Multi-Channel, Multi-Amplifier Audio System Using Software Crossover and Multichannel-DAC

[dualazmak]

So much equipment! I wonder what the room acoustics are like?
I am looking forward to getting an OKTO DAC8PRO!

Hello Descartes, welcome to this thread!

As for the room acoustics, you would please find some info in my post #17 through #24, especially post #22, where I briefly measured the room acoustics using REW's wavelet analysis with a measurement microphone.

And my post #311 is sharing the actual listening room environments. Post #318 shows the Fq response measured at the listening position, post #321 shows Fq responses in digital level, line level and amplifier's SP out (or pre-out, headphone out) level.

 

[dualazmak]

"Near Ultrasound - Ultrasound" Noises in Improperly Engineered/Processed HiRes DSD Music Tracks, and EKIO's DSP-EQ Configuration to Cut-Off Such Noises

Hello friends,

I assume many of you are aware of amirm's recent nice thread entitled "Comparison: PCM DXD DSD (Sound Liaison High Res Format Comparison)" and his YouTube video clip pasted in his very first post.

Since I also have some download-purchased suspicious music albums and tracks having possible "Near Ultrasound - Ultrasound" Noises (let me say UHF noises in this post), I analyzed some of them using the nice software MusicScope, and also prepared software crossover EKIO's configuration to cut-off such UHF noises.

You may find license-free download link of MusicScope (Final version 2.1.0 for Windows 10 64 bit) here. It looks the Company Xivero developed MusicScope has been closed down and they are kindly tentatively (?) offering the license-free download link. This means that there will be no version-up nor support on MusicScope; you may use the software at your own risk. In case if you have any difficulty in getting the installer file of license-free MusicScope, and you are still really serious in testing/using it, please simply PM me of your thoughts.

One of my "UHF noise" suspicious albums, which I really love the performance and music, is "While She Sleeps" by Art Lande (piano) available from Blue Coast Records. I download-purchased the DSD128(2x) HiRes album in dff format in February 2011.

Even though the album is still available at Blue Coast Records site in many formats, please note that my current issue and analysis are on their DSD128(2x) dff album which I downloaded in February 2011.

At that time, I also converted this album into DSD128(2x) dsf format, and also converted into 192 kHz 24 bit AIF format, for flexible tag configurations.

As repeatedly shared, my standard EKIO crossover/EQ configuration in my multichannel multi-driver multi-amplifier system is like this;

Whenever I play this album by "on-the-fly" DSP down-sampling by JRiver MC into 192 kHz 24 bit, or by simply play the converted 192 kHz 24 bit tracks, I always "see" very unusual, consistent and inaudible UHF noises in high (HI) and Super-High (SH) VU meters as shown here;

(SL: Super Low, LO: Low, MD: Mid, HI: High, SH: Super High)

I already well understand that these rather high level UHF noises exists in 20 kHz and higher Fq region which are almost inaudible to our ears and brain, but I always feel quite uncomfortable seeing the four VU meters of HI and SH consistently stay at considerably high levels.

Amirm's recent nice post, therefore, encouraged me to analyze this "UHF noise" suspicious album with MusicScope.

I first analyzed the intact downloaded first track "Tendery" in DSD128 (x2) dff format by MusicScope;

This is the screen capture of MusicScope almost at the end of the track (i.e. at 3:50 of total 4:03), and the green bar level shows the 3:50 Fq spectrum and the yellow dot-line shows the accumulated Fq peak spectrum over 0:00 to 3:50.

Really surprisingly, but as expected, the UHF noise level at 22.0 kHz is rather high in -88 dB followed by wide and higher level of UHF noises over up to ca. 88 kHz. The level of UHF noise at 50 kHz reached -62 dB. Even though I do not know up to how high frequency EKIO's VU meters detect the "integral level" of these UHF noises, I now can well understand the quite unusual inaudible consistent VU meter positions while playing this track/album.

Next, I analyzed the AIFF converted same track (first track) in 192 kHz 24 bit;

The AIFF conversion into maximum 96 kHz (in 192 kHz 24 bit stereo format) enhances the relative level of UHF noises against the true music signals in 20 Hz - 20 kHz, and the UHF noise at 90 kHz reached -48 dB.

Since I really do not like to "see" the inaudible abnormal high levels of VU meters for HI and SH, and I also would like to prevent these UHF noises go into DAC8PRO and amplifiers covering Be-Squawkers, Be-Tweeters and Horn Super Tweeters, I quickly prepared a unique EKIO crossover EQ configuration by adding high-cut (low-pass) 48 dB/Oct LR (Linkwitz–Riley) filters at 22.049 kHz (which is the highest Fq value EKIO can accept) to MD, HI and SH regions as shown here;

Using this "UHF-noise cut-off" EKIO configuration, I can play the track/album quite nicely without any sound deterioration, and of course VU meters in HI and SH are moving up-and-down normally in conformity with the actual audible sound. This is mentally and psychologically comfortable playing these music tracks/albums having relatively high level of UHF noises even almost inaudible.

(SL: Super Low, LO: Low, MD: Mid, HI: High, SH: Super High)


I also analyzed one recently download-purchased DSD256 (4x) dsf track/album entitled "Passage" (Liszt, Mozart, Schumann, Chopin) by Mao Fujita (piano);
https://www.e-onkyo.com/music/album/nycc27306/
This recording has been said to be, in Japan, one of the world best HiRes recording of piano performance, as reviewed here.

The result of MusicScope analysis of the 5th track of the album, Chopin Piano Sonata No. 3 OP.581 Allegro Maestoso in DSD256 (4x);

Even this wonderful DSD256(4x) track/album has certain level of UHF noises in which we can find a little peak at 67 kHz in -97 dB consistently seen in entire track. I assume this tiny peak maybe due to some improper or inevitable mastering technique during the processing from the raw 352.8 kHz 24bit DXD recorded format. Other widespread UHF noises in 22 kHz to 140 kHz, however, are reasonably or acceptably low enough in less than -100 dB; for this album, therefore, I have no need to apply the EKIO configuration of UHF noise cut-off filters, I believe.


In any way, I could confirm amirm's concerns on possible UHF noises in some of improperly engineered/processed HiRes DSD tracks/albums, and if needed, I can apply suitable EKIO crossover/EQ configuration to cut-off these UHF noises on-the-fly during the listening sessions using my multichannel multi-driver multi-amplifier system.

 

[gene_stl]

I have wondered about this noise shaping and whether the various players filter it. As Amir says in his videos it could trigger oscillation or amplifier problems such as intermodulation , aliasing down, using up headroom, possibly even baking a tweeter voice coil.

not good.

 

[dualazmak]

Hello gene_stl,

I essentially agree with you.

Many of my playlists now consist of mixture of various formats;

16-bit/44.1kHz CD ripped aif,
24-bit/192kHz down-sampled or up-sampled aif,
24-bit/96kHz flac,
24-bit/192kHz flac,
1-bit/DSD64(1x) 2.8MHz dsf,
1-bit/DSD128(2x) 5.6 MHz dsf,
1-bit/DSD256(4x) 11.2 MHz dsf,
1-bit/DSD512(8x) 22.4 MHz dsf

and JRiver MC feeds all of the tracks in 192 kHz 24 bit (i.e. 96 kHz Fq window in 2-ch stereo) by on-the-fly conversion into EKIO for crossover/EQ processing.

As you may guess and agree, I am now intensively testing and evaluating the EKIO configuration with 48 dB/Oct L-R high-cut filters at 22.049 kHz;

So far I found no audible negative effect with this EKIO configuration, and therefore it would be highly possible to use this configuration as my standard one from now on.

Fortunately, my listening capability is still fairly above the average level of people in my age, and I still can hear well up to 17 kHz pure sine-wave tone, self-checked by PC-based software audiometer using headphone, even though the sensitivity over 10 kHz is slightly declining (depending on my age); this is why I prepared the EKIO configuration with slightly increased total level in 8 kHz to 19 kHz zone.

 

[dualazmak]

Hello Gene_stl and friends,

Regarding the UHF-Noise Cut-Off Filters in software crossover EKIO's configuration which we discussed in the above three posts, i.e. post #362, post #363 and post #364...

Today, I prepared three EKIO configurations with High-Cut LR (Linkwitz–Riley) filters at 22.049 kHz with different cut-off slopes of -12 dB/Oct, -24 dB/Oct and -48 dB/Oct. As shared in post #362, the High-Cut filters are added to Mid Range Be-Squawkers, High Range Be-Tweeters and Horn Super-Tweeters, shown below;

The three EKIO configurations are separately saved with very long, look similar, file names for our quasi-blinded ABCX comparative listening sessions;

64136d2537eca4a89743aaba9127b6d5a56c60db2d8b1721a387ab17b920eeb4.ekio
64136d2537eca4a89743aaba9127b6d5a56c60db2d8b1721a387ab17b92oeeb4.ekio
64136d2537eca4a89743aaba9127b6d5a56c6odb2d8b1721a387ab17b920eeb4.ekio

As long as I do not see the above crossover graph-diagram screen in EKIO, even myself cannot identify which is which, and another listener (my wife) knows nothing about what is going on. All the audio gears are always carefully warmed-up for more than one hour before the listening sessions.

We just started our careful comparative listening sessions using our latest Audio Sampler Playlist consists 65 tracks in various formats, aif, dsf, flac, and in various sampling rates.

We will continue the quasi-blinded ABCX comparative listening sessions during the coming two weeks or three.

Of course, all the other settings and parameters remain strictly unchanged as shown here;

For the details of these diagrams, you would please visit my post #311 where you can find the larger ones.

I will get back to you when we would decide which EKIO configuration out of the three should be the standard/regular one in my system.

 

[sebackman]

Hi dualazmak,

There is no doubt that a DSP XO @22k will filter out whatever is there. I think my question is if there is anything there after the DAC.

But I do think you need an analogue oscilloscope to see if there is any analogue signal above 20k from the DAC when fed high sample rate material that does showsignal above 20k in MusicScope. Or maybe the sound card ADC's are good enough to take high frequency analogue signal into the computer. I would guess that any sound card with an ADC would limit the function to about 20-30k to avoid adding garbage from the signal that we can't hear and avoid oscillation & other problems.

I was under the impression the filters in the DAC filters that in the digital domain so there is in reality no signal above 20k (or maybe 25k) on the analogue side. Or at least like 80db down.

 

[dualazmak]

Hi sebackman,

I believe DAC8PRO would "nicely" output the HF noises after the DAC process, at least up to 50 kHz or higher.
Otherwise DAC8PRO's spec of "up to 192 kHz 32 bit and up to DSD128(2x) handling" performance would be meaningless...

I assume that amirm will hopefully properly respond to your inquiry with some evidences.
You would please also ask amirm in his thread regarding your points.

For myself,,, I could successfully roop back the OKTO DAC8PRO's crossovered channels, still in digital, within PC. And I could also roop back the analog channels after DAC into PC, as shared in my post #321.

This means, if I would use nice HiFi grade ADC like KORG DS-DAC-10R, I may digitally "record" the analog out of DAC8PRO into 192 kHz 24 bit WAV or AIF, or even in 2.8 KHz, 5.6 mHz or 11.2 MHz DSF.

Unfortunately, however, currently I do not have such nice HiFi grade ADC, and have no plan to purchase it only for current issues we are discussing.

As you suggested, I may tentatively use my PC's sound chip Realtek-HD (not so HiFi), which has 192 kHz ADC capability, (as I did in post #321), for HiRes digital recording using AUDACITY. If I could find time, I will try this in a few days.

I believe amirm has a lot of gears including nice ADC with which he may easily see and confirm your points...

Of course, we need to carefully check the base (blank) noise level of ADC and HiRes recording process with nul or 44.1 kHz/16-bit analog input from DAC8PRO.

In any way, mentally and psychologically, (and also maybe physically and electronically better to avoid), I would like to avoid the high level and high mass of UHF noises would go into DAC8PRO, and this is why I am preparing and testing EKIO's configurations with digital high-cut filter at 22.049 kHz before sending the digital signals into DAC8PRO.

I am really afraid of the possible damage on DAC8PRO (or damages on ES9028PRO DAC chip and other IC chips in DAC8PRO) given by the high level and high mass of UHF noises.

Furthermore, the UHF noises in analog line level may have unfavorable negative effects (or damages) to our amplifiers and SP drivers, especially highly sensitive tweeters and super-tweeters (and also to our ears and brain?). Not good, and we should avoid it as far as we can.

 

[dualazmak]

Hi sebackman,

By using PC's Realtek-HD audio chip (not so HiFi), I quickly checked and confirmed that DAC8PRO "nicely and correctly" outputs the UHF noises into analog sides.

In EKIO configuration, I tentatively feeded all the L-Channels (CH1, CH3, CH5, CH7, CH9) into DAC8PRO's CH1, and also R-Channels (CH2, CH4, CH6, CH8, CH10) into DAC8PRO's CH2. Then DAC8PRO's CH1(L)+CH2(R) stereo analog signal was given back into PC's Realteck-HD line-in for analysis and recording by AUDACITY.

Of Course, AUDACITY's recording features were set in 192 kHz 24 bit with max real-time sampling quality.

I first played this "UHF-noise containing track-1 Tenderly" in 192 kHz 24 bit aif using JRiver into EKIO with no High-Cut filters;

and DAC8PRO's analog out was recorded by AUDACITY in 192 kHz 24bit.

In the recorded AUDACITY tracks, I selected 2:41 - 3:11 time area for its Fq spectrum analysis;

As you can see, in the analog output, the UHF noises still "perfectly" exist 30 kHz - 96 kHz zone. Please note that the recording was in 192 kHz 24 bit in stereo, so the Fq width of the recorded track is 0 - 96 kHz.

Then I played the same track using the EKIO configuration with LR High-Cut filters of -48 dB/Oct at 22.049 kHz for CH5 to CH10 (for MID, HI, Super-HI);

The recorded analog output of DAC8PRO does not have the UHF noises, as well expected;

Let me just compare the Fq spectrum of almost the same portion of the analog recorded tracks, without and with High-Cut filters in EKIO's EQ configuration;

Now let me say again...

In any way, mentally and psychologically, (and also maybe physically and electronically better to avoid), I would like to avoid the high level and high mass of UHF noises would go into DAC8PRO, and this is why I am preparing and testing EKIO's configurations with digital high-cut filter at 22.049 kHz before sending the digital signals into DAC8PRO.

I am really afraid of the possible damage on DAC8PRO (or damages on ES9028PRO DAC chip and other IC chips in DAC8PRO) given by the high level and high mass of UHF noises.

Furthermore, the UHF noises in analog line level may have unfavorable negative effects (or damages) to our amplifiers and SP drivers, especially highly sensitive tweeters and super-tweeters (and also to our ears and brain?). Not good, and we should avoid it as far as we can.

 

[dualazmak]

Hello friends,

As I just wrote here in amir's thread,,,,

Even though over 22.049 kHz is inaudible Fq range, we, myself and my wife, can hear very very subtle audible differences between these three slopes.

I once read that 22 - 50 kHz sound cannot be heard with simple sine-wave-tone tests, but when mixed with high quality music sound of 10 - 20 kHz, the 22 - 50 kHz sound may have unproved "audible preferable effects". Someone discussed this magic effect in terms of our "bone conduction", but I am rather doubtful since we usually cannot hear over 15 kHz through "bone conduction."

I assume that the "very very subtle audible differences" would be related to the "phase rotations" or "phase complications" in 20 - 50 kHz range, but I have no objective tool to measure them. (I have only one measurement microphone BEHRINGER ECM8000 which can hear up to 20 kHz.)

I would like to decide, therefore, the choice of the high-cut slope out of the three by our very careful quasi-blinded ABCx listening sessions still going on for the coming two weeks or three. When decided, I will get back to you in this thread.

 

[thewas]

Even though over 22.049 kHz is inaudible Fq range, we, myself and my wife, can hear very very subtle audible differences between these three slopes.

The nominal cutoff filter frequency of 22.049 kHz doesn't mean that they don't act below that which can be seen from your above plots which show different responses even at 10 kHz.

 

[dualazmak]

The nominal cutoff filter frequency of 22.049 kHz doesn't mean that they don't act below that which can be seen from your above plots which show different responses even at 10 kHz.

Hello thewas, thank you for your kind post.

Yes, your observation on the "theoretical total response curve" in white is nice and correct, and the "very subtle audible difference" may be also due to the little differences (including the phase natures) in 10 kHz to 20 kHz caused by the Linkwitz-Riley high-cut slopes at 22.049 kHz.

BTW, today I contacted with LUPISOFT (who develops and distributes EKIO) inquiring possibility of higher Fq high-cut crossover point setting, like anywhere between 22 kHz to 50 kHz.

 

[sebackman]

Good work!

Worrying that we have that amount of garbage coming out from the DAC. I thought the DAC would limit all of that but obviously not, and as @amirm pointed out, good DAC's only do their job and pass it on to the analogue side.

I will make a test on my BSS DSP later in the week to see if I get the same overshoot when applying a 20kHz filter . I don't remember having that overshoot on the curve when applying the 48db filters as you curves suggests.

The frequency variation following the 22kHz 48db filter can probably be cured by a HS DSP filter.

Is this mainly the result for DSD files or do we have the same amount of "garbage" in all up-sampled files?

This is puzzling and exciting.

 

[dualazmak]

Good work!
.....
.....
Is this mainly the result for DSD files or do we have the same amount of "garbage" in all up-sampled files?
This is puzzling and exciting.

I already analyzed many of HiRes tracks in my huge digital music library using MusicScope, and found the high level UHF-noise "garbage" not in all of them, but in the tracks/albums improperly (or poorly) engineered and processed. I believe this the issue of "quality control".

Very much surprisingly, it looks many of the recording and mastering engineers in minor labels, and sometimes in rather major labels, do not know, do not care about, the possibilities of the "garbage" in Fq range over 22.05 kHz; and they even do not have any analytical tools like MusicScope, Audacity, Adobe Audition. Really disappointing aspect of HiRes music business...

In any way, I strongly feel that we, as end consumers and HiFi audio enthusiasts, need to have "self-defense measures" against this really serious problem. I believe my approach to have nice High-Cut digital filters before DAC would be (one of?) the best measures/solutions.

 

[dualazmak]

Another (Ultimate?) self-defence solution for the HiRes Tracks/Albums having UHF-noises

Hello sebackman, gene_stl, and friends,

As I shared in the specific post here, the day before yesterday, I download-purchased "Sound Forge Pro 15 Suite", including "Steinberg SpectraLayers Pro7", since if needed we may very much flexibly "edit" the music tracks having UHF-noises.

I briefly tested and evaluated the Steinberg SpectraLayers Pro7 to cut-off(delete) all the UHF-noises in the specific "Track 01 - Tenderly" which I shared in my post #362 and #368;

As shown above, I could successfully cut-off (delete) the UHF-noises over 25.5 kHz by Steinberg SpectraLayers Pro7. Please note that I deleted the selected area for 4 times, since I found a little remaining HF-noises at a few time points after only one delete procedure. The cleaned-up tracks can be saved in the original format of 192 kHz 24 bit AIFF.

Of course, I also checked and confirmed "Before and After", using MusicScope;

Yes!, we can rather easily cut-off (delete) the UHF-noises by using nice music editing software like Steinberg SpectraLayers Pro7.

This self-defence measure of editing music tracks/albums would be another (ultimate?) solution for undesirable UHF-noises in the HiRes digital files, if you have not so many tracks/albums having the issue.

In my case, since I already have so many HiRes tracks/albums in my huge digital library and my playlists, "to edit all the UHF-noise accompanying files using SpectraLayers Pro7" would not be a practically feasible solution.

I still would like to use, therefore, the High-Cut digital filters in the EKIO crossover configuration before DAC8PRO, as shared in my post #362 through #365.

 

[KSTR]

Yes!, we can rather easily cut-off (delete) the UHF-noises by using nice music editing software like Steinberg SpectraLayers Pro7

There is no need to spend any money on that IMHO. With simple freeware tools like SoX (command-line) or Audition 3.0 you can get the exact same results. In the simplest case: downsample, upsample, done.

 

[dualazmak]

There is no need to spend any money on that IMHO. With simple freeware tools like SoX (command-line) or Audition 3.0 you can get the exact same results. In the simplest case: downsample, upsample, done.

Hello KSTR,

Thank you, I agree with you for simple cut-off (delete) of all of the UHF noises over a certain frequency throughout the track.

I am also interested in, however, SpectraLayers Pro7's various sound editing features including delete and copy-paste of free-hand selected portion on the spectrum, though not directly related to the issue we are discussing.

Adobe Audition 3.0 is not a free software but we need to join Adobe's (rather expensive) monthly payment plan, right?

 

[KSTR]

Adobe Audition 3.0 is not a free software but we need to join Adobe's (rather expensive) monthly payment plan, right?

No. Adobe publicly offered it in 2013 for free download (including the serial keys!) for a limited time, actually the full CS2 package from 2007, and you can find the resources still today. Adobe officially says that it is only legal to use if you own an Adobe CS2 license but that's nonsense, obviously (IMHO).

 

[dualazmak]

Oh, thank you for the nice info!
I will soon try to get Audition 3.0 even though I have purchased "Sound Forge Pro 15 Suite", including "Steinberg SpectraLayers Pro7".

 

[dualazmak]

Hello KSTR, gene_stl and friends,

Yesterday in late evening, after my 15 min web search, I found the Audition 3.0 (not 3.01) installer package and the serial number to activate it. I could successfully install and activate it.

This morning, I smoothly applied the "Audition 3.01 update patch" (which one of my ASR friends kindly provided) to the already installed Audition 3.0 making it to 3.01, and I found Audition 3.01 is more stable than 3.0.

It is very interesting for me, even not so practically important, that the rectangle-deletion (rectangle-erase) process with Audition 3.01 gave slightly different results compared to that with SpectraLayers PRO7 as shown below;

As you can see, SpectraLayers PRO7 has completely erased/deleted the selected rectangle area to -infinity dB (no signal there). On the other hand, Audition 3.01 has erased the area, but still having residual very low level of UHF noise (ca. -118 dB) after the process.

 

[dualazmak]

Just for my interest and curiosity,,,
I found that Audition 3.01 is filling the erased area with "a kind of pattern" of very low level UHF noises;