I use Audirvana on my iMac and was looking at some setting and decided to give upsampling to DSD a try. What surprised me is that it sounded really good, better to my ears than the bit perfect PCM at 24/96. Of course that doesn't make sense that a converted file would sound better than original, but maybe the DAC just does a better job with DSD. I know a lot of people on these forums swear by DSD and SACD so maybe there is something to this.
Anyway, I'm using an iMac with quad core i5 3.3ghz processors, the latest Audirvana, and UPnP into an RPi with Allo Boss DAC running Volumio. When first starting a track the iMac goes into about 75% utilization, that's 3 processors worth, and it does a look ahead and converts the next track on the list as well. As each track starts it will then convert the next. I'm using Qobuz as the music source. If you switch tracks there is some hesitation as it has to start over again, but consecutive tracks still play gapless. I tried Tidal MQA but it won't upsample that.
I've also experimented with the bit depth, at DSD256 it will play smoothly but only if I don't touch the iMac, and when I start typing the music will break up on occasion. At DSD128 it will work without skipping. I haven't tried DSD64 because I started high and settled on 128. I haven't tried 512 either but not sure my DACs can handle it anyway. You need good wifi and a fast processor to do this, which was already in place.
Bass is a bit tighter, and the highs more sparkly. We're not talking a huge difference but still noticeable. And it just sounds more alive.
Anyway, I'm using an iMac with quad core i5 3.3ghz processors, the latest Audirvana, and UPnP into an RPi with Allo Boss DAC running Volumio. When first starting a track the iMac goes into about 75% utilization, that's 3 processors worth, and it does a look ahead and converts the next track on the list as well. As each track starts it will then convert the next. I'm using Qobuz as the music source. If you switch tracks there is some hesitation as it has to start over again, but consecutive tracks still play gapless. I tried Tidal MQA but it won't upsample that.
I've also experimented with the bit depth, at DSD256 it will play smoothly but only if I don't touch the iMac, and when I start typing the music will break up on occasion. At DSD128 it will work without skipping. I haven't tried DSD64 because I started high and settled on 128. I haven't tried 512 either but not sure my DACs can handle it anyway. You need good wifi and a fast processor to do this, which was already in place.
Bass is a bit tighter, and the highs more sparkly. We're not talking a huge difference but still noticeable. And it just sounds more alive.
Last edited:
What you're saying would be quite plausible ...
except that the ALLO Boss DAC does not accept DSD !
It is doing something to accept DSD since it is receiving it and playing music. Just looking around I can't find much info on how it is doing it as nobody really has a clear understanding. It may be running DoP, it may be converted to PCM before conversion, but whatever it does it sounds better than native PCM, which I already said doesn't make sense.
I also have an iFi nano iDSD to try, maybe tomorrow. And a Hifiberry too, but that is also a DAC Hat, which doesn't natively support DSD but can support DoP, and it also works in my kitchen system with DSD. I can't say I hear any improvement there over the usual kitchen noise.
And in any case this isn't costing any money, which is my goal now that I've retired. I've done a lot of playing around in my amp, cleaning sockets, rewiring/resoldering where things are messy, and even that improved the sound as at least the dirty sockets would create less than ideal contact.
I also have an iFi nano iDSD to try, maybe tomorrow. And a Hifiberry too, but that is also a DAC Hat, which doesn't natively support DSD but can support DoP, and it also works in my kitchen system with DSD. I can't say I hear any improvement there over the usual kitchen noise.
And in any case this isn't costing any money, which is my goal now that I've retired. I've done a lot of playing around in my amp, cleaning sockets, rewiring/resoldering where things are messy, and even that improved the sound as at least the dirty sockets would create less than ideal contact.
Last edited:
It is not the DAC doing this... Look at your settings in Volumio...
I think you will find that you are upsampling to DSD on your Mac and then Volumio is converting DSD using DoP.
So two resamplings....and yes it may sound 'different' and if you like the new sound then all power to you....🙂 although there is a lot of cpu usage happening so that overhead may be more affecting than the conversion..??
DoP is just a trick to get DSD through interfaces meant for PCM. It doesn't decimate or otherwise change the DSD data in any way.
I had an expensive CD player that had an upsample option. It sounded brighter but weird. I got rid of it.
Why do you think upsampling can add any digital bits that aren't in the file that will somehow give you more?
Why do you think upsampling can add any digital bits that aren't in the file that will somehow give you more?
Allo Boss uses PCM5122. There is no mention of DSD in that DAC's datasheet. Also no mention of DSD on Allo's page for the Boss, logically.
Conversion PCM -> DSD (before Volumio) -> PCM (automagically by the Volumio chain) may introduce some noise. People like listening to noise, look at the popularity of filterless NOS DACs which generate lots of noise. I tried too, the sound is brighter, raw ("genuine"), just like reported here.
MPD player used in Volumio converts from DSD to PCM automatically User’s Manual — Music Player Daemon 0.22 documentation
Conversion PCM -> DSD (before Volumio) -> PCM (automagically by the Volumio chain) may introduce some noise. People like listening to noise, look at the popularity of filterless NOS DACs which generate lots of noise. I tried too, the sound is brighter, raw ("genuine"), just like reported here.
MPD player used in Volumio converts from DSD to PCM automatically User’s Manual — Music Player Daemon 0.22 documentation
Even Volumio is reporting it as DSD although it doesn't say what it does with it.
I have a handful of true DSD files too, just the free samples from NativeDSD. I'll see what it does with those.
I really have no interest in purchasing DSD files either. In fact I never bought any SACDs when I finally had a player that could play them, by then they were going away and the one or two stores I knew that sold them only had a few left.
I'll continue to play with the Audirvana settings to see what sounds good to me, but so far this was the best. I had SOX upsampling from 16/44.1 to 16/176.4 and 16/48 to 16/192 on for a long time, and never felt like it added anything. I always left higher bitrates to play natively.
The problem with iFi iDSD is that it is a portable unit that doesn't have RCA out, just headphone, and my silver cables won't work with it without an adapter. It just wouldn't be apples to apples, plus I never liked the iFi anyway. I'll try it anyway.
I have a handful of true DSD files too, just the free samples from NativeDSD. I'll see what it does with those.
I really have no interest in purchasing DSD files either. In fact I never bought any SACDs when I finally had a player that could play them, by then they were going away and the one or two stores I knew that sold them only had a few left.
I'll continue to play with the Audirvana settings to see what sounds good to me, but so far this was the best. I had SOX upsampling from 16/44.1 to 16/176.4 and 16/48 to 16/192 on for a long time, and never felt like it added anything. I always left higher bitrates to play natively.
The problem with iFi iDSD is that it is a portable unit that doesn't have RCA out, just headphone, and my silver cables won't work with it without an adapter. It just wouldn't be apples to apples, plus I never liked the iFi anyway. I'll try it anyway.
Volumio is a linux distribution with javascript (node.js) control engine, communicating with MPD for playback Technical Overview * Volumio and displaying HTML UI in browser. If MPD is told to play a DSD track, it reports DSD. But before sending to the soundcard (alsa layer), it converts the output to a format supported by your specific card - PCM only.
The exact params of the stream being passed to the soundcard is reported by the alsa driver in file /proc/asound/cardX/pcm0p/sub0/hw_params :
Code:
pavel@precision:~$ cat /proc/asound/card4/pcm0p/sub0/hw_params
access: MMAP_INTERLEAVED
format: S24_3LE
subformat: STD
channels: 2
rate: 48000 (48000/1)
period_size: 20000
buffer_size: 40000Just access your volumio via ssh SSH - Volumio Documentation
I tried the iFi vs. Allo in both DSD and PCM on the same track (Norah in 24/192 as the source) and I will honestly say I couldn't tell the difference. I'm not sure I could say the silver cables made any difference either. It wasn't any kind of scientific test for sure but it was all I could do in a spare half hour. I had to switch the upsampling at the computer in the next room that runs Audirvana but I could switch the DAC using my phone via Volumio sitting in the same position and use the remote selector on the pre-amp. I tried to keep volume the same but wasn't measuring and just using my ear.
At least this is the result I expected once I did a test. You are always hearing something "better" when you change something, but when you are swapping back and forth sometimes it isn't so obvious. And the tracks in 24/192 are already ultra hi-res. I've always doubted anyone could hear a difference between 24/96 and up. Some say they can't hear a difference between redbook and hi-res. Some said they couldn't hear a difference between MP3 and CD too.
According to the iFi manual, the iDSD does bit-perfect decoding of DSD (hence the name). Maybe a higher-end DAC would produce different results, but I'm not so sure on that either. I won't get to try as I don't have anything truly high-end. I wanted to go the the latest NYC audio show a few weeks ago to hear some of the latest DACs, streamers etc... but went to Burning Amp instead.
At least this is the result I expected once I did a test. You are always hearing something "better" when you change something, but when you are swapping back and forth sometimes it isn't so obvious. And the tracks in 24/192 are already ultra hi-res. I've always doubted anyone could hear a difference between 24/96 and up. Some say they can't hear a difference between redbook and hi-res. Some said they couldn't hear a difference between MP3 and CD too.
According to the iFi manual, the iDSD does bit-perfect decoding of DSD (hence the name). Maybe a higher-end DAC would produce different results, but I'm not so sure on that either. I won't get to try as I don't have anything truly high-end. I wanted to go the the latest NYC audio show a few weeks ago to hear some of the latest DACs, streamers etc... but went to Burning Amp instead.
Upsampling is a bit of a fools errand...
For example, when upsampling from 16 bit to 24 bit you end up 16 bits of audio data in a 24 bit frame. A good codec will either silence or noise-fill the remaining 8 bits at the low volume (LSB) of the frame... but you are still listening to 16 bit audio.
When pushing from (for example) 48k to 96k samples per second you run into the problem of having more samples than data, the intermediate samples have to be filled with something so the system will either use duplicate samples or try to extrapolate the differences between samples. In the end, you are still listening to 48k audio.
It's pretty simple ... you can't extract what simply isn't there.
If there is an audible difference it will come from the way the DACs and Codecs set themselves up during the conversion. 24bit decoders require a larger dynamic range, for example. In a proper design these differences will be as minor as possible in order to remain true to the original source.
Converting to another format is, most times, just a matter of decoding then re-encoding the data, something like translation... If you hear a difference, the new encode is not being true to the original.
Now if you want to talk about downsampling ... converting DSD to MP3128 for example, there is lots to talk about since you are moving your music into a lossy format with definite limits on it's performance.
For example, when upsampling from 16 bit to 24 bit you end up 16 bits of audio data in a 24 bit frame. A good codec will either silence or noise-fill the remaining 8 bits at the low volume (LSB) of the frame... but you are still listening to 16 bit audio.
When pushing from (for example) 48k to 96k samples per second you run into the problem of having more samples than data, the intermediate samples have to be filled with something so the system will either use duplicate samples or try to extrapolate the differences between samples. In the end, you are still listening to 48k audio.
It's pretty simple ... you can't extract what simply isn't there.
If there is an audible difference it will come from the way the DACs and Codecs set themselves up during the conversion. 24bit decoders require a larger dynamic range, for example. In a proper design these differences will be as minor as possible in order to remain true to the original source.
Converting to another format is, most times, just a matter of decoding then re-encoding the data, something like translation... If you hear a difference, the new encode is not being true to the original.
Now if you want to talk about downsampling ... converting DSD to MP3128 for example, there is lots to talk about since you are moving your music into a lossy format with definite limits on it's performance.
Yes, we can't get something from nothing. That is understood. Most of what I was using though was already 24/96 or higher.
NP. In the meantime I've turned all the conversions off again. Ultimately I couldn't hear a difference.