[460414]

 
When you do sample rate conversion the quality of that conversion, especially the digital filtering varies.  The conversion in Windows is not a terrific algorithm.  The filters in the player I suggested and also some other players is of far higher quality. 
 
http://src.infinitewave.ca/
 
Look here.  Though these are down-conversions you can see quite a range of quality in the conversion depending upon what is used. 
 
In the case of HQ player the filtering is complex and of high quality using many taps.  If you upsample to the max rate your DAC can do then it does less internal upsampling and filtering.  This lets the player software use better filtering for conversion of 44 khz material.  Less aliasing, less noise, better results.  If your DAC has a very poor analog section it may swamp the benefit, but most decent ones are good enough. 

Looking at that link the signalyst one: "signalyst 2.9.1 (sync)" seem to definitely be the one of the best overall (not that any DACs have a noise floor down at sub 180db, but still..). 
Do they have the default windows up-sampler tested in there? Couldn't see it?
 
Some of them are just terrible, eg:

 
Mentioned signalyst 2.9.1 (sync) for reference
 

With 44.1khz sampling rate there should get a bit of stuff above 22khz, judging from the impulse test it shouldn't be too bad.
 
Do these test applying when up-sampling using the same method as well?

 

[spruce music]

  Looking at that link the signalyst one: "signalyst 2.9.1 (sync)" seem to definitely be the one of the best overall (not that any DACs have a noise floor down at sub 180db, but still..). 
Do they have the default windows up-sampler tested in there? Couldn't see it?
 
Some of them are just terrible, eg:

 
Mentioned signalyst 2.9.1 (sync) for reference
 

With 44.1khz sampling rate there should get a bit of stuff above 22khz, judging from the impulse test it shouldn't be too bad.
 
Do these test applying when up-sampling using the same method as well?

 
Yes, another good free up or down sampler is SOX.  If you use Audacity, it uses Sox for resampling. You also can get Sox plugins for Foobar.
 
The results show more problems with downsampling than you get with upsampling.  And no I don't think there is the Windows resampler in that group of tests. 
 
HQ player allows quite a few settings.  If nothing else it is useful for listening to that for yourself.  You can do minimum phase, or his preferred filters and several variations in between.  You can download a free trial which is fully functional.  No reason not to give it a test.  The fellow who writes this is Jussi Laako.  Seriously good filter designer.  He works for Intel, worked for Nokia, Finnish Defense on sonar systems.  Unlike many audiophile products his is not some BS, he really designs better filters.  I have doubts of audibility, but in a measured sense they are better.  This playback software is preferred by many people who can afford the best of the best. 

 

[460414]

   
The fellow who writes this is Jussi Laako.  Seriously good filter designer.  He works for Intel, worked for Nokia, Finnish Defense on sonar systems.  Unlike many audiophile products his is not some BS, he really designs better filters.  I have doubts of audibility, but in a measured sense they are better.

I'm all for technical improvements when it comes to SS amps/dacs within cost/practicality, even if its not audible.. An audible difference is just a bonus.(headphones/speakers I only care what they sound like)
 
With standard CD quality 16/44.1; from a technical point of view what would the best bit rate and sample rate to up sample it to? No dacs can do up to or over 24 bits so I assume there's not any point on going for higher than that? But what about sample rate?
 
HQ player says "sample rate (/limit)" does that mean it will only use the same sample rate as the source?

 

[spruce music]

  I'm all for technical improvements when it comes to SS amps/dacs within cost/practicality, even if its not audible.. An audible difference is just a bonus.(headphones/speakers I only care what they sound like)
 
With standard CD quality 16/44.1; from a technical point of view what would the best bit rate and sample rate to up sample it to? No dacs can do up to or over 24 bits so I assume there's not any point on going for higher than that? But what about sample rate?
 
HQ player says "sample rate (/limit)" does that mean it will only use the same sample rate as the source?

The best sample rate if using HQ player is whatever the highest your DAC can do.  Often 192 khz, though some do high rate DSD or 384 khz.  Some DACs do now have 32 bit operation.  Though analog noise will limit the resulting output.  HQ player isn't limited by the rate of the source material.  It can upsample and digitally filter as it plays to whatever your DAC can use.

 

[castleofargh]

I can't say that I have much love for non free audio players. I get that people need to eat but the free ones do a pretty fine job already. so it somehow annoys me when it doesn't bother me to pay for other types of software(humans and their weird way of thinking

). ATM the only thing that annoys me with foobar is that I can't use the VST3 and/or 64bit version of some stuff.

 

[460414]

 
The best sample rate if using HQ player is whatever the highest your DAC can do.  Often 192 khz, though some do high rate DSD or 384 khz.  Some DACs do now have 32 bit operation.  Though analog noise will limit the resulting output.  HQ player isn't limited by the rate of the source material.  It can upsample and digitally filter as it plays to whatever your DAC can use.

My DAC (Auralic Vega) can revive a maximum of 32/384. From a technical point of view is it any better to run say 32bit over 24bit? Considering the nose floor is at -140/-145db with a 50khz sine at 0dbfs going off bibliophiles measurements. Say its -130db at more normal frequencies. That equates to ~21.67 bits for 6db =1 bit. 21.67 bits is ​

3,336,721 different values, 24bit is 5 times that at 16,777,216.00. From that 24 bits should be sufficient with no benefit of 32bit?​

 

As I said I'm talking technical here, not audible...​

 

[castleofargh]

whatever works. if you have a clean input you should try RMAA or the likes with different settings. why wonder when you could know? ^_^
asio by default will use the highest sample rate available for the DAC, while wasapy will by default use whatever it's playing. both are called "bit perfect".
sampling makes little difference on my DACs, it only improves my 30$ dac/amp/adc startech crapolio device, but then it's failing pretty bad at 44.1 so anything is logically and effectively better(I wouldn't go as far as saying it's good though ^_^). 
while going to 24bit improves measurably my odac(but I honestly can't hear it). I let it on max bit depth available because it gives me good conscience when I change loudness from the computer which I always do.

 

[spruce music]

 

My DAC (Auralic Vega) can revive a maximum of 32/384. From a technical point of view is it any better to run say 32bit over 24bit? Considering the nose floor is at -140/-145db with a 50khz sine at 0dbfs going off bibliophiles measurements. Say its -130db at more normal frequencies. That equates to ~21.67 bits for 6db =1 bit. 21.67 bits is ​

3,336,721 different values, 24bit is 5 times that at 16,777,216.00. From that 24 bits should be sufficient with no benefit of 32bit?​

 

As I said I'm talking technical here, not audible...​

I really don't know.  If the software and DAC do 32 bit then I would use it as it probably doesn't help, but also surely doesn't hurt.  As suggested you could do RMAA or some other tests and see if anything improves.  I doubt it would. 
 
I have not seen bibliophiles measurements.  I doubt the noise floor is -145 db.  If you know this, then ignore it.  One thing to keep in mind when looking at FFT's is how they let us look deeper into noise floors.  But don't confuse that with the measured noise floor of the device over the whole bandwidth.  Let us say I measure white noise over 0-20 khz at -10 db.  If I split that into two bands 10 khz wide, then each band will be at a level of -13 db.  If I now split that 20 khz band into 4 bands 5 khz wide the level will be -16 db in each of those 4 bands.  That is what an FFT is doing it is splitting bands narrower.  However a sine tone will read its actual level because all of the energy will be inside one band of the FFT.   So for instance a device with a -100db noise floor (assuming it is close to white noise) would graph the noise at around - 145 db using a 64K FFT. 
 
The Vega is excellent in terms of noise and does lie near -130 db at near 21 bit for the who audible band.  Which under the right conditions on an FFT would show at something below -150 db.   So yes one thinks there would be likely no benefit in 32 bit over 24 bit. 

 

[mindbomb]

Upsampling is to reduce transport jitter or to make filtering easier.

 

[pila405]

How come nobody mentions dbpoweramp when it comes to SRCs?
Seems like more than a decent option. Even the best, from the info on http://src.infinitewave.ca/

 

[GuyUnder]

Unless you are using a NOS DAC, it's going to oversample your PCM input much higher than your software will handle anyway.

Modern Sabre and AK449x DACs can bypass the USB input and PCM conversion stage using an I2S input, thereby eliminating the self-noise generated by those steps entirely. Software that can convert PCM to DSD and bitstream the DSD over USB or DoP over spdif and then into an I2S port via an I2S bridge will straight into the DAC chip for upconversion to some super high rate DSD stream -- THAT makes a large improvement in sound. I've seen it described as "transformational", but my own experience is more in the realm of "significant improvement". I couldn't tell the difference between HQ Player and JRiver in that setup.

Indeed, the reduction in digital grain/haze is so pronounced I will never consider another DAC that doesn't have an I2S port. That puts me at odds with the Yggdrasil fans, but the benefit is just so compelling...

 

[460414]

I have not seen bibliophiles measurements. I doubt the noise floor is -145 db. If you know this, then ignore it. One thing to keep in mind when looking at FFT's is how they let us look deeper into noise floors. But don't confuse that with the measured noise floor of the device over the whole bandwidth. Let us say I measure white noise over 0-20 khz at -10 db. If I split that into two bands 10 khz wide, then each band will be at a level of -13 db. If I now split that 20 khz band into 4 bands 5 khz wide the level will be -16 db in each of those 4 bands. That is what an FFT is doing it is splitting bands narrower. However a sine tone will read its actual level because all of the energy will be inside one band of the FFT. So for instance a device with a -100db noise floor (assuming it is close to white noise) would graph the noise at around - 145 db using a 64K FFT.

The Vega is excellent in terms of noise and does lie near -130 db at near 21 bit for the who audible band. Which under the right conditions on an FFT would show at something below -150 db. So yes one thinks there would be likely no benefit in 32 bit over 24 bit.

Here's Stereophiles measurement:

(~-140dbfps noise floor)

Here is from another source:

 

[460414]

  How come nobody mentions dbpoweramp when it comes to SRCs?
Seems like more than a decent option. Even the best, from the info on http://src.infinitewave.ca/

 
Compared to the Signalyst ones the dbpoweramp looks superior measurement wise IMO (top signalyst bottom dbpoweramp)

 
 
I looked into RMAA. It looks like it's meant primarily meant for purely digital analysis? For testing anything serious via an analogue input you would be limited by the on-board ADC?
I'm away from home but I can try any of the mentioned on 29th or later and post any results.

 

[castleofargh]

Unless you are using a NOS DAC, it's going to oversample your PCM input much higher than your software will handle anyway.

Modern Sabre and AK449x DACs can bypass the USB input and PCM conversion stage using an I2S input, thereby eliminating the self-noise generated by those steps entirely. Software that can convert PCM to DSD and bitstream the DSD over USB or DoP over spdif and then into an I2S port via an I2S bridge will straight into the DAC chip for upconversion to some super high rate DSD stream -- THAT makes a large improvement in sound. I've seen it described as "transformational", but my own experience is more in the realm of "significant improvement". I couldn't tell the difference between HQ Player and JRiver in that setup.

Indeed, the reduction in digital grain/haze is so pronounced I will never consider another DAC that doesn't have an I2S port. That puts me at odds with the Yggdrasil fans, but the benefit is just so compelling...

welcome to headfi, in this particular sub section of the forum we're not big on subjective impressions from the guy who knows a guy. we go for blind tests or measurements instead when we can.

  ....
I looked into RMAA. It looks like it's meant primarily meant for purely digital analysis? For testing anything serious via an analogue input you would be limited by the on-board ADC?
I'm away from home but I can try any of the mentioned on 29th or later and post any results.

I suggested RMAA because everybody can do it without any measurement gear. you have a bunch of signals in a wave file, you send them however you like from whatever source, and record them also however you like(directly in RMMA in a loop or any recording method and then import). but it has quite a few limitations, it's the go to software because it free and is easy to use, not because it's the best tool. if you have actual means to measure stuff, RMMA probably won't amaze you. ^_^