HQ Player

[Miska]

1 hour ago, asdf1000 said:

Is anyone out there successfully using macOS NAA 4.11  ?

 

Yes... On newer macOS versions you need to run it with root privileges, so through for example "sudo ./networkaudiod".

 

Reason is that it doesn't have GUI, so it cannot trigger the OS prompt for audio device and network access.

 

[Yviena]

13 minutes ago, Miska said:

 

It is technically difficult for various reasons.

Oh I see, is it windows thread scheduling, and some missing functionality  that is the issue? 

 

I wonder if WSL2, or future improvements  there would allow HQplayer to run via it.

[k6davis]

On 2/10/2021 at 6:14 PM, glitchesbrew said:

I've just tried with my Mini M1 16GB and with 4.10.0 for the first time (I believe) I was able to do Sinc-L, DSD256x48 but I had to use ASDM5EC. I tried the following file formats: DSD256, DSD128, DSD64, PCM 192/24, PCM 96/24, PCM 48/24 and PCM 44.1/16. I didn't have a single dropout.

To my ears it sounds great. For reference I've been using mostly DSD256 and ASDM7EC so far.

 

Wow!! I hadn't heard of anybody getting that to work without a CUDA card.

 

The M1 chip continues to impress me.

 

 

 

 

[Miska]

6 hours ago, Kirbyt said:

And there it is k6davis, thanks for that. 
still a bit foggy on this though. For the sake of clarity I guess my question is can hqplayer OS run on a Mac mini m1. I have desktop running right now and I send signal direct to dac via a icron USB ranger (fibre media converter) so no naa in my case. As well I am not using roon anymore just strait HQPlayer and the HQPlayer app for remote control. 
thanks 

 

No, the current HQPlayer OS can run on x64 and Pi4 hardware. There is no possibility of running it on M1. And Apple is not particularly open about other OS than macOS on their M1 hardware.

 

[Miska]

3 hours ago, k6davis said:

The M1 chip continues to impress me.

 

Somehow it seems to have very low penalty for cross-core work distribution / communication.

 

[Miska]

5 hours ago, Yviena said:

Oh I see, is it windows thread scheduling, and some missing functionality  that is the issue? 

 

Windows has multiple annoying aspects that makes it different from any other platform.

 

5 hours ago, Yviena said:

I wonder if WSL2, or future improvements  there would allow HQplayer to run via it.

 

If you use NAA and disable UPnP support, it likely works already. You can easily install Ubuntu 20.04 on WSL from Microsoft Store. And then HQPlayer Embedded AMD optimized package on it. But I know the UPnP support is broken, likely partially because Windows itself already binds to UPnP related things and WSL shares same IP as the main Windows. So you cannot have overlapping services.

 

[k6davis]

1 minute ago, Miska said:

Somehow it seems to have very low penalty for cross-core work distribution / communication.

 

I have high hopes, given that this is just the first gen and lowest end version. More powerful Apple Silicon system will certainly be forthcoming. 

 

But knowing you, you'll write something that taps them out too!! 😂

[Miska]

2 minutes ago, k6davis said:

But knowing you, you'll write something that taps them out too!! 😂

 

When HQPlayer 1.0 was released, there was no way to do something we can do today. And it is likely the case in future too.

 

If the new silicon would enable running ASDM7EC at DSD512, that would be already great! Also doing DSD1024 even with non-EC ASDM7 is challenging and needs fastest CPU models. Reminds me that I actually need to check how that would (if) work on M1.

 

[Geoffrey Armstrong]

19 minutes ago, Miska said:

 

Somehow it seems to have very low penalty for cross-core work distribution / communication.

 

I think I read somewhere recently that someone got linux working on a Mac M1

[Miska]

3 minutes ago, Geoffrey Armstrong said:

I think I read somewhere recently that someone got linux working on a Mac M1

 

Yes, that means "kernel boots up enough that you can run console shell".

 

For example graphics support is completely missing still.

 

It is looong way for such to become officially supported Linux platform. And I'm not going to touch it until it is officially supported platform.

 

[semente]

On 12/9/2020 at 10:03 AM, Miska said:

 

Yes, it arrived. At least for me, with native version, upsampling with EC modulators to DSD256 doesn't work (from RedBook, with poly-sinc-ext2). Output to a NAA. With regular modulators it does work.

 

Fan is running, but it is still fairly quiet, so could be used in a listening space.

 

So the performance is roughly what I expected, similar to equivalent U-series Intel CPUs. Over time Apple will likely come up with bigger and more powerful versions to match the bigger Intel CPUs in performance.

 

 

What would be a better Mac option regarding fan noise and upconverting Redbook to DSD256 (perhaps DSD512)?

 

a Mac Mini M1 2020

 

a Mac Mini Core i7 3.2GHz 2018

a Mac Pro Quad Core Xeon 3.7GHz 2013

[Confused]

I have noticed a few posts over the last couple of months asking about the relationship between HQPlayer filters and DAC hardware filters.  I was wondering if there is any guidance on this?  Such as any possible combinations of HQPlayer filters with DAC hardware filters that, from a technical perspective, should work well, and also cases where the filters might somehow be in conflict, and perform poorly.  OK - I realise that it is easy enough for anyone to try different filters and select based on subjective preference, I have done this myself, but I am interested in improving my own understanding of how this works.

 

Maybe this has been covered elsewhere?  If anyone knows of anything relevant and could provide a link, that would be great.

[Miska]

6 minutes ago, Confused said:

I have noticed a few posts over the last couple of months asking about the relationship between HQPlayer filters and DAC hardware filters.  I was wondering if there is any guidance on this?  Such as any possible combinations of HQPlayer filters with DAC hardware filters that, from a technical perspective, should work well, and also cases where the filters might somehow be in conflict, and perform poorly.  OK - I realise that it is easy enough for anyone to try different filters and select based on subjective preference, I have done this myself, but I am interested in improving my own understanding of how this works.

 

Maybe this has been covered elsewhere?  If anyone knows of anything relevant and could provide a link, that would be great.

 

Usually when used intended way, the DAC oversampling filters are bypassed. But when you cannot do that, you should select some sharp linear phase filter from the DAC.

 

[Confused]

14 minutes ago, Miska said:

 

Usually when used intended way, the DAC oversampling filters are bypassed. But when you cannot do that, you should select some sharp linear phase filter from the DAC.

 

In part my query was general, just to improve undersanding.  The above makes perfect sence, but in some cases users may not have the option to change DAC filters.  

 

To give some specific examples, I  have a desktop DAC that has a range of filters that can be selected.  In terms of Linerar Phase filters it offers "half-band" and "apodising", I am not sure if either of those are "sharp"?

 

Then I have my Devialet.  This is a very unusual case as the DAC is fully integrated with Devialet's ADH amplification, and no filter options are offered, you get what you are given.  I do know that Devialet use PCM1792 DAC chips, and will upsample any digital inputs below 24/192, but to be honest, I have no idea what filters they might be imposing.

[Miska]

28 minutes ago, Confused said:

To give some specific examples, I  have a desktop DAC that has a range of filters that can be selected.  In terms of Linerar Phase filters it offers "half-band" and "apodising", I am not sure if either of those are "sharp"?

 

If you use HQPlayer filters before the DAC, use half-band filter. But for many modern DACs you can input 352.8k or higher rate that bypasses the built-in filters.

 

29 minutes ago, Confused said:

Then I have my Devialet.  This is a very unusual case as the DAC is fully integrated with Devialet's ADH amplification, and no filter options are offered, you get what you are given.  I do know that Devialet use PCM1792 DAC chips, and will upsample any digital inputs below 24/192, but to be honest, I have no idea what filters they might be imposing.

 

There you can just always send 192k and there won't be problem with the combination of filters. Likely the sharp filter is used there. Most source material is 44.1/48k which leaves a lot of space between the source content band and the filters operating from 192k up. There is plenty enough space also for 88.2/96k inputs.

 

Practically only cases where you may have overlap/interaction between filters:

1) When doing 1:1 (or close to) rate conversion in HQPlayer, for example 192 -> 192 or 176.4 -> 192

2) When DAC uses minimum-phase filter

 

[Nikko1960]

I have placed an order for a Lampizator Baltic 3, and it will be one of the first models to be produced with a volume control in around a month from now.

 

The dealer has dropped off a loaner Baltic 3, which I have connected directly to my power amp.  I am enjoying the sound very much and I use the digital volume control in HQP, and its integration with Roon, to control it.

 

The Lampi I have on order costs a fair bit more than a standard unit, so I am asking myself whether having control of volume in the analog domain will offer superior sound compared to the current setup using HQP volume control in the digital domain.

 

Anyone have any experience please?

 

Thx

 

Nick

[perpetualapprentice]

5 hours ago, semente said:

 

What would be a better Mac option regarding fan noise and upconverting Redbook to DSD256 (perhaps DSD512)?

 

a Mac Mini M1 2020

 

a Mac Mini Core i7 3.2GHz 2018

a Mac Pro Quad Core Xeon 3.7GHz 2013

 

I don’t have personal experience with each model on the list but have used Minis exclusively since they were first introduced over a decade ago. Of the four models I have owned/continue to use, the M1 has, by far, been the quietest and up to this point, run the coolest. No spinning drive is probably a very big factor.

 

Upsampling Redbook to DSD1024 (via UPboard NAA>Holo Spring 2) on the M1 running a full array of applications and their background processes is consistently doable without dropouts (mostly AMSDM7+fs w/ wide variety of filters). Mine has 16GB of RAM but that does not seem to be a requirement - it was a ‘future proofing’ decision more than anything.

 

My experience is not as deep or varied as many here but can objectively say that at least with wintertime ambient temps the first gen M1 runs cooler and is quieter than any computer I’ve ever owned.

[Andrey Boldakov]

Hello, @Miska

 

Many greetings from Russia, St.Petersburg

I'm in love with HQ Player, and favorite to ASDM7EC & poly-sync-long-lp for DSD256 conversion.

But now it tooks too long to compute something before sound starts playing.

Even on my Ryzen 9 3900 with Nvidia 1650 (but GPU doesn't seems to be busy or involved somehow at all).

 

Am I the only one?

Does it possible to include some pre-calculated cash for the most common formats into the HQ Player distributive?

As commonly we have only 44.1/48/88.2/96/176.4/192 sample rates and 16/24 bits at the input

Those 12 variants covers 99.9 percent of Tidal music playing through the Roon to HQ Player worldwide..

And 99.9 percent of people needs DSD128/256/512 on the output for DSD and 24/32 & 192/384/768 for PCM.

Anything lower can be computed on the fly easily, so doesn't requires any cash.

 

Am I crazy suggesting that or you can think it over?

Maybe it would be good to calc all that data needed for specific user from the settings by clicking some buttons?

[Nikko1960]

On 2/10/2021 at 11:14 PM, glitchesbrew said:

 

I've just tried with my Mini M1 16GB and with 4.10.0 for the first time (I believe) I was able to do Sinc-L, DSD256x48 but I had to use ASDM5EC. I tried the following file formats: DSD256, DSD128, DSD64, PCM 192/24, PCM 96/24, PCM 48/24 and PCM 44.1/16. I didn't have a single dropout.

To my ears it sounds great. For reference I've been using mostly DSD256 and ASDM7EC so far.

 

 

Hey Glitches - how much memory in your M1 Mini?  I have a basic spec model (8gig/256gig) and it can't run Sinn-L at DSD256x44.1 without drop outs when upsampling redbook.

[Miska]

1 hour ago, Andrey Boldakov said:

As commonly we have only 44.1/48/88.2/96/176.4/192 sample rates and 16/24 bits at the input

Those 12 variants covers 99.9 percent of Tidal music playing through the Roon to HQ Player worldwide..

And 99.9 percent of people needs DSD128/256/512 on the output for DSD and 24/32 & 192/384/768 for PCM.

Anything lower can be computed on the fly easily, so doesn't requires any cash.

 

Am I crazy suggesting that or you can think it over?

Maybe it would be good to calc all that data needed for specific user from the settings by clicking some buttons?

 

You can already see how many rate combinations that makes with number of filters...

 

poly-sinc-long is among the slowest to initialize, but if you use -2s version and stick to the same rate family it becomes quicker.

 

[fmzip]

10 hours ago, semente said:

 

What would be a better Mac option regarding fan noise and upconverting Redbook to DSD256 (perhaps DSD512)?

 

a Mac Mini M1 2020

 

a Mac Mini Core i7 3.2GHz 2018

a Mac Pro Quad Core Xeon 3.7GHz 2013

My M1 256GB 8MB whispers upconverting to DSD512. That’s with HQPlayer and ROon both running on it. Processing speed isn’t an issue at all

 

 

[Andrey Boldakov]

4 hours ago, Miska said:

 

You can already see how many rate combinations that makes with number of filters...

 

poly-sinc-long is among the slowest to initialize, but if you use -2s version and stick to the same rate family it becomes quicker.

 

 

Yes, I forgot about filters count (((

It's really pity there is no way to pre-calculate some cache and store it for next usage

Playing Tidal playlist having tracks with different sample rate is painfully due to pre calculations already done for example two tracks earlier..

 

Another question is there: do you have plans or is it even possible to set up HQ Player as WASAPI audio device in Windows?

For lot's of settings (PCM 24/192 and Adaptive output rate) the latency is so small...

it makes possible to watch video but not YouTube, iTunes, Roon free Tidal and so on.

Maybe it's a stupid question. sorry..

Just wondering to have a great feature

[Miska]

33 minutes ago, Andrey Boldakov said:

Yes, I forgot about filters count (((

It's really pity there is no way to pre-calculate some cache and store it for next usage

Playing Tidal playlist having tracks with different sample rate is painfully due to pre calculations already done for example two tracks earlier..

 

Filters can take a lot of memory, like 600 MB for example. It could blow the memory consumption quite a bit.

 

34 minutes ago, Andrey Boldakov said:

Another question is there: do you have plans or is it even possible to set up HQ Player as WASAPI audio device in Windows?

 

Technically possible, but not planned. Because I would need to make three such drivers, for Windows, macOS and Linux. And that is a lot of maintenance burden. It is endless uphill battle with OS updates breaking things, especially on macOS.

 

But it already works through various other ways, especially if you don't need automatic input sample rate switching (for many things it is not necessary). Including some 3rd party loopback drivers.

 

For automatic input side sample rate switching, you need a Linux-based NAA and suitable hardware.

 

[Andrey Boldakov]

10 minutes ago, Miska said:

3rd party loopback drivers

Wow!!!! Thanks a lot for great idea!!!!

I'm running to try it immediately!)))

[Andrey Boldakov]

11 minutes ago, Miska said:

Filters can take a lot of memory, like 600 MB for example.

It's nothing. I have 64Gb of DDR4 and easily can manage hardware with tons of memory...

Here in St.Petersburg we have lots of oldschool servers like dual E5-2680v4 with 256Gb of memory from Finland

 

Listening Tidal with playlist tracks like: 16/44.1, MQA 24/88.2, 16/44.1, MQA 24/192, 16/44.1 is:

1. Absolutely normal and quite ordinary

2. Really pain with those pre calculations

 

Settings like filters, output mode and rate are typically constant to user preferences

Input formats in Tidal also very common and 99% of Tidal content is fit in 3-4 formats

like in example above

 

But I agree I can use fast filters...