HQ Player

[Miska]

On 9/12/2019 at 4:38 PM, adamthebrave said:

Can you reccomend me a good GB switch?

 

Unmanaged:

16-port version on Zyxel GS1100, if you want optical support, get the 24-port version, NOT the 10-port version.

With this, enable 802.3az (EEE) support using the button on front panel.

 

Managed:

HPE 1820 series 24+2-port (24G) model. Supports also optical networking.

I'm also about to test Cisco SG350-10 switch. Should be good, but been too busy to do it yet.

With any managed switches remember to check that 802.3x Flow Control is enabled in settings! Also check that 802.3az support with cable length detection/link power adjustment is enabled, this way it doesn't unnecessarily blast at full power over short cables, but instead uses only the amount of signal power needed (lower noise).

 

All these switches are very reasonably priced but high quality.

 

[ThenewGearPPK]

Finally got HQplayer started & yikes, is there a lot of options to choose.

 

Has there been anyone who tested all the options to give them simple descriptions of what they sound like & what they do/how much power they need?

The combinations to try is in the millions.

[Yviena]

@Miska have you ever experimented with filters too see if there are actually gains to SQ by increasing the stop band attenuation to 340/360- dbfs like Rob watts from chord dave mentions?

[AnotherSpin]

1 hour ago, ThenewGearPPK said:

Finally got HQplayer started & yikes, is there a lot of options to choose.

 

Has there been anyone who tested all the options to give them simple descriptions of what they sound like & what they do/how much power they need?

The combinations to try is in the millions.

 

Look back through the thread, it's only 6 hundred pages to read...😎

[LoryWiv]

2 hours ago, ThenewGearPPK said:

Finally got HQplayer started & yikes, is there a lot of options to choose.

 

Has there been anyone who tested all the options to give them simple descriptions of what they sound like & what they do/how much power they need?

The combinations to try is in the millions.

The manual gives some insight into the filters, but there is no substitute for a bit of trial and error in your system. Actually once you decide if you want to output DSD versus PCM, begin to try modulator / filter combinations and see what both sounds good in your chain and keeps cpu load manageable, the choices narrow quickly. Then every now and then it's fun to reassess, but I think many users settle into some standard settings for most listening. 

 

Currently I'm outputting everything to DSD128 with ASDM7EC modulator & poly-sinc-xtr-mp-2s filter, and results are excellent. YMMV.

[Miska]

4 hours ago, Yviena said:

@Miska have you ever experimented with filters too see if there are actually gains to SQ by increasing the stop band attenuation to 340/360- dbfs like Rob watts from chord dave mentions?

 

Where does he talk about such? So far Chord has been using filters with 120 dB stop-band attenuation.

 

For example if you output 32-bit PCM, using such filter would be pointless because there's no output resolution to represent it... So for example for something like MScaler it would be totally pointless.

 

[asdf1000]

4 hours ago, Yviena said:

@Miska have you ever experimented with filters too see if there are actually gains to SQ by increasing the stop band attenuation to 340/360- dbfs like Rob watts from chord dave mentions?

 

Here is the background of this number:

 

https://www.head-fi.org/threads/chord-electronics-dave.766517/page-315#post-12859578

 

3 minutes ago, Miska said:

So far Chord has been using filters with 120 dB stop-band attenuation.

 

Yep:

 

https://www.head-fi.org/threads/chord-electronics-dave.766517/page-71#post-12172918

[Miska]

2 minutes ago, asdf1000 said:

Here is the background of this number:

 

https://www.head-fi.org/threads/chord-electronics-dave.766517/page-315#post-12859578

 

OK, I have seen that. There he is talking about the SD modulator...

 

3 minutes ago, asdf1000 said:

Yep:

 

https://www.head-fi.org/threads/chord-electronics-dave.766517/page-71#post-12172918

 

Yes, that's about the filter (WTA).

 

[asdf1000]

2 hours ago, Miska said:

OK, I have seen that. There he is talking about the SD modulator...

 

I know. Just shared the background, in case @Yviena thought this ~ 350dB number was about the WTA filter.

[Yviena]

Ahh ok my mistake I thought it was the actual filter that was at -360 dbfs...

[ThenewGearPPK]

 

 

 

So the options for input & output Backend need to be the same? I set Output Backend to ASIO & input Backend to none.

Is this right?

Digital Output Mode is set to SOURCE, instead of PCM or SDM . I won't get a performance boost switching to PCM or DSD specifically??

 

Then there are two columns; PCM & SDM & with their respective options/filters.

Are they both active? How would I know which of them is active or how would I know if I am upsampling PCM or DSD?

 

My DAC can play 512 externally & has an option for NOS mode & I only play FLAC files. Would I have to enable NOS mode to get the the benefits of upsampling & would I need to manually select Digital Output Mode to SDM to have the SDM filter column be active?

 

 

 

[Miska]

1 hour ago, ThenewGearPPK said:

So the options for input & output Backend need to be the same? I set Output Backend to ASIO & input Backend to none.

Is this right?

 

If you don't need input support, you can turn it off by setting input backend to "none".

 

1 hour ago, ThenewGearPPK said:

Digital Output Mode is set to SOURCE, instead of PCM or SDM . I won't get a performance boost switching to PCM or DSD specifically??

 

Usually you would have either PCM or SDM selected, with settings to output at format that gives best performance with your DAC. Very few DACs natively support both and for even fewer it makes sense to switch between the formats. So "[source]" would be extremely rarely used.

 

1 hour ago, ThenewGearPPK said:

Then there are two columns; PCM & SDM & with their respective options/filters.

Are they both active? How would I know which of them is active or how would I know if I am upsampling PCM or DSD?

 

If you set output mode to "[source]" PCM settings are active for PCM sources, and SDM settings for DSD sources. If you set output mode to PCM, then only PCM settings are active and if you set output mode to SDM, then only SDM settings are active.

 

1 hour ago, ThenewGearPPK said:

My DAC can play 512 externally & has an option for NOS mode & I only play FLAC files. Would I have to enable NOS mode to get the the benefits of upsampling & would I need to manually select Digital Output Mode to SDM to have the SDM filter column be active?

 

Yes, that would likely be the preferred way.

 

[Yviena]

16 hours ago, Miska said:

 

OK, I have seen that. There he is talking about the SD modulator...

 

 

Just curious but what's the performance of your modulators, does it also reach 60+ bits like mentioned by rob watts?

[Miska]

7 hours ago, Yviena said:

Just curious but what's the performance of your modulators, does it also reach 60+ bits like mentioned by rob watts?

 

I don't want to go more into the idiotic game of totally meaningless numbers Chord is playing. At what bandwidth? Rob has gone for simplified modulator and just maximum dynamic range. Down side is that it produces annoying amount of horrible spurious tones.

 

Here's wide band output of 0 - 22.05 kHz sweep from Chord Mojo:

There some strange stuff especially just above 3 MHz frequency.

 

Here's same spectrum with 1 kHz test signal:

You can better see the forest of spurious tones. And although he says the modulator outputs 100 MHz, noise floor jumps up already at 100 kHz.

 

Just like digital filters, modulators cannot be chacterized simply with one single number. DSD256 from proper modulator gives you better in- and out-of-band noise without spurious tones. Too bad you cannot utilize that with DACs like Mojo, because it converts DSD to PCM and then puts it through it's own modulator again, spoiling the performance.

 

[asdf1000]

1 hour ago, Miska said:

Here's same spectrum with 1 kHz test signal:

 

Would be interesting to see if optical input is better for this test...

 

He has openly said Mojo is mpre sensitive to incoming RF than his other DACs, for cost reasons.

 

Optical is immune to RF interference, so would be interesting if this measurement is better or same.

 

[Miska]

14 minutes ago, asdf1000 said:

He has openly said Mojo is mpre sensitive to incoming RF than his other DACs, for cost reasons.

 

Yes, it is very sensitive, that is probably why HiFi-News is using it for measuring differences between different USB sources. But this input signal dependent picket fence is feature of the modulator, RF interference is not correlated with the data.

 

The oversimplified modulator in TI/BB DAC chips have a similar feature, but to lesser extent. So if you send PCM to those chips output is somewhat similar. However, they luckily can convert DSD natively, so you can get rid of the problem by using DSD mode.

 

 

[AnotherSpin]

On 9/14/2019 at 4:06 PM, Miska said:

 

[...]

STP/FTP cable types are absolutely no-no for NAA use. So at least check that the RJ45 connector on the cable is full plastic-body, not the metal body type.

 

It absolutely shouldn't make any sound quality difference. If there is difference, it needs some research to find out why to eliminate it. If it makes difference, it is negating point of NAA.

[...]

 

 

I was experimenting with different types of ethernet cables between my mac mini and NUC endpoint, from cheap generic no-names to Supra CAT8. Never tried exotic very expensive cables though. 

 

As a result I should fully confirm Miska's advice about no STP type connectors. After many trials I went as far as removed metal cover of Supra CAT8 RJ45 connector and it makes this particular cables sound better in my set. I am not suggesting doing the same with this very good cables, but UTP cables should be surely tried and used in NAA systems.

[Yviena]

12 hours ago, Miska said:

 

I don't want to go more into the idiotic game of totally meaningless numbers Chord is playing. At what bandwidth? Rob has gone for simplified modulator and just maximum dynamic range. Down side is that it produces annoying amount of horrible spurious tones.

 

Here's wide band output of 0 - 22.05 kHz sweep from Chord Mojo:

There some strange stuff especially just above 3 MHz frequency.

 

Here's same spectrum with 1 kHz test signal:

You can better see the forest of spurious tones. And although he says the modulator outputs 100 MHz, noise floor jumps up already at 100 kHz.

 

Just like digital filters, modulators cannot be chacterized simply with one single number. DSD256 from proper modulator gives you better in- and out-of-band noise without spurious tones. Too bad you cannot utilize that with DACs like Mojo, because it converts DSD to PCM and then puts it through it's own modulator again, spoiling the performance.

 

Ahh okay so it's basically marketing talk/boasting...

 

On another note I think I remember you saying somewhere that you wanted to create a standard sending multi-bit DSD, has there been any progress on that or did you decide to scrap that?

[Quadman]

I just installed HQP 4.  I was then resetting AO 3 and with service tool I was trying to do a shell replacement, Roonserver and HQPlayer.  But AO gives me message HQP cannot be found.  I have it (HQP) installed in the default directory.  So I then tried to set just HQP as default shell and got this message  "Install HQP in the default directory: C:\Program Files\signalyst\HQPlayer Desktop 3\"

 

Well it is in that directory, but it is now HQPlayer 4, not 3.   How can I change this so AO can see it?  Any ideas?

[Miska]

18 minutes ago, Yviena said:

On another note I think I remember you saying somewhere that you wanted to create a standard sending multi-bit DSD, has there been any progress on that or did you decide to scrap that?

 

Well, there are good enough ways already, and moving over from USB to networked solutions solves that for good. But OTOH, I don't see any pressing need for such either. So practically it is on hold, because there are more important things to work on.

 

[Paetrick]

26 minutes ago, Quadman said:

I just installed HQP 4.  I was then resetting AO 3 and with service tool I was trying to do a shell replacement, Roonserver and HQPlayer.  But AO gives me message HQP cannot be found.  I have it (HQP) installed in the default directory.  So I then tried to set just HQP as default shell and got this message  "Install HQP in the default directory: C:\Program Files\signalyst\HQPlayer Desktop 3\"

 

Well it is in that directory, but it is now HQPlayer 4, not 3.   How can I change this so AO can see it?  Any ideas?

 

You could use the custom shell for that. I would copy the content of the roonserverhqp_shell_x64 file into the custom_shell file, modify the HQP path there and use the custom shell. AO 3.0 has HQP3 paths preconfigured, but it does not offer the path for the new HQP version.

 

Old one is C:\Program Files\Signalyst\HQPlayer Desktop 3\HQPlayer-desktop.exe

New one is C:\Program Files\Signalyst\HQPlayer 4 Desktop\HQPlayer4Desktop.exe

in your case.

 

Further help and instructions can be found here: custom shell as shell replacement

 

[Quadman]

1 hour ago, Paetrick said:

You could use the custom shell for that

 

Thank you Patrick.  I didn't think of that.  I tried renaming the folder in program files Signalyst to HQPlayer Desktop 3, but of course that still left the actual program HQP 4 and I still got the error message.

[k6davis]

On 9/12/2019 at 4:53 PM, dminches said:

Are you using CUDA off-load now?

 

Sorry for the late reply. No, I'm not using CUDA anymore. In my stereo, non-convolution setup, I don't need it.

[Yviena]

@Miska just noticed there's a new filter in the 4.3 pro release, is it just a longer poly-sinc filter, or has there been other changes to it?

 

 

On another note miska, remember that i said the newest release has drop outs when multithread DSP is set to enabled  without touching, affinity settings which i need to get smooth playback,

i been monitoring the threading behavior, and it seems it's due to scheduling/threading conflicts, load is actually more spread between cores, on 4.1.0.1 build which never hit 100% thread usage.

 

While it seems that Ryzen has something called favored cores which is a firmware thing which tells windows which cores are likely to boost/perform best, i think it's basically like core affinity but baked into firmware, and i can see it heavily favoring core 2 which leads to 100% thread usage on the newest build.

i can confirm though that it's only on windows not linux that it happens.

[Miska]

1 hour ago, Yviena said:

@Miska just noticed there's a new filter in the 4.3 pro release, is it just a longer poly-sinc filter, or has there been other changes to it?

 

Yes, it is just longer/steeper version. I was actually asked for such two years ago, but didn't have a good idea what kind of spec such should have. Now I finally decided it while working on some other things. So it sort of completes the basic poly-sinc set, which is now three filters; poly-sinc-short, poly-sinc and poly-sinc-long. These three are closely related. Other poly-sinc filters have bigger differences.

 

1 hour ago, Yviena said:

On another note miska, remember that i said the newest release has drop outs when multithread DSP is set to enabled  without touching, affinity settings which i need to get smooth playback,

i been monitoring the threading behavior, and it seems it's due to scheduling/threading conflicts, load is actually more spread between cores, on 4.1.0.1 build which never hit 100% thread usage.

 

With multicore enabled the work split is same in those versions. But unless you disable core pinning, latest version pins threads to specific cores so that OS doesn't get bad ideas about moving threads around and suddenly putting too much on a single core which would make playback stall.

 

AMD was doing aggressive power management in their Ryzen Gen 3 CPU firmware, which they have now backed out a bit in their latest version because it was causing problems in certain conditions. This should eventually come through a BIOS update.

 

Not hitting 100% thread usage with uneven per-core loads can be also indication of threads bouncing around to different cores. Those load figures are always averages for some time window. When the CPU is doing work it is always at 100% and the load percentage is just figure how many percents of the measurement time interval it was doing work instead of running idle loop.

 

1 hour ago, Yviena said:

While it seems that Ryzen has something called favored cores which is a firmware thing which tells windows which cores are likely to boost/perform best, i think it's basically like core affinity but baked into firmware, and i can see it heavily favoring core 2 which leads to 100% thread usage on the newest build.

i can confirm though that it's only on windows not linux that it happens.

 

I'm thinking about way to let you tell which cores you want to get allocated. Extending the environment variable is easy, doing a nice GUI for it is not so straightforward.