HQ Player

[Confused]

2 hours ago, firedog said:

Okay,so have you tried one of the external GUI’s written for HQP?

The best aspect of using Roon with HQPlayer is that it is more or less plug and play, hassle free, and 100% robust in use.

 

Are there any other options that can offer similar?

[DancingSea]

40 minutes ago, Miska said:

 

Low enough that "Limited" counter stays zero. -3 dBFS is a good starting point, but in the end it depends on your filter selection and source material. I have content where -3 dBFS is not enough, but that is quite rare in the end.

 

 

Themes/skins/styles provide look-and-feel, but they cannot change the functionality, only how things are presented / look like.

 

 

Thus is -3 minimum and 0 max the recommended setting, with the understanding there may be exceptions?

[scintilla]

I've been a bit confused for some time over using the gain comp in the convolution setup.  Since my filters really cut the volume a lot, I used 20db in my last set up and I have tried to use a number related to what acourate is telling me the gain loss is when I run the filter macro. I built a filter that included 6db of gain rather than leaving it at 0.0db and it clipped hard on some Bill Frisell that I played. So does HQplayer always soft-clip? Maybe the gain comp explanations in the manual could use a bit more explanations?

[Miska]

44 minutes ago, scintilla said:

I've been a bit confused for some time over using the gain comp in the convolution setup.  Since my filters really cut the volume a lot, I used 20db in my last set up and I have tried to use a number related to what acourate is telling me the gain loss is when I run the filter macro. I built a filter that included 6db of gain rather than leaving it at 0.0db and it clipped hard on some Bill Frisell that I played. So does HQplayer always soft-clip? Maybe the gain comp explanations in the manual could use a bit more explanations?

 

"Gain loss" is probably drop in average level due to peak boosts. It sounds more quiet, but still the boosted peak frequencies are at higher level and you need to account for these.

 

If you use negative gain compensation, for example -10 dB, then the convolution is run with 10 dB lower volume than when convolution is disabled. This is what you should do if you have any boosts and the convolution filter has levels of over 0 dB at any frequency.

 

If you use positive gain compensation, then the amount is applied as negative gain when convolution is disabled. This can be used when the convolution filter already takes the peaks into account and doesn't have higher than 0 dB level at any frequency. Then it is easier to compare the effect when switching the filter in/out on the fly because you get same perceived/average volume in both cases.

 

So the gain compensation can be used in two ways depending on your filters.

 

And yes, HQPlayer always soft limits and small overs of one or two dB may even go unnoticed by listening. But if the peak is high enough, particularly at low frequencies, then the limiting will still sound nasty, because it is relatively quick compared to the waveform cycle length. The limiting algorithm is fancier when you have volume control enabled in HQPlayer and in that case adjusting volume will also reset the limiter.

 

[Miska]

3 hours ago, DancingSea said:

Thus is -3 minimum and 0 max the recommended setting, with the understanding there may be exceptions?

 

I use -6 to 0 myself, in which case the volume knob is usually at 12 o'clock position (-3 dB) and can be easily adjusted couple of steps down or one step or so up if needed.

 

This range and setting is also reflected on Roon.

 

[copy_of_a]

12 hours ago, Miska said:

And yes, HQPlayer always soft limits and small overs of one or two dB may even go unnoticed by listening. But if the peak is high enough, particularly at low frequencies, then the limiting will still sound nasty, because it is relatively quick compared to the waveform cycle length. The limiting algorithm is fancier when you have volume control enabled in HQPlayer and in that case adjusting volume will also reset the limiter.

Miska, just to double-check ...: so if I set the max gain to -5db the the limiter will effectively be "bypassed" (assumed -5db is enough even with sharp mixes to avoid any intersample peaks due to upsampling)?

Any chance to implement an on/off setting for the limiter (I for one would prefer to set it off)?

 

As far as your library goes: apart from a more userfriendly GUI (that you may or may not implement) you should at least improve metadata handling. HQPlayer favours "artist" over "album artist"... which is a real mess. A compilation of "Various Artists" contains, say, 10 different tracks of 10 different Artists. HQPlayer shows the "track"-Artist as the "Album Artist". It should show the "Album Artist" (and only in case this tag is missing should refer to "Artist").

I've just batch-retagged a copy of my entire library and copied the "Album Artist" filed into the "Artist" field. This way the library of HQPlayer actually works for me. Problem is the metadata of this retagged library now is incorrect and therefore not really usable in softwares that respect common tagging rules.

[Miska]

3 hours ago, copy_of_a said:

Miska, just to double-check ...: so if I set the max gain to -5db the the limiter will effectively be "bypassed" (assumed -5db is enough even with sharp mixes to avoid any intersample peaks due to upsampling)?

Any chance to implement an on/off setting for the limiter (I for one would prefer to set it off)?

 

If there's nothing to do for it, it does nothing. There's no point in "turning it off", because if you run out of sample value range something needs to be done. Clipping is just another limiter, just a stupid one. The internal pipeline has practically no limit on sample value range, but regardless if you output PCM or SDM both have limited output value range.

 

3 hours ago, copy_of_a said:

As far as your library goes: apart from a more userfriendly GUI (that you may or may not implement) you should at least improve metadata handling. HQPlayer favours "artist" over "album artist"... which is a real mess. A compilation of "Various Artists" contains, say, 10 different tracks of 10 different Artists. HQPlayer shows the "track"-Artist as the "Album Artist". It should show the "Album Artist" (and only in case this tag is missing should refer to "Artist").

I've just batch-retagged a copy of my entire library and copied the "Album Artist" filed into the "Artist" field. This way the library of HQPlayer actually works for me. Problem is the metadata of this retagged library now is incorrect and therefore not really usable in softwares that respect common tagging rules.

 

There are now three different views of the library already in HQPlayer Desktop... ?

 

This has been discussed several times before.

 

The metadata specs  that FLAC uses (Vorbis comments) don't define concept of "album artist". HQPlayer's library is built around what is specified for those. Other content types that use ID3v2 (WAV/AIFF/DSF) are mapped to those using a scheme described earlier. ID3v2 doesn't define a tag called "album artist" either, but I'd like to hear which one of those is what you call "album artist", see section 4.2.2 in the specification for the tags defined for ID3v2:

http://id3.org/id3v2.4.0-frames

 

Note that the tags defined by the two standards don't match accurately. Then there's a third standard - APEv2 for tagging that is used by WavPack which is something I need to deal with if I add support for WavPack... List of APE tags can be found here:

http://wiki.hydrogenaud.io/index.php?title=APE_key

These are very easy to map to Vorbis comments used by HQPlayer. Composer -> Artist, (Performing) Artist -> Performer

 

For the compilation albums I have, I just have artist tagged as "Various Artists" and that's it. Luckily such albums are rare, I have maybe one or two. DJ sets I have are tagged under the DJ name, rest is in the song name.

 

[copy_of_a]

9 minutes ago, Miska said:

If there's nothing to do for it, it does nothing. There's no point in "turning it off"

Thanks Miska!

So the limiter only affects the signal when it exceeds 0db ... at -0.1db it isn't active?

 

9 minutes ago, Miska said:

This has been discussed several times before

I wasn't aware of that ... sorry (didn't follow the whole 462 pages ?)

 

As to the Artist field I'd say it's pretty self-explanatory.

"Album Artist" is the Artist the whole Album is labled under.

For instance "Dire Straits".

Album: Brothers in Arms.

The Aritst of the track "Money for Nothing" is Mark Knopfler & Sting (or maybe "Dire Straits" & Sting ... don't know, since I don't have that album).

Finally we can also have a composer of a certain track who is not the Album Artist (say Trent Reznor is the composer of "Hurt" performed by Johnny Cash - who is the Album Artist - on the Album "American IV: The Man Comes Around")

 

Audirvana, JRiver and Roon display the Album Artists / Artists / Composers according to these tags... all within the same Album

 

 

 

 

[Miska]

7 minutes ago, copy_of_a said:

So the limiter only affects the signal when it exceeds 0db ... at -0.1db it isn't active?

 

Exactly, it is always "active" meaning that it is always watching the data flowing through. But it doesn't touch it unless there's a need.

 

8 minutes ago, copy_of_a said:

As to the Artist field I'd say it's pretty self-explanatory.

"Album Artist" is the Artist the whole Album is labled under.

For instance "Dire Straits".

Album: Brothers in Arms.

The Aritst of the track "Money for Nothing" is Mark Knopfler & Sting (or maybe "Dire Straits" & Sting ... don't know, since I don't have that album).

Finally we can also have a composer of a certain track who is not the Album Artist (say Trent Reznor is the composer of "Hurt" performed by Johnny Cash - who is the Album Artist - on the Album "American IV: The Man Comes Around")

 

Yes, I don't have a problem understanding that. But if you look at the specifications I was referring to you see what I'm talking about.

 

ID3v2 for example doesn't have a thing called "album artist". It only has things called TPE1, TPE2, TPE3, TPE4, TOPE, TEXT, TOLY, TCOM, TMCL, TIPL and TENC.

 

FLAC has three things specified, called ARTIST, PERFORMER and ORGANIZATION.

 

[Yviena]

Hmm I see that most DAC measurement's show  less distortion when ran at  -8 or -10dbfs signal level would setting hqplayer volume to -8 or -10db do the same thing?

 

 

[copy_of_a]

2 hours ago, Miska said:

 

Exactly, it is always "active" meaning that it is always watching the data flowing through. But it doesn't touch it unless there's a need.

 

 

Yes, I don't have a problem understanding that. But if you look at the specifications I was referring to you see what I'm talking about.

 

ID3v2 for example doesn't have a thing called "album artist". It only has things called TPE1, TPE2, TPE3, TPE4, TOPE, TEXT, TOLY, TCOM, TMCL, TIPL and TENC.

 

FLAC has three things specified, called ARTIST, PERFORMER and ORGANIZATION.

 

Thanks Miska!!

 

I have no idea how Audirvana, JRiver and Roon handle ID3v2 tags and read the (optional) „Album Artist“ tag... but they do.

 

Maybe your library has to „root“/„map“ and/or prioritize the tag (something like „prefer ‚albumartist‘ as main artist if ‚albumartist’ is present“ ...).

 

Maybe this is of help?

 

https://help.mp3tag.de/main_tags.html

 

[Miska]

2 hours ago, copy_of_a said:

I have no idea how Audirvana, JRiver and Roon handle ID3v2 tags and read the (optional) „Album Artist“ tag... but they do.

 

Maybe your library has to „root“/„map“ and/or prioritize the tag (something like „prefer ‚albumartist‘ as main artist if ‚albumartist’ is present“ ...).

 

Maybe this is of help?

 

https://help.mp3tag.de/main_tags.html

 

That is already being done.

 

So in MP3Tag it is used as name for the TPE2 field which is described in the ID3v2 specification as:

 

Quote

  TPE2
   The 'Band/Orchestra/Accompaniment' frame is used for additional
   information about the performers in the recording.

 

HQPlayer uses this TPE2 field to populate "Performer", in case TCOM wasn't defined together with TPE1 in which case TPE1 would be "Performer". If neither of these is the case, then TPE3 is used as "Performer" if defined. I think this is closer to what the ID3v2 specification says.

 

Of course this doesn't apply to FLACs, because they use different tagging system (Vorbis comments).

 

For FLACs and APEv2 the MP3Tag document just says:

Quote

Please note that fields of VorbisComments (used with FLAC, OGG, SPX) and of APEv2 tags are not mapped internally by Mp3tag but are displayed with their actual name unless they have a user-defned mapping.

 

 

P.S. At this point you may have realized what kind of mess and disaster the entire tagging thing is, in addition to all the incorrect information in the internet databases. So the best is just to try to deal with the most simple things "artist", "album" and "song". Nothing else. Even artist seems to be hard to get right. I've got files where artist says "Mark Knopfler" (most) and then one album has "Knopfler, Mark" as artist...

 

P.P.S. Luckily, in HQPlayer, you can non-destructively edit the library metadata. So no need to modify files if you want to change artist/performer/album name.

 

[copy_of_a]

2 hours ago, Miska said:

P.S. At this point you may have realized what kind of mess and disaster the entire tagging thing is

Sure... agreed.

However ... I really don't want to bother you - not at all!! - but said other softwares do handle/display (and sort by) these sensitive tag fields correctly.

They also work as supposed to after conversion to AIFF, ALAC or AAC through XLD ... so it's not FLAC-specific.

I've tagged all my files by hand (with Yate), so there is no mess at all in my library due to incorrect data from the internet or so.

I only have trouble with the library of HQPlayer.

For the time being I can live with the retagged copy of my library for HQPlayer. But it would be nice if you could look into this at some point...

 

Another little "feature request" ?

Library -> full screen mode.

Switching views by flick gestures on an i-device display works fine. However, on a regular display with mouse or trackpad it's really somewhat cumbersome and it would be extremely handy if you could ad either keyboard-shortcuts to swipe the 3 views left/right or - alternatively or even in addition - add buttons on the 3 screen to swipe left/right.

 

BTW...: im on Mac OS X ...

 

Thanks ?

[Miska]

33 minutes ago, copy_of_a said:

Sure... agreed.

However ... I really don't want to bother you - not at all!! - but said other softwares do handle/display (and sort by) these sensitive tag fields correctly.

They also work as supposed to after conversion to AIFF, ALAC or AAC through XLD ... so it's not FLAC-specific.

I've tagged all my files by hand (with Yate), so there is no mess at all in my library due to incorrect data from the internet or so.

I only have trouble with the library of HQPlayer.

For the time being I can live with the retagged copy of my library for HQPlayer. But it would be nice if you could look into this at some point...

 

Well, different file formats use completely different tagging systems so they are file-format specific as you can see also from the MP3Tag link you provided. When it comes to that "album artist" and using ID3v2's TPE2 field specified for the band/orchestra name for the "album artist" purpose is misuse/misrepresentation of the information and outright wrong - goes against the standard.

 

OK, so here's the deal:

- I can add custom "ALBUMARTIST" tag as a priority tag for artist name for Vorbis comments (this covers FLAC files)

- Misusing TPE2 of ID3v2 is plain outright wrong and breaks the spec, so it is a definite no-go (so won't happen for WAV/AIFF/DSF/DSDIFF files)

 

Reason is that former, although not being formally specified in any standard, shouldn't either break anything. While latter is clearly breaking the standard specification.

 

33 minutes ago, copy_of_a said:

Another little "feature request" ?

Library -> full screen mode.

Switching views by flick gestures on an i-device display works fine. However, on a regular display with mouse or trackpad it's really somewhat cumbersome and it would be extremely handy if you could ad either keyboard-shortcuts to swipe the 3 views left/right or - alternatively or even in addition - add buttons on the 3 screen to swipe left/right.

 

You may then like the 3.22 release I just made that has precisely all this! It has clickable "flick-zones" on screen edges you can use to quickly switch between the views. I also gave a bit of facelift for the overall look-and-feel.

 

Keyboard shortcuts for switching between the views should be also easily doable, I'll look into that for next release.

 

I know macOS doesn't even support touch screens. On Windows and Linux it works quite nice with touch screens, but there are now less monitors on the market with touch support. I'm personally quite used to flick gestures with mouse - a reason why I originally started using Opera browser and still stick to it because I cannot live without it's mouse gestures (right button + flick left/righ is back/forward action), makes navigating web pages a lot easier! Opera has a pile of other nice cool features, like unit conversion by painting text on a page, etc.

[DancingSea]

1 hour ago, Miska said:

 

You may then like the 3.22 release I just made that has precisely all this! It has clickable "flick-zones" on screen edges you can use to quickly switch between the views. I also gave a bit of facelift for the overall look-and-feel.

 

 

 

Moi,

 

How are we notified of HQP updates?  I've owned since January and have not been prompted to update....

[shadowlight]

1 hour ago, DancingSea said:

 

Moi,

 

How are we notified of HQP updates?  I've owned since January and have not been prompted to update....

 

I normally just do a quick glance at Signalyst website to see a new version is available but I believe the official announcements are made on HQPlayer Facebook Page.

[lucretius]

6 hours ago, Miska said:

 

Well, different file formats use completely different tagging systems so they are file-format specific as you can see also from the MP3Tag link you provided. When it comes to that "album artist" and using ID3v2's TPE2 field specified for the band/orchestra name for the "album artist" purpose is misuse/misrepresentation of the information and outright wrong - goes against the standard. 

 

OK, so here's the deal:

- I can add custom "ALBUMARTIST" tag as a priority tag for artist name for Vorbis comments (this covers FLAC files)

- Misusing TPE2 of ID3v2 is plain outright wrong and breaks the spec, so it is a definite no-go (so won't happen for WAV/AIFF/DSF/DSDIFF files)

 

Reason is that former, although not being formally specified in any standard, shouldn't either break anything. While latter is clearly breaking the standard specification.

 

 

The thing is, most softwares have hijacked the TPE2 tag for Album Artist. Why fight it?

[copy_of_a]

7 hours ago, Miska said:

OK, so here's the deal:

- I can add custom "ALBUMARTIST" tag as a priority tag for artist name for Vorbis comments (this covers FLAC files)

- Misusing TPE2 of ID3v2 is plain outright wrong and breaks the spec, so it is a definite no-go (so won't happen for WAV/AIFF/DSF/DSDIFF files)

For me this would be a really helpfull change - many thanks for considering this!

 

7 hours ago, Miska said:

You may then like the 3.22 release I just made

 

I love it! These little helpers improve the user experience appreciably. Great - many, many thanks!!!

 

[DancingSea]

I'm having some odd behavior.  On the PCM side, if the filter is set to FIR, or Minring FIR (and mp), HQP is not upsampling to 352K, its just passing through the original sample rate.  On poly sinc it is upsampling like normal...  its weird because the FIR filters have always behaved normally to this point.  Ideas?

[Miska]

6 hours ago, shadowlight said:

 

I normally just do a quick glance at Signalyst website to see a new version is available but I believe the official announcements are made on HQPlayer Facebook Page.

 

There's also Twitter account, but I occasionally retweet and like other kind of things - audio and computer security/privacy related. 

[Miska]

3 minutes ago, DancingSea said:

I'm having some odd behavior.  On the PCM side, if the filter is set to FIR, or Minring FIR (and mp), HQP is not upsampling to 352K, its just passing through the original sample rate.  On poly sinc it is upsampling like normal...  its weird because the FIR filters have always behaved normally to this point.  Ideas?

 

Strange indeed, I just re-tested on all three platforms, on Linux with Mytek Brooklyn DAC+, on Windows with T+A DAC8 DSD and on macOS with iFi iDAC2 and for me it is working correctly.

 

Can you email me a HQPlayer log of this happening?

 

[copy_of_a]

@Miska

 

while we're at it...

There's one particular thing in Roon's convolution engine I'd consider really well implemented with regard to "audiophile" (whatever) sound quality:

When you have a correction curve applied to IR files at different sample rates (all common rates from 44.1kHz to 384kHz) you can store them in wav format and pack them all into a zip file. In Roon's convolution engine you simply select said zip file containing your correction at different sample rates and Roon will pick the correction file that corresponds to the source file sample rate you are playing back. So there's 1.) no need to internally adapt the IR files. 2.) better audio quality since convolution works best when the IR correction files match the sample rate of the source file the convolution is applied to.

Maybe something to consider for a future update of HQPlayer.

 

I'm sure you know this, but just in case here's the description re Roon:

https://kb.roonlabs.com/DSP_Engine:_Convolution

Quote

Filter resampling

If no filter is available that exactly matches the playback sample rate, Roon will resample the provided impulse response file to match.

You can avoid filter resampling by providing a separate filter for each sample rate. This can improve performance--both CPU performance and sonic performance--so you should try to provide a filter for each rate when possible.

______________

 

If you have a set of impulse response files, one per sample rate or channel layout:

1. Place all of the impulse response files in the same directory

2. Create a .zip file from that directory

3. Browse to the .zip file and select it in Roon

4. Enjoy the music.

 

[Miska]

48 minutes ago, copy_of_a said:

When you have a correction curve applied to IR files at different sample rates (all common rates from 44.1kHz to 384kHz) you can store them in wav format and pack them all into a zip file. In Roon's convolution engine you simply select said zip file containing your correction at different sample rates and Roon will pick the correction file that corresponds to the source file sample rate you are playing back. So there's 1.) no need to internally adapt the IR files. 2.) better audio quality since convolution works best when the IR correction files match the sample rate of the source file the convolution is applied to.

Maybe something to consider for a future update of HQPlayer.

 

That is completely unnecessary exercise to provide multiple IR files. Just provide one at highest sampling rate (from Acourate at 384k) and you are fine. Convolution in HQPlayer is processed at native content source rate, except for the case when you are doing DSD-to-PCM conversion in which case it is processed at the 1/16th of DSD rate in PCM (for much lower CPU load). This way you get best sound quality.

 

None of the tools to produce IR files are capable of producing those for DSD rates anyway... So when you play let's say content from DSD64 to DSD512 output, you'd need to provide a special variant of the IR at 2.8 MHz sampling rate.

 

Acourate has the nice Brickwall extension feature, so you don't need to use the HQPlayer's HF Expansion feature:

 

[copy_of_a]

22 minutes ago, Miska said:

That is completely unnecessary exercise to provide multiple IR files.

I'm a bit confused now since this is from page 7 of your current manual ...?!

Quote

Note! When source material sampling rate differs from the impulse sampling rate, impulse responses will be scaled to the source material's sampling rate. This can have a huge impact on CPU load, and with large impulse responses will require significant amount of CPU processing power.

 

--

 

22 minutes ago, Miska said:

Just provide one at highest sampling rate (from Acourate at 384k) and you are fine. Convolution in HQPlayer is processed at native content source rate...

Ok, this is new to me. Thanks! Does this also apply when I upsample PCM in "auto rate" family (so by power of 2)?

So when upsampling 44.1kHz to 176.4kHz I simply use a 192kHz IR correction and HQPlayer does not have to resample the IR file?

 

[Miska]

10 minutes ago, copy_of_a said:

I'm a bit confused now since this is from page 7 of your current manual ...?!

 

Yes, externally providing different IRs for different sampling rates don't change anything. If you look at IR for 176.4 it is likely about 4x bigger than IR for 44.1k. Rate is higher and the IR is bigger -> higher CPU load.

 

10 minutes ago, copy_of_a said:

Ok, this is new to me. Thanks! Does this also apply when I upsample PCM in "auto rate" family (so by power of 2)?

So when upsampling 44.1kHz to 176.4kHz I simply use a 192kHz IR correction and HQPlayer does not have to resample the IR file?

 

It is scaled ("resampled") to the needed rate, that is not a problem. If you make a room measurement at for example 48k, then same thing happens for all the other IR output rates you generate. It is just math, nothing special about that.

 

HQPlayer anyway does bunch of operations on the IR to prepare it for use, scaling is part of that.