what they claim) add a much more precise clock for sending data out to the SPDIF interface.

 

Hence got me thinking that: would improving the clock improve sound?

 

We all know that different music player client (e.g. iTune, Pure Music, ...) makes different sound. I think its because of different filter scheme? different timing in sending date out the USB/FW interface? So would a device that can control timing better than software improve sound?

 

For the record: I'm using a MacMini with 8G RAM feeding a Weiss Minerva via FW. I pinged Daniel Weiss about the different sound. His comments is that if each music player is bit perfect (they are) and the fact that the DAC is "driving" the clock, it should not make any difference. But we all know - they are different.

 

[Miska]

Now with these definitions here, any computer player is a memory player!?

 

And even the cheapest PC mainboard audio interface will read it's data asynchronously out of the computer RAM based on it's own clock. With USB interfaces there's a bit more work to put audio data into USB packets and send it off. But still it is all played from some memory buffer in any playback software.

 

Now I'm a bit confused of this thread...

 

...Regarding idle or processing time, at least for HQPlayer it is not so straightforward, since all the work is distributed across different processor cores. It can utilize up to roughly 24 cores, in parallel, depending on situation...

 

 

[The Computer Audiophile]

So maybe we should try to define a memory player.

 

With the PWT it seems like the audio simply flows through the 64MB buffer. If this is a memory player then what about players with 32MB, 16MB, 8MB ... 640k?

 

 

 

[hobieboy]

I am definitely simplifying, but the key actions to getting bits stored somewhere to the input of a DAC involves: reading the bits, storing-then-"forwarding" or streaming it towards the output. And somewhere in between, apply the upsampling / filtering / "magic" to form the output stream.

 

Now, with typical transport (CD/SACD/DVD/...), the old school is really stream them to the output - everything that needs to be done (read/process/output) happens in real time.

 

Recently, a crop of players claims that they either read the data off the source (disc, USB stick, NAS, ...), store them in memory first. To me this is what I called "memory player". This indeed is exactly like what a computer does. Except - they all claim to use a highly precise clock; plus potentially their secret upsample/filter/... mechanism & that they only do one thing.

 

I'll be the 1st one to admit I don't know where the secret sauce is within PM, Decibel, Amarra, Foobar, ... but I'm simply curious if timing is a major factor that is affecting what we hear?

 

[bottlerocket]

Interesting thread. I've thought about this while trying out Amarra 2.3.1. It occurs to me that Amarra's memory management is the source of a lot of its bugs and glitches. Other players like iTunes, Bitperfect, Fidelia don't talk about memory play, but of course they read and buffer, maybe just in smaller amounts. Fidelia appears to pretty much read in a whole track at the start of each track and then do very little disk I/O in between. So what is the difference? Both Amarra and Fidelia show fairly frequent writes to disk during playback. Why? I have no idea.

 

And then there is memory for the app itself. Amarra takes a whopping 535 meg just for the app itself and another 80 meg for iTunes. Wow. Good thing memory is cheap.

 

[Paul R]

All CD, DVD, Bluray, etc. players read from the optical media to an internal buffer, and from there process the data to whatever format it needs to be, digital or analog.

 

The few I have seen touting this are merely trying to get a marketing leg up on the computer audio revolution that is happening around them, and threatening to make them extinct.

 

I would say a memory player needs to be able to read an entire track into memory, then do any all processing with no further access to the optical media.

 

Of course, that begs the question of whether or not flash based devices, like iPods, are memory players or not.

 

This seems to be a can of worms.

 

-Paul