Jump to content
IGNORED

HQ Player

freesteve

By freesteve
in Software

Recommended Posts

dean70

dean70

Posted
Posted
Just downloaded 3.10 (Mac OSK 10.11) and tried out the new Closed Form Interpolator with both DSD 7 and ASDM 7, up sampling Red Book CD to 2xDSD. There's an underlying low-level popping noise that's continuously there every second or so. Anyone else hearing that? (Feeding Geek Pulse via USB)

 

That sounds similar to what I experience when outputting to DSD256 to W4S DAC2SE. DSD128 works, so put it down to a bandwidth or CPU limitation.

Alchemy Desktop

http://www.origen.net.au/Alchemy/

Link to comment
57gold

57gold

Posted
Posted
How come I can not adjust the buffer in Preferences? It is greyed out.

 

Still have dropouts...Miska's only suggestion is that I have latency getting data to Mac Mini so replacing a 2X 1.5TB OWC Raid1 with USB2 out with a 4TB LaCie Thunderbolt fanless HDD. May solve dropouts and fan noise, which are both incredibly annoying.

 

Initial play back with LaCie HDD providing music files to HQPLayer (it's a USB3 and Thunderbolt 2 capable and fanless unit, made out of a very solid hunk of aluminum that acts as a heat sink) seems to have eliminated the dropouts and actually let me play with a non -2 filter.

 

Thought I was going to use the Thunderbolt as the link to the Mac Mini, as there is one on the unit, but in 2011 the Thunderbolt was a video screen output, not a fully functioning Thunderbolt jack. Learned this the hard way, when Mini did not recognize the LaCie attached with Thunderbolt. First strike on me.

 

Resorted to USB, and of course the 2011 Mini's USB ports are second generation and the LaCie has USB3 ports...strike two for me.

 

But, when playing music files from the LaCie using the USB3 output to USB2 input on 2011 Mini I do not get dropouts, at least not in first hour, nor do I get nasty fan and disc noise. And, for some reason, HQPlayer requires 10% less CPU to play CD rip files at DSD128 with a -2 filter versus files served up by the NewerTech HDD with USB2 output that I had previously used.

 

Curious, like when I sourced files from my onboard SSD on the 2011 Mini for HQPlayer playback, the CPU usage went up 10% versus files served up by the outboard NewerTech USB2 HDD?

 

Just placed order for a new, old stock 2012 Mac Mini with 2.6GHz Quad Core processor, 16GB RAM, 256 GB SSD, USB3 and Thunderbolt 1 ports...

 

My next adventure, if I can figure out how to set up the new/old Mac Mini, will be to see if I can use my 2011 Mini as a NAS or get a DSD256 capable DAC...or maybe I should just stop messing around with this crap and listen to the music, now that I will not have dropouts nor disc/fan noise!

 

More time to play my guitars.

Tone with Soul

Link to comment
Chodi

Chodi

Posted
Posted

I had the same issue with ticks in one channel on higher dsd rates. I solved it by using a usb 3 port. Even though the hardware (dac) and the drivers only support usb 2 I still got more bandwidth through the usb 3 port and solved the problem. It was not a problem for dsd 128 only dsd 256 or 512. So I guess there is some advantage to using a usb 3 port for dsd if you have one available.

 

Eventually the dacs will have usb 3 receivers built in and drivers that work with usb 3. For now I know of only one dac that has that but there may be more. There are sure to be more coming.

Link to comment
Miska

Miska

Posted
Posted
what are the characteristics of close form interpolation?

 

It is reversible back to original to data, some could call it "bit-perfect". That also means that it is not apodizing, so it doesn't remove any ringing that may already exist in a recording.

 

Most of all, when is it recommended? What hardware, what recording?

 

I don't really like it myself, or recommend it for anything. But since Schiit recently made much noise about such I thought it would be good addition and nice for comparison purposes. And someone could possibly like it.

 

It is also application where high number of points is useful, so since Chord on the other hand made much noise about number of taps, this is good place to use high number. So to tout about numbers in response to Chord, this one has 16k taps, input-relative. So multiply that with the conversion factor and you get total number. 44.1k to 11.3 MHz output that means 4M taps...

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
bibo01

bibo01

Posted
Posted
Just downloaded 3.10 (Mac OSK 10.11) and tried out the new Closed Form Interpolator with both DSD 7 and ASDM 7, up sampling Red Book CD to 2xDSD. There's an underlying low-level popping noise that's continuously there every second or so. Anyone else hearing that? (Feeding Geek Pulse via USB)

I, instead, have a problem after about 1min of playback - sound stops, time counter still moving but sound hiccups. Unusable.

Setting: closed-form, dsd7, dsd128 on ArchLinux.

Back to poly-sinc it's fine.

How curious are you?
Link to comment
Miska

Miska

Posted
Posted
I, instead, have a problem after about 1min of playback - sound stops, time counter still moving but sound hiccups. Unusable.

Setting: closed-form, dsd7, dsd128 on ArchLinux.

Back to poly-sinc it's fine.

 

What CPU do you have? It is certainly heavy affair for high oversampling ratios. Doing 8x to 352.8/384k PCM is light, but doing something like 256x is heavy.

 

My Xeon E5v3 (Haswell-EP) can do DSD256 to iFi iDSD Micro (on Ubuntu Studio) without problems, but DSD512 is out of question.

 

I could make a "light" version of the conversion by relaxing quality requirements (now the target is 32-bit PCM worth of resolution). Relaxing it to 24-bit level or so would already make it quite a bit faster.

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
bibo01

bibo01

Posted
Posted
What CPU do you have? It is certainly heavy affair for high oversampling ratios. Doing 8x to 352.8/384k PCM is light, but doing something like 256x is heavy.

 

My Xeon E5v3 (Haswell-EP) can do DSD256 to iFi iDSD Micro (on Ubuntu Studio) without problems, but DSD512 is out of question.

 

I could make a "light" version of the conversion by relaxing quality requirements (now the target is 32-bit PCM worth of resolution). Relaxing it to 24-bit level or so would already make it quite a bit faster.

i7-4790

I should not have upsampling problems with such cpu in this condition. I think the problem resides in the filter itself.

 

Are you saying that closed-form is more demanding than poly-sinc, with which I have no problem, for doing 44.1>DSD128?!

How curious are you?
Link to comment
Miska

Miska

Posted
Posted
i7-4790

I should not have upsampling problems with such cpu in this condition. I think the problem resides in the filter itself.

 

I think it should be fine. Not sure where the problem could be, since it is working fine for me on Linux too. Do you have pipeline SDM enabled? I do... I also have Buffer Time set to 100 ms.

 

Could be some incompatibility with ArchLinux, especially since I moved to C++11 in 3.9.

 

Are you saying that closed-form is more demanding than poly-sinc, with which I have no problem, for doing 44.1>DSD128?!

 

I didn't compare it to that one, but it is heavier than poly-shrt-mp-2s I'm using normally for going to DSD512, so I had to drop to DSD256.

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
Miska

Miska

Posted
Posted
Are you saying that closed-form is more demanding than poly-sinc, with which I have no problem, for doing 44.1>DSD128?!

 

Here are some load figures (with some small load increasing improvements to the closed-form)...

 

This is poly-sinc to DSD256:

hqp-poly-sinc.png

 

This is closed-form to DSD256:

hqp-closed-form.png

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
brightonjel

brightonjel

Posted
Posted

On my Macbook laptop (2.4 Ghz i5, 8GB, running OS X 10.11), cutting the DSD output rate so it was up sampling RB CD to 1xDSD instead of 2xDSD made no difference - the periodic low-level popping noise is still there with the Closed Form filter option. Even copying a music track to the desktop and having HQP play this instead of getting the track from my NAS via the network still produced the same result.

Link to comment
bibo01

bibo01

Posted
Posted
I think it should be fine. Not sure where the problem could be, since it is working fine for me on Linux too. Do you have pipeline SDM enabled? I do... I also have Buffer Time set to 100 ms.

 

Could be some incompatibility with ArchLinux, especially since I moved to C++11 in 3.9.

 

 

 

I didn't compare it to that one, but it is heavier than poly-shrt-mp-2s I'm using normally for going to DSD512, so I had to drop to DSD256.

With no NAA my buffer time it's fine at 5 ms with poly-sinc/short/mp filter (pipeline enabled) RB->DSD128.

With closed-form filter increasing buffer time to 50 ms (it may work with less) was enough to playback fine RB->DSD128.

How curious are you?
Link to comment
Miska

Miska

Posted
Posted
On my Macbook laptop (2.4 Ghz i5, 8GB, running OS X 10.11), cutting the DSD output rate so it was up sampling RB CD to 1xDSD instead of 2xDSD made no difference - the periodic low-level popping noise is still there with the Closed Form filter option. Even copying a music track to the desktop and having HQP play this instead of getting the track from my NAS via the network still produced the same result.

 

OK, I get this also on my MacMini when Pipeline SDM is disabled. Enabling it makes the problem go away...

 

I'll hunt down the problem and fix it.

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
bibo01

bibo01

Posted
Posted

Miska,

 

there are various reports that v3.10 (Windows, OSX) has improved sonically compared to the past and it is also a touch more optimized in terms of resources.

Have you made any changes that could explain it? Different compiler perhaps???

I noticed that 3.10 is slightly smaller in terms of KB compared to the previous three versions.

How curious are you?
Link to comment
Chodi

Chodi

Posted
Posted
Miska,

 

there are various reports that v3.10 (Windows, OSX) has improved sonically compared to the past and it is also a touch more optimized in terms of resources.

Have you made any changes that could explain it? Different compiler perhaps???

I noticed that 3.10 is slightly smaller in terms of KB compared to the previous three versions.

 

Interesting you should mention that. I also thought I heard a small improvement in sound quality on Windows 10 in v3.10. I also noticed somewhat less demand on resources.

Link to comment
elcorso

elcorso

Posted
Posted
How would you describe the differences ?

 

Hard to me because on my Spanglish :)

 

Less background noise, much better instrument definition and presence, much better bass (extension, punch, with zero phase shift), thanks to this bass the lovely soundstage?

 

Roch

 

PS/ Regarding "bass phase shift" I previously attributed certain problems on bass resonance caused by room resonances and subs & main speakers matching, but Miska solved, thanks to HQ P, the problem.

 

Roch

Link to comment
zackthedog

zackthedog

Posted
Posted
Mac OS X Yosemite, HQP 3.9.0.1, PCM192 FLAC file.

 

I just noticed poly-sinc-ext filter doesn't work. I mean playout doesn't even start. Not sure if it's a bug or not.

 

Have a nice day, Massimiliano

 

Poly-sinc-ext only does 2x or 4x upsampling, I believe. When I use it I notice the upsampling rate changes from 192 to 176, and my DAC does not do 176 so no sound. You can set upsampling to 88 and it should work.

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Go to topic listing


sonore

aurender

Lumin

HEAP

Schiit

SOtM

Mytek

mutec

uptone

jcat

sotm usa

audioquest

Taiko Audio

Teddy Pardo

Eversolo




×
×
  • Create New...