PS Audio Music Server In The Pipeline?

Just a comment regarding price.

I am quite comfortable with the given ball park price.

I would even be happy to pay a bit more.

I don’t really care about the prices that other people are charging.

I want everything to be just right:

the proposed software;

the upgradeable software; and

the hardware that I know PS Audio will get just right!

Looking forward to getting one.

Meanwhile I am eagerly awaiting every scrap of information offered.

Bruce

Bruce, if you were referring to the prices I posted, it was with regard to the Raspberry, a mini computer that operates off a micro SD card, and sells for around $35.

It is for diy people. There is a whole subculture out there based around the Raspberry. It’s not just for audio.

rmueller1 said

I would question your interest in building your own interface if only because this would seem to be such a massive undertaking and distraction from what you do best which is great hardware. Frankly, you can’t help but fail at it in someone’s view. Why not let customers just pick their favorite 3rd party flavor and pay for it as an integration? As you point out, these companies and their products are quite ephemeral. A few years ago JRiver was all the rage, now its Roon, tomorrow someone else.


A good thought in the abstract, though not one I can agree with in the practical. First, there are no great interfaces with the single exception of Roon. Everything else out there sucks. Big time. So it’s not like choosing amongst a number of candidates.

Second, we recognize over the long term that the future of our industry increasingly lies in the user interface and hardware will become less of a focus to consumers. Now, before everyone jumps up and runs for the door because that sounds like we’re backing off on hardware, let me explain. As you point out, we’re good at hardware and I will not disagree. We’ve honed our hardware chops over the last 45 years and just recently added another asset to the hardware team, making the company strong as hell in this area. That said, I am committed to climbing higher on the mountain rather than resting on our laurels.

The worst thing we can do at any one time is stop moving forward. We have always strived to change the world. We’re rarely followers. Now that we have hardware under control and where we want it to be, software and user interface become the bigger focal points. And this should not be a surprise. Just look what we did with our DAC, DirectStream. Ted Smith’s wonderful work on both hardware and software say it all.

How we access our music and interface with our machines to enhance the pleasure of music is the next frontier, one we intend to be at the front of. That’s important to us and why we have been building and investing in our team. Over the last few years, we have steadily added more software programmers, and we’re not yet done.

I apologize for the long ramble, but this is hopefully a means of explaining why we aren’t intending to build just another hardware box and let people choose amongst an empty shopping cart of half-baked user interfaces or a single one that is too expensive and doesn’t sound good. If we’re climbing a mountain, let’s keep moving up the hill until we can see the peak.

1 Like
magister said

Paul, you’ve set yourselves worthy goals, both for the upcoming server and for bringing the price down on other products. I wish you every success (and would be glad to help test), particularly with metadata for classical music – that’s really tough.


Indeed. The toughest and I think we’ll make good progress but it’s doubtful we’ll be able to claim victory for classical. We’ve engaged a few experts on classical to help us sort through the challenges, among them Richard Murrison of Bit Perfect and Larry Schenbeck (our professor Schenbeck of Copper).

The challenges with classical are enormous and the metadata thin. One of the fundamental precepts of Octave will be honoring the work users perform. We know that most classical music buffs devote hundreds of hours cataloging their libraries the way they want and our software honors that. Using a lot of complex fuzzy logic strings, our challenge is to allow the users their cataloging and at the same time, have the software know what it is so we can then add further metadata value. That’s a really tough challenge because not everyone catalogs in the same way.

More to come as progress develops.

cxp said

How far out is the Stellar Server?

Will it have the same galvanic isolated outputs? Will it have HD as well or just streaming?


It’s sometime out, hopefully this calendar year. It is completely galvanically isolated and it will have an internal had drive and a front panel disc drive for ripping.

First time in this thread so sorry if this is a silly question:

Is there a chance that you also could provide a software used on a PC just to retag files for whatever software on whatever OS people use for other reasons? Like Songkong, but better?as I understood such a component is part of a music server anyway.

Songkong’s problem is, the media libraries used for automatic tagging deliver only 30% of information and don’t recognize most files.

Seeing the music server as a whole: am I right to assume it competes with all the functionality of Jriver, Roon, Songkong at the same time and additionally ads the HW integration this complete service aims for?

I understand how difficult this is and would be interested if there is any area you don’t plan to go in this scenario.

Not sure I understand the question. I can tell you that one of the biggest challenges in metadata is song ID. In order for the program to do a good job it has to correctly identify the track or album - not how you named it. That’s a challenge because it means we cannot rely upon your data for our ID, yet we must honor your data in the way that data is represented.

To do this we will use something called acoustic ID where a snippet of the track is gathered and sent to a cloud server to be identified based on matching it with a massive library. Our track record so far runs in the high 90 percent, so that’s not bad. Once we know for sure what the track is, then however you labeled it is fine. At least we know that Davis, Miles is really Miles Davis and even if you have him listed as a classical conductor by mistake, and your data will display him as such, all the other things that connect to him like associated artists and tracks and history will magically be correct.

Paul McGowan said

… We’ve engaged a few experts on classical to help us sort through the challenges, among them Richard Murison of BitPerfect and Larry Schenbeck (our professor Schenbeck of Copper).

The challenges with classical are enormous and the metadata thin. One of the fundamental precepts of Octave will be honoring the work users perform. We know that most classical music buffs devote hundreds of hours cataloging their libraries the way they want and our software honors that. Using a lot of complex fuzzy logic strings, our challenge is to allow the users their cataloging and at the same time, have the software know what it is so we can then add further metadata value. That’s a really tough challenge because not everyone catalogs in the same way…


When it comes to the state of classical music metadata in 2017, Mick Jagger of all people summed it up best. “You can’t always get what you want. But if you try, sometimes you get what you need”.

If anybody is interested in contributing to the internal discussion on Classical Music metadata, and how this will be implemented in the forthcoming Octave Server, please PM me. I would be happy to hear from you.

Paul, thanks very much for the strategic insights into how you think about the industry and PSAudio’s positioning. I found it interesting that you see software as the direction for the future. You have far deeper vision into the nature of the business than I do but it seems to me that the switching costs between software alternatives is usually quite low compared to hardware and if you try to lock customers into a proprietary solution they will revolt.

I must agree that all the current user interface software out there sucks (even Roon although that is what I currently use as it’s the best among poor choices.)

I will be interested to listen to what you come up with. I was thinking about putting together one all PSAudio solution in one of my two listening spaces. Currently one has the Directstream DAC and a PSAudio power unit. I had the Memory player but switched to a Musicvault server which is good but probably not the best pairing with the Directstream. I would like to see what your server solution with the BHK amp would sound like. (The other space is all my own home built tube amps and horn speakers.)

In addition to letting us keep our tags the software should allow us to override the song/album identification. Most identification failures are likely to be just that–can’t identify the song/album, but it’s more than possible that an occasional song will be misidentified. I use dBpoweramp to rip and it is quite good at identifying albums but once in awhile it gets the identification completely wrong. No software is perfect.

Thanks Paul,

yes the one question was song identification and how you can get it better than Songkong which uses musicbrainz and discogs but recognizes very few anyway. MP3tagger uses also freedb and amazon and is quite good but manual. DBpoweramp is good but only tags while ripping, not afterwards.

So the problem is until now I haven’t seen a good automatic tagging of albums (not single files alone).

The other question was if you only offer your SW within the bundled HW of your appliance or also as just the software to be used with the bridge instead of Jriver or Roon or even just modular to just use it for retaking without using the server instead of Jriver or Roon.

Richard Murison said
Paul McGowan said .... We've engaged a few experts on classical to help us sort through the challenges, among them Richard Murison of BitPerfect and Larry Schenbeck (our professor Schenbeck of Copper).

The challenges with classical are enormous and the metadata thin. One of the fundamental precepts of Octave will be honoring the work users perform. We know that most classical music buffs devote hundreds of hours cataloging their libraries the way they want and our software honors that. Using a lot of complex fuzzy logic strings, our challenge is to allow the users their cataloging and at the same time, have the software know what it is so we can then add further metadata value. That’s a really tough challenge because not everyone catalogs in the same way…

When it comes to the state of classical music metadata in 2017, Mick Jagger of all people summed it up best. “You can’t always get what you want. But if you try, sometimes you get what you need”.

If anybody is interested in contributing to the internal discussion on Classical Music metadata, and how this will be implemented in the forthcoming Octave Server, please PM me. I would be happy to hear from you.


Kudos for the Jagger reference. It will be interesting to see how this pans out. The differences between many classical recordings of the same piece are not trivial, so I applaud the efforts to bring some sanity to this process. It will be interesting to watch the Octave integration of user & server data. I’ve seen this on Roon, where the system “pulls a Wiki” and mistakenly adopts erroneous user submitted data and replaces valid information. Doesn’t happen to often, but if I have seen it more than once, then it seems reasonable to assume hundreds or thousands of others have as well.

Thanks for the contributions – all good points. It seems that Paul & Co. have thought carefully about this, and the notion of honoring the user’s tags while at the same time having another method to accurately identify a track seems very appropriate. I looked at Roon shortly after it came out and didn’t try it because it would make no use of the tagging that I have so many hours invested in. I would just add two things.

As others have said, no software is perfect. So the PSA server should allow users to adjust the tagging to suit their preferences or correct for errors (not often, we hope). Even if the software finds accurate information while ripping a CD, what it wants to use may not match the user’s preferences. Could it have a program like mp3tag built in? If not, at least the user should be able to access the server’s disc from a networked PC to make adjustments.

Second, there should be reasonable flexibility in terms of how users can access the music and how the metadata are displayed. For example, I can choose music using any of the tags I’ve defined, in any order. In other words, I can go composer --> work --> album if I have a specific performance in mind. If I’m in the mood for some violin music, I can start with my Instrument tag and then choose by composer or historical period or subgenre (concerto vs solo piece). I can store composers as lastname, firstname but have the ‘now playing’ screen display ‘firstname lastname.’ If you haven’t looked carefully at MinimServer, you should – I know of no other software that provides this kind of flexibility. Indeed, Minim may well provide more options than you will want to in your server software, but at least take a look at what can be done.

Yes, I think these are all good and well reasoned requests. As I mentioned, Octave’s “prime directive” is to honor existing metadata on the one hand, but spend quality time identifying the track in its own parlance. This combination of the machine knowing your track in its way, and displaying metadata your way, will make for a robust music server experience IMHO.

Keep the suggestions coming. They are all very helpful. And hopefully a few of you have PM’d Richard to offer your help. Those that do will be first in line to test Octave when we start asking for volunteers - probably next month. We’re getting close.

I would add that Jazz (my favorite genre) is very much like Classical music as described above. ( Charlie Mingus didn’t like the term “Jazz” and called it “Black Classical Music”). It’s a bit like the paradigm in Minimserver where you have composer, orchestra(perhaps arranger in jazz) and soloists. The paradigm of album/tune doesn’t work as there are dozens of versions of standard tunes that are completely unique in performance.

I think you will find that opinions on music library interfaces will be highly nuanced depending on who you are talking to. I wish you the best of luck in addressing this.

In addition, regarding Jazz as one genre (as some software does - or maybe they recognize “smooth jazz” - 2guns_gif) is about as sensible as grouping all classical one genre and forgetting medieval, renaissance, baroque, romantic etc. Tags, need a “period”, like New Orleans, Kansas City Blues, Bop, Post Bop, etc.

Paul McGowan said

Yes, I think these are all good and well reasoned requests. As I mentioned, Octave’s “prime directive” is to honor existing metadata on the one hand, but spend quality time identifying the track in its own parlance. This combination of the machine knowing your track in its way, and displaying metadata your way, will make for a robust music server experience IMHO.

Keep the suggestions coming. They are all very helpful. And hopefully a few of you have PM’d Richard to offer your help. Those that do will be first in line to test Octave when we start asking for volunteers - probably next month. We’re getting close.


Hi Paul,

I’m sure it’s already in the plans, but (1) I would want easy access to change/upgrade the hard drive, and (2) I would like an easy way to link up an external HDD to back up the internal drive every so often.

Looking forward to this next creation. Keep up the great work.

By far the toughest challenge facing the developer of a Music Server is dealing with genres. The main reason is quite simply that there is no consensus as to what specific genre tags should be applied to any given piece of music. There are seemingly gazillions of genre tags in use, there is huge overlap in their meaning and application, and they can change both geographically and over time. On one level all this is intuitively obvious to most people, but on another level the problem gets progressively more intractable the deeper you look into it. For this reason, genre support in the major on-line music databases varies from spotty to non-existent. MusicBrainz, for example, refuses to touch genres with a 10-foot pole.

Until recently this has not been a huge problem. Individual users have generally tagged their own music with genre tags that meet their own requirements, and our user experience has been limited only by the ability of the music server software in use - be it iTunes or whatever - to make use of those tags. But now the landscape is changing. With services like TIDAL able to deliver high quality lossless music, we will be dealing with large music libraries whose genre tags (in fact, whose tags in general) probably won’t be at all consistent with the tags in our own local libraries, and (for all practical purposes) cannot be manually tagged. The server software will be more or less obliged to live with that.

While the Octave server is expected to do a sound job of honoring the user’s pre-existing tags, this will be limited by its ability to recognize the user’s tags and parse them accurately. For example, if you have created a genre tag ‘soul & funk’, should Octave treat this as a single genre ‘soul & funk’ or should it treat it as two separate genres ‘soul’ and ‘funk’? You can nit-pick over the specifics, but if whatever Octave chooses is not what you originally had in mind, would you feel that Octave was not honoring the user’s tags? Ultimately, the decision on handling genres is going to be a compromise between a comprehensive and detailed treatment, and an intuitive and uncomplicated user interface.

JM said

Hi Paul,

I’m sure it’s already in the plans, but (1) I would want easy access to change/upgrade the hard drive, and (2) I would like an easy way to link up an external HDD to back up the internal drive every so often.

Looking forward to this next creation. Keep up the great work.


(3) a shuffle program akin to Roon’s “radio” that kicks in after the selected song or album finishes playing. It does a good job of tracking and playing similar songs. Used often.

Richard Murison said

For example, if you have created a genre tag ‘soul & funk’, should Octave treat this as a single genre ‘soul & funk’ or should it treat it as two separate genres ‘soul’ and ‘funk’?


I would expect any database to treat this as a single tag. One could, however, allow multiple tags in a comma delimited format; i.e., “soul, funk.”

Given how random genre tags have become, perhaps develop a solid list of a reasonable number of tags (30?) into which most music would fall and offer this as a drop-down. Then, if the user has already used additional tags, or adds more, the drop-down expands.

I appreciate this could get nuts if a user employs 100’s of genre tags, but at this point perhaps they deserve it. :slight_smile:

Paul McGowan said Keep the suggestions coming. They are all very helpful. And hopefully a few of you have PM'd Richard to offer your help. Those that do will be first in line to test Octave when we start asking for volunteers - probably next month. We're getting close.
Yeesh, classical, like soundtrack scores, are complicated. There's composer, conductor, soloist (instrument 1 to N and/or vocals 1 to N and/or data for each soloist (instrument, age, category) where N can be as high as 30), Orchestra, Opus number, Common title, track number, and those are just the ones off the top of my head.

–SSW