Overthinking DSP & how to build an audiophile PC

[Phobic]

For my next cripplying round of overthinking I will mostly be focusing on DSP.

I've pretty much made my mind up that I'm going to build a few subs as my next upgrade step and couple them with DSP,  but I'm not sure which DSP route to go down, what boxes to shortlist, what are important features, pros and cons to look out for, trade offs to make, impact on signal chain to to consider.

Here's where my shortlist is so far, am I missing any key options? any other pros/cons to consider?





DSP


URL


Pro


Con




PC running something like Acourate and HQPlayer


https://www.audiovero.de/en/acourate.php


Stupid number of taps "ultimate solution?". Can be combined with a streamer and server


It's a PC! Need to figure out how to build a high quality silent, low EMI/RFI solution




Roon or Jriver with convolution filters made with REW or Audiolense 


https://juicehifi.com/


cheap with REW


limited functionality




miniDSP - SHD


https://www.minidsp.com/products/streaming-hd-series/shd


Dirac "auto" configure with a mic


Less functionality




Roomperfect


https://lyngdorf.steinwaylyngdorf.com/roomperfect/


Epic reviews, auto mic config


Less functionality




Linea research - ASC48


https://linea-research.co.uk/asc48/


Studio grade (?)  lots of features & power


Cost, complexity, not as many taps as a PC





I've read a few articles and posts about PC based options being "the best solution" due to the software, options and quality that can be created from effectively unconstrained taps.

As a computer nerd I'm quite drawn to a PC option, however I'm a bit paranoid about being able to build a PC that I'm happy to connect directly to my hifi - most likely this would be via a USB connection to an external DAC. however it will connect into the same power bar as the hifi so need to make sure the PSU is clean, and the USB connection is perfect, clean & noise free.

I'm mindful of all the conversations about clocking (and reclocking) and the differences between USB data transmission clock versus DAC sample clock, even so I know that generally PCs have bloody awful clocks. I don't know if I will ever try reclocking with something like a mutec but having the option to do so wouldn't be a bad thing.

Building a crazy "audiophile" grade pc like this one seems a step too far https://audiopc.shop/en/winkel/audiophile-music-servers/reference-audiophile-music-server-coffee-lake-fc5-ppa-studio-paul-pang/

but what high quality options & considerations should you give to a PC in the audio chain if you are striving for high end sound reproduction? making it silent running is a given. Audiopc sell "audio grade" USB cards, is that key? What PSU options are out there to make sure it's not dumping loads of crap onto the hifi power bar?

 

[jamster]

For my next cripplying round of overthinking I will mostly be focusing on DSP.

Haha that sounds familiar!

I've pretty much made my mind up that I'm going to build a few subs as my next upgrade step and couple them with DSP,  but I'm not sure which DSP route to go down, what boxes to shortlist, what are important features, pros and cons to look out for, trade offs to make, impact on signal chain to to consider.

Here's where my shortlist is so far, am I missing any key options? any other pros/cons to consider?

Forgive the self-interest, but what about:



This gives some idea of what it can do,  and it is much more energy efficient to run a silicon based DSP than a PC:

https://www.hifiberry.com/blog/announcing-the-dac-dsp/

I've read a few articles and posts about PC based options being "the best solution" due to the software, options and quality that can be created from effectively unconstrained taps.

The challenge is that DSP gets insanely complicated, very quickly.   A PC might be overkill.  What is it that you need specifically that a hardware based DSP won't give you?

As a computer nerd I'm quite drawn to a PC option, however I'm a bit paranoid about being able to build a PC that I'm happy to connect directly to my hifi - most likely this would be via a USB connection to an external DAC. however it will connect into the same power bar as the hifi so need to make sure the PSU is clean, and the USB connection is perfect, clean & noise free.

Yeah exactly - much easier to clean up power to a low power DC device using an aftermarket PSU if that's important to you.  Or just use an SPDIF into the dac for isolation.

I'm mindful of all the conversations about clocking (and reclocking) and the differences between USB data transmission clock versus DAC sample clock, even so I know that generally PCs have bloody awful clocks.

I don't know if I will ever try reclocking with something like a mutec but having the option to do so wouldn't be a bad thing.

Seems to be some mixed reports on this, when running without a DSP. 

My own, possibly simplistic, view is that use of a DSP mitigates this; hardware DSPs run at a fixed sample rate so effectively re-clock the source as part of the process.  Putting a hardware based DSP in between source and DAC may actually help not hinder. 

Whether you hear a difference in the re-clocked sound will be down to your ears, and your susceptibility to confirmation bias (let's settle the debate once and for all and agree that "hearing" includes interpreting sound as transmitted AND as perceived ). 

It certainly cleaned up my SONOS when I tried it.

The maker of the DSP noted above, Hifiberry, is adament that the performance characteristics of the DSP will better the analogue stage of any DAC, but does concede that some DACs perform better at certain input frequencies.   In my highly unscientific test, my ESS Sabre dacs appear to prefer multiples of 44khz rather than 48khz but my AKM-based DAC doesn't seem to care.  

Building a crazy "audiophile" grade pc like this one seems a step too far https://audiopc.shop/en/winkel/audiophile-music-servers/reference-audiophile-music-server-coffee-lake-fc5-ppa-studio-paul-pang/

Haha yes! I assume that no music in the world was mastered on a PC like that.  

but what high quality options & considerations should you give to a PC in the audio chain if you are striving for high end sound reproduction? making it silent running is a given. Audiopc sell "audio grade" USB cards, is that key?

I have never seen any proof that add-on USB cards offer benefit over onboard ones. In fact, on some motherboards where some USB ports are connected directly to the chipset, and others are via an additional controller over the PCI(e) bus, general advice is to use the directly connected ones for critical tasks. Shorter signal path, lower latency and less likely to suffer driver bugs. Frankly, I suspect that the differences are not measurable, and money is better deployed elsewhere.

What PSU options are out there to make sure it's not dumping loads of crap onto the hifi power bar?

If you do really want to go down the PC route, then perhaps use a laptop?  It'll be silent or at least much quieter than a desktop.  Unplug it during critical listening sessions, so that it runs on battery = no mains interference.  Nice screen that can live atop the hifi kit, for cover art.  Add a USB IR receiver and a remote (per my listing above  - would split) and you have an excellent streamer.

 

[tuga]

I have opted for your option A but the light version. By that I mean that HQPlayer is using very simple EQ filters generated on REW.

When you are upconverting to DSD256 or above then you need massive processing power, probably combining CPU and GPU. For my old 2012 i7 Macbook Pro, jumping from DSD128 to DSD 256 has meant that the computer now needs the noisy fan on because passive cooling is no longer suficient. Which means that I will have to think about replacing it...

So first concern is processing power. A laptop is not the best option...

Best CPU for hqplayer – https://audiophilestyle.com/forums/topic/37775-best-cpu-for-hqplayer

HQPlayer version 4 requirements – https://audiophilestyle.com/forums/topic/57259-hqplayer-version-4-requirements/

Second concern is computer-generated "noise" travelling along with the signal. One of the advantages of using HQPlayer is that you can use a low processing power "quiet" NUC as a buffer at the end of the network feeding into your DAC. I use a CuBox-i (same as Sonore Sonicorbiter SE) with a linear PSU running the minimalist dedicated HQP OS and software. But Jussi Laako has tested quite a few DACs which he claims are able to deal with the "noise" coming down the USB cable includig the RME.

HQPlayer's Network Audio Adapter – https://audiophilestyle.com/forums/topic/13649-hqplayers-network-audio-adapter

Other features offered by HQPlayer are the possibility of using multiple endpoints into different systems along the house or running the processing Desktop computer headless, and it can drive multi-channel (f.e. 2 main + 2 subs) or active multi-way systems, although I would expect massive demands on processing.

Ideally you should be using HQPlayer with a NOS DAC which will take DSD. Here's a list:

- Aqua DACs

- Armature (some models): Asterion (similar to Holo Spring) and Cronos (similar to Denafrips Ares)

- Audiobyte Vox

- Denafrips

- Esoteric (some models?)

- Holo Audio (some models)

- iFi micro (some models)

- Metrum Acoustics

- Phasure NOS1 DAC

- Singxer SDA-x family

- RME ADI-2 DAC, ADI-2 Pro ADC/DAC

- Rockna DACs

- T+A (some models): DAC 8 DSD, SD 3100 HV, SDV 3100 HV, HA 200

- Teac UD-50x family

- Topping (some AKM models): E30, T70, T90

 

[tuga]

In terms of upconversion performace I believe that HQPlayer is a step above but its GUI is very simplistic bare-bones IT stuff... That is why I use iTunes for browsing and then drag-drop into HQP Desktop's window:

.

Audirvana which uses SoX for resampling and has a better interface (almost as good as iTunes) would be a decent alternative, perhaps Roon too. I haven't followed JRiver for a long time now but I didn't particularly enjoy its GUI either.

.

Alternatively you could upconvert with a box, for example a dCS Upsampler or an MScaller.

 

[Phobic]

This gives some idea of what it can do,  and it is much more energy efficient to run a silicon based DSP than a PC:

https://www.hifiberry.com/blog/announcing-the-dac-dsp/

The challenge is that DSP gets insanely complicated, very quickly.   A PC might be overkill.  What is it that you need specifically that a hardware based DSP won't give you?

I don't doubt that a Pi will be a viable option, the reason for looking at a PC is that I really want a very high end solution and I don't think a PC will be overkill at all. Hardware solutions have constraints on processing power, memory and number of taps (which will directly drive quality).

Yeah exactly - much easier to clean up power to a low power DC device using an aftermarket PSU if that's important to you.  Or just use an SPDIF into the dac for isolation.

I'm far from an expert and others will no doubt chip in but I think the perceived wisdom is that SPDIF (and TOSLink) aren't the best interfaces to be using, USB I think is the prefered route.

I have never seen any proof that add-on USB cards offer benefit over onboard ones. In fact, on some motherboards where some USB ports are connected directly to the chipset, and others are via an additional controller over the PCI(e) bus, general advice is to use the directly connected ones for critical tasks. Shorter signal path, lower latency and less likely to suffer driver bugs. Frankly, I suspect that the differences are not measurable, and money is better deployed elsewhere.

I think the issue with RAW PC based USB is that it's designed for PC comms and not hifi, we're just using it for our own evil means.

no idea if these "audiophile" USB cards make a difference but I'd be willing to give them a try

If you do really want to go down the PC route, then perhaps use a laptop?  It'll be silent or at least much quieter than a desktop.  Unplug it during critical listening sessions, so that it runs on battery = no mains interference.  Nice screen that can live atop the hifi kit, for cover art.  Add a USB IR receiver and a remote (per my listing above  - would split) and you have an excellent streamer.

If I go down the PC route it will be a silent headless unit which I keep in another room, just need a USB cable to connect through the wall to the DAC

I have opted for your option A but the light version. By that I mean that HQPlayer is using very simple EQ filters generated on REW.

I'm not sure if REW EQ filters is a viable option, I'm likely to do Geddes sub integration as a 1st step, and possibly Dual Bass Array in the future. I think both are going to require manipulation of delay and phase.

Having played around with rephase and REW I hate it, it's a real ball ache to do, I'd happily pay for a better solution!

I have opted for your option A but the light version. By that I mean that HQPlayer is using very simple EQ filters generated on REW.

When you are upconverting to DSD256 or above then you need massive processing power, probably combining CPU and GPU. For my old 2012 i7 Macbook Pro, jumping from DSD128 to DSD 256 has meant that the computer now needs the fan on because passive cooling is no longer suficient. Which means that I will have to think about replacing it...

So first concern is processing power. A laptop is not the best option...

I've used HQplayer already with DSD conversion, my current PC has an i9 9980hk and 1080ti, it didn't break a sweat. I think I've got a decent spare i7 kicking around somewhere as well which I could reuse.

I could take a look at just using the current PC as an option but length of USB cable will be an issue without moving the PC or Hifi. anyway I'd rather have a dedicated hifi PC I think - you can never have enough computers!

Second concern is computer-generated "noise" traveling along with the signal. One of the advantages of using HQPlayer is that you can use a low processing power "quiet" NUC as a buffer at the end of the network feeding your DAC. I use a CuBox-i with a linear PSU. But Jussi Laako has tested quite a few DACs which he claims are able to deal with the "noise" coming down the USB cable includig the RME.

Other features offered by HQPlayer are the possibility of using multiple endpoints into different systems along the house or running the processing Desktop computer headless, and it can drive multi-channel (f.e. 2 main + 2 subs) or active multi-way systems, although I would expect massive demands on processing.

I really like the Nuc idea, I could parent this off my main high powered PC.

I don't get your point on RME though, isn't that for an SPDIF/toslink card? I thought it was better to use USB instead? I can get that an optical connect will help address the noise over USB, but what is better overall?

I do like the idea of a PC running HQplayer more and more

Ideally you should be using HQPlayer with a NOS DAC which will take DSD. Here's a list:

I've done a fair bit of looking at DACs and honestly I think I'm spoilt for choice! there's some great options on the market, looking forward to trying a few

 

[Phobic]

In terms of upconversion performace I believe that HQPlayer is a step above but its GUI is very simplistic bare-bones IT stuff... That is why I use iTunes for browsing and then drag-drop into HQP Desktop's window:.

HQplayer is the right answer, it integrates with Roon as well!

just thinking on this some more but if I understand things correctly I can put the core processing on my existing main PC, buy a cheap NUC, and run HQPlayer client on the NUC which will offload to the main PC, and will dump the sound out from the NUC locally.

The main PC and NUC will be connected via Ethernet.

What I don't understand is if the NUC HQplayer can output over local USB to the DAC, I assume it can?!

 

[tuga]

The miniDSP SHD is an interesting option all-in-one option but only upconverts to 96kHz and uses only 32-bit processing (compared to HQP's 64/80-bit floating point processing and 1,536MHz rates).

Otherwise it should be a reasonable single-box alternative, and probably a better option if you are looking at active speaker DSP'ing.

.

I owned a KRK Ergo DAC some years ago, my first attempt at digital EQ room correction. The Ergo uses Lyngdorf's RoomPerfect below 300Hz.

Unfortunately its DAC side was so poor that whatever improvement it may have brought it just wasn't sigificant to outweigh the mash coming out out my speakers...

I heard a couple of really nice demos with the Lyngdorf Millennium, one of them driving Lyngdorf's open baffle DP1s and BW1 boundary woofers.

I was a bit taken aback by the measured performance of the TDAI-3400...

 

[tuga]

I don't doubt that a Pi will be a viable option, the reason for looking at a PC is that I really want a very high end solution and I don't think a PC will be overkill at all. Hardware solutions have constraints on processing power, memory and number of taps (which will directly drive quality).

As far as I know, the Pi is not an option for a processing computer. It is underpowered and at least until the Pi3 both USB and Ethernet shared the same bus.

.

The way I would go at this:

- set budget

- decide what you value the most: convenience vs performance, room correction, D/A conversion?

- which speakers: active vs.passive, commercial vs. DIY?

- which DACs?

 

[tuga]

HQplayer is the right answer, it integrates with Roon as well!

just thinking on this some more but if I understand things correctly I can put the core processing on my existing main PC, buy a cheap NUC, and run HQPlayer client on the NUC which will offload to the main PC, and will dump the sound out from the NUC locally.

The main PC and NUC will be connected via Ethernet.

What I don't understand is if the NUC HQplayer can output over local USB to the DAC, I assume it can?!

I can plug both my processing computer (runs HQP Desktop) and my endpoint (runs HQP NAA) to my DAC with either USB or Toslink.

The enpoint is only a buffer so it won't work without the processing computer.

 

[tuga]

I'm not sure if REW EQ filters is a viable option, I'm likely to do Geddes sub integration as a 1st step, and possibly Dual Bass Array in the future. I think both are going to require manipulation of delay and phase.

Having played around with rephase and REW I hate it, it's a real ball ache to do, I'd happily pay for a better solution!

I've used HQplayer already with DSD conversion, my current PC has an i9 9980hk and 1080ti, it didn't break a sweat. I think I've got a decent spare i7 kicking around somewhere as well which I could reuse.

I could take a look at just using the current PC as an option but length of USB cable will be an issue without moving the PC or Hifi. anyway I'd rather have a dedicated hifi PC I think - you can never have enough computers!

I really like the Nuc idea, I could parent this off my main high powered PC.

I don't get your point on RME though, isn't that for an SPDIF/toslink card? I thought it was better to use USB instead? I can get that an optical connect will help address the noise over USB, but what is better overall?

I do like the idea of a PC running HQplayer more and more

I've done a fair bit of looking at DACs and honestly I think I'm spoilt for choice! there's some great options on the market, looking forward to trying a few

I have no problem with REW, it's only complicated the first time and lots of friendly Wammers here to help. It is limited in scope, so I am considering a more powerful option.

Mitchco's https://accuratesound.ca service is very tempting. :mrgreen:

.

I am sorry if I confused you regarding the RME. What I meant to say is that some recent DACs are in practice immune to USB "noise", making the use of and enpoint for "noise" reduction purposes moot.

A high sample-rate NOS DAC would be the way to get the best performance out of Redbook files but it is only viable if you use analogue amplification.

If you are intent on using digital amplifiers or digital-active speakers then the advantage of using HQPlayer's upconverting/filtering algoritms becomes less significant.

That is why I put DACs last in my list above. Have you had a thought about it yet?

 

[tuga]

Changing the enpoint is easy, you go to settings and select which of the available outputs you wish to stream to:

 

[Phobic]

I have no problem with REW, it's only complicated the first time and lots of friendly Wammers here to help. It is limited in scope, so I am considering a more powerful option.

Mitchco's https://accuratesound.ca service is very tempting. :mrgreen:

.

I am sorry if I confused you regarding the RME. What I meant to say is that some recent DACs are in practice immune to USB "noise", making the use of and enpoint for "noise" reduction purposes moot.

A high sample-rate NOS DAC would be the way to get the best performance out of Redbook files but it is only viable if you use analogue amplification.

If you are intent on using digital amplifiers or digital-active speakers then the advantage of using HQPlayer's upconverting/filtering algoritms becomes less significant.

That is why I put DACs last in my list above. Have you had a thought about it yet?

REW frankly is easy, it has a quite a steep learning curve but it's not that bad. the issue is more rephase, and I don't know how to manage delay with this kind of a solution, is it just random experimentation?! no idea. Either way I think life's just to short for this crap

Would be good to understand how DACs are immune to USB noise, that implies every DAC has USB filtering etc to remove USB carried noise as an issue. I might be being paranoid but I simply don't trust that all PC USBs and DAC USBs are created equally. I'd like to understand what they're doing to mitigate the transfer of noise in a bit of detail to gain confidence that it's not detrimental

I'm still confused about RME, I thought it was a card that you put in a PC to support SPDIF, does it mean something else?

I guess you've also made me realise that I don't quite understand the full benefits of a NOS DAC, something I need to read more about.

I've not thought much about amplification in the context of DAC, streamer and DSP,  not sure I get the dependency to be honest. I think I'm moving away from "digital" active speakers/subs i.e. colocated DAC with active speakers, but I still intend to go with an active setup. DAC > AMP > Speaker. not sure yet on what kind of amp or speaker I'll use but for now it's what I have access to.

 

[tuga]

REW frankly is easy, it has a quite a steep learning curve but it's not that bad. the issue is more rephase, and I don't know how to manage delay with this kind of a solution, is it just random experimentation?! no idea. Either way I think life's just to short for this crap

Would be good to understand how DACs are immune to USB noise, that implies every DAC has USB filtering etc to remove USB carried noise as an issue. I might be being paranoid but I simply don't trust that all PC USBs and DAC USBs are created equally. I'd like to understand what they're doing to mitigate the transfer of noise in a bit of detail to gain confidence that it's not detrimental

I'm still confused about RME, I thought it was a card that you put in a PC to support SPDIF, does it mean something else?

I guess you've also made me realise that I don't quite understand the full benefits of a NOS DAC, something I need to read more about.

I've not thought much about amplification in the context of DAC, streamer and DSP,  not sure I get the dependency to be honest. I think I'm moving away from "digital" active speakers/subs i.e. colocated DAC with active speakers, but I still intend to go with an active setup. DAC > AMP > Speaker. not sure yet on what kind of amp or speaker I'll use but for now it's what I have access to.

It’s this RME that I bought:

https://www.rme-audio.de/adi-2-dac.html

 

[tuga]

Most DACs will apply a form of upconversion to the input signal via an off-the-shelf SRC chip or the one embedded in the D/A chip and use the filter or filters bundled with said chip.

HQPlayer only needs the D/A conversion because it performs the upconversion, the filtering and the dithering (and EQ, crossover filtering as well if needed). It does so using far more refined algorithms which require a lot of processing power, much more even than what is available for example in an FPGA-based device like the MScaler.

By removing upconversion and filtering from inside the DAC you are also removing sources of EMI/RFI noise as well as the demands that they impose on the power supply.

 

[tuga]

Also, in most DACs the Redbook PCM signal is upconverted by an off-the-shelf SRC chip and then gets modulated into the D/A stage (is turned to DSD).
DSD signal goes straight into the D/A stage.

 

[jamster]

As far as I know, the Pi is not an option for a processing computer. It is underpowered and at least until the Pi3 both USB and Ethernet shared the same bus.
.
The way I would go at this:
- set budget
- decide what you value the most: convenience vs performance, room correction, D/A conversion?
- which speakers: active vs.passive, commercial vs. DIY?
- which DACs?

I wasn't suggesting the PI to do the DSP. I was suggesting a Hifiberry DSP+DAC board. Very similar to the SHARC unit in the MiniDSP. It contains a very powerful standalone dsp IC. It can run without the PI, simply uses the PI to configure it, and as an optional streamer.

Custom integrated circuits can easily exceed general purpose silicon (pc processors) at the same power input. That's why bitcoin miners use them rather than high power PCs.

I am asking the OP what particular advantage a higher power PC will give over an IC (hifiberry, minidsp etc.). The number of IIR and FIR taps on both is plenty, afaik.

Genuinely interested.

 

[tuga]

I wasn't suggesting the PI to do the DSP. I was suggesting a Hifiberry DSP+DAC board. Very similar to the SHARC unit in the MiniDSP. It contains a very powerful standalone dsp IC. It can run without the PI, simply uses the PI to configure it, and as an optional streamer.

Custom integrated circuits can easily exceed general purpose silicon (pc processors) at the same power input. That's why bitcoin miners use them rather than high power PCs.

I am asking the OP what particular advantage a higher power PC will give over an IC (hifiberry, minidsp etc.). The number of IIR and FIR taps on both is plenty, afaik.

Genuinely interested.

I missed that sorry.

I used the HiFi Berry DAC + Pro on a Pi3 with LPSU. Even fed 24/192 by HQPlayer it was still far from Premier League performance.

That may have improved a bit but not enough I would expect.

We are talking of a different level of clarity here.

That is why I was asking what the priorities were. With an endless budget the sky is the limit but that is hardly ever the case.

 

[jamster]

I missed that sorry.
I used the HiFi Berry DAC + Pro on a Pi3 with LPSU. Even fed 24/192 by HQPlayer it was still far from Premier League performance.
That may have improved a bit but not enough I would expect.
We are talking of a different level of clarity here.
That is why I was asking what the priorities were. With an endless budget the sky is the limit but that is hardly ever the case.

No worries.

For this use case it's the DSP chip I'm recommending not the DAC chip.

I guess to take a step back... OP wants to build a new PC, I get that. Nothing wrong with another PC in the house.

But what exactly does he want to do with the PC? For some simple REW filters, a PC is possibly overkill. For anything more advanced, then you either need a lot of skill, or to buy something off the shelf e.g. Dirac.

 

[tuga]

No worries.

For this use case it's the DSP chip I'm recommending not the DAC chip.

I guess to take a step back... OP wants to build a new PC, I get that. Nothing wrong with another PC in the house.

But what exactly does he want to do with the PC? For some simple REW filters, a PC is possibly overkill. For anything more advanced, then you either need a lot of skill, or to buy something off the shelf e.g. Dirac.

I understand now.

 I agree that unless he wishes to DSP with HQPlayer they he probably won’t need a dedicated super computer. But I’d like me he thinks that HQP will provide the best performance then he has no alternative but to put together a powerful machine and I see no harm in dealing with EMI/RFI issues.