Miscellany: FlexASIO for 384kHz, Philips 2001 SACD/DSD64 test signals (thanks Black Elk!), and Roon network multicast. [And Coltrane's A Love Supreme Live Hi-Res, Music Industry Crystal Ball, Magic Quantum Fuses!]

A scene from The Simpsons 1996...

Hey everyone, for this week, let's talk about a few "miscellaneous" topics which I've either wanted to mention over the last few months or have just come up as interesting tidbits I think worth documenting but not necessarily large enough as topics in individual posts. The main topics are:

A. Instead of ASIO4All, we can use FlexASIO with the E1DA Cosmos ADC for 384kHz samplerate measurements in Room EQ Wizard.

B. A look at some "standard" SACD test signals from Philips back in 2001. With many thanks and great discussions with Black Elk.

C. Roon needs network multicasting. Check this out if you're running into network issues; I had some problems initially with my ASUS ROG GT-AX11000 and relatively complex home network.

 

A. FlexASIO, up to 384kHz using REW and ADCs.

Instead of ASIO4All, I saw a tip on E1DA's site to try out FlexASIO which will allow operations up to 384kHz (instead of 192kHz max). Yup, the current 1.8 version works quite well and go install FlexASIO_GUI v0.3 while you're at it.

A quick peek at the control panel for set-up:

FlexASIO is a "wrapper" that will allow you to use the ASIO interface with devices through one of the other driver types on your computer system. In Windows, there's the option of using WASAPI, WDM/Kernel Streaming, MME, and DirectSound (ASIO4All uses WDM/Kernel Streaming). I've come across situations where the driver refuses to load due to parameters being incompatible; for example, I've had issues using the WDM/Kernel Streaming setting with the E1DAC Cosmos ADC. In these situations, you can manually delete the configuration file:

C:\Users\<User name>\FlexASIO.toml

And restart the driver config in REW. Or use the FlexASIO_GUI to change the settings.

With those settings above, here's a look at a 1kHz -3dBFS signal from the RME ADI-2 Pro FS XLR, +19dBu level (the older AK4490 DAC model, not the more recent Pro FS R Black Edition) to E1DA Cosmos ADC. With this DAC, -3dBFS yields about the highest THD+N result:

There you are, frequencies all the way to 192kHz. We can of course look at basically any signal of interest including for example the "1/10 Decade Multitone 32":

Within 15Hz to 30kHz, we see better than 118dB distortion/noise-free range (>19-bits) on the ADI-2 Pro FS.

While this works, of course I still think it would be nice for REW to support WASAPI natively, and prefer that the E1DA Cosmos ADC's ComTrue ASIO driver worked in REW!

B. A look at Philips' 2001 SACD DAC Test Disc ("Ichimura & Sekii") - Scarlet Book 1.2 standard.

I mentioned in the previous post using SoX-DSD that I would start using that to encode my DSD test signals, primarily "sdm-8" noise shaping as my "standard". Reader Black Elk who in the past had worked with Sony at the dawn of SACD was able to provide a few ripped samples from this disc for me to examine:

Cool! An actual test disc from one of the creators of SACD based on "version 1.2" dated May 2001 of the standard (you might recall the term "Scarlet Book" referring to the SACD specs). When ripping the DSD64 data, the CD-TEXT disc title was labelled as "Ichimura & Sekii" which I assume must be the names of the engineers who created these synthetic signals; you'll find both these names either in the DSD literature or Sony's 1-bit digital patent.

As you can see in the disc contents, there are various test tones on the disc. So let's compare this "SACD standard" with SoX-DSD sdm-8. First, let's have a look at the noise floor:

Based on the filename, the Philips Test Disc track appears to be using a 7th-order filter. The noise level is similar through the audible spectrum, with the sdm-8 setting keeping noise lower a bit further out from the audible range until 25kHz. The Philips encoding allows noise to rise relatively quickly just above 20kHz. However, notice that the Philips noise floor appears just slightly lower in the audible range compared to SoX-DSD sdm-8.

Let's compare the 1kHz 0dBFS tone (I encoded the SoX-DSD at -0.1dBFS to prevent any clipping):

 

Interesting. As per the noise floor comparison above, we see that indeed the SoX-DSD encoded signal pushes ultrasonic noise a bit further out. It has better THD, but lower THD+N on account of the "N" component with a slightly higher noise level than the Philips v1.2 Test Disc signal.

All in all, we see differences but we're looking at just +/-1dB variations with better than -100dB THD+N so the variance would not really be of significance for music. Note that this is with an actual DAC playback. We could just analyze the signal in the digital domain with DSD-to-PCM conversion in the computer; but that's not as fun as playing it on an actual DAC. ;-)

Here's the Philips SACD Test Disc encoding of the 19 & 20kHz tones, commonly used for intermodulation (CCIF/ITU-R) testing:

Looks clean through the ADI-2 Pro FS. We're looking at about -116dB on that 1kHz "d2L" intermodulation product.

It's quite possible that similar signals were used by John Atkinson back in 2001 when he reviewed the Sony SCD-1 (he talked about using a "provisional Test SACD" from Sony) and got measurements like Figure 15 which is a 1/3-octave smoothed graph using the -60dBFS and -120dBFS 1kHz signals. I see these on the Philips Test Disc as well.

I don't know why Atkinson chose the low-resolution 1/3-octave smoothing back then with the Audio Precision System One; maybe limitations of the resolution and rising noise floor with that equipment when looking at bandwidth >20kHz.

These days, more than 20 years later, we can just grab a hi-res 1M-point FFT and show the accuracy of the DSD signal, 192kHz frequency bandwidth, no smoothing:

No harmonic distortion of concern as expected at these low output levels of -60/-120dBFS.

Thanks again to reader Black Elk for sharing these Philips Test Disc signals from the early days of "hi-res audio". It's great virtually "meeting" folks like him with backgrounds and experiences in the technology over the years through this blog.

We also talked about this 2002 "SACD Surround Sound Reference Disc" which he worked on:

With his permission, this is what he said about it when I mentioned that the disc had some great-sounding tracks on it:

"... It was largely put together by ex-Philips Classics balance engineer/producer Hein Dekker. His own recordings are excellent, but Philips Electronics developed close relationships with a number of top Dutch recording engineers (Ronald Prent for rock/pop; the team at Polyhymnia [formerly Philips Classics Recording Centre] but especially Erdo Groot). They recorded most of the stuff on the disc, and it is reference/demo quality.

One of the many missed opportunities with SACD came with the recording of works by Jerry Goldsmith. While the recordings on that disc (which were released by Philips on some promos, and later by Telarc) are good, they could have been so much better. They were made at Abbey Road with legendary engineer/producer Bruce Botnick (The Doors and a gazillion others). I love Bruce, he's a really sweet guy, but he and Jerry were used to working on major movie soundtracks (either in LA or London), and they opted to work at Abbey Road with Bruce's tube mics.

However, those recordings only came about by chance. We were in Budapest with Dekker (2-ch) and Groot (M-ch) to record Dvorak symphonies with Ivan Fischer and the Budapest Festival Orchestra. We knew the orchestra was going to be rehearsing with another conductor, so the guys were only able to set up their main pick-up (for 2-ch that was 4 mics across the front of the stage, and for M-ch a 5-mic array at 0 deg, +/-30 deg and +/- 110 deg. The two mics. at +/-30 were common to both microphone arrays.

As the orchestra layout was going to be completely different, they could not rig any spot microphones. With the limited work done, we decamped to the M-ch control room, which was the old control room set up by Hungaroton. We could hear the orchestra tuning up through the open door, and Erdo said, "Well, we may as well listen to what they are going to rehearse." So, he opened up the 5 microphones on his array.

As soon as the orchestra had settled they launched into the theme from Star Trek (the movie scored by Goldsmith, but containing part of the TV theme). That got our attention, and, after the piece broke down, we went into record and recorded the rest of the rehearsal. It sounded amazing. Just 5 mics., and the orchestra in a concert hall with a natural acoustic.

When they had a break Dekker ran down and introduced himself, explained what had happened, what we were doing with SACD, and invited Goldsmith to come hear a playback, with the condition that we would erase everything on his say so.

We played a few pieces with him in Erdo's chair, and I remember watching his face and thinking, 'Uh-oh, he hates it.' but he loved it. In subsequent breaks Dekker thrashed out a plan to record all the pieces we heard, but he opted to do the work at Abbey Road with the LSO instead. Abbey Road doesn't have the acoustics of a concert hall, and Polyhymnia have customized all of their microphone electronics to make them state-of-the-art, so we lost a lot when we did it for real.

I think Gus Skinas still has some of the raw recordings in his archive. They are fantastic 'you are there' recordings, but also you hear the musicians pick up and drop things and make noises since they were not in a recording situation.

Still, we had some fun demo'ing Basic Instinct at shows! :D"

As for DSD beyond DSD64:

"With regard to high speed DSD, I don't know if anyone is recording at anything higher than DSD256, and there's likely only a handful of people recording at DSD128 and DSD256. I'm sure DSD512 and DSD1024, etc. are just gimmicks like the oversampling factor on CD players back in the 80s. DSD designers like Andreas Koch (Playback Designs) and Ted Smith (PS Audio) will tell you that you get nothing for nothing, and while DSDX may look great on paper in practice you may hit other problems due to the higher clock jitter requirements. Andreas thinks DSD128 is the sweet-spot, while Ted sees some benefit to DSD256 (can't remember exactly what now). Just like in the days of PCM DDCs (where users would upsample all CDs to 24/192) there may well be audible differences, but you may not be getting exactly what you think you are getting."

Yup, that certainly resonates with the technical data. I think DSD64 if we're looking for best fidelity is a bit too "tight" in terms of the rising noise level being close to the 20kHz upper potentially audible range. DSD128 is better, and DSD256 would be IMO the highest that we should need assuming one has a good DSD playback DAC.

As for high ultrasonic noise in DSD64:

"Sony fried a few high-end amps in the very early days of DSD because they had no output filter (Philips must have learned their lesson when they introduced Bitstream DAC based players in the mid-80s). Tube amps also didn't usually respond well to the SDM noise. We used to term this distortion 'birdies'."

So the "technical" term is "birdies", eh? ;-) As tempting as it is to leave the signal alone and not apply any filtering, I agree that with a DAC like the RME ADI-2, I'd keep the filter at 50kHz rather than 150kHz for listening just in case that DSD noise doesn't play well with the amp! 

Great to hear these kinds of background insights from folks who were "there"! And the discussion was too good not to share with you (with BE's permission of course). ;-)

As audiophiles, I think more and more we need to be educated with the experience and know-how of the professional audio world. We cannot remain isolated in a little "bubble" of audiophilia talking about often questionable beliefs ("myths" even) when in truth what we are listening to are products of the audio engineers out there working with state-of-the-art equipment and the actual artists themselves!

If you want to read more from Black Elk, make sure to check out this Steve Hoffman Forum thread on early SACDs with great insights and anecdotes. I didn't know about the "visual watermarking" feature that was promised for example, and this is the best summary of what happened to SACD within Sony/Philips I have ever read.

C. Roon needs multicast.

Here's an excellent post on the Roon forum discussing lots of details which you might find very interesting. This issue with multicast came up when I was setting up my ASUS ROG Rapture GT-AX11000 back in April.

I was noticing that on occasion, some of the Roon endpoints on my network would randomly stop playing. Depending on the time of the day and how many devices were on, the stoppage could be after an hour or within minutes of playback. Here's the network architecture diagram at my home as a reminder:

I run RoonServer on my Windows 2019 Server machine and I can check the Roon log at:

C:\<Username>\AppData\Local\RoonServer\Logs\RoonServer_log

Every time the playback stopped on whichever endpoint, it looked like an error of this nature:

Notice the "Unable to read data from the transport connection" issue. Well, the way to resolve this with the ASUS router is to turn on multicast which on mine defaulted to being off:

 

Also check that "Enable Fast Leave" stays "Disabled", I've noticed occasional issues with this enabled.

If you have a managed switch, you might also want to check out the multicast settings. Turning on IGMP snooping and blocking unknown multicast address to reduce denial-of-service (DoS) attacks might be something you'll want to do as well. So for example, here's my "audiophile network switch", the Netgear Nighthawk S8000 settings:

Before we throw caution to the wind and turn on all kinds of network parameters, note that there are some which could shutdown the communication between Roon and endpoints. For example on my mixed 10/2.5GbE QNAP QSW-M2108-2C switch, don't turn on "Multicast flood blocking" or this would prevent the Google Chromecast and AirPlay devices (like Yamaha receiver or AppleTV) from being seen by the Server:

It's good to be systematic when dealing with network issues, changing one parameter at a time so as to discover nuances like this.

One other thing to check with Roon is to make sure the Windows Firewall is allowing the RoonServer to get through:

As you can see, "roon.exe", "RoonAppliance", "roonappliance.exe", "RoonServer" have all been allowed through. Make sure you also check for "RAATServer" and "raatserver.exe" items on the list.

One last thing I checked on my network was latency since I allow large buffers for the 10GbE connection, especially between my main Workstation computer and Server. You can do this with something like "ping -t <RoonServer computer address>" (my server is at 192.168.1.8) which will continuously ping the computer every second and you can get a reading of latency while doing stuff like playing music streamed from the Roon server, or doing something like a large file transfer as shown below:

In the picture above, I'm doing a massive 70GB file transfer between the Roon Server computer to my Workstation while I'm also playing music through the Roon app. Notice the file transfer between SSDs on the two machines is running at 340MB/s (the limit is the SSD read/write speeds, not the 10GbE network which is even faster). The "ping" looks good, <1ms with the highest result being merely a single 2ms round-trip latency throughout the whole 3+ minutes large file transfer.

High "network jitter" will show up as variations in that "ping" time when the system is busy and not responding quickly. If you see significant issues, you might want to play with the transmit and receive buffers on your network device driver.

For the record, with the Windows 2019 Server computer (32GB RAM), I've set it at 2048 receive and 4096 transmit buffers these days on the 10GbE ASUS XG-C100C card - a 2:1 ratio for  transmit:receive buffers has been said to be optimal for many machines; well, it works for me even though I haven't done any systematic testing. Be mindful that an excessive number of buffers could worsen latency in the form of "bufferbloat". Smaller buffer sizes like 512 and 1024 would be fine with 1GbE networks.

(Addendum - May 2022) One last setting to consider if you still have issues, turn off "QoS Packet Scheduler" helped in one of my installs:

Settings above have resulted in very stable playback with Roon through my network these days across various audio endpoints. No problem with concurrent streaming to 3 or 4 devices around the house. No more "Unable to read data..." log issues.

Check out Roon's "Networking Best Practices" page for more if you have issues.

------------------

My new album this week is John Coltrane's A Love Supreme: Live In Seattle (1965) (2021 Release, DR13). It's available on most major streaming sites (no Tidal). An interesting historical live performance (see Rolling Stones review) and worth a listen if you like A Love Supreme (studio recording). Note that this album is definitely not for everyone, nor is the "avant-garde" jazz genre in general appreciated by many. ;-)

For those thinking about the hi-res 24/192 version, I would not bother:

As you can see above (Track 1), as an old analogue capture, this is not a high-resolution recording with relatively high noise floor and ultrasonic stuff of high levels. Notice the -35dBFS 77kHz tone for example which sometimes will go up to -30dBFS like on Track 3. IMO, as discussed before, it's best to discard this stuff and potentially improve sound quality plus save storage space. Just get the CD or downsample to 16/48.

Coltrane died about 2 years after this recording in 1967 at the young age of 40 from liver cancer.

A friend sent me an interesting blog piece the other day titled "12 Predictions for the Future of Music" by Ted Gioia who has written a number of books about music history (especially jazz). Pretty dire stuff which considers music (especially pop music) as a generic, basically disposable, commodity these days driven by price-conscious consumers fed by the tech giants with a tremendous volume of music in existence already (and tens of thousands of albums released each week!). I think the predictions are certainly not unreasonable given the trends we have been seeing not just in the mechanism of music distribution (big tech), but also in the quality of the music these days and the short-lived artiste du jour.

Lots of pretty safe predictions there I think even if worrisome for the music industry and for artists. Items 11 and 12 seem about right as well:

11. Record labels will increasingly pursue a bare-bones, low investment approach. A few execs will be able to maintain lavish lifestyles even in a declining industry, but there won’t be much cash for those at lower levels in the organization. Most people working in the music business will be poorly paid, with few opportunities for advancement. New openings at the top will typically go to family members and close friends of the old school bosses—making nepotism the one record business tradition that will survive no matter what.

12. But the greatest dream of the music execs will be. . . to get out (of) the music business. They will try to sell NFTs or promote audiobooks or finance biopics or sell music-themed apparel or open music-themed casinos, etc., etc.. And who can blame them? After collapsing the economics of the record business, they clearly need a new field to destroy.

As dire as the direction of the Music Industry might be heading, as consumers, we have tons of material to enjoy. More stuff, at lower cost, and even lossless quality than at any time in history.

Finally, for ENTERTAINMENT ONLY.

Psst... Wanna buy some "Magic" Quantum Fuses? Check out the review from the shaman, and follow-up from the supposed skeptic(s). Wow, is this some kind of "good cop / bad cop" tactic to get fence-sitting audiophiles to fork over >US$700 on some purple magic fuses?

Need I remind everyone that fuses are passive components (like cables/wires) and it's perplexing how a few centimeters is supposed to help in any way improve sound since it's not even located along the audio signal path and typically in the (directional?!) A/C power circuitry. Let's see the "(quantum) science".

Remember the old saying: "Do not be so open-minded that your brains fall out". Alas audiophiles, I think we're witnessing a few billion neurons on the floor. Clearly this is a good example of "Class A" snake-oil. I suppose snake-oil salesmen have to eat also!

I don't know who these Stereo Times people are and have never read their reviews, but dang, that's some good psychedelics they've been passing around in the form of apparently directional fuses thru USPS over the last year through the pandemic! I guess that's where Synergistic "Research" is advertising these days.

Hope you're all staying safe and having fun with your favourite music.