Article: Reality Quest: Going to Extremes with the Taiko Audio SGM Extreme (Part 3 of 5)

[ray-dude]

View full article

[taipan254]

Quote

Shifting to some favorite performances on shoddy recordings, Shamas-Ud-Doha Bader-Ud-Doja (from “Shahen-Shah” by Nusrat Fateh Ali Khan) is absolutely vivid and present and (incredibly) real! Dynamics off the chart, making any noise in the recording just not matter any more. A glorious performance channeling the angels, brought to vivid tangible life...wow!

 

I love this track!!!

[Marcin_gps]

@ray-dude what a great and comprehensive report! And great to read that you're enjoying  JPLAY  Looking forward to the next part. 

[matthias]

@ray-dude

Very nice review 🙂

 

Maybe I missed it, but anyway:

 

Did you compare directly bit-perfect (no upsampling) with vs. without HQPlayer with the Extreme as a single-box solution (not as endpoint)?

 

Thanks

 

Matt

[DuckToller]

"The Extreme is quite simply a Next Gen Tesla Roadster in a Gen 1 Prius world."
Love your work and your automotive analogy ... 😉

[ray-dude]

13 minutes ago, matthias said:

@ray-dude

Very nice review 🙂

 

Maybe I missed it, but anyway:

 

Did you compare directly bit-perfect (no upsampling) with vs. without HQPlayer with the Extreme as a single-box solution (not as endpoint)?

 

Thanks

 

Matt

 

Very much so!  HQPlayer does all the good things that we love about HQPlayer. At least for my prefered settings (detailed in the HQPlayer section), Extreme barely breaks a sweat.  My demands on HQP are pretty modest though

[AudioDoctor]

Another great write up.

[bbosler]

I'm worn out just reading it. You must be exhausted after doing it.

 

You mention the Hugo Upscaler but I don't see it in your Part 1 diagram. Are you not using it or does that come in later?

 

I'm curious why you would choose a Monoprice USB extender to connect the Extreme to the Dave, and more fundamentally why USB at all? Since both devices will handle digital via BNC or AES, or at least the Extreme will if you add that option,  it seems to me doing the digital conversion to USB out of the Extreme and back in the DAVE can't improve anything? In my mind, the use of USB is for when you have to, not the first choice when you do have choices. Eager to get your  and anyone else's take on it.

 

While we are at it... a shout out for Milcho Leviev on MA recordings. As a huge Art pepper fan I discovered him on the Live at Ronnie Scott's "Blues for the Fisherman" sets and was excited to see him on MA. I only have the one available to download with the other 2 set to arrive in a few days. 

 

BTW the Ronnie Scott stuff is superb with the Pure Pleasure vinyl box set a delight.

 

https://www.analogplanet.com/content/art-pepper-ronnie-scott-stand-issued-complete-0

 

 

 

 

[The Computer Audiophile]

 

3 minutes ago, bbosler said:

it seems to me doing the digital conversion to USB out of the Extreme and back in the DAVE can't improve anything?

 

What do you mean conversion? It's data going from the Extreme to the DAVE. There is no intermediate conversion to an audio format then back to USB data. 

 

 

3 minutes ago, bbosler said:

In my mind, the use of USB is for when you have to, not the first choice when you do have choices.

 

Why do you feel that way? I'm usually the opposite. 

[ray-dude]

1 minute ago, bbosler said:

You mention the Hugo Upscaler but I don't see it in your Part 1 diagram. Are you not using it or does that come in later?

 

I used to have both a Chord Blu2 and a Hugo mScaler (both do upsampling to 1M taps).  I've since moved to HQPlayer for upsampling.  I use the sinc-M filter (1M taps linear phase) with the 15th order noise shaper (close to what is implemented in the Hugo mScaler), but I continue to experiment with upsampling parameters with HQPlayer (esp. as Jussi keeps adding new goodies).  At this point, I'm fully on addicted to the clarity and reality that comes from 1M since filters, so they're going away any time soon.

 

4 minutes ago, bbosler said:

I'm curious why you would choose a Monoprice USB extender to connect the Extreme to the Dave, and more fundamentally why USB at all? Since both devices will handle digital via BNC or AES, or at least the Extreme will if you add that option,  it seems to me doing the digital conversion to USB out of the Extreme and back in the DAVE can't improve anything? In my mind, the use of USB is for when you have to, not the first choice when you do have choices. Eager to get your  and anyone else's take on it.

 

DAVE does support 384kHz on a single BNC input, but needs dual BNC connections for 768kHz input.  The Chord upsamplers (Blu2 and HMS) both have dual BNC outputs, but that is not a broadly adopted standard (at least I'm not aware of it if it is).  USB is the best way I've found to deliver 768kHz 16fs audio to DAVE, although I'm sure we will all be dancing in the street when a new standard for digital audio emerges.

 

7 minutes ago, bbosler said:

While we are at it... a shout out for Milcho Leviev on MA recordings. As a huge Art pepper fan I discovered him on the Live at Ronnie Scott's "Blues for the Fisherman" sets and was excited to see him on MA. I only have the one available to download with the other 2 set to arrive in a few days. 

 

 

Milcho's "Man From Plovdiv" is one of those very special albums that everyone should experience...incredible performance and artistry, other worldly recording.

[matthias]

43 minutes ago, bbosler said:

In my mind, the use of USB is for when you have to, not the first choice when you do have choices. Eager to get your  and anyone else's take on it.

 

Please have a look at Mike Moffats post here:

 

Matt

[Account Closed]

3 hours ago, ray-dude said:

 

Very much so!  HQPlayer does all the good things that we love about HQPlayer. At least for my prefered settings (detailed in the HQPlayer section), Extreme barely breaks a sweat.  My demands on HQP are pretty modest though

HQPlayer does have very modest system demands for PCM but for SDM (DSD) the demands are much greater especially if one wishes to use the EC modulators. At DSD 256 using ASDM7EC there has to be at least two cores capable of running at 4.0 Ghz continuously. The filters can be distributed among multiple cores but the modulation needs to run on a single core per channel (all according to Jussi's posts that I have followed). I have not looked at the specs for the CPUs running here but if one wishes to use EC modulation at DSD 256 (anything higher like 512 EC is not possible right now as there is no CPU that can run at 8 Ghz) then it would be necessary to verify this capability.

[bbosler]

49 minutes ago, The Computer Audiophile said:

What do you mean conversion? It's data going from the Extreme to the DAVE. There is no intermediate conversion to an audio format then back to USB data. 

 

"Conversion" was the wrong word. Maybe I know just enough to be dangerous, but USB was never designed with audio in mind, it is like ethernet where packets of data are sent with no real regard to timing. If the data file gets there intact it is a success.. don't really care when the packets get there, just that they do. On the other hand with audio it is all about timing which has led to various schemes to overcome the inherent problems with USB timing, like the asynchronous mode your Rossini uses. So you are correct, it is not converted, but broken into packets and sent that way. On the other hand spdif and AES are streams of data with embedded timing. Just seems like a better way to go than breaking down into packets then reassembling at the other end.

 

That said, perhaps the problems associated with overcoming timing issues with spdif are greater than those inherent with USB. I'm thinking with extremely stable clocking solutions like the Mutec REF10 to control the data transfer or what I would hope would be extremely stable clocks in a $26K server, that spdif might be a better solution.  So at this point I am officially in over my head, but given the extreme measures audiophiles are taking with USB (Sotm, ISO Regen, the new Innous, and others) I don't see where USB would be the goto solution to transfer data.

 

I now see the Dave will do 384K on BNC but only 96K on AES which is a bit odd, but realistically... is there any performance difference from upscaling to 768 you can transfer via USB vs  384 via BNC? That's a question, not a statement of fact,

 

Like I said, maybe know just enough to be dangerous, but 

 

 

 

 

[matthias]

3 hours ago, ray-dude said:

 

Very much so!  HQPlayer does all the good things that we love about HQPlayer. At least for my prefered settings (detailed in the HQPlayer section), Extreme barely breaks a sweat.  My demands on HQP are pretty modest though

 

I asked because the Extreme has been designed particularly for bitperfect playback and not primarily for use with HQPlayer.

 

Matt

[ray-dude]

26 minutes ago, bobflood said:

HQPlayer does have very modest system demands for PCM but for SDM (DSD) the demands are much greater especially if one wishes to use the EC modulators. At DSD 256 using ASDM7EC there has to be at least two cores capable of running at 4.0 Ghz continuously. The filters can be distributed among multiple cores but the modulation needs to run on a single core per channel (all according to Jussi's posts that I have followed). I have not looked at the specs for the CPUs running here but if one wishes to use EC modulation at DSD 256 (anything higher like 512 EC is not possible right now as there is no CPU that can run at 8 Ghz) then it would be necessary to verify this capability.

 

Absolutely Bob, and an important consideration for folks that lean into DSD.  I'm a 98% PCM to PCM guy, but I've been happy with sinc-M and DSD7 256+fs for DSD content (I upsample to DSD256).  I'll give the ASDM7EC modulator a go but I haven't played with the EC modulators.  The Extreme has a pair of Xeon Silver 4210 processors (40 logical cores, base speed of 2.2GHz, burst of 3.2Ghz)

[bbosler]

1 hour ago, The Computer Audiophile said:

1 hour ago, bbosler said:

In my mind, the use of USB is for when you have to, not the first choice when you do have choices.

 

Why do you feel that way? I'm usually the opposite. 

 

That is my next experiment. I am using an Antelope Eclipse 384 for my DAC which is superb. Currently USB with an ISOregen. I have a Mutec 3+ USB coming that will take the USB and reclock it into an AES stream for the DAC. I'm hoping that with my REF10 also feeding the Mutec reclocker it will be even better... we'll see. Still starting with USB but my Mac Mini server has no other options. 

 

 

 

[ray-dude]

24 minutes ago, matthias said:

 

I asked because the Extreme has been designed particularly for bitperfect playback and not primarily for use with HQPlayer.

 

Matt

 

Thank you Matt, this is an important point.  Things are obviously not bit perfect after going through HQP (although with the sinc-M filters, there is a case to be made that it is bit perfect ++...see Whittaker-Shannon https://en.wikipedia.org/wiki/Whittaker–Shannon_interpolation_formula )  There are enough cores on the Extreme that you can segregate the HQP processing away from the rest of the system.

 

Difficult to asses SQ impact though.  HQP is improving SQ through upsampling, but it may be stressing other processes with the extra compute, adversely impacting SQ.  I can say that as one gives higher priority to the HQP process, SQ (dynamics, etc) improve.  I think there is still opportunity to improve OS tuning when running HQP (and I've been iterating on this), but on an absolute basis, I prefer to have HQP running to not running (and my gut telling me there is room to improve more still)

 

 

[Account Closed]

12 minutes ago, bbosler said:

 

"Conversion" was the wrong word. Maybe I know just enough to be dangerous, but USB was never designed with audio in mind, it is like ethernet where packets of data are sent with no real regard to timing. If the data file gets there intact it is a success.. don't really care when the packets get there, just that they do. On the other hand with audio it is all about timing which has led to various schemes to overcome the inherent problems with USB timing, like the asynchronous mode your Rossini uses. So you are correct, it is not converted, but broken into packets and sent that way. On the other hand spdif and AES are streams of data with embedded timing. Just seems like a better way to go than breaking down into packets then reassembling at the other end.

 

That said, perhaps the problems associated with overcoming timing issues with spdif are greater than those inherent with USB. I'm thinking with extremely stable clocking solutions like the Mutec REF10 to control the data transfer or what I would hope would be extremely stable clocks in a $26K server, that spdif might be a better solution.  So at this point I am officially in over my head, but given the extreme measures audiophiles are taking with USB (Sotm, ISO Regen, the new Innous, and others) I don't see where USB would be the goto solution to transfer data.

 

I now see the Dave will do 384K on BNC but only 96K on AES which is a bit odd, but realistically... is there any performance difference from upscaling to 768 you can transfer via USB vs  384 via BNC? That's a question, not a statement of fact,

 

Like I said, maybe know just enough to be dangerous, but 

 

 

 

 

The answers to all these questions is very long and detailed but can best be summed up by saying that everything in audio is a compromise. SPDIF/AES are legacy protocols with many issues to overcome and were not well designed in the beginning either. USB was not designed for audio. It is the Universal Serial Bus so it was designed to get all kinds of data into and out of computers. It replaced a bunch of really antiquated interfaces. In this day and age, if SPDIF comes out of a computer it likely has gone thru a USB data bus somewhere along the way. All the external format converters start with USB. Audio is not big enough to get teams of highly paid engineers working on on an interface just for high quality audio data transfer. We have to borrow from what is available.

[ray-dude]

21 minutes ago, bbosler said:

I now see the Dave will do 384K on BNC but only 96K on AES which is a bit odd, but realistically... is there any performance difference from upscaling to 768 you can transfer via USB vs  384 via BNC? That's a question, not a statement of fact,

 

 

I only did this experimentation with the Chord mScaler products (Blu2 and HMS).  When using the 500k upsampling filter, output is 8fs (384kHz) and can connect via a single BNC.  Jumping up to 1M taps (16fs, 769kHz) required the dual BNC connection.  The extra 500k taps had a huge impact for me (although admittedly, I'm definitely all in for upsampling with sinc filters for reconstruction)

 

I have not played with HQP with different length sinc filters and different output frequencies.  I can't speak to 1M taps at 8fs vs 1M taps at 16fs

[bbosler]

14 minutes ago, bobflood said:

everything in audio is a compromise.

 

and there you have it... well said

 

so with everything needing a Band-Aid,  we have all of these companies jumping in to sell us the next, best Band-Aid. You have to love capitalists preying on the phobias of audiophiles

 

Alas, I can't afford any $26K Taiko Band-Aids so I'll play at a lower level

 

Disclaimer: As a Johnson and Johnson employee this was a shameless plug for Band-Aids.

[Account Closed]

Ahh, I know what you mean about $26K Band-Aids. This has reminded me of English Lit class many many years ago (in a place far far away) and a book I had to read called "A Tale of Two Cities" by Charles Dickens. I know that this is cutting edge technology but I don't want to lose my head over it. I don't know what I would have to put on the chopping block to get one. But, it is a far better thing I will do today by ending this post.

[Archimago]

4 hours ago, bobflood said:

The answers to all these questions is very long and detailed but can best be summed up by saying that everything in audio is a compromise. SPDIF/AES are legacy protocols with many issues to overcome and were not well designed in the beginning either. USB was not designed for audio. It is the Universal Serial Bus so it was designed to get all kinds of data into and out of computers. It replaced a bunch of really antiquated interfaces. In this day and age, if SPDIF comes out of a computer it likely has gone thru a USB data bus somewhere along the way. All the external format converters start with USB. Audio is not big enough to get teams of highly paid engineers working on on an interface just for high quality audio data transfer. We have to borrow from what is available.

 

I'm really quite unclear of what everyone's worried about with USB these days... A high speed asynchronous, packet-based transmission protocol like ethernet or USB is great. Both are mature, can be implemented without great fuss or expense, and "universal" as in highly compatible. What would you prefer?

 

Perhaps one concrete example of concerns that can be repeated and confirmed to be a problem might be helpful here to cut through the vague speculations?

 

[bbosler]

2 hours ago, Archimago said:

Perhaps one concrete example of concerns that can be repeated and confirmed to be a problem might be helpful here to cut through the vague speculations?

 

Ok, my concern is very concrete and not at all vague. If you believe what you read in these reviews, one needs to spend many thousands if not 10's of thousands of dollars to overcome the limitations inherent in music servers with a USB output. The USB data stream coming from a 10 year old Mac Mini you can buy for a few hundred dollars contains the exact same data as the stream coming from this $26K server. So if asynchronous is the answer, and that is all controlled by the receiver, our friend Ray-Dude is wasting a tremendous amount of money (and time), and everyone investing in any type of USB re-clocker/buffer/ whatever you want to call it, high dollar audiophile ethernet switch, etc.  is also wasting their money.

 

If they are not wasting their money then there is something these $$$ devices are doing to overcome whatever problems these data schemes inherently have.

 

That said, I don't have a concrete example of what that problem is. Just saying if a $26K server sounds better than the Mac Mini, there must be some problem with USB that it is addressing.

 

So my question to you... are they wasting their money ?

 

 

 

 

 

 

[Archimago]

26 minutes ago, bbosler said:

 

Ok, my concern is very concrete and not at all vague. If you believe what you read in these reviews, one needs to spend many thousands if not 10's of thousands of dollars to overcome the limitations inherent in music servers with a USB output. The USB data stream coming from a 10 year old Mac Mini you can buy for a few hundred dollars contains the exact same data as the stream coming from this $26K server. So if asynchronous is the answer, and that is all controlled by the receiver, our friend Ray-Dude is wasting a tremendous amount of money (and time), and everyone investing in any type of USB re-clocker/buffer/ whatever you want to call it, high dollar audiophile ethernet switch, etc.  is also wasting their money.

 

If they are not wasting their money then there is something these $$$ devices are doing to overcome whatever problems these data schemes inherently have.

 

That said, I don't have a concrete example of what that problem is. Just saying if a $26K server sounds better than the Mac Mini, there must be some problem with USB that it is addressing.

 

So my question to you... are they wasting their money ?

 

 

Let's not detract from this series of interesting "extreme" posts on a rather unique piece of hardware by asking if I think anyone is "wasting" money. That can ultimately only be determined by the buyer. My wife thinks I "waste" money all the time :-).

 

Dual Xeon computers are cool as is and I'm glad that @ray-dude is having fun with his system and writing about it. Arguments can be had another place, another time.

 

Through the years, that idea of USB being "bad for audio" has been repeated countless times. How do we know it's true in 2020? Sure, back in the day when USB first came out it wasn't all that great. I ran into compatibility issues with USB 2.0 in 2000 and my first USB 3 card I bought in Japan back in 2009 was very finicky and could barely maintain anything close to advertised speed.

 

I honestly don't see an issue now with USB after all these years. Except for very cheap DACs, certainly by 2013, reputable USB devices have been reliable and the interface has provided designers/developers with the ability to extend hi-res bitrates, DSD, both ADC/DAC function through a single cable.

 

I was mainly responding to your post suggesting that "Audio is not big enough to get teams of highly paid engineers working on on an interface just for high quality audio data transfer". Not sure what engineering interface issue you're referring to here... If there is an engineering issue, then what concrete concerns do you think engineers should fix?

[bbosler]

8 minutes ago, Archimago said:

I was mainly responding to your post suggesting that "Audio is not big enough to get teams of highly paid engineers working on on an interface just for high quality audio data transfer". Not sure what engineering interface issue you're referring to here... If there is an engineering issue, then what concrete concerns do you think engineers should fix?

 

that was Bob

 

9 minutes ago, Archimago said:

honestly don't see an issue now with USB after all these years. 

 

So that answers the question.. which despite all of the fun we all may be having playing around with this stuff and the idea of "wasted money" may not be the exact way to say it, if USB has no issues as you say then anybody spending money to solve a non-existent issue is by definition,  spending money they don't have to to achieve the same result. If USB has no issues, then one way to say it would be they are wasting money to solve this non-existent issue. If there isn't an engineering issue, then this highly engineered product is not needed.

 

BTW I'm not arguing. I'm asking legitimate questions that I obviously can't answer or I would be making statements, not asking.