HQplayer 4 and Mac-mini M1

[Miska]

8 minutes ago, Schafheide said:

I have no idea - I just do it. Maybe someone else can chime in?

 

Depends on what your rate limit is set to. Unchecked adaptive rate means that the output rate is fixed. With most DACs this is necessary to DSD output, since they only support DSD at multiples of 44.1k. So the rate limit would be also set to some multiple of 44.1k and 48k DSD unchecked.

 

So typical case would be to have 48/96/192k content converted to 44.1k x256 or 44.1k x512 fixed rate. There is very slight advantage of going from 44.1/88.2/176.4 to 48k x256 or 48k x512, but it comes with increased CPU / GPU load. In best case the CPU load difference is negligible.

 

[Miska]

1 hour ago, LondonDan said:

why is greyed out multicore more efficient than ticked multicore if anyone knows?

 

Greyed means "auto" where HQPlayer is trying to figure out most efficient combination of settings based on the detected hardware.

 

Checked means "all on" which just blindly enables all possible parallel paths.

 

[Miska]

3 minutes ago, dericchan1 said:

So should multi-core and cuda offload (if you have a gpu) checked or greyed? What is better?

 

Multicore greyed is the default. If you encounter problems, try with checked. On dual-core machines, it is best leave it cleared.

 

CUDA offload checked means all possible algorithms are offloaded. Greyed means only convolution algorithm in offloaded.

 

If you hover mouse over these settings, small help text is displayed.

 

[Miska]

1 hour ago, copy_of_a said:

may I ask why?

 

Because the output rate is slightly higher.

 

1 hour ago, copy_of_a said:

My understanding is that integer upsampling should be advantageous ??

 

Why would it be?

 

[Miska]

23 minutes ago, copy_of_a said:

I thought integer has less rounding errors.

Is this a wrong assumption?

 

Yes, it is wrong assumption. Both are equally accurate. Only difference is that simpler ratios are less demanding for CPU/GPU and RAM.

 

[Miska]

2 hours ago, paulmckwan said:

I also "feel" that integrate upsampling sound signature is closer to original music, and non integrate upsample ratio result in bigger difference from original.

 

There's just no technical reason for such...

 

[Miska]

I can hear fan of my Mac Mini from 1m - 50 cm away when under full load, in otherwise quiet environment. But it is really quiet and doesn't have any annoying distinctive tone.

 

It is as quiet as my quietest fan cooled big PC's. (which is pretty good given the size/performance ratio)

 

[Miska]

1 hour ago, Kalpesh said:

And I have been missing that for years !! Why is it ? 8x6 ? octets packages in USB ?

 

8x6?

 

Because the output rate is almost 10% higher.

 

[Miska]

19 hours ago, sworksone said:

I just installed 4.22 and have been outputting DSD512 from 44.1 up to 192 pcm using poly-sinc-gauss-hires-lp / AMSDM7EC 512+fs on an 8GB M1 MacBook for 30min with zero stutters (through Roon running on the same laptop).  I’m so confused.  Does this seem possible?

 

Yes, AMSDM7EC is the only EC modulator where that works. ECv2 doesn't, unfortunately.

 

[Miska]

2 minutes ago, Kalpesh said:

You mean AMSDM7EC 512+fs ? Does it suit Holo May well ? (technically, subjectively I dare say yes)

 

Yes, it should be fine as long as you run it at DSD512 or DSD1024.

 

[Miska]

19 hours ago, Tihon said:

Which is better from an objective point of view? ASDM7EC 512+fs at 512 or ASDM7EC v2 at 256? 

 

Note that it is AMSDM7EC vs ASDM7ECv2. These are very different and produce very different kind of data.

 

Now there is no straightforward answer, because it depends on your DAC.

 

AMSDM improves more as function of increased output rate. So it is better at DSD1024 than at DSD512 - if the DAC can keep up.

 

[Miska]

On 4/21/2023 at 4:32 AM, Kalpesh said:

From samples I had tested I found 512 (played Direct SDM) to be the sweet spot on my system. Going up to 512, it was getting better and better. 1024 was not a step up but down. Because of :

 

the samples ? 

the Holo May ?

My limited RAM ? (could see paging at 1024) 

No electronics (clocks what have you) can keep up ?

My bad taste ?

?

 

Simplifying the answer quite a bit, I would say combination of Holo May, your electronics after the May, your speakers and your personal hearing, combined with the source material used.

 

Going beyond DSD256 there is no universal answer. With certain few DACs it could be even about going beyond DSD128.

 

In digital domain, it is more straightforward. But when you combine the produced data with real world hardware it is not so straightforward anymore.

 

I run my Holo DACs at DSD256 or DSD512. There are technical aspects that get better at DSD1024, but for me personally, and my systems, it has not been beneficial to go there. It doesn't mean it wouldn't be for some other system / person.

 

[Miska]

On 4/22/2023 at 5:39 AM, Tihon said:

Oh, thank you for noticing. I've just re-read the manual. That "M" means "pseudo-multi bit"? What is this?

 

Yes, that is a very special arrangement to produce the data. It utilises the highest possible rates in a special way.

 

On 4/22/2023 at 5:39 AM, Tihon said:

I used to think that ASDM7ECv2 is the best possible modulator (for pure DSD DAC) and all others are worse and should be used only when you don't have enough power for the best one. Is it so?

 

At the moment that is the case, but the AMSDM is a special case that is sort of completely separate parallel track. Some people asked me to create EC version of these, so I did.

 

On 4/22/2023 at 5:39 AM, Tihon said:

At highest rates like DSD1024 which modulator should be better ASDM7ECv2 (wish my laptop had such power 🤣) or AMSDM7EC?

 

That largely depends on the D/A hardware and the hardware following it. There is no very straightforward answer to this.

 

On 4/22/2023 at 5:39 AM, Tihon said:

I've got Holo Red for experiments with higher rates. So the chain is Red (NAA OS image) -> Holo May. My MBP (M1 Max) can do either ASDM7ECv2 at 256 or AMSDM7EC at 512.

 

I don't think there is huge difference between these two choices. So you could try and compare which one you think works better for you.

 

On 4/22/2023 at 5:39 AM, Tihon said:

Tried to compare 512x AMSDM7EC with 256x ASDM7ECv2 and realised that I don't know which one is better. To me AMSDM7EC sounds more dynamic but less natural tonally, maybe even slightly "shouty". ASDM7ECv2 sounds more real in terms of tonal balance, but feels slightly "slower".

 

This is exactly what I mean, it is not very straightforward one is better than the other. It is more like what is more optimal for your particular case. May not be even the same for all source material. It is more about putting more weight on some aspect.

 

[Miska]

11 hours ago, Schafheide said:

I have just added a second Mini to my system. Both Mini's are running the latest macOS and HQP desktop.  I have two problems.

 

#1 With Mini #1, when I open Client, I can see Mini#1 and "Localhost".  But with Mini #2, I can only see "Localhost".

 

#2 HQPDControl can see Mini #1 (after selecting it's IP address) but not Mini #2 (after selecting it's IP address).

 

Do you have allow network control -button pressed on HQPlayer's toolbar on #2? Another possible reason is firewall blocking access to the application. macOS asks initially whether you want to allow HQPlayer to accept incoming connections.

 

[Miska]

2 hours ago, Kalpesh said:

@MiskaDoes the mapping you provide to listen to Mch sources in Stereo work as well with Mch SACD sources ?

 

Yes it does. You may need to adjust LFE level in case seems to be wrong, but everything else should fit.

 

2 hours ago, Kalpesh said:

Is it the same channels mapping for PCM & SACD ?

 

Yes, it is ITU standard.

 

2 hours ago, Kalpesh said:

Is it hit & miss with PCM ? proportion ?

 

Mostly LFE level varies especially for PCM, since there is no specific rule for PCM files.

 

2 hours ago, Kalpesh said:

Is it hit & miss with SACD ? proportion ?

 

It should be generally fine.

 

2 hours ago, Kalpesh said:

Is there a Mac able to convert SACD Mch sources into Stereo ? which settings ?

 

Which stereo output rate?

 

[Miska]

On 4/25/2023 at 5:24 PM, Kalpesh said:

Since I don't have any output at all, any would be a huge change. I guess 128 would be the minimum target with convolution applied.

 

Well, DSD -> DSD convolution makes it a lot heavier than just producing downmix or similar.

 

On 4/25/2023 at 5:24 PM, Kalpesh said:

I'm under the impression that Quad does not follow the ITU std and I'd be mainly interested in hearing the Columbia Quad mixes of Miles Davis, Santana...

 

For source formats with proper channel mapping data, HQPlayer remaps such to ITU standard. So for example DSF or DSDIFF sources with the necessary data provided (marked as quad arrangement) are mapped to ITU 5.1 layout.

 

On 4/25/2023 at 5:24 PM, Kalpesh said:

I found the attached for Quad ; the matrix after ITU does not seem to be a match

 

 

You can do that easily in matrix and save it as a profile. But remember that the source channel arrangement is ITU standard. So channels 3 and 4 are empty (silence) for quad sources.

 

That 0.707 factor is pretty close to -3 dB. But remember that you need to compensate mix gains more, otherwise you get overs. (unity gain mix of two channels requires -6 dB mix gain for both)

[Miska]

1 hour ago, Kalpesh said:

I'm much afraid you overestimate my poor ol' brain but thank you !

does the attached looks correct ? should I dub 5 & 6 as 3 & 4 ? (would make sense since as much as 3&4 don't exist as center and LFE per ITU, 5 &6 don't exist if the encoding is SQ ???)

 

I wouldn't do it that way. The diagram you posted earlier is encoding quad to stereo recording device and then decoding it again from stereo back to quad. But that is not applicable to digital quad recordings which stay quad throughout.

 

[Miska]

2 hours ago, Kalpesh said:

I feel like a loosy student ! Would it be better, assuming that either, but not both, of the (3,4) or (5,6) couple codes for Lb, Rb ??

 

You can do for example that yes.

 

You can then compare which downmix sounds better for you, that or the one I've provided in the manual. Easy to switch the profiles on the fly.

 

[Miska]

On 4/28/2023 at 7:47 PM, Kalpesh said:

Is it possible to make up a kind of benchmark of HQP performance for the recent Mac ?

 

There is such already. You can select Null output backend in HQPlayer settings and play something through it. At the end of the playback, HQPlayer shows how much time was spent on processing. You can compare this to the playback time of the content in question for relative speed.

 

[Miska]

12 hours ago, James Stephens said:

I hearing a couple of small popping sounds at the start of each track when doing SDM conversion.

 

I am also hearing occasional pops during playback .. maybe once every other track or so and always from the left speaker.

 

Have you tried with the NAA OS on Holo Red? Does it change anything?

 

What kind of network infrastructure do you have?

 

[Miska]

2 hours ago, James Stephens said:

Network-wise I have some DSD256 downloads from NativeDSD and they have always played to my previous streamer (an Aries G2.1) without any issues so I don't suspect my network is the problem (at the moment).

 

Just in case, please make sure 802.3x flow control is actually active on your network...

 

[Miska]

2 hours ago, James Stephens said:

(*) There is a strange behavior when playing back DSD64 ripped from SACD's (but not purchased and downloaded DSD content from other sources such as NativeDSD) I have tried a few albums and they all produce a pop soon (a second or two) after playback commences when playing native DSD but this does not occur when I set SDM Pack to DoP. This was true also with Red OS running as an NAA. Not sure what to make of that ......

 

At least at one point some versions of the extraction tools had some bugs, so it needs a careful extraction process to avoid such.

 

[Miska]

19 minutes ago, MarcMarc said:

And I have been very interested in the DSD upsampling Sony does with devices like the TA-ZH1ES DAC.

 

Is there a similar Modulator in HQPlayer?

 

Yes, there are multiple modulators to choose from, M1 can upsample to DSD256 or DSD512 depending on the modulator and filter selection.

 

23 minutes ago, MarcMarc said:

I guess since I used the core audio of the Mac I couldn't get DSD to work.

 

Through CoreAudio you can get DSD working by setting "SDM Pack" to "DoP" in HQPlayer output device Preferences.

 

[Miska]

6 minutes ago, MarcMarc said:

Is there a particular filter that is close to Sony's upsampling? Another review mentioned Sony's upsampling Introduces some noise that measures similar to a 300B tube amp.

 

I don't have much more information about their functionality than what their web page says.

 

6 minutes ago, MarcMarc said:

So MacBook and Mac mini = HQPlayer desktop? 

 

Yes, HQPlayer Desktop is for Macs.

 

HQPlayer Embedded needs a device running Linux, typically some PC. Or one can use the self contained HQPlayer OS on a suitable PC.

 

6 minutes ago, MarcMarc said:

Did you already look more into the Gustard A26 ? I'm only following the posts in the roon forum where the wide and narrow filter was mentioned as not working.

 

I'm hoping Gustard will eventually release firmware update that would fix it. But since they don't respond to my emails, I don't know if that is ever going to happen.

 

Now it is sitting in my rack waiting for such to happen. I occasionally listen through it, it sounds pretty nice as it is already, running at DSD512. But I would really prefer that it's features would actually work.

 

[Miska]

Now DSD512 with EC modulators is also properly possible on M1/M2 Macs, this running right now on my MacBook Pro with M1Max:

 

No DSD256 limitations anymore.

 

(that higher efficiency core load earlier is Opera browser for reading/writing on this forum)