HQ Player

[austinpop]

Folks,

 

Is there a current cheat sheet or guide to the favorite settings for HQPlayer? I've only seen guides that are 3+ years old.

 

The problem is I don't even understand what the naming convention is. I think I've worked out that mp means minimum phase. What are modulator settings, and what changes from DSD5 to DSD7? What is 256fs? What are all the dither settings? What changes from NSx to NSy settings? I would be happy to study, but I need pointers.

 

Here is what I've done so far. My DAC is the Ayre Codex. Its max rates are 384k for PCM, and 128 for DSD (via DoP). I've already determined that for DSD sources, I hear no benefit from any upsampling, so I've set the DirectSDM flag to just pass through.

 

For PCM, I do hear a benefit, by upsampling to either 352.8 or 384, based on the rate family of the source (44.1 or 48). So I set "Auto rate family."

 

Finally, for filters, I was suggested to use the poly-sinc-xtr filters. I do find I like the mp variant better. On the few occasions I listen to MQA content from Tidal, I switch to poly-sinc-mqa-mp.

 

My dither is set to NS5, as I have zero idea how to vary this.

 

That's the extent of my exploration. Is it the case that at any given point, there is a small subset of settings that are "the latest" state of the art, and most people are using them? If so, a bit of reorganization to make that clear would be nice.

 

Just to be clear, I'm a paid-up user, not just kicking the tires.

[austinpop]

4 hours ago, Miska said:

 

Just, checking... Have you checked the included manual, or is it not descriptive enough?

 

 

Hi Jussi,

 

I think it all centers around providing guidance around these 4 key settings:

 

Yes, the manual does describe each available choice for each of these 4 pull-downs. But that is 4-dimensional cross-product of hundreds of combinations. it doesn't tell me how I should navigate. I'm looking for guidance. Maybe a lot of these settings are older, and have been superseded. Maybe the best current options are a small subset. How about some guidance that says - this is the most preferred one, but try these other 2 as well. Or - this is recommended, but if it's too CPU-intensive, fall back to this.

 

How about this? Pick a set of source types. Say:

• DSD
• PCM
• MQA

For each of these, come up with the "Top 3" recommended combinations of filter/dither or oversampling/modulator. If 3 is too limiting, OK, make it top 5. But give me a set of manageable choices to try. This is no different than filter settings on DACs. There's usually a small handful.

 

Then the rest of the choices can be for people who really have specific reasons to go beyond.

 

I know several folks who liked the sound of HQPlayer, but could never enjoy it, because the sheer volume of the multi-dimensional parameter space caused stress, which is the complete opposite of why one listens to music!

 

Please know that this is sincere customer feedback, and not meant to criticize.

 

Thanks.

[austinpop]

I rest my case.  

[austinpop]

7 hours ago, Miska said:

 

I think the default settings are quite good. Then you can start playing with different filters which depend on type of music you listen, and your sonic preferences and sensitivities. You can also try between ASDM7 and ASDM5 modulators.

 

You can then expand it from there, or just stick to the default and only change the output rate limits as necessary.

 

 

 

Thanks. I've been reading the section on Filters and Dither very carefully.

 

Since I am only doing PCM > PCM upsampling, I'm not even getting into the SDM column.

 

For Dither, if I am reading correctly, since I am upsampling to 352/384, it looks like NS5 is recommended, especially for 44/48 recordings, since that is 8x upsampling. But what about 24/96 recordings (which is only 4x)? Should I be switching to NS9 - the manual says that is suitable for 4x. What about 24/192 recordings? That is only 2x.

 

Then for filters: there are currently 23 options:

 

Should I cycle through all 23 to see what my ears like best? Or at this point in time, are most people using a smaller subset? That's really all I'm asking - should I treat these as 23 equally valid options, or can they be partitioned into 2 sets: most frequently used currently, and rare/deprecated.

 

I hope that's not an unreasonable question?

[austinpop]

3 minutes ago, ted_b said:

Since you're staying in PCM, I would test both NS5 (since you are staying at 352/384) and regular TPDF.  In my PCM-to-PCM world I do TPDF (Holo Spring Level 3 via SU-1). 

 

The most popular filters are the Poly-Sinc and Poly-Sinc-XTR filters in both minimum phase (mp) and not. 

 

So that gives you 8 combinations to listen to seriously for awhile.

 

Thank you Ted.

 

Would you add Poly-sinc-mqa(-mp) if the stream happens to be an MQA track on Tidal? Or have you not gone there? Obviously, this is for a non-MQA DAC.

 

[austinpop]

On this volume min/max setting thing... Since I don't use SW volume control at all, I had been setting min = max = 0 to bypass.

 

Based on the discussion here, it sounds like I need to pay closer attention to this. If I am doing ONLY PCM->PCM upsampling, is there any reason to lower these settings?

 

Further, if the answer above is yes, should I do something like min = max = -3? Or is there a reason for min to be lower than max?

[austinpop]

9 minutes ago, ted_b said:

My min/max is -4 FYI.

 

2 minutes ago, Dr Tone said:

Set yours to -6db for min and max, it's the safest setting in my experience.

 

Thank you both for the advice. I'll do as you suggest.

[austinpop]

10 minutes ago, Dr Tone said:

@austinpop On a side note which filter/settings do you prefer for PCM->PCM upsampling over the internal Ayre filtering?

 

When it comes to settings, I'm a set it and forget it kind of guy. I hate playing with filters and dither every time I listen to a song. I am really struggling with the multitude of choices on HQPlayer. I don't have the patience to try the 207 different combinations (23 x 9) of filter and dither.

 

Based on recommendations, I've tried poly-sinc-xtr and poly-sinc-xtr-mp. I like the -mp. For dither, I tried NS5 and TPDF (per @ted_b) and can't really hear a difference.

 

So my default for now is poly-sinc-xtr-mp and NS5. And I upsample to the max of 352.8 or 384, based on rate family.

[austinpop]

22 minutes ago, d_elm said:

@austinpop when you upsample PCM is it worth the trouble over what the Codex will do ?

 

Just barely, and I think that's why it's all seeming like a chore to me.  

 

Don't get me wrong - there IS an improvement, but it is small. That's not a surprise - I know the improvements vary with DAC. In fact, the first time I evaluated HQPlayer, it didn't do anything for me, so I let the eval period expire. More recently, with the improved resolution in my chain with the whole trifecta, I tried again. This time I did hear a modest improvement - nice, but the earth didn't move.

 

I am sure if I move to one of the specific DACs that appear to benefit greatly - like the T+A DAC 8 DSD or the iFi MicroDSD Black - and upsample everything to DSD512, I might hear more of the magic.

 

With my DAC, if I'm being totally honest, I hear a similar improvement by just using Roon upsampling. Heresy on this thread, I know, but there it is.

 

I don't regret buying an HQPlayer license, as one day I might upgrade to a DAC that benefits more. Until then, I'm not sure yet whether it'll stay in my chain or not.

[austinpop]

9 minutes ago, Miska said:

 

Yes, because you need to leave headroom for material that contains digital clipping / inter-sample-overs (most modern RedBook material). HQPlayer will anyway force the signal to stay within the bounds through it's special limiter functionality and you may not notice anything, but it is better not to trigger the limiter. So keep eye on the "Limited" counter in main window and the volume knob. The limited counter should stay 0 at all times. For most (but not all) material -3 dB is enough.

 

I have my range set to min=-6 max=0, allowing me to fine tune if necessary with normal setting knob pointing up.

 

 

 

Thanks, that is really great to know. I'll do as you suggest, and keep an eye on the Limited count. Luckily, mine is showing 0 right now, but I'm guessing it's only since the last time I started the process.

[austinpop]

On 1/3/2018 at 12:44 PM, arglebargle said:

@Miska Do you have any sense yet of the performance hit to HQPlayer from upcoming patches to address these CPU-level security problems that are in the news? 5-30% is a big range and obviously this is a developing story but as you have some expertise! 

 

Link for anyone who hasn't heard the story: https://www.theregister.co.uk/2018/01/02/intel_cpu_design_flaw/

 

 

 

 

 

On 1/3/2018 at 1:24 PM, Miska said:

 

It should be low, but testing will show the specific impact. So probably closer to that 5%.

 

 

That makes sense, as the KPTI (Kernel Page Table Isolation) fix essentially runs the kernel in a separate virtual memory address space, which slows down system calls - i.e. calls an application makes to the OS kernel. The performance hit to an application depends on the total number and frequency of OS system calls it makes.

 

I'm guessing HQPlayer, by virtue of its computationally intensive nature, spends the bulk of its time in user mode (Ring 3 in Windows parlance), so the impact of slower system calls should be low on overall execution time.

[austinpop]

Could someone point me to a link/post/doc that explains what "convolution" means, and how it relates to SQ?

 

Much appreciated.

[austinpop]

5 minutes ago, lmitche said:

Put on your thinking cap!

 

https://www.google.com/url?sa=t&source=web&rct=j&url=http://www.analog.com/media/en/technical-documentation/dsp-book/dsp_book_Ch6.pdf&ved=0ahUKEwiP_8OL99fYAhXC24MKHf7JAIwQFgiGATAQ&usg=AOvVaw2J1gGdD-Xc9jNgZqueYirP

 

Thanks Larry! That's a great link.

 

I do understand the mathematical meaning of convolution, but need/want to understand it's use in HQPlayer. 

 

Clearly, these Matrix settings are not necessary for basic HQPlayer operation. I'm curious how people are using the convolution settings. As the link explains, in DSP, one uses an impulse response to convolve with the input signal, to get the output.

 

What is the use case in HQPlayer? Where does one get an IR file?

 

Stuff like that.

[austinpop]

Ah OK - room correction. Got it.

 

Thanks!

[austinpop]

On the premise that the magic of HQPlayer really starts to happen when upsampling to DSD512... is this achievable in an NAA architecture, or are people only doing this with direct attach only?

 

Just curious if DSD512 is achievable with NAA endpoints. If so, which ones?

[austinpop]

3 minutes ago, rickca said:

Sure you can do it with an ultraRendu, for example ... but your DAC has to have DSD512 support on Linux.  This is what has been driving T+A DAC8 DSD users crazy.  It's a whole lot easier with a Windows NAA endpoint. 

 

Yeah I meant Linux - *Rendu, SOtM, Zenith etc.

 

I was wondering about the imminent iFi Pro iDSD.

[austinpop]

4 hours ago, Miska said:

Hmmh, there's not a word about the product on iFi web pages. AFAIK, it has not been launched yet. So it is mostly vaporware/speculation to talk about the features...

 

"Described" starting with this post, in this thread:

https://www.computeraudiophile.com/forums/topic/29373-ifi-audio-pro-idsd-official-thread-its-happening-folks/?do=findComment&comment=757744

 

[austinpop]

Here's probably the most relevant post:

https://www.computeraudiophile.com/forums/topic/29373-ifi-audio-pro-idsd-official-thread-its-happening-folks/?do=findComment&comment=766113

 

[austinpop]

I guess the key question wrt the iFi Pro iDSD is this: like other iFi DACs, is there magic in using HQPlayer to upsample all content to DSD512?

 

OR - is it better to feed it native resolutions, and use the internal upsampling to DSD1024?

 

No way to know this until people try it!

[austinpop]

3 hours ago, Miska said:

 

I have to admit, with HQPlayer use I have not noticed any difference in practice... This is probably just because of the type of load HQPlayer is.

 

 

I assume HQPlayer spends the bulk of its time in compute-intensive functions or methods in user-mode.

 

The Meltdown/Spectre fix penalizes code that that has a high frequency of system calls. I doubt HQPlayer fits this profile.

[austinpop]

33 minutes ago, The Computer Audiophile said:

I don’t remember how the Chord handles incoming high rate PCM. 
 

Perhaps @austinpop can chime in, as he knows these DACs pretty well. 

 

@CheapSplurge

 

The TT2 can accept a max PCM rate of 768kHz or DSD512 native when using the Chord ASIO driver on Windows.

 

As for HQPlayer settings, I can't help you, as I use am not an HQPlayer user — I use a Chord M-Scaler — but I know people like @ray-dude are getting excellent results with HQPlayer. As I recall, they are using PCM-to-PCM upsampling to 705.6/768 using the 1M taps Sinc-M filter and 15th order noise shaper (LNS15). You can look at Ray's recent article on the SGM Extreme to get his HQPlayer settings. See https://audiophilestyle.com/ca/reviews/reality-quest-going-to-extremes-with-the-taiko-audio-sgm-extreme-part-3-of-5-r909/

 

[austinpop]

32 minutes ago, CheapSplurge said:

Thank you so much. 

I changed all I could as instructed but the other stuff seems complex.. Like dither? Change that to what? 

 

Cheap Splurge,

 

You're asking for help, but are you actually listening to the answers being given? I pointed you to an article a few posts back. Did you actually look? If you had, you would have seen that the author has posted his HQPlayer settings for DAVE in 2 screenshots, that I will replicate below to make it easier for you:

  

 

These same settings should be a great starting point for the TT2. Why not start here, and then once you've gotten more familiar, tweak further? 

 

Be aware — there is no single MAGIC setting for anything. Everyone has different ears, and they find their own favorites by tweaking.

[austinpop]

Hi folks,

 

I'm checking out HQPlayer to drive my DAVE directly. I am finding the DSD Settings and Settings menus very confusing, but think I'm beginning to understand. The key appears to be that not all filter/oversampler/dither/modulator options apply for any particular scenario. Could y'all experts and @Miska please advise if my understanding is correct? I want to look at 3 scenarios:

1. PCM to PCM
2. DSD to PCM
3. DSD to DSD.

Here are my DSD Settings and Settings:

  

 

Scenario 1: PCM to PCM Upsampling

 

Only the following settings apply:

 

Scenario 2: DSD to DSD upsampling

 

Only these settings apply:

  

 

Scenario 3: DSD to PCM upsampling

 

  

 

Obviously, other flags and options apply, but my focus is on the filtering, dithering, and upsampling options. 

 

Is my understanding correct? Thanks for your guidance.

[austinpop]

9 minutes ago, ericuco said:

For Scenario 3 (DSD to PCM), you need to change Default Output Mode from [source] to PCM. Otherwise PCM stays PCM and DSD stays DSD.

 

Thanks, excellent point. I was actually aware of that, but should have indicated correctly in the diagram!

[austinpop]

@Miska

 

Hi Jussi,

 

I am really enjoying playing with HQPlayer 4 Desktop. Could you please clarify a few more things for me?

1. I discovered albums from the label TRPTK in 32-bit floating point DXD format. Let's call it 32f/352.8. When I do PCM to PCM upsampling using sinc-M/LNS15/768 max, does HQPlayer output 32 bit floating as well? I see this:
   
   Since the output is not 32f but 32, does that mean the upsampled output is 32-bit integer? From an SQ perspective, I feel like bit-perfect 32f sounds better with my DAC (DAVE). DAVE can natively handle 32f, according to Chord.
2. Speaking of bit-perfect, is there an easy way in HQPlayer to toggle between bit-perfect (pass though source to output) and upsampled for PCM input? I guess something like a "PCM Direct" flag analogous to "SDM Direct?"
3. More generally, is it possible to change filter and shaper on the fly, without having to open "Settings?"