HQ Player

[Quadman]

6 hours ago, micheloupatrick said:

the apodizing counter reflects the content of the track before it is upsampled

According to Jussi's quote above in my post if a track has 100 indications then it has some clipping but probably does not need an apodizing filter.  Almost every track I looked at had into the 100's or even thousands and this was using apodizing filters.  Like Ted says, what are we to do with this data?  I briefly tried a non-apodizing filter and the counter moved at about the same rate as it did with an apodizing filter.  The music sounded so good, but I should have played Track A with apodizing filter then repeat Track A with non-apodizing and see how many indications each produced.  I take it that the apodizing filter should minimize these instances.  If not then what is the reason for this?

[Account Closed]

2 hours ago, Quadman said:

According to Jussi's quote above in my post if a track has 100 indications then it has some clipping but probably does not need an apodizing filter.  Almost every track I looked at had into the 100's or even thousands and this was using apodizing filters.  Like Ted says, what are we to do with this data?  I briefly tried a non-apodizing filter and the counter moved at about the same rate as it did with an apodizing filter.  The music sounded so good, but I should have played Track A with apodizing filter then repeat Track A with non-apodizing and see how many indications each produced.  I take it that the apodizing filter should minimize these instances.  If not then what is the reason for this?

I think he has it setup so that the APO analysis is continuous. Using an apodizing filter will correct the output but have no effect on the analysis which is probably happening at the input stage and is only there.

[Quadman]

1 hour ago, bobflood said:

but have no effect on the analysis which is probably happening at the input stage and is only there.

That makes sense, and probably how it's is.  Should be written a bit more clearly in the manual so this confusion would be avoided.

[toddrhodes]

Trying out the new Sinc-Mx and using it on PCM384/LNS15 (also tried NS5 but LNS15 seems more clear and more well-defined to me?) and really digging it. 

 

Reference material is none other than the original CD of Megadeath - Rust in Peace - Tornado of Souls.

[AudioDoctor]

9 minutes ago, toddrhodes said:

Trying out the new Sinc-Mx and using it on PCM384/LNS15 (also tried NS5 but LNS15 seems more clear and more well-defined to me?) and really digging it. 

 

Reference material is none other than the original CD of Megadeath - Rust in Peace - Tornado of Souls.

 

Have you previously tried Sinc-L and LNS15? If so, what are the differences with Sinc-Mx?

[bibo01]

39 minutes ago, Quadman said:

That makes sense, and probably how it's is.  Should be written a bit more clearly in the manual so this confusion would be avoided.

What is confusing in my view is that for the clipping indicator, you adjust the volume and it is fixed; for the apodizing detector, you choose an appropriate filter and nothing changes. 

[toddrhodes]

5 minutes ago, AudioDoctor said:

 

Have you previously tried Sinc-L and LNS15? If so, what are the differences with Sinc-Mx?

 

I have used that combo before, yes, but IIRC I was using it on DSD256 so Sinc L was doing a lot more taps than even Sinc-M would have done at that point. It was a very nice sound but was a bit, polite? 

 

I recently switched to battery power for my Pi and my D2D converter (Ian Canada parts). Part of that switch was using an Amanero Combo384 USB input and, to my ears, I prefer the muscle and immediacy of PCM384 on this setup, so it's a bit of an apples and oranges comparison I think. Because even converting 44.1 to 384, Sinc L isn't doing a whole heck of a lot different from Sinc-M at that point in terms of taps and such. And since I believe that lends to the overall sound - I switched to using Sinc M for both PCM options, on 4.10.3.

 

To me, and this is completely unscientific, just trying to pull out sonic memories from the brain jelly, Mx + LNS15 seems to be a bit stronger sounding versus Sinc-M + NS5. Stronger in the upper bass and lower midbass, which is where I find a lot of the meat and foundation of the music I prefer. And since it seems to marry that up with the detail and imaging I really enjoyed from the DSD256 Sinc-L days - it seems like a pretty nice combo to me.

 

Listening now to Art Blakey & the Jazz Messengers - Mosaic. This was a track I really grew to love through that DSD256/Sinc L combo. But man does it tax even a very good PC - but I recall how vivid the title track sounded, especially when Art pivots to his solo. It's one of those moments where you can be chatting with buddies or reading liner notes or something and that part comes on and you just kinda snap to attention and stay that way til it's done. Maybe it's just a really good recording. But with this Mx + LNS15 combo, I kinda have that same feeling throughout the whole track, not just the solo. 

 

 

[AudioDoctor]

8 minutes ago, toddrhodes said:

 

I have used that combo before, yes, but IIRC I was using it on DSD256 so Sinc L was doing a lot more taps than even Sinc-M would have done at that point. It was a very nice sound but was a bit, polite? 

 

I recently switched to battery power for my Pi and my D2D converter (Ian Canada parts). Part of that switch was using an Amanero Combo384 USB input and, to my ears, I prefer the muscle and immediacy of PCM384 on this setup, so it's a bit of an apples and oranges comparison I think. Because even converting 44.1 to 384, Sinc L isn't doing a whole heck of a lot different from Sinc-M at that point in terms of taps and such. And since I believe that lends to the overall sound - I switched to using Sinc M for both PCM options, on 4.10.3.

 

To me, and this is completely unscientific, just trying to pull out sonic memories from the brain jelly, Mx + LNS15 seems to be a bit stronger sounding versus Sinc-M + NS5. Stronger in the upper bass and lower midbass, which is where I find a lot of the meat and foundation of the music I prefer. And since it seems to marry that up with the detail and imaging I really enjoyed from the DSD256 Sinc-L days - it seems like a pretty nice combo to me.

 

Listening now to Art Blakey & the Jazz Messengers - Mosaic. This was a track I really grew to love through that DSD256/Sinc L combo. But man does it tax even a very good PC - but I recall how vivid the title track sounded, especially when Art pivots to his solo. It's one of those moments where you can be chatting with buddies or reading liner notes or something and that part comes on and you just kinda snap to attention and stay that way til it's done. Maybe it's just a really good recording. But with this Mx + LNS15 combo, I kinda have that same feeling throughout the whole track, not just the solo. 

 

 

 

Interesting. I am listening now and I can almost see the polite-ness in the sound, or how someone could get there. But god its beautiful on the right music. And to me, the most natural. Perhaps I will give Mx a try this weekend. Thanks for the reply.

 

I remembered that I can change it remotely again, so am listening to Mx now. I see what you're saying.

[toddrhodes]

1 minute ago, AudioDoctor said:

 

Interesting. I am listening now and I can almost see the polite-ness in the sound, or how someone could get there. But god its beautiful on the right music. And to me, the most natural. Perhaps I will give Mx a try this weekend. Thanks for the reply.

 

That's the thing - if I could, I'd will it to be that Roon would allow integration with HQP to a level that some genres and styles use this format and filter/shaper setting, and others use different settings. For example - metal and prog and hard rock? PCM all day. Jazz, Classical and Acoustic music? DSD. 

 

As it is, I just have to kind of go with what I listen to most. But I do occasionally stop and change over the settings and go on a classical/jazz/classical guitar binge.

 

But as I listen to some Tool right now? I just prefer the immediacy of PCM that I get in my room/system, and I don't necessarily need the space and 3D feel that DSD adds at the expense of some of that leading edge muscle. 

 

But that's also why I really like this whole setup - it can adapt and it's not that inconvenient to do.

 

Cheers!

[AudioDoctor]

Right now I am listening to Irish electronica/dance artist Roisin Murphy and the Mx filter suits this very well.

 

 

[Miska]

On 5/11/2021 at 2:37 PM, Jean Paul D said:

M1 has 8 cores and can't do it at a mere 128( but/7EC + convo)

 

It has 4 high power cores which go to the modulator and 4 low power cores which are left for running filters. It is so called ARM big.LITTLE architecture.

 

[Miska]

On 5/11/2021 at 8:41 PM, GoldenOne said:

Well, got the NUC set up with windows yesterday. Switched to linux (ubuntu 20.04) today. Won't run 1.536mhz via NAA :(
768khz is fine, but 1.536mhz is a no go.

 

You'll need my custom kernel, stock Ubuntu kernel won't do.

 

On 5/11/2021 at 8:41 PM, GoldenOne said:

Additionally with the HQP NAA OS installed on another intel x5 mini pc, that 'works' but I get pauses every few seconds. Whereas with windows installed on it it's fine.

 

Sounds like some networking issue....

 

[The Computer Audiophile]

7 hours ago, toddrhodes said:

 

That's the thing - if I could, I'd will it to be that Roon would allow integration with HQP to a level that some genres and styles use this format and filter/shaper setting, and others use different settings. For example - metal and prog and hard rock? PCM all day. Jazz, Classical and Acoustic music? DSD. 

 

As it is, I just have to kind of go with what I listen to most. But I do occasionally stop and change over the settings and go on a classical/jazz/classical guitar binge.

 

But as I listen to some Tool right now? I just prefer the immediacy of PCM that I get in my room/system, and I don't necessarily need the space and 3D feel that DSD adds at the expense of some of that leading edge muscle. 

 

But that's also why I really like this whole setup - it can adapt and it's not that inconvenient to do.

 

Cheers!

That would be next level stuff. A way to tag albums to identify filter choices and have the player adjust. Would be awesome. 

[GoldenOne]

3 hours ago, Miska said:

 

You'll need my custom kernel, stock Ubuntu kernel won't do.

 

 

Sounds like some networking issue....

 

So I replaced the device with this: https://www.amazon.co.uk/gp/product/B08T6CT5VL
Fanless, LAN, and is now working 100% perfectly with the NAA Image :D

No idea what the bottleneck in the previous device was but this is working wonderfully.
If anyone else is looking for an up-board alternative, this works great. nice and compact too!

@Miska I don't suppose there's a way to get roon bridge running alongside on the NAA image?
Would be handy to be able to have both. That was one of the benefits of (attempting to run) ubuntu.

Thank you!

[Miska]

On 5/12/2021 at 8:40 AM, LoryWiv said:

I may be too unsophisticated to fully understand your explanation above so I'll just ask: are the variable # taps with constant length at each multiple applicable to PCM sources whether they are upsampled tp higher rate PCM or output to SDM / DSD?

 

Number of taps depends on the output rate. So the length is about 1.5 seconds and delay thus about 0.75 seconds.

 

[Miska]

On 5/12/2021 at 2:46 PM, Flextreme said:

At 4 or more logical CPUs, HQP threads seem to start jumping cores and it this seems to leads dropouts. 2 CPU's/Threads via masking is not enough for ASDM7EC. Ultimately,. something looks wrong with the scheduling of threads of HQP on AMD...

 

This is one of the Windows annoyances... HQPlayer sets affinity for each thread, but Windows doesn't always honor these but instead moves threads to other cores ruining performance especially when it decides to put multiple heavy load threads on the same core.

 

[Miska]

On 5/12/2021 at 7:38 PM, k6davis said:

Jussi, would you consider allowing DSD64 content to be converted with the 1x filter setting? I understand that DSD64 is a higher bit rate than 1x PCM, but it's a light load for the DSP - lighter than Redbook. Plus the vast majority of my (and maybe most people's ??) DSD content is SACD based DSD64. 

 

To ask it in another way, could the difference between 1x and Nx be more like "easy to process" (including DSD64) and "more difficult to process" ?

 

Neither 1x nor Nx filter applies to DSD sources...

 

[Yviena]

6 minutes ago, Miska said:

 

This is one of the Windows annoyances... HQPlayer sets affinity for each thread, but Windows doesn't always honor these but instead moves threads to other cores ruining performance especially when it decides to put multiple heavy load threads on the same core.

 

Hmm maybe you should open a bug issue, and forward the info to windows so they can get It fixed as it seems like a bug that windows is not honoring affinity.

[Miska]

On 5/12/2021 at 8:23 PM, JTS said:

Are there other dithers that would be appropriate to try when upsampling to 96K or 192K? I love all the improvements that upsampling are bringing to the table but want to chase a bit more of the "natural" quality the 2514 exhibits without upsampling. These comes from midrange tone and all the little round edges in micro details when listening without upsampling.

 

Especially with 96k and less I would stick with TPDF or Gauss1. At 192k you can try NS9 noise shaper too.

 

[Miska]

On 5/13/2021 at 3:07 AM, bobflood said:

2. Had to restart HQP to get play started (reported above)

 

Log file snippets are needed for these kind of cases, otherwise it will remain mystery what happens.

 

On 5/13/2021 at 3:07 AM, bobflood said:

3. Track faded before finish (reported above

 

Same thing for this...

 

Please use email for logs to avoid adding lot of noise to this thread.

 

[Miska]

On 5/13/2021 at 8:03 PM, micheloupatrick said:

As I understand it, the apodizing counter reflects the content of the track before it is upsampled : so it seems that at least for these tracks, you’re right about using an apodizing filter (although using an apodizing filter on a file that doesn’t need it is OK too).

 

Yes, this is the case. In screenshots, limited count is 0, but "need apodizing filter" counter is high which is analysis of the source content before any DSP is applied.

 

[Miska]

On 5/14/2021 at 1:31 AM, ted_b said:

My problem is I have NO idea what one is supposed to do with this data?  And I assume the apodiing filters are the ones with the "X" in the APO column, so if we use ext2 are we set, regardless?

 

If you get high figures for a track, you should be using an apodizing filter. When you do so, you are fine.

 

[Miska]

On 5/14/2021 at 1:46 AM, austinpop said:

So if output rate is set at 16FS, does that mean Sinc-Mx and Sinc-M will be equivalent, both at 1M taps? Irrespective of input sample rate, between 1FS and 8FS?

 

Yes, that's the case. And also including 16x, 32x, 64x or such source rate.

 

[Miska]

On 5/14/2021 at 2:11 AM, Quadman said:

According to Jussi's quote above in my post if a track has 100 indications then it has some clipping but probably does not need an apodizing filter.  Almost every track I looked at had into the 100's or even thousands and this was using apodizing filters.  Like Ted says, what are we to do with this data?  I briefly tried a non-apodizing filter and the counter moved at about the same rate as it did with an apodizing filter.  The music sounded so good, but I should have played Track A with apodizing filter then repeat Track A with non-apodizing and see how many indications each produced.  I take it that the apodizing filter should minimize these instances.  If not then what is the reason for this?

 

Analysis is done before any DSP processing, on the source content. So filter choice doesn't affect the result.

 

[Miska]

On 5/14/2021 at 5:51 AM, Quadman said:

That makes sense, and probably how it's is.  Should be written a bit more clearly in the manual so this confusion would be avoided.

 

I'll try to reword the manual description to make it more clear.