L.K.S. Audio MH-DA004 Dual ES9038PRO DAC - Van Damme's double impact?
Dec 6, 2020 at 7:14 PM Post #3,016 of 4,419
Hello boys. As I am experimenting with HQPlayer, comparing it to Foobar and upsampling DSD512 (... to the limit of my computer ...:scream:), I notice that HQP has a more liquid, brighter sound and Foobar has a darker, heavier sound , thicker. So, in the face of the terrifying multiconfiguration of HQP filters, could one of you please tell me which you think are the most suitable filters in search of a thicker, fuller, darker sound in this impressive HQP ...? Thanks for your opinion. Greetings to everyone from Spain ....:relaxed:
poly-sinc-ext2 filter
ASDM7 DSD modulator

Leave the rest at default

Screencap is from embedded version of HQPlayer, find the corresponding setting if you are using HQPlayer Desktop
hqpe_settings.png
 
Last edited:
Dec 6, 2020 at 7:23 PM Post #3,017 of 4,419
Hi B0bb, what can I say?
Where would this forum be without you. Great job!!!
Can you please share the exact part number of the tantal bead?
Thanks

The tantalums are generic parts just find the lowest leakage device on offer.
Avoid devices with ESR of 5 ohms of greater. Typical ESR values are for this application is between 1-3 ohms.

Example below from AVX has 8uA leakage@25C, this is small enough to not cause Vcom reference errors.

https://www.mouser.com/ProductDetail/AVX/TAP107K010SRW?qs=NORa6xUSEah7%2BvryW8kR/w==
 
Last edited:
Dec 7, 2020 at 7:37 AM Post #3,019 of 4,419
Dear b0bb, I just changed the filters and the modulator that I had before in HQP and I have put the ones that you have kindly indicated. Well, right, right now I am listening to a dsd256 upsampled file to dsd512 and there is one thing that strikes me, that I notice a lower consumption of gpu on my pc, much less than with my previous configuration which was:
-Oversampling filter: poly-sinc-xtr-mp-2s
-Modulator: AMSDM7 512 + fs
Please, I read something that this configuration of mine with that "xtr" filter, is very demanding in resources, what do you think ...?
Thanks for your always valuable help b0bb. :relaxed:
 
Dec 8, 2020 at 1:08 AM Post #3,020 of 4,419
Dear b0bb, I just changed the filters and the modulator that I had before in HQP and I have put the ones that you have kindly indicated. Well, right, right now I am listening to a dsd256 upsampled file to dsd512 and there is one thing that strikes me, that I notice a lower consumption of gpu on my pc, much less than with my previous configuration which was:
-Oversampling filter: poly-sinc-xtr-mp-2s
-Modulator: AMSDM7 512 + fs
Please, I read something that this configuration of mine with that "xtr" filter, is very demanding in resources, what do you think ...?
Thanks for your always valuable help b0bb. :relaxed:
Those settings I recommended reduces processor and gpu computation load, xtr filters does not seem like something your computer can comfortably handle.
 
Dec 8, 2020 at 9:51 AM Post #3,022 of 4,419
Very impressive mods b0bb!

A few questions:

1) Accusilicon AS318-B-100, where can this component be purchased from? This part is not available anywhere from what I searched.
2) Is anyone using I2s or is it best to use the upgraded Amanero USB in stock form and simply upgrade the power supply as noted? The Ebay link to the power supply board that was posted is dead, here is a link to the same https://www.ebay.com/itm/274403969704
3) b0bb, are you a dedicated headphone user or is this DAC being used in a 2 channel setup? I'm curious as I would think the audible differences with these changes would seem to be more noticeable in a more critical listening headphone environment
3) Is the better route, USB or I2S? Do the all the mods listed, other than the upgraded power supply to the USB, pertain to improvement in the I2s chain as well?
4) I am not too familiar with HQ player but I thought I saw someone say they enabled this on the Sonictransporter and didn't need a separate PC? I am a set it and forget it kind of person, is it as simple as checking off a couple of boxes in the sonictransporter to enable HQplayer or is it more involved?


My current setup is ROON Sonictransporter I5>Sonore Ulratrendu>Ultradigital I2s>LKS>Anthem D2V Pre>Emotiva XPA-1L>B&W Nautilus 803

I was using the USB originally but was limited to DSD128 with ROON which is why I added the Ultradigital to the mix. The Ultradigital allowed me to go up to DSD256. Now I see that if I were to update the firmware on the USB, I could have accomplished the same thing.

I work at a company that assembles electronics printed circuit boards so I am fortunate in that regard to have the pro's with the proper equipment do mods. The B&W 803's are very transparent so I'd like to remove the glare and brightness whichever way I can. Tighter bass and more punch is what I am after. Would the 950x crystal be the better choice for me or the Accusilicon?

Thanks to all for contributing to this amazing thread!
 
Last edited:
Dec 9, 2020 at 11:40 AM Post #3,023 of 4,419
I'm going to buck the trend here and say the best route into the LKS is via AES/EBU. I played a lot with I2S and ultimately found it a dead end due to constant pops and bangs and a harsh sonic character. USB is disappointing from a resolution point of view.

I use AES because I input into it from a Mutec MC-3 reclocker, itself driven by USB from my network bridge/streamer. Every single time I have improved the performance of my digital chain, the LKS just rises with it, seemingly having no limit to performance via AES.

I recently replaced my Asus Tinker Board S running Volumio network bridge with a Sonore ultraRendu running Sonicorbiter. The LKS responded with better soundstage and greater depth and note decay. Similarly when I added an external master clock to the EtherREGEN and Mutec.

I also replaced the AES cable from an Audioquest Carbon to a Wireworld Micro Eclipse Platinum and that improvement was also evident.

So far, AES is giving me by far the best performance I've ever had in my system. Take care to feed the LKS with a fine quality signal and it responds in kind.
 
Dec 9, 2020 at 1:24 PM Post #3,024 of 4,419
Martin,

What did you use when you were dabbling with the I2S? I've had no issues sending 256DSD to the DAC via SonicOrbiterRoon/UltraDigital/UltraRendu at DPLL 7.

My system has always been quite bright as the surroundings cannot be changed due to the WAF (see below) . I am hoping some mods to the DAC will help with it. I am quite pleased with the DAC already, just trying to discern what's the best mods to do it, or is adding other pieces before it like the Regen more useful? I have Synergistic Research Red fuse of the way to add that into the mix next

28811693127_dd23ed85b8_k.jpg
 
Last edited:
Dec 9, 2020 at 2:51 PM Post #3,025 of 4,419
I tried two I2S solutions, the first was an Audiophonics LVDS streamer and the second a Pi3 with Kali I2S board. The second was better in resolution but the LKS was too bright with both. I'm convinced its I2S input is suboptimal. The clicks when it goes into and out of signal lock became very annoying since there is no mute circuit.

AES input has a much better balance and properly mutes when unlocked.

If you want to tame the LKS a little more, try an SR Blue fuse.
 
Last edited:
Dec 9, 2020 at 10:09 PM Post #3,026 of 4,419
Very impressive mods b0bb!

A few questions:

1) Accusilicon AS318-B-100, where can this component be purchased from? This part is not available anywhere from what I searched.
2) Is anyone using I2s or is it best to use the upgraded Amanero USB in stock form and simply upgrade the power supply as noted? The Ebay link to the power supply board that was posted is dead, here is a link to the same https://www.ebay.com/itm/274403969704
3) b0bb, are you a dedicated headphone user or is this DAC being used in a 2 channel setup? I'm curious as I would think the audible differences with these changes would seem to be more noticeable in a more critical listening headphone environment
3) Is the better route, USB or I2S? Do the all the mods listed, other than the upgraded power supply to the USB, pertain to improvement in the I2s chain as well?
4) I am not too familiar with HQ player but I thought I saw someone say they enabled this on the Sonictransporter and didn't need a separate PC? I am a set it and forget it kind of person, is it as simple as checking off a couple of boxes in the sonictransporter to enable HQplayer or is it more involved?


My current setup is ROON Sonictransporter I5>Sonore Ulratrendu>Ultradigital I2s>LKS>Anthem D2V Pre>Emotiva XPA-1L>B&W Nautilus 803

I was using the USB originally but was limited to DSD128 with ROON which is why I added the Ultradigital to the mix. The Ultradigital allowed me to go up to DSD256. Now I see that if I were to update the firmware on the USB, I could have accomplished the same thing.

I work at a company that assembles electronics printed circuit boards so I am fortunate in that regard to have the pro's with the proper equipment do mods. The B&W 803's are very transparent so I'd like to remove the glare and brightness whichever way I can. Tighter bass and more punch is what I am after. Would the 950x crystal be the better choice for me or the Accusilicon?

Thanks to all for contributing to this amazing thread!

1)
Accusilicon 100MHz AS318-B availability has become very limited, Magnahifi stills sells them
https://magnahifi.com/audio-gd-accusilicon-as318b-100mhz-femto-clock-upgrade/

CCHD950X is readily available from the usual suppliers (Mouser, Digikey, Arrow, Farnell etc)

2 and 3)
USB vs I2S really depends on whether your specific setup is quiet enough RF wise to outperform SPDIF.
Given your description of the bright sonic presentation, I am thinking that RF noise is affecting your DAC.

The Amanero is pretty good, bested by the Singxer SU6 and then only with very careful isolation of RF noise and preparation of the music source

SPDIF performance is the the baseline.
If you are prepared to do the work on I2S, it gives the best performance, however be aware the LKS has zero RFI protection on the I2S input.
USB input is isolated, unfortunately LKS leaves the jitter removal caused by the isolator to the 9038, this overburdens the 9038 especially the DPLL

My source preparation for low noise operation is thus:
- It starts with a FCC Class B certified machine running the audio endpoint to keep below 1mV conducted noise for the music player
Forget about DIY unless you can afford FCC style EMI testing for your specific box configuration.
- Separate the player and audio renderer into different machines.
- Drive down the remainder of noise with the SU6 taking care to break the chassis ground conection between the SU6 and the PC
- Use a minimal Linux install (no X11 GUI, all non essential processes removed) as the audio player, the one I use has a memory footprint of 256-350MB when running
This reduces the non audio activity on the machine to its absolute minimum.
Linux kernel is a low-latency build.

A PC going full bore can easily send 100A (100W@1V) of current into the ground planes creating a lot of ripples as it interacts with the non zero ground plane impedance.
No amount of filtering and isolation can fix this, the recourse is to reduce the number of active processes.
The biggest CPU hog on Windows is the GUI and associated background processes
Getting rid of the GUI on Windows requires expensive options like Server Core, that is almost 2/3rd the cost of the LKS004.
(A lot of people overlook the fact that the CPU is the single biggest noise source in the system)

I recently bought a new DDC, Singxer SU2, this is a Windows only device, running Win10 20H2 with Fidelizer and the Singxer ASIO driver
There was little sonic difference between it the Amanero and the SU6.
(This is one of the reasons a few have questioned the usefulness of the more expensive SU6)

Switching back to barebones Linux made the differences between I2S (SU6) and USB (Amanero) much clearer.
SU6 vs SU2 differences not obvious on Win10, benefits of SU6 only weakly felt on Win10 eg bass slam is mildly hinted at on Win10.
On Linux SU6 was clearly ahead, bass was very fast and hit with an impact that neither SU2 or Amanero could match.
SU2 un-officially runs in DSD mode on Linux about the same level as the Amanero.

Similar performance to Linux can be expected Win Server Core 2019 or Nano Server from Windows Server 2016.
Neither has the GUI and can be configured to run very lean.

Performance overall was pretty poor on Windows Desktop versions (Home/Pro) , bass was lacking and fast transients smeared, cymbal clash was quite painful to listen.
String quartets were a wooden affair lacking musicality.

In this situation, SPDIF/AES is actually a better option as @MartinWT mentioned especially if one is using a single box for playback and audiorendering.
LKS put some effort in isolating the SPDIF inputs.

This however limits the DAC to 192kHz sampling/DSD64 at best, the full potential of the LKS is not realized here.

It boils down to whether you can get RF noise down on I2S and USB such that performance exceeds what you get on SPDIF
(AES is just SPDIF on a differential transport)

4)
I do the initial baselines on 3 sets of headphones, Sennheiser HD800, Stax SR009 and Audeze LCD3F.
The results are directly translatable to my speaker listening environment.

My speaker setup:
Martin Logan Summit powered by Aragon Palladium 1Ks Monoblocks, differential/balanced drive to MLs.
Bass is handled by an old school Velodyne HGS18 with analog motion feedback on the cone, this is one of the very few subwoofers that can keep up
with the MLs and not muddy the sound with lots of DSP help from HQPlayer

The room has the necessary acoustic treatment (early reflections, DSP assisted time alignment via HQplayer)
I generally do not discuss the results of speaker listening due to the fact that the room exerting a large influence on the presentation.
It also depends on how tightly the acoustic environment is controlled

5)
I have not used sonic transporter with HQPlayer
I would suggest you consider using a NUC which is FCC certified or one of the audioendpoints like the uRendu or SMS200

Ultradigital is closely related to the Singxer SU1, check the metal shells on the usb inout and HDMI output are not connected.
If they are connected you might want to experiment taping the over the metal shell of the USB connector going into the PC


Before you spend time and effort on the hardware mods see if you can clean up the RF noise environment per the above.

The success criteria is a sonic improvement over SPDIF.
 
Last edited:
Dec 9, 2020 at 11:00 PM Post #3,027 of 4,419
One way to get and retain performance when using the LKS I2S hdmi input.
Use a very short cable.

68DCFABD-1EB6-40DD-AC43-44430B1A27F7_1_105_c.jpeg


LKS hdmi input impedance is out of whack with the rest of the industry.

The commonly accepted LVDS signal termination is 100ohms.
LKS decided to use 120 ohms, note the resistors below with the orange bands
This is a recipe for signal reflection on the cable, making finding the right I2S cable an exercise in frustration.

The difference in termination on the transmitter (100ohms) and receiver end (120ohms) means signal energy can bounce back and forth causing signal interference with the new data coming out of the transmitter.

This is an additional source of jitter in the signal going to the dac.

Shortening the cable reduces the delay and reduced the effect of the reflection problem

FCF7247C-D099-418B-A16C-3DF2636BF651_1_105_c.jpeg



On to the LVDS receiver...
This is the DS26LV32A made by National Semiconductor
Notice that the max toggle frequency is 32MHz.
There a 2 transitions/toggles per I2S clock, potentially 16MHz max clock.

DSD 512 operates at 22/24MHz we are just beyond those limits.
I am wondering if this might be one of the items for the coming round of modifications.

DS26LV32AT.png
 
Last edited:
Dec 9, 2020 at 11:06 PM Post #3,028 of 4,419
Coming back to the hardware mods....

I am experimenting with reducing the leakage of the cap bypassing the Vcom reference chain.

The silver component in the middle of the picture is a wet tantalum unit, leakage is 1uA vs 8uA from the yellow bead.

The initial presentation is more focused imaging, bass got a lot faster, most notable on stringed instruments ( bass guitar, double bass) and percussion instruments like the kick drum.

Avoid surplus sources and Ebay as older caps of this type many have reliabllity and operating constraint issues.
C9704C08-71DA-4072-91BD-A9BD13F60A27_1_105_c.jpeg
 
Last edited:
Dec 9, 2020 at 11:18 PM Post #3,029 of 4,419
Finally got rid of the metallized mylar/polyester caps on the analog final/post filter. Yay!

Replaced with polyproplyene caps.
Black ones are Vishay KP1830 film/foil, blue ones are Kemet PHE426 metallized polypropylene

LKS used polypropylene for the analog post filter in the LKS003.
I was disappointed when they went with metallized mylar for the 004.

The problems this caused was predictable, transient smearing (kickdrums and cymbals), loss of top end control on highly compressed tracks and a diminished sound stage.

Improvements were as expected, namely the reversal of the above, no surprises here just predictable gains

I used components with a 1% tolerance to improve the consistency.

D326DDB2-082D-4825-BC75-F95046CD435F.jpeg



Wider shot of the mods
852BEEE0-0225-40C5-A20E-2E69830E9536_1_105_c.jpeg



Original blue metallized mylar/polyester before mod
11374189.jpg
 
Last edited:
Dec 10, 2020 at 11:03 PM Post #3,030 of 4,419
Thanks b0bb.

I took the cover off tonight to update the USB firmware, twist the pairs on the torrids as well. Noticed that the grounding screw for the torrid to the chassis was very loose like barely on, like the ring lug was wobbly loose! All the screws holding the torroids down were not too snug either.

Thought I'd share the link as the firmware has changed since it was last posted, it's 2006be11:

CPLD_1081 or CPLD_1081_SWAPPEDSDnative DSD512 on Linux/Windowsfirmware_2006be11oem tool 117Async DAC


Instructions once again:
https://www.head-fi.org/threads/l-k...van-dammes-double-impact.840938/post-14595255

https://amanero.com/combo384_firmware.htm

If I go down the route of soldering the socket in for crystal swapping, does it ultimately need to then be removed? Do I also have to swap something else out for this mod with the socket? Attached are some photos for reference
 

Attachments

  • IMG_3043.jpg
    IMG_3043.jpg
    462.3 KB · Views: 0
  • IMG_3053.jpg
    IMG_3053.jpg
    777 KB · Views: 0
  • IMG_3055.jpg
    IMG_3055.jpg
    667.4 KB · Views: 0
  • IMG_3052.jpg
    IMG_3052.jpg
    765.1 KB · Views: 0
  • IMG_3054.jpg
    IMG_3054.jpg
    519.7 KB · Views: 0
Last edited:

Users who are viewing this thread

Back
Top