Hazard, I was attracted by the prospect of 'out of the box' operation and the apparent tight integration of HQPlayer. I have also seen some positive reports. And, yes, it is supposed to have support for native DSD for the JL Sounds board.
I'll report back in due course. First though, I want to use the existing set up to test the mute function on one my filter boards.
Ray
Despite my first resply which said that everything was working fine at DSD256 - I had an extended listening session on the weekend, I was a having a fine time listening to remastered Led Zep 2 (HDTracks 24 bit 96k version) and I noticed a few glitches. Reset to DSD128 and everything seemed OK.
So I did a bit more testing. At DSD256, task manager was showing CPU usage in the range 50-54%. So there should be plenty of head room, I don't know why I would get any glitches. At DSD128, CPU useage was around 35%.
I might need to upgrade PC after all?
Bad news for those waiting for one of my boards. I opened the package to build one up yesterday and found I have the wrong boards; it seems there has been a mix up. Replacements will take a week or two to arrive. Sorry for testing your patience.
Ray
Hi Hazard. Although upsampling will be a largely CPU bound process and it appears there is plenty of CPU capacity available computer performance isn't that simple. The machine is multi-tasking and the upsampling is only one part of its workload so the processor has to timeslice its resources so that everything gets done. If you are running, say, a word processor you're unlikely to be aware of the processor timeslicing but with music, where timing is everything, it only takes a tiny 'anomaly' to cause a glitch. The more load on the processor, the more likely an 'anomaly' occurrence. I hope that isn't patronising you.
Does your Linux distribution have the realtime kernel?
I'm assuming that I will have to upgrade my computer hardware to achieve smooth upsampling to DSD256.
Ray
Ray
This is what made me wonder, why? Most of the time the data stream ends up in the same DAC.
Is there something in the conversion process? Most audio is recorded in PCM and converted to DSD for distribution on DSD media.
Is DSD inherently sounding better than PCM? Even though recorder in PCM.
Would it then be better to always convert to DSD, so precede your DAC with an PCM to DSD converter and convert all incoming data first to DSD and then feed it to the DAC (making DAC's with an PCM input obsolete)?
If you are running a dual core system, a 50% load might mean that one core is running at full load while the other one is almost idle. That would definitely cause audio glitches.
It does require a solid computer server-side (if you're using a smaller network-attached audio device near the DAC like in the ideal situation) or your main computer for higher-rate DSD real-time up-conversion.
You can play with a few different settings to find something comfortable, in particular using different filters and modulators can use less of your processor. If you have many cores, there's a setting for 'pipeline SDM' that can be useful too.
I can only speak for myself and our system here (I have a DAC which does native Quad DSD) and my previous DAC did PCM only:
1. Native DSD material on PCM DAC with conversion in between with Audirvana+ - sounds great, better than equivalent PCM recording!
2. PCM (Redbook and above) up-converted in real-time or offline to DSD128 and sent to native DSD DAC - sounds great! Sounds better that if I listen to the PCM on my PCM only soundcard.
3. Native DSD in native DSD playback sounds the absolute best, especially DSD256 (I have to resort to Win + Foobar2000 for native DSD256 otherwise I'm using DSD128 with DoP on Mac OS X most of the time)
To make a fuller comparison, I would certainly need to compare to PCM playback on pure multi-bit DACs and R2R DACs or NOS DACs.
But even so, the following can be concluded in my system:
Most probably the filtering used in recording and playback for PCM in most DACs is audible and lack quality.
The fact that even Redbook up-converted to DSD128 sounds better than pure PCM playback shows audibility of the output section when using DSD.
I end up listening to all my collection re-ripped in lossless, and up-converted offline to DSD128 as it's the sweet-spot in my system.
Pairing that output with my SET Tube Amp (PCB by George @ Tubelab! + Big Elcors + Electro-Harmonix KT-88) is the ultimate in SQ for now.
Hallo,
I've been following this thread silently until now. The idea of filtering the DSD stream passively is quite interesting.
Today I finally had time again to do some tinkering. I connected a spare amanero board with two sowter 3575 transformers and directly feed a Torpedo parafeed headphone amp (also transformer coupled). I have to say I'm quite surprised: The sound is definitely very good. Not perfect though, very neutral, and very detailed from the bottom to the top end. On some tracks I heard details I never encountered before.
I had some problems first. The DSD feed from Jriver requires a large buffer (500 ms) to run smoothly. DSD128 was used, since with higher samplingrates, the processor (core i5) could not cope (as soon as one core is at 100%, artefacts are produced).
The bass level is quite there, the sowters present an easy load to the amanero output. at least easier than a lundahl ll1527. I inserted some 4,7 uF Wima mkp4 between the amanero and the transformers to take care of the DC voltage of around 1.5 V at the digital output. This certainly improved the sound, since the 3575 can not cope with DC on the primary.
Some thoughts on this: A fast digital buffer at the amanero output would be certainly a good idea. I assume a flip flop also just functions as a buffer, but the clocked switching of the outputs probably is not needed. It might just be an additional jitter source. I thought about using a high speed digital buffer instead. VDSL-line buffers have low output impedance and large current capacity, hence they can drive also low impedance loads like a LL1527 (which I also have spare) and filter networks. They work over 100 MHz, so the bandwith is no problem, even for DSD512...
Another problem are the artifacts produced by PC playback. Maybe it is a jriver problem or a processing problem, I don't really know. But it would be nice to have a "perfect" DSD signal. Maybe once a song there are short ticks, that probably come from the on the fly conversion... Do other persons have this problem, too?
There are no high frequency artifacts, as described elsewhere, I assume the isolation by the transformer is beneficial in this matter. I had artifacts with the amanero and sabre DACs before since it has no galvanic isolation.
An interesting thing is the inability to play silence in the tracks over DSD. For example, the silence between a track and a hidden track is just skipped right now. With a "regular" DAC there are always a few minutes of silence in between....
Cheers
Florian
I've been following this thread silently until now. The idea of filtering the DSD stream passively is quite interesting.
Today I finally had time again to do some tinkering. I connected a spare amanero board with two sowter 3575 transformers and directly feed a Torpedo parafeed headphone amp (also transformer coupled). I have to say I'm quite surprised: The sound is definitely very good. Not perfect though, very neutral, and very detailed from the bottom to the top end. On some tracks I heard details I never encountered before.
I had some problems first. The DSD feed from Jriver requires a large buffer (500 ms) to run smoothly. DSD128 was used, since with higher samplingrates, the processor (core i5) could not cope (as soon as one core is at 100%, artefacts are produced).
The bass level is quite there, the sowters present an easy load to the amanero output. at least easier than a lundahl ll1527. I inserted some 4,7 uF Wima mkp4 between the amanero and the transformers to take care of the DC voltage of around 1.5 V at the digital output. This certainly improved the sound, since the 3575 can not cope with DC on the primary.
Some thoughts on this: A fast digital buffer at the amanero output would be certainly a good idea. I assume a flip flop also just functions as a buffer, but the clocked switching of the outputs probably is not needed. It might just be an additional jitter source. I thought about using a high speed digital buffer instead. VDSL-line buffers have low output impedance and large current capacity, hence they can drive also low impedance loads like a LL1527 (which I also have spare) and filter networks. They work over 100 MHz, so the bandwith is no problem, even for DSD512...
Another problem are the artifacts produced by PC playback. Maybe it is a jriver problem or a processing problem, I don't really know. But it would be nice to have a "perfect" DSD signal. Maybe once a song there are short ticks, that probably come from the on the fly conversion... Do other persons have this problem, too?
There are no high frequency artifacts, as described elsewhere, I assume the isolation by the transformer is beneficial in this matter. I had artifacts with the amanero and sabre DACs before since it has no galvanic isolation.
An interesting thing is the inability to play silence in the tracks over DSD. For example, the silence between a track and a hidden track is just skipped right now. With a "regular" DAC there are always a few minutes of silence in between....
Cheers
Florian
Cool experiments, Florian.
What would be awesome is to get DSD out of the computer without a packetised interface or protocol if we could find a way to do it.
Some other thing you can try now that you've done initial testing is replace JRiver with HQ Player and have it up-convert all your files to the max DSD rate possible. Also, do a client-server config with HQ Player and the networkaudiodaemon running on the NAA (a smaller network attached audio device). The NAA should ideally be just a buffer combined with your DSD D/A circuit. On top of that, you could use fiber optic.
Keep it up!
DCDC CONVERTER FOR THIS FANTASTIC DSD IN CAR
se qualcuno desidera installare questo ottimo progetto in car, io posso progettare e realizzare un ottimo DC DC Converter di estrema qualità e piccole dimensioni ?
Sinceri complimenti a tutti.
JEDY
Cari tutti,
se qualcuno desidera installare questo ottimo progetto in car, io posso progettare e realizzare un ottimo DC DC Converter di estrema qualità e piccole dimensioni ?
Sinceri complimenti a tutti.
JEDY
DCDC CONVERTER FOR THIS FANTASTIC DSD IN CAR
if someone desires to install this good project in your car, I can realize an excellent DC DC Converter of extreme quality and small dimensions.
Sincere compliments to all.
JEDY
Dears All ... I'm SORRY !!!
if someone desires to install this good project in your car, I can realize an excellent DC DC Converter of extreme quality and small dimensions.
Sincere compliments to all.
JEDY
oh, I just looked at HQPlayer, it cost over 100 Euros...
After some more listening I have to say that there is indeed some harshness to the top end. I will try some filtering to get rid of that...
THE BEST SOLUTION !!!
I am sure that for eliminate completely this problem is possible only if utilize a preamplifier with JFet or valves with EL86 ... you see my electric scheme that I insert in my space.
We can have a lot solutions, but the best is filtrate a lot also if happen a strong limitation of high frequencies and to lift the attenuate high frequencies to increasing the value of capacitor connected to the cathode.
With this system the linearity return normal.
If you desire the electric scheme with JFet or JFet/Mosfet ... say to me.
Good Listening.
JEDY
Dears ALL,
I am sure that for eliminate completely this problem is possible only if utilize a preamplifier with JFet or valves with EL86 ... you see my electric scheme that I insert in my space.
We can have a lot solutions, but the best is filtrate a lot also if happen a strong limitation of high frequencies and to lift the attenuate high frequencies to increasing the value of capacitor connected to the cathode.
With this system the linearity return normal.
If you desire the electric scheme with JFet or JFet/Mosfet ... say to me.
Good Listening.
JEDY
There are a lot of modulators/filters combinations you can try with HQ Player.
My thinking is to use a BBB or RPi with an isolation/reclocking board, they obviously wouldn't keep up with on the fly conversion, but have I2S interface, which cuts out the USB interface.
Not sure I'd use a Raspi as they usually use a combined Ethernet/USB circuit...
You could use a BBB as NAA.
I'm using the BBB as a NAA in my main system at the moment, the PCM/DSD play out happens via it's I2S interface which doens't use packetisation, That's what your comment was about, or did I mis understand?
It's just DSD (or PCM) modulation in the time domain in one continues bitstream, I then use Acko's S03 isolation and reclocking board.
I'm doing the same with a RPI B in combination with a Soekris DAC for PCM, which works equally well.
It's just DSD (or PCM) modulation in the time domain in one continues bitstream, I then use Acko's S03 isolation and reclocking board.
I'm doing the same with a RPI B in combination with a Soekris DAC for PCM, which works equally well.
Last edited: