HQ Player

[Bob Stern]

If anyone else is bothered by this bug, please notify Miska so he'll prioritize fixing it:

 

For comparison purposes I sometimes want to listen to a passage in the middle of a track. If I click the pause button, drag the playback position marker, then resume playback, the first few seconds of music is garbled. This is very annoying at full volume.

 

(It sounds like the shift register for the upsampling filter has a mix of samples from the old and new playback positions.)

 

This is an old bug, unrelated to the 3.16 update.

[Theobetley]

Tried to install 3.16 and got errors. I kept inputting ignore and now I have a 3.16 Manual and an About indicator of 3.5. I did uninstall 3.15 before trying to install 3.16.

[Samuel T Cogley]

A big thank you to Miska for the new max sample rate pulldown. I don't have to memorize 44100 or 48000 x (128, 256, 512, etc.) anymore!!!!

[Solstice380]

Tried to install 3.16 and got errors. I kept inputting ignore and now I have a 3.16 Manual and an About indicator of 3.5. I did uninstall 3.15 before trying to install 3.16.

 

What OS? I did a Win10 3.15.1 uninstall / 3.16 install this morning. No problem.

 

Do you reboot after uninstall? Usually a good idea.

[ted_b]

A somewhat theoretical question, and one that is likely quite dac-dependent, but will ask anyway:

* is it better to strive to get to the dac's max sample rate even if it involves a lighter weight filter (say, poly-sinc-mp-2s) than doing less-than-maximum sample rate with heavier filters (say poly-sinc-mp)? This question also has a slight variation: is sample rate family as important (do 48k-based stuff to 384k even if it requires lighter filters than instead doing 352k for everything)? I know the answer should include listening impressions, but asking anyway, given several factors that are not easily changed on the spot (the biggest one being compute horsepower).

[Jud]

is sample rate family as important (do 48k-based stuff to 384k even if it requires lighter filters than instead doing 352k for everything)? I know the answer should include listening impressions, but asking anyway, given several factors that are not easily changed on the spot (the biggest one being compute horsepower).

 

 

Miska has said there really is not more computer horsepower involved in conversion to non-integer multiples than to integer multiples.

[Samuel T Cogley]

Miska has said there really is not more computer horsepower involved in conversion to non-integer multiples than to integer multiples.

 

Empirically, I find this to be true. A 44100 rate -> 48000 rate doesn't put any more stress on the CPU than 44100 -> 44100 or 48000 -> 480000.

[ted_b]

Jud, I agree, but then why do families at all, unless maybe it's purely for those 24/96 dacs that skip certain sample rates? Moreover, after I wrote the question, I was unsure if the families issue was even an issue, so you got me. I'm still very interested in the bigger issue....push to the max rate and back off the filters or visa versa. In the worst case scenarios, like my Holo DAC ASIO driver, it shows capability to 768k but I have not spent much time trying to find a filter that will even play those rates, as it is somewhat futile, and I have an i7 6700k with 32GB RAM and pretty decent CUDA offloading.

[GUTB]

Can a HQPlayer license be transferred to a different system or a new OS install?

 

 

Sent from my iPad using Computer Audiophile

[craighartley]

Miska has said there really is not more computer horsepower involved in conversion to non-integer multiples than to integer multiples.

 

I'm pretty sure he has said the opposite.

 

"Upsampling 44.1k to 12.3M (or 48k to 11.3M) with poly-sinc is very heavy operation. You can ease up the load significantly for that case by using poly-sinc-2s instead."

 

Sent from my iPhone using Computer Audiophile

[lmitche]

Miska has said there really is not more computer horsepower involved in conversion to non-integer multiples than to integer multiples.

My expectation is that internally HQplayer holds a curve that is higher in resolution than the output sample rate and is therefore largely indifferent to the input sample rate. When rendering that curve it can be expressed at a multiple of either base frequency based on the output rate required.

[Jud]

I'm pretty sure he has said the opposite.

 

"Upsampling 44.1k to 12.3M (or 48k to 11.3M) with poly-sinc is very heavy operation. You can ease up the load significantly for that case by using poly-sinc-2s instead."

 

Sent from my iPhone using Computer Audiophile

 

See http://www.computeraudiophile.com/f6-dac-digital-analog-conversion/are-sampling-dacs-bad-thing-13909/#post186086

 

The point is that it is not more resource intensive to use 44.1k-based sample rates with 48k-based sources, or vice versa.

[bogi]

A somewhat theoretical question, and one that is likely quite dac-dependent, but will ask anyway:

* is it better to strive to get to the dac's max sample rate even if it involves a lighter weight filter (say, poly-sinc-mp-2s) than doing less-than-maximum sample rate with heavier filters (say poly-sinc-mp)? This question also has a slight variation: is sample rate family as important (do 48k-based stuff to 384k even if it requires lighter filters than instead doing 352k for everything)? I know the answer should include listening impressions, but asking anyway, given several factors that are not easily changed on the spot (the biggest one being compute horsepower).

 

I will answer non theoretically based on my listening impressions. With poly-sinc family I preferred 44.1k->5.6M with non 2s filters against 44.1->6.1 with 2s ones. The sound of non 2s filters was somehow fuller with a bit darker background. But with the new xtr filters I didn't found such a difference. Now I prefer going to 6.1 with 2s filters. Just my opinion...

[ted_b]

Since I can't edit my response anymore, I forget that auto-rate-family is also good for those filters that do not like to convert to/from 44 and 48k families, i.e only the ones that like 1:1 ratios.

 

Here is an excerpt from my Holo HQP log, NAA running Windows Server 2012 and Holo ASIO driver:

2017/03/16 13:36:40 Discovered 1 Network Audio Adapters

2017/03/16 13:36:40 Set channels: 2 (2)

2017/03/16 13:36:40 DAC bits: 32

+ 2017/03/16 13:36:40 Connect to 192.168.1.161:43210

2017/03/16 13:36:40 Network format: 44100/32/2 [pcm]

2017/03/16 13:36:40 Network format: 48000/32/2 [pcm]

2017/03/16 13:36:40 Network format: 88200/32/2 [pcm]

2017/03/16 13:36:40 Network format: 96000/32/2 [pcm]

2017/03/16 13:36:40 Network format: 176400/32/2 [pcm]

2017/03/16 13:36:40 Network format: 192000/32/2 [pcm]

2017/03/16 13:36:40 Network format: 352800/32/2 [pcm]

2017/03/16 13:36:40 Network format: 384000/32/2 [pcm]

2017/03/16 13:36:40 Network format: 705600/32/2 [pcm]

2017/03/16 13:36:40 Network format: 768000/32/2 [pcm]

2017/03/16 13:36:40 Network format: 2822400/1/2 [dsd]

2017/03/16 13:36:40 Network format: 5644800/1/2 [dsd]

2017/03/16 13:36:40 Network format: 11289600/1/2 [dsd]

2017/03/16 13:36:40 Network format: 22579200/1/2 [dsd]

2017/03/16 13:36:40 Volume limit: -3

2017/03/16 13:36:40 Automatic rate family disabled

2017/03/16 13:36:40 DirectSDM disabled

 

Does this really mean my Holo does not accept 48k DSD rates? And where is it getting the fact that the Holo will do 768k in PCM? Is it simply a text in an ASIO table somewhere?

[ted_b]

I will answer non theoretically based on my listening impressions. With poly-sinc family I preferred 44.1k->5.6M with non 2s filters against 44.1->6.1 with 2s ones. The sound of non 2s filters was somehow fuller with a bit darker background. But with the new xtr filters I didn't found such a difference. Now I prefer going to 6.1 with 2s filters. Just my opinion...

 

Thanks Peter. Your feedback is always valuable to me.

[Lionelh2]

Miska,

 

I found these remarks you made last year and I was wondering what you meant by "disabling" the Mojo volume control.

 

Do you mean setting it to full volume or invoking the "line out" level preset?

 

 

 

 

"I only have Mojo, have not tried Hugo. But Mojo clearly performs best when it's volume control is disabled and it is fed with 705.6/768k PCM data. Volume can then be controlled for example form HQPlayer as necessary.

 

Of course if you prefer the sound/performance of the Chord's WTA filter, then you prefer it and there's nothing to argue. The most similar sonic signature you get from HQPlayer with poly-sinc-hb or closed-form filters.

 

For me, the second biggest improvement with Mojo was to disable it's volume control and use external headphone amp (Schiit)"

 

 

Sent from my iPhone using Computer Audiophile

[arglebargle]

Empirically, I find this to be true. A 44100 rate -> 48000 rate doesn't put any more stress on the CPU than 44100 -> 44100 or 48000 -> 480000.

 

With my modest cpu I do find there is a very slight difference in initialization time going from 44.1 to DSD256x48 vs DSD256x44. It can lead to the occasional half-second hitch on track changes. During actual playback the difference in load is negligible.

 

Ted's question touches on something I've been wondering about. The materials for HQPlayer make it sound like there is very little difference in the 2s filter, like it's all upside. My system can handle the full filters with some material, but not well enough to form real world impressions. Comparisons would be welcome.

 

I have noticed that when playing back x44.1 material, if you watch the real time information of the stream in alsa, momentary frequencies jump back and forth between several numbers to maintain this rate. With powers of two 48/96 etc, momentary frequency is rock solid. Now maybe knowing this enforces my suspicion that 44.1->48 sounds better? I do recall Miska commented on a similar question in a thread on specific settings for the iDSD Micro:

 

This may be related to how sampling rate fits the USB packet rate. At 44.1k the packet size varies because the packet interval is not in sync with the sampling rate' date=' while 48k is in sync with the packet interval and thus the sizes vary less. This causes different patterns on the [/color']USB interface interference pattern.

 

On the other hand, someone over on the Roon forums said integer upsampling 44.1 is better because 22 and 45 MHz clocks have less jitter than 49mhz clocks, but that it's largely dac dependent. There was no further discussion and I don't know enough to gauge the truth of that.

 

Any information even tangentially related to these issues would be welcome! As a novice, this stuff is scattered in many places and difficult to piece together.

[Jud]

Can a HQPlayer license be transferred to a different system or a new OS install?

 

 

Sent from my iPad using Computer Audiophile

 

 

 

OSes: Mac and Linux are one license. Windows is a different one. However, if you contact Miska, I believe he may provide some consideration for the purchase you've already made if you want to switch from Mac or Linux to Windows, or vice versa.

 

Different system, I don't recall the answer. I have run HQP on both Mac and Linux computers at home, though not simultaneously.

[craighartley]

See http://www.computeraudiophile.com/f6-dac-digital-analog-conversion/are-sampling-dacs-bad-thing-13909/#post186086

 

The point is that it is not more resource intensive to use 44.1k-based sample rates with 48k-based sources, or vice versa.

 

Sorry Jud, I don't get your point. No doubt my fault. But I thought Ted's question was essentially about whether it's better to strive to upsample everything to 48k based rates, and I thought you were contradicting the idea that it is more resource intensive to upsample 44.1k files to 48k based rates, or 48k files to 44.1 based rates. So I provided a quote from Miska confirming that that is indeed the case - and of course it is the reason for Miska providing the autorate family in HQPlayer. So how does that relate to the post that you link? What am I failing to see?

 

 

Sent from my iPhone using Computer Audiophile

[Jud]

Sorry Jud, I don't get your point. No doubt my fault. But I thought Ted's question was essentially about whether it's better to strive to upsample everything to 48k based rates, and I thought you were contradicting the idea that it is more resource intensive to upsample 44.1k files to 48k based rates, or 48k files to 44.1 based rates. So I provided a quote from Miska confirming that that is indeed the case - and of course it is the reason for Miska providing the autorate family in HQPlayer. So how does that relate to the post that you link? What am I failing to see?

 

 

Sent from my iPhone using Computer Audiophile

 

I could certainly be wrong. It just seemed to me Miska was commenting "purely" about converting between 44.1k and 48k-based sample rates in the post I linked, and mixing in -2s and non-2s filters in the post you quoted. But it could also be that Miska was commenting at my link on an operation that wasn't very resource intensive in the first place, and in your quote about one that was, where converting between 44.1k and 48k-based rates might possibly matter. So I don't know the answer.

 

I can say that for me it doesn't make a practical difference, since my 2010-vintage computer will do either 44.1k-based or 48k-based rates at DSD256 with the -2s filters, but won't do either with the non-2s filters, and won't get to DSD512 even using -2s.

[bogi]

I'm happy that 3.16.0 combined the sq of 3.15.0 with faster filter initialization of 3.15.1.

I can now do 44.1k to 6.1M with poly-sinc-xtr-2s on my i5-4300M. I couldn't do it with 3.15.0 because of endless initialization.

 

Miska, I observed that [x] PipelineSDM changed to MultiCore DSP, which can be now set in 3 levels. Could you please explain the middle level meaning? I have enabled hyperthreading on my i5-4300M, which provides 4 threads on dual core CPU. Is this related?

 

Previously I observed clear sonic difference between Pipeline processing enabled and disabled. I liked to have it enabled on my system because it brought me better low level detail resolution. When it was disabled, sound was brighter on my system and some detail was less pronounced. I know it's "only" about difference in computer generated noise...

 

Now with MultiCore DSP the difference disappeared on my system. I still have feeling that ticked or filled (not the cleared one) setting could bring a very very small addition of low level detail, but ... I'm not sure I would like to do blind test for that.

[louisxiawei]

I compared 3.15 and 3.16 back and forth several times. (Lots of uninstallation and installation)

 

Still find the harshness existed in the 3.16 version. Much better than 3.15.1 but still worse than 3.15.

 

I'm using Roon with HQplayer, upsampling Tidal 44.1 and local DSD files to DSD512. Modulator ASDM 512+FS, filter: poly-sinc-xrt-2s. DAC: T+A DAC8 DSD.

 

Anyone feels the same with me?

[MarcinGD]

MultiCore DSP

 

In manual you can find that middle is an auto.

[Marco]

Kudos for this new 3.16 version. I`m not yet really sure because I didn`t check the exact processes running in the background, but Task Manager with the same settings just went down in CPU utilization from around 38% to a really steady 30%. That`s awesome!

 

44k1 -> Closed form, ASDM7, 11.2896 on I7 [email protected]

 

Would very much like to hear confirmation of lower cpu utilization with the new 3.16 version!

[Theobetley]

I compared 3.15 and 3.16 back and forth several times. (Lots of uninstallation and installation)

 

Still find the harshness existed in the 3.16 version. Much better than 3.15.1 but still worse than 3.15.

 

I'm using Roon with HQplayer, upsampling Tidal 44.1 and local DSD files to DSD512. Modulator ASDM 512+FS, filter: poly-sinc-xrt-2s. DAC: T+A DAC8 DSD.

 

Anyone feels the same with me?

 

I do. I'm letting it go for several hours before I go back to 3.15.