[bogi]

if you dont mind could you explain how OS and NOS DACs tie in here aswell?

Trouble with NOS multibit R2R DACs is the ability to process high enough bit resolution on digital input in an acceptable price. More the bits processed higher the price. The delta sigma design is much cheaper in the regard. But it needs lot of treatment.

The solution I prefer is using HQPlayer to convert all music content on the fly (during playback) to the highest rate DSD the DAC is allowing on its digital input. DSD128 as a minimum suitable for this purpose. Then to use a DAC which does minimal digital processing with DSD signal input. The goal is to complete skip the hardware oversampling process within delta sigma DAC chips. In the case of some DAC chips (from TI Burr Brown, Cirrus Logic) it is also possible to skip the delta sigma modulator and to use the DAC chip purely for digital to analog conversion. No volume control is then possible on DAC chip level with such digitally non processed DSD input. You can find an example of such direct DSD path on picture for CS4398.

When using DSD input with Sabre DAC chips, volume control within DAC chip is possible, because the input DSD signal always goes through delta sigma modulator, but without any sample rate change. No direct DSD path is available within Sabre DAC chips, but no traditional DSD to PCM conversion occurs within Sabre DAC chips to allow volume control. Series of n bits of DSD signal is most probably processed as 'moving window' (one bit in, one bit out) on the original DSD signal rate, so the lossy effect of DSD to PCM conversion, called decimation, is not present in this case. Regarding the volume control itself, HQPlayer does it in much higher precision (using 64bit floating point resolution for computing), so it is better to do it in software if it is needed. Some people think that Sabre DACs sound better with DSD input when no on chip volume control (0 dB setting) is used. I'm able to hear many other differences but I'm not convinced I could distinguish this one in blind test.

I never listened music with high quality multibit DAC like Yggdrasil. I consider it to be too expensive. I compared by listening only delta sigma designs and the HQPlayer / DSD way sounded me always better. With some USB input treatment (not using +5V from computer, as short USB cable as possible, not leading ground wire near DATA wires) I reached better result than with USB to SPDIF converters and DAC coax input. I'm using HQPlayer and DSD solution about 5 years already.

 

[carlmart]

Bogi,

Your comments sound quite intriguing, for my limited knowledge on some areas, and also exciting because I'm looking for the best sounding options on the audio network i'm assembling.

Next week I will receive the Raspberry Pi 3 B+ that i bought to manage my audio network. So the first question: how can I implement HQPlayer with it?

The network will start on the file server by my PC computer, and via ethernet cable connect to the RPB, and then USB out to the D50, not SPDIF. Then power amplifier & speakers, never headphones. The power amp will have a dedicated RCA input for the DAC. What do you think of this arrangement? Would you change anything?

This will also be my first time dealing with DSD files, and I was advised to stick to lower DSD files, never to exceed DSD128, because of increased jitter. You seem to have a different opinion. Would you please elaborate on that?

 

[439598]

Trouble with NOS multibit R2R DACs is the ability to process high enough bit resolution on digital input in an acceptable price. More the bits processed higher the price. The delta sigma design is much cheaper in the regard. But it needs lot of treatment.

The solution I prefer is using HQPlayer to convert all music content on the fly (during playback) to the highest rate DSD the DAC is allowing on its digital input. DSD128 as a minimum suitable for this purpose. Then to use a DAC which does minimal digital processing with DSD signal input. The goal is to complete skip the hardware oversampling process within delta sigma DAC chips. In the case of some DAC chips (from TI Burr Brown, Cirrus Logic) it is also possible to skip the delta sigma modulator and to use the DAC chip purely for digital to analog conversion. No volume control is then possible on DAC chip level with such digitally non processed DSD input. You can find an example of such direct DSD path on picture for CS4398.

When using DSD input with Sabre DAC chips, volume control within DAC chip is possible, because the input DSD signal always goes through delta sigma modulator, but without any sample rate change. No direct DSD path is available within Sabre DAC chips, but no traditional DSD to PCM conversion occurs within Sabre DAC chips to allow volume control. Series of n bits of DSD signal is most probably processed as 'moving window' (one bit in, one bit out) on the original DSD signal rate, so the lossy effect of DSD to PCM conversion, called decimation, is not present in this case. Regarding the volume control itself, HQPlayer does it in much higher precision (using 64bit floating point resolution for computing), so it is better to do it in software if it is needed. Some people think that Sabre DACs sound better with DSD input when no on chip volume control (0 dB setting) is used. I'm able to hear many other differences but I'm not convinced I could distinguish this one in blind test.

I never listened music with high quality multibit DAC like Yggdrasil. I consider it to be too expensive. I compared by listening only delta sigma designs and the HQPlayer / DSD way sounded me always better. With some USB input treatment (not using +5V from computer, as short USB cable as possible, not leading ground wire near DATA wires) I reached better result than with USB to SPDIF converters and DAC coax input. I'm using HQPlayer and DSD solution about 5 years already.

Interesting stuff. Is DSD necessary to prevent hardware oversampling? I dont like how great the CPU usage is for DSD conversion if you want to use the higher rates with the best filters, so, among other reasons, I upsample to 705.2kHz PCM instead. it being the highest supported PCM rate I thought it should prevent internal oversampling but DSD has much higher sample rates.

Since both the Sabre DACs I had could control DSD volume, while very convenient, seemed like there was PCM conversion happening making it mostly pointless but your saying that isnt the case after all.
if upsampling PCM to 705.2 can also prevent HW oversampling then PCM and DSD are both viable options for Sabre DACs, depends what you think sounds better.
The recent AKM DAC chips have the direct DSD output, thats as close we'll get to Sabre level performance with unprocessed DSD right now... too bad there are no options for cheaper DAC options with the AKM chip (there is a diy kit from DIYinHK however).

I use XXHE and its a PCM only player. for me any benefit of using DSD, even native DSD files with a direct DSD DAC, is lost because its means using HQPlayer instead, HQP has really good filters but as a player XXHE is on another level. you need to listen to both with all upsampling/processing disabled to see what I mean

 

[bogi]

Next week I will receive the Raspberry Pi 3 B+ that i bought to manage my audio network. So the first question: how can I implement HQPlayer with it?

Raspberry Pi 3 B+ is not sufficient to run HQPlayer with it's digital filters, but it is quite suitable to run Network Audio Adapter (NAA), which is optional part of HQPlayer solution. NAA is nothing more than simple buffer, acting between network input (Ethernet, Wifi) and audio output (typically USB, but it is not a must, it works with any type of digital audio output). NAA intentionally does no processing on input data, the goal is to minimize computer noise generated by NAA. NAA fulfills 2 functions: a) to isolate the audio device (DAC) from noise generated by HQPlayer computer and b) multiroom support - in HQPlayer you can choose which NAA you want to play into. Or if your player PC is located distantly from your DAC within the same room, that's happens quite often with loudspeaker setup. People report that using WiFi on HQPlayer PC and ethernet on NAA improves sound because it provides additional level of isolation from PC/HQPlayer computer noise and solves also possible ground issues between them. I'm using such setup too, after some comparisons this sounded to me clearly the best.

Ready to use NAA images for Raspberry Pi 3 and other platforms appear here: https://www.signalyst.eu/bins/naa/images/
And one level up from that folder you can find NAA installations for Windows, Linux and Mac for people with other hardware.
Try to serach RPi3 information in this thread: https://audiophilestyle.com/forums/topic/13649-hqplayers-network-audio-adapter/
For example this one confirms RPi 3 B+ working as NAA on DSD512 rate: https://audiophilestyle.com/forums/...-audio-adapter/?do=findComment&comment=873655

What do you think of this arrangement? Would you change anything?

HQPlayer on your PC and NAA on your RPi3 could be a suitable arrangement. Your PC has to be enough powerful to be able to upsample to the highest DSD rate your sigma delta DAC allows.

This will also be my first time dealing with DSD files, and I was advised to stick to lower DSD files, never to exceed DSD128, because of increased jitter. You seem to have a different opinion. Would you please elaborate on that?

I don't know about relation of DSD sample rate and increased jitter. Regarding DSD sample rates and DSD vs. PCM comparisons, the highest rate DSD produces the least ultrasonic content, like you can see on wide band measurements:
https://audiophilestyle.com/blogs/entry/428-ifi-idsd-micro-measurements/
https://audiophilestyle.com/blogs/entry/343-playback-formats-on-a-teac-ud-501-ti-pcm1795/
HQPlayer users mostly convert all their PCM or DSD content to max. DSD rate their DAC allows, regardless on PCM/DSD sample rate of the original content played. The conversion is performed in real time during playback. Therefore it is not so important if DSD64 audio file (for example ripped SACD), or DSD128 or DSD256 is played.

 

[carlmart]

My PC is GA-Z97X, with Intel i7-4790K, 32GB SDRAM DDR3. Is that powerful enough?

About the other things, can I ask you as they come along?

 

[bogi]

I dont like how great the CPU usage is for DSD conversion if you want to use the higher rates with the best filters

The so called 2s filters (the DSD target case) require much less computational resources than their non 2s siblings. 3GHz or higher processor clock is recommended. i5 level processor is recommended for DSD128 or DSD256 target rate. For DSD512 i7 operating on at least 3GHz is recommended. Requirements are stronger (for example for number of processor cores) if you want to use digital room correction (or other convolution type) with HQPlayer. Convolution is possible for both PCM and DSD outputs.

I am going up to DSD128 with my DACs and I am able to do easy tasks like browsing internet and processing emails while HQPlayer is running. The same for i5 and i7 based computer. I tried DSD256 with borrowed DAC few years ago and it worked with my i5 notebook too, but only if nothing other was running. But stereo DSD512 with digital room correction, or multichannel processing with digital room correction require much more computational resources. Digital room correction processing can be optionally offloaded to graphic card processor.

I upsample to 705.2kHz PCM instead. it being the highest supported PCM rate I thought it should prevent internal oversampling but DSD has much higher sample rates.

Right point. Delta sigma DAC chips always internally oversample the PCM input. Look at it as to 2 stages of oversampling. The first stage is typically up to the highest PCM input sample rate the DAC chip allows, it can be for example 705.6/768 kHz. Then the next stage goes up to for example 11.3/12.3 MHz. These 2 stages are typically differently implemented within DAC chip. The implementation of the 2nd stage is typically much simpler because of much higher target sample rate - that means much shorter time for any processing, so there is no much time for complicated algorithms within simple DAC chip and additionally a DAC chip doesn't have enough computational power for high quality DSP. It is known that Burr Brown TI DAC chips use simple Sample And Hold (SOH) solution for the 2nd oversampling stage. SOH means simple repeating of the previous value n times. That is the weakest possible solution how to perform oversampling (or interpolation). Some information from ESS indicate that Sabre chips do interpolation on higher level, but it is still impossible for a DAC chip to do it on level of HQPlayer type of solution running on powerful computer. HQPlayer is able to substitute that 2 level up/oversampling with one tap processing from source PCM or DSD sample rate to target PCM or DSD sample rate.

Is DSD necessary to prevent hardware oversampling?

if upsampling PCM to 705.2 can also prevent HW oversampling

Currently I don't know about any other way to skip also that 2nd oversampling step than to provide DSD signal on DAC input. You are skipping the complete PCM oversampling with DSD input, although in the case of Sabre DACs you are not skipping the delta sigma modulator. There are discussions how much is lost against (for Sabre chips impossible) direct DSD path, but ... When I compared effect of upsampling to max. PCM vs. max. DSD rate with different DAC chips, the difference is of the same type with direct DSD path (I tried some Burr Brown and Cirrus solutions) and Sabre. It could look that avoiding the 2nd stage of oversampling may bring more benefit than avoiding the delta sigma modulator, but it's only my speculation.

I use XXHE and its a PCM only player. for me any benefit of using DSD, even native DSD files with a direct DSD DAC, is lost because its means using HQPlayer instead, HQP has really good filters but as a player XXHE is on another level. you need to listen to both with all upsampling/processing disabled to see what I mean

Your indication that filterless XXHE may sound better than filterless HQPlayer is interesting. If you prefer to use XXHE, then maybe Yggdrasil or some solutions from Chord could be best for you. I don't know much about Chord DACs implementation, but I know they don't use traditional DAC chip solutions, so there may be a chance they do internal DSP with PCM input better.

 

[bogi]

My PC is GA-Z97X, with Intel i7-4790K, 32GB SDRAM DDR3. Is that powerful enough?

Users report that DSD512 is easily reachable with i7-4790K for so called 2s poly sinc filter family. Other filters require more computational resources. In fact I am using only those 2s filters. If you want to trial the solution your HW setup seem to be suitable.

About the other things, can I ask you as they come along?

I'm not sure how much time I will have during work week but sooner or later I will answer you.

 

[ggillies]

New guy coming into this thread.

I got my D50 about 2 weeks ago after researching the hell out of DACs. I wanted a standalone DAC to go with my Darkvoice 336SE, rather than a DAC/Amp combo unit, on the theory that if DAC technology changes quickly, I can upgrade the DAC without junking my Amp investment.

My entire chain looks like this:

Intel Skull Canyon NUC with 32 GB RAM, 2 X 512GB SanDisk M.2 SSD drives ----> Audioquest Cinnamon USB cable out to Topping D50 ----> Audioquest Cinnamon RCA cables out to DarkVoice 336SE ----> Focal Clear and Oppo Digital PM3 headphones.

After playing around with a few tube combinations, I settled on a NOS (1960s vintage) Tung-Sol 5998/421A for the Power tube and a Sylvania 6SN7W short envelope phenolic base (1940s vintage). That combination works best for my tastes.

So far, the D50/DarkVoice 336SE combination is phenomenal. The main reason I went with the D50 was the full support for DSD files as well as MQA. I have been listening to some DSD files and MQA files I downloaded and have on my local drives, as well as Tidal's MQA offerings, and my entire CD collection ripped in FLAC format to my local drives as well.

For local listening, I have been using Audirvana (plays DSD and MQA) and am so far really impressed with how they sound through Audirvana.

For helping me choose which Tubes sound best, I have been using Tidal to play MQA versions of the Who "Live at Leeds" album and MAN is that a freaking powerhouse of a live album.. And, it lets me hear various differences between tubes (I write all my observations down as I am listening, then change tubes and listen again).

Anyway, thanks for all the great advice and the informative posts on this Forum; great place, great information.

Cheers!

 

[Ripper2860]

I'm glad you are enjoying your D50! I know I am enjoying the heck out of mine, but the D50 does not support MQA decoding. Tidal may do level 1 unfolding with the Tidal player, but the D50 is not licensed by MQA and does not do the 2nd level unfolding.

Still a wonderful DAC and in my case, MQA is not something I care about anyway.

Enjoy!!

 

[ggillies]

I'm glad you are enjoying your D50! I know I am enjoying the heck out of mine, but the D50 does not support MQA decoding. Tidal may do level 1 unfolding with the Tidal player, but the D50 is not licensed by MQA and does not do the 2nd level unfolding.

Still a wonderful DAC and in my case, MQA is not something I care about anyway.

Enjoy!!

Yes, I keep forgetting that the D50 doesn't do MQA. That's one of the original reasons I went with Audirvana (now that I remember my decision process when I chose the D50); Audirvana seems to do a decent job with MQA, even if it only partially unfolds MQA, and that a DAC is necessary to do the complete job.. From their Website:

• Non MQA capable audio devices to benefit from the high resolution (doubled sample rate compared to the encoded file one) thanks to the MQA decoder integrated in Audirvana Plus 3. In this case, Audirvana Plus brings, in addition to the general Sound Quality improvement, the decoding of the MQA file that would be played only at little above CD quality otherwise, losing all its high resolution benefits. Note that decoding the signal beyond twice the sampling rate of the encoded file (for the few rare recordings actually made above 96kHz) can only be done in a DAC MQA.

 

[carlmart]

What's MQA and what am I missing by not having it?

 

[DjBobby]

What's MQA and what am I missing by not having it?

You can read about it here: https://arstechnica.com/gadgets/201...ything-you-need-to-know-about-high-res-audio/

Some swear on MQA, some hate it. No middle opinion.

 

[TheSnafu]

Well, some MQA mixes are good, some not. SQ wise, can't tell whats SQ whats mastering needed for MQA...

D50 plays Tidal masters 96kHz/32bit (at least with my Asus laptop, and no D50 doesn't support MQA).

 

[bogi]

What's MQA and what am I missing by not having it?

Others may have a different opinion, but my opinion is: Stay away from it. Use hi res downloads or FLAC streaming services, as these are not lossy and they don't restrict you as customer in any way. With WAV, FLAC etc. you can do anything: any format conversions and any DSP you wish. With MQA any DSP is rather impossible. Especially in the world of free software. MQA is attacking the world of free formats, free software, small hardware designers. MQA attacks customer freedom in using DSP including basic format conversions like 5.1 to stereo or digital room corrections. MQA is attempt to grab money from every part of music chain by the MQA company: from music recording through DACs, software players etc. MQA is attempt to force you to buy once again the music content you already bought and then to restrict how you can use it. There is strong propaganda behind MQA and DAC producers are providing it many times only to avoid the disadvantage of not providing it. What is the most important: With high quality hires recordings (FLAC or DSF) you get better sound than with MQAized recordings. My recommendation is to have a look here: https://audiophilestyle.com/forums/topic/30381-mqa-is-vaporware/

I am not asking anyone to share my opinion and I respect any other opinion. Since this is not a thread about MQA, I used it only to briefly express my opinion and I will not make any further discussions here on the MQA topic.

 

[lacknothing]

Others may have a different opinion, but my opinion is: Stay away from it. Use hi res downloads or FLAC streaming services, as these are not lossy and they don't restrict you as customer in any way. With WAV, FLAC etc. you can do anything: any format conversions and any DSP you wish. With MQA any DSP is rather impossible. Especially in the world of free software. MQA is attacking the world of free formats, free software, small hardware designers. MQA attacks customer freedom in using DSP including basic format conversions like 5.1 to stereo or digital room corrections. MQA is attempt to grab money from every part of music chain by the MQA company: from music recording through DACs, software players etc. MQA is attempt to force you to buy once again the music content you already bought and then to restrict how you can use it. There is strong propaganda behind MQA and DAC producers are providing it many times only to avoid the disadvantage of not providing it. What is the most important: With high quality hires recordings (FLAC or DSF) you get better sound than with MQAized recordings. My recommendation is to have a look here: https://audiophilestyle.com/forums/topic/30381-mqa-is-vaporware/

I am not asking anyone to share my opinion and I respect any other opinion. Since this is not a thread about MQA, I used it only to briefly express my opinion and I will not make any further discussions here on the MQA topic.

Thanks a lot for that !!!

Seems to be the same money making machine idea like with Lucas / THX, Dolby Laboratories and others, who's stuff is 80% lawyers .