HQ Player

[Bob Stern]

On 4/15/2017 at 4:21 PM, zoltan said:

I totally agree with this observation. I've done tests now p/s/xtr vs. p/s/shrt vs. p/s and p/s shrt is the most detailed and realistic at high frequencies. At the same time also more tiring as some commented.

 

Those of you who enjoy extended treble but find your present system a bit relentless in the treble might want to try the exaSound e32 DAC (with 30-day return privilege).

 

I replaced my Metric Halo LIO-8 with the exaSound e32 a few months ago, and the difference is surprisingly dramatic.  The exaSound treble is both more extended and much smoother (i.e., absence of harshness) than the LIO-8.

 

(Apologies if I sound like a paid shill, but I'm just a happy customer.)

[Bob Stern]

6 hours ago, ted_b said:

Bob, being that the LIO-8 was PCM only, do you upsample only to exaSound's DSD, or both?

 

With the exaSound e32, I always upsample to DSD-256 with HQP.  (I do not like Audirvana with the exaSound.  I should edit my signature to show that I'm no longer using Audirvana.)

 

With the LIO-8, the opposite was true: Audirvana was the best sounding music player to me.

 

I did not find HQP beneficial with the LIO-8 and I did not use it except for occasional trials to see whether my impression had changed.  Perhaps this relates to my judgment that the LIO-8 sounds best at sample rates of 96 KHz or less, with or without HQP.

[Bob Stern]

6 hours ago, jtwrace said:

Are you using the PlayPoint as well?  The measurements of the e32 are stellar so I'm not surprised by your comments.  It's one of the dacs on my list to hit up at Axpona this week.

http://www.exasound.com/e32/e32DACMeasurements.aspx

 

No, I'm using the setup shown in my signature: Mac Mini connected directly to exaSound via USB, using exaSound's proprietary ASIO driver for Mac.

 

If you have questions regarding the e32/PlayPoint combination, you might ping "orgel", who replaced his Mytek 192 with that in December or January.

[Bob Stern]

16 hours ago, lucretius said:

 

In my expereience with amateur radio, a "cheap" clock (used for frequency determination) should stabilize (be relatively drift free) within an hour ...

 

Apparently you never owned a Heathkit "Mohawk" receiver (circa 1960).  That thing never stabilized!  In fact, it was better when I replaced the temperature-compensating capacitor in the local oscillator with a non-compensating capacitor.

 

(Apologies for the ludicrously OT post, but the amateur radio reference was nostalgic.)

[Bob Stern]

6 hours ago, auricgoldfinger said:

I am interested in parametric equalization of my headphones.  Is there a plug-in or some other way to do parametric EQ in HQPlayer?

 

See my post two days ago.

 

(Too bad the new forum platform omits post numbers to facilitate referring to a specific prior post.  Even date and time is dicey because most of us use local time instead of GMT.)

[Bob Stern]

On 4/19/2017 at 4:29 PM, pkane2001 said:

Sure! I use REW to create a set of filters, then export them as impulse response WAV files for both channels. These are then loaded directly into HQPlayer convolver. The filters are applied automatically when HQPlayer is playing, as long as you have the convolver enabled. Let me know if you need any specific step described in more detail.

 

I've never used REW before, but I figured out the following procedure.  I'd be grateful for your advice on simplifications or improvements: 

 

Prefs > View > clear “Enable mouse wheel zoom” if using a trackpad.

 

Open EQ window by clicking EQ button.

• Gear button:  Clear “Invert Filter Responses”.  Click Gear button again to close dialog.

 

Open "EQ Filters" window by clicking EQ Filters button (top center of EQ window).

• Clear “Always on top”.

• Control > Manual.

• Type > PK for parametric EQ.

• enter Q=1.4 for BW=1 octave.  (0.67 for BW=2.)  http://www.rane.com/note170.html

• close EQ Filters window.

 

Close EQ window, or else bring the main REW window to the front via cmd-tilde.

 

File menu > Export > "Export filters impulse response as WAV".

• Clear “Normalize”.  Mono.  32-bit.  Sample rate same as imported impulse response.

• No minus sign or other non-alphanumeric characters in filename.

 

Optional: Verify frequency response of exported IR WAV file:

• Open the WAV file:  File menu > Import impulse response.

• Click the “SPL & Phase” button above the graph.

• Hover mouse near lower right of graph to reveal 20–20,000 X-axis button and click it.

• Upper left/right zoom icons separately zoom Y-axis of amplitude & phase curves.

[Bob Stern]

@Miska:  Convolution question:

 

The user guide says: "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 ..."

 

It seems to me that HQP would resample the impulse response before it begins playing the audio, save it in memory or disk, and then begin playback by convolving the saved (resampled) impulse response with the music source.  This would cause a delay before playback begins, but it would not affect the CPU load during playback.  Am I missing something?

[Bob Stern]

49 minutes ago, Miska said:

 

It [the impulse response resampled to match the music source sample rate] is not saved anywhere as that operation is really fast...

 

So, the real-time convolution of the music during playback does not load the CPU differently depending on whether the impulse response file matches the sample rate of the music source?  In other words, there is no advantage for the user to create a separate impulse response file for each sample rate?  

 

(If so, the user guide is in error.)

[Bob Stern]

Thanks, Miska — that answers my questions.

 

I would suggest splitting the paragraph I quoted from the user guide into two paragraphs to clarify that the warning regarding CPU load is based only on the sample rate of music source, not the impulse response:

 

Original single paragraph:  

"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 ..."

 

Suggested two paragraphs:

 

"When source material sampling rate differs from the impulse response sampling rate, impulse responses will be scaled to the source material's sampling rate.  For maximum filter resolution, it is best to save an impulse response at the highest possible sample rate and let HQP downsample it as needed to match the source material sampling rate."

 

"Note! The CPU load greatly increases with the source material sampling rate and the length of the impulse response."

[Bob Stern]

@wwaldmanfan:  Thanks for digging up that list of shortcut keys.

 

For commands that don’t already have F-keys, you can use the OS X "Automator" program to create an OS X "service" that sends the command to the HQP command line utility.  You can assign a keyboard shortcut to the service in OS X Prefs > Keyboard > Shortcuts.

 

For example, for Stop, use Automator to create a service that runs the following shell script:

/Applications/HQPlayerDesktop.app/Contents/MacOS/hqp-control localhost --stop

 

To see the available commands, paste the following into Terminal:

/Applications/HQPlayerDesktop.app/Contents/MacOS/hqp-control help

 

Note that "help" is the only HQP command that is not preceded by localhost.

 

Tutorials on Automator:

http://www.macosxautomation.com/automator/

 

I do not believe there is an un-mute command — i.e., there is no built-in way to return the volume to its level before the mute command.

 

 

[Bob Stern]

6 hours ago, JohnDonaldson said:

Add me to the list of those having playback stop 1 sec. before the end of a file when using HQP/Roon for playback. Not all the time but often. For me the problem started with the release of HQP V16.

 

I have that problem without Roon — HQP Desktop playing a local file.  It sometimes stops just before the end of a track and fails to advance to the next track.  MacOS 10.11.6.  No messages in the OS X Console log.

[Bob Stern]

6 hours ago, jimdukey said:

Is the 3.16.3 Update mainly for Windows, or are there changes/improvements for Mac/Sierra?

I'm happy with 3.16.1!

 

I don’t know about other OS's, but on the Mac you can simply copy both versions of HQP somewhere on your hard drive and launch one or the other.  After comparing them, you can trash whichever one you don’t like.

 

(fwiw, 3.16.3 is fine in El Capitan.  I don’t have Sierra.)

[Bob Stern]

I have to admit I haven’t tried the XTR filter (I use Poly-Sinc Short MP), but I wonder whether the popularity of XTR is due to the mystique of anything "extreme".  I don’t recall Miska ever recommending it.

 

When audiophiles are given a choice between lots of ringing and minimum ringing, most instinctively choose minimum ringing.  In HQP, the XTR filters have the most ringing, and the Short filters have the least ringing.

[Bob Stern]

7 hours ago, Jud said:

I'm not almost all audiophiles then.   But as I've mentioned before, I think this may be due to the fact that my speakers rely on linear phase and time alignment for imaging and soundstage, and minimum phase filters may mess that up.

 

You got it backwards.  Your Vandersteen speakers (and my Thiel speakers) are designed to be minimum phase.  Linear phase requires symmetrical pre-ringing and post-ringing.  A system with pre-ringing is called "non-causal" because it produces a output prior to the impulse that caused it.  Pre-ringing (a non-causal system) it is impossible to produce using only capacitors, inductors and resistors, i.e., without DSP.

[Bob Stern]

8 hours ago, Jud said:

Minimum phase filters have less pre-ringing but more post-ringing unless they are apodizing.

 

Minimum phase filters have zero pre-ringing.

 

Apodizing has nothing to do with pre or post ringing.  It means the filter has a cutoff frequency "much" lower than 22 KHz for the purpose of attenuating the ringing caused by analog low-pass filters used by the recording engineers when recording at the 44.1 KHz redbook sample rate.  This is equivalent to using a cutoff parameter "much" less than 1.0 in Audirvana's iZotope upsampling.

 

"Much" is not clearly defined, but the design approach is that if the recording engineer used a 20 KHz low-pass filter which produced ringing at 20 KHz, then your apodizing filter should have a sufficiently low cutoff frequency and a sufficiently steep slope to attenuate 20 KHz sufficiently to be inaudible.  You can use a less steep filter slope (less ringing) if you use a lower cutoff frequency.

[Bob Stern]

1 hour ago, Jud said:

Apodizing minimizes both.

 

No, you're confusing two types of ringing:

(1) The ringing embedded in a Redbook recording caused by the analog low-pass filter used in the recording process.

(2) The ringing at half the sample rate of the D/A used for playback.

 

Meridian's apodizing was the first attempt to minimize (1).  Whether the chosen upsampling filter also minimizes (2) is orthogonal.

 

Also, keep in mind that Meridian coined the term "apodizing" for audio playback as a fanciful extrapolation of its original meaning in the field of optical image processing.  "Apodizing" has no well-defined meaning in professional or academic audio engineering, notwithstanding the uncredited Wikipedia definition.  So you can say that it means whatever Meridian wants it to mean for marketing purposes at various times.

[Bob Stern]

2 hours ago, Jud said:

A filter that maintains the phase of the original must be linear phase, I thought.  No?

 

Here's some hairy theory:

 

https://www.dsprelated.com/freebooks/filters/Zero_Phase_Filters_Even_Impulse.html

 

https://www.dsprelated.com/freebooks/filters/Filters_Preserving_Phase.html

 

https://www.dsprelated.com/freebooks/filters/Minimum_Phase_Filters.html

[Bob Stern]

52 minutes ago, Miska said:

Analog filters may of course also ring, but since all modern ADCs are oversampling, the analog filter is to large extent non-issue.

 

I think the problem intended to be solved by Meridian's apodizing filter is early digital recordings with non-oversampling A/D's.  A steep low-pass "anti-aliasing" analog filter was required to avoid aliasing distortion caused by frequencies above the Nyquist frequency being fed to the A/D.  Of course, any steep filter causes ringing at the corner frequency.  Because this analog filter preceded the A/D, its ringing became encoded in the digital recording.

 

I apologize if this discussion is OT.

[Bob Stern]

On 5/14/2017 at 5:05 PM, Jud said:

I've read the articles and note linear phase filters and the special case of them called a zero phase filter are classified as filters preserving phase (what I mentioned), while minimum phase filters are not so classified.

 

You're right.  I’m embarrassed that I’ve been perpetuating a faulty understanding of “minimum phase”. 

 

Proponents of minimum phase digital filters extol the absence of pre-ringing, but they either ignore or only obliquely mention waveform distortion, so I never figured that out.

[Bob Stern]

"Apodizing" — Since there was some disagreement above as to its meaning, here's what Charles Hansen (Ayre owner/designer) posted in 2009 in the MeridianUnplugged web forum:

 

Quote

 

In Peter Craven’s 2004 AES paper where he coined the term “apodising filter”, he defined it as a filter that had a stop-band frequency low enough to eliminate the ringing (pre- and post-) from *any* filters upstream in the entire record/playback chain.

 

Of course with the CD standard, a filter that eliminates ringing at 22.05 kHz and yet is flat to 20 kHz is going to be a steep son-of-a-gun, which will introduce a lot of ringing itself.

 

Craven’s solution to this was to use a *minimum phase* filter instead of the usual *linear phase* filer. An MP filter works just like an analog filter — there cannot be any output before the input. So all of the ringing happens *after* the transient event.

...

 

[Meridian’s] apodizing filter had a steep rolloff so that it was dead flat to over 19.7 kHz and yet had over -100 dB of stopband rejection at 22.05 kHz (where all of the ringing occurs in the “upstream” filters in the record/playback chain). The apodizing filter was *extremely* close to the filter in the Meridian 808.2. I know this because the review of the Meridian in Hi-Fi News last year showed the impulse response of the filter (which basically describes all of the attributes of the filter).

 

 

[Bob Stern]

18 minutes ago, firedog said:

It's just about what sounds better to you. Who cares about the waveform?

 

When I said minimum phase distorts the waveform, I was merely describing it mathematically, not expressing an opinion as to whether this is audibly good or bad.

 

I strongly believe that people hear things differently and vary in their sensitivity to different types of imperfections because the mind is an integral part of hearing.  Therefore, I would never argue that one audio design approach is universally superior.

[Bob Stern]

18 hours ago, Miska said:

[Minimum phase] doesn't [distort the waveform] any more than linear phase. There are many views and angles to linear vs minimum phase and there's no simple clear statement about the subject. If you look at step response of of the two, first in linear amplitude scale and then in logarithmic amplitude scale you understand what I'm talking about.

 

Scientists say the ear consists of thousands of very narrow band filters.  So perhaps dispersion (phase delay versus frequency) actually is a more helpful way to estimate the audible effect of a filter rather than the time-domain waveform.

 

I recall reading a long time ago that the brain's perception of the timbre of a note (e.g., whether the note sounds like a flute or a clarinet) depends entirely on the frequency spectrum during the initial attack of the note.  (I wish I remembered the supposed duration of that period.)  The spectrum of the sustained note after the attack contributes almost nothing to the brain's perception of timbre.

[Bob Stern]

One thing that mystifies me about the different PCM upsampling filters is why they should sound different when you are upsampling by a factor of 4 or more.  Shouldn’t such filters have essentially all their phase shift between 20 KHz and 80 KHz?  If the phase shift is essentially zero from 20 Hz to 10 KHz, I don’t understand why differences in phase shift above 20 KHz would be audible.

 

On the other hand, I can imagine that if you are converting from PCM to DSD (which I do), perhaps certain PCM upsampling filters interact more favorably with the DSD modulator.  I don’t understand the latter at all, so this is mere speculation.

 

Miska, can you shed light on this?

[Bob Stern]

11 hours ago, Miska said:

the linear- and minimum-phase variants of my filters are the same filter, but just phase adjusted. One can move the phase anywhere on the minimum - linear - maximum axis.

 

Since poly-sinc and poly-sinc-shrt are your most recommended filters, it would be nice if you offered versions of those filters intermediate between minimum and linear phase, similar to "asymFIR".

[Bob Stern]

On 7/18/2017 at 7:24 AM, AnotherSpin said:

Drag and drop into lower area [of HQ Player window] any folder which contains FLACs. It should work.

 

Note to HQP neophytes:  On the Mac you cannot drag music files or folders onto the HQP icon in the Dock or the Finder.  You can only drag them to the lower half (queue) of the HQP window.