The new beta enables CUDA assist so also depends on GPU. The answer isn't entirely straightforward because different filters have different requirements. For DSD512 a 3.4 Ghz x 4 core machine or better is recommended. DSD256 uses roughly 2 cores, generally. Processor speed is important.
Wow, that is crazy. Resampling of 2 channels to 384/32 if done right takes like 2% CPU on a single core of an old 3.1 GHz i5.
Where are the guys that claim that high CPU and GPU load are bad for audio quality? 😛
Where are the guys that claim that high CPU and GPU load are bad for audio quality? 😛
Last edited:
Has he gone?
Has he gone?
I've been experimenting with some of HQPlayer's filters this afternoon (all DSD265 12MHz); rather taken by DSD7 256+fs though CPU loading is around 55% (around 20% higher than DSD5).
Ray
Has he gone?
An externally hosted image should be here but it was not working when we last tested it.
I've been experimenting with some of HQPlayer's filters this afternoon (all DSD265 12MHz); rather taken by DSD7 256+fs though CPU loading is around 55% (around 20% higher than DSD5).
Ray
Last edited:
Does anyone have DSD test tone, like 1kHz at 0 dB damping.
I need it to evaluate the gain through a passive LC filter.
I need it to evaluate the gain through a passive LC filter.
I've given convolution in HQPlayer a try. I'm seeing no additional CPU load, my i7 Mavbook stays comfortably at 10%CPU, only running DSD5 as the windows driver doesn't support a higher bitrate. Convolution solves a few problems I was having with my speakers and acoustics, which I have no other easy fixes for. Sounds much better than the convolution engine in Foobar2000.
But, unfortunately I have an issue. Initially the relay (Ray’s pcb), kept cutting the signal seemingly at random between 5 to 10 sec. This suggested pin 7 on the JLSounds board, was pulled high/low. However, after taking off the relay, I still got drops in the signal (not the loud codec change pops) roughly at the same random intervals. This indicates two things, the relay was trigger by something other than pin 7 going high/low and secondly something other than the relay is disrupting the signal. (maybe a fault on the PCB?)
So I looked at the system statistics, no spikes or anomalies in CPU/mem/disk/IO which correlated with the drop-outs. I tried different driver buffer settings, no changes either.
My next option is to try Foobar->TUSBaudio driver -> JLSounds USBtoI2S card. Which might allow me to pin point the problem in HQPlayer.
If that fails, DSD from Windows via the JLSounds USBtoIS2 card, might be a nonstarter for the time being.
But, unfortunately I have an issue. Initially the relay (Ray’s pcb), kept cutting the signal seemingly at random between 5 to 10 sec. This suggested pin 7 on the JLSounds board, was pulled high/low. However, after taking off the relay, I still got drops in the signal (not the loud codec change pops) roughly at the same random intervals. This indicates two things, the relay was trigger by something other than pin 7 going high/low and secondly something other than the relay is disrupting the signal. (maybe a fault on the PCB?)
So I looked at the system statistics, no spikes or anomalies in CPU/mem/disk/IO which correlated with the drop-outs. I tried different driver buffer settings, no changes either.
My next option is to try Foobar->TUSBaudio driver -> JLSounds USBtoI2S card. Which might allow me to pin point the problem in HQPlayer.
If that fails, DSD from Windows via the JLSounds USBtoIS2 card, might be a nonstarter for the time being.
Last edited:
Sorry to hear you're still having problems Stijn.
For what it is worth my setup is working perfectly with the same filter/mute board.
Have you attached a probe to pin7 while the problem is occurring? If it shows the pin going hi/lo/hi/etc then you'll know the issue is upstream?
Have you considered making up the second board?
I suspect this is a software or network problem.
Ray
Last edited:
nice movie from 2010 that demonstrate DSD or one bit DAC with FPGA.
https://www.youtube.com/watch?v=qqf89Cbju9s
https://www.youtube.com/watch?v=qqf89Cbju9s
Yes, this 'no-DAC' approach has been around for quite a while; the late Allan Wright was using it in his very positively reviewed SACD player mods more than a decade ago. Would have been nice if the video had included a bit more information.
In essence our current USB boards have replaced his FPGA dev board.
Ray
Not to worry..
I would need to order parts for the second board, so I might as well wait to do so for the new board.
It's not a network problem, the content was on the local drive..
I had a good longer session playing with HQPlayer and bits and bobs, last night.
First off, I fixed the dropped out issue, by setting the buffer time in HQPlayer to 250ms. This does sounds simple enough, but there were a lot a variables, so I give myself some slack for not having tried that earlier.🙄
I also did a side by side comparison with HQPlayer between my PCM only DDDAC and the no-dac design. Mostly upsampling various bitrates to, PCM (poly-sinc-shrt-mp & NS5 @ 384000) versus, DSD (poly-sinc-shrt-mp & DSDv2 @5644800). To be honest, I heard very little sonic differences, the DSD had a little bit more air in the soundstage at the expense of sounding a little thinner.
What I did find very interesting is the nice sounding convolution engine in HQPlayer. Together with REW/UMIK-1 this allows me to tweak the frequency response to the infinite degree.
First off, I fixed the dropped out issue, by setting the buffer time in HQPlayer to 250ms. This does sounds simple enough, but there were a lot a variables, so I give myself some slack for not having tried that earlier.🙄
I also did a side by side comparison with HQPlayer between my PCM only DDDAC and the no-dac design. Mostly upsampling various bitrates to, PCM (poly-sinc-shrt-mp & NS5 @ 384000) versus, DSD (poly-sinc-shrt-mp & DSDv2 @5644800). To be honest, I heard very little sonic differences, the DSD had a little bit more air in the soundstage at the expense of sounding a little thinner.
What I did find very interesting is the nice sounding convolution engine in HQPlayer. Together with REW/UMIK-1 this allows me to tweak the frequency response to the infinite degree.
Merry Christmas Stijn.
Good news that you ironed out the wrinkles. So the direct DSD solution costing just a few tens of $ compares favourably with a DAC costing hundreds of $.
I've not experimented with the convolution engine yet so would be interested in finding out more.
We're having a quiet Christmas morning before travelling to see family this afternoon. I got up really early and spent a couple of hours sorting out the layout for my flip-flop build and I've just ordered a chassis for it so it should come together quite quickly in the new year.
Ray
...glad to read that things are working, @stijn001... I messed with REW years ago and never was able to improve the overall presentation. Since then I've conversed with others who shared the experience: the more you tweak the poorer the result - better to treat the room. Perhaps this set of filters is the key.
Ray, promising news over on the Botic thread! ...will be interested to hear of the sonic results and the CPU load to run NAA on BBB (using 'top') if/when it works. My future implementation of 'no-dac' could take any of several configurations. I think I will have PCM performance nearly maxed-out when all is assembled and running. That's why the BBB-as-NAA option would be a great test platform. I'll be interested in different user's results and how they inform theories regarding the cause/effect of any HQPlayer sonic benefits. I am curious to learn if theoretical differences in "speaker cone micro-distortion" is supported by any evidence! 😀
Frank
We'll see Frank. I'll give it a try but I'll not hold my breathe as I believe I've tried something along those lines before.
Ray
Hi
My XMOS U8 256 DSD arrived.
So strange to connect the FPGA straight to the amp 🙂
The sound is very good, but it's hard for me to compare with the delta sigma DAC (pcm1976, I2S), its take 10 sec to replace the setup and play again the same song. Still trying to test.
I have some constant noise(-40db), maybe white noise after the LPF, Suggestions?
Usb power
No isolation yet
Only LPF 2.2nf 1khom
Tube pre amplifier 12ax7
Class A with FR speaker
Jriver, Foobar (DSD256)
My XMOS U8 256 DSD arrived.
So strange to connect the FPGA straight to the amp 🙂
The sound is very good, but it's hard for me to compare with the delta sigma DAC (pcm1976, I2S), its take 10 sec to replace the setup and play again the same song. Still trying to test.
I have some constant noise(-40db), maybe white noise after the LPF, Suggestions?
Usb power
No isolation yet
Only LPF 2.2nf 1khom
Tube pre amplifier 12ax7
Class A with FR speaker
Jriver, Foobar (DSD256)
Attachments
My family know that I wanted a new computer for my 'music stuff' but were unsure of the specifics so I received several Christmas gifts of money towards it and I've now ordered an Intel NUC E3815 for use as an HQPlayer NAA. When it arrives I plan to try and install the Debian Stretch distro that I've been using onto its eMMC (I'll document and share).
Which got me thinking about the ongoing effort to get Beaglebone Black/Botic working as an NAA.
I've thinking that, apart from the specific instance of being able to use my existing DAC in an HQP context, there is not much value in progressing this issue. Why? Because to make BBB/Botic effective you need to add external clocks and isolation along with the requisite additional power supplies so there isn't an enormous cost differential between the BBB/Botic solution and a USB-based one using a NUC/JLSounds. I remain to be convinced that there will be any SQ difference between these two digital solutions.
Ray
Which got me thinking about the ongoing effort to get Beaglebone Black/Botic working as an NAA.
I've thinking that, apart from the specific instance of being able to use my existing DAC in an HQP context, there is not much value in progressing this issue. Why? Because to make BBB/Botic effective you need to add external clocks and isolation along with the requisite additional power supplies so there isn't an enormous cost differential between the BBB/Botic solution and a USB-based one using a NUC/JLSounds. I remain to be convinced that there will be any SQ difference between these two digital solutions.
Ray
Hi Frank, I'm about to go over to the Botic's thread to check out this discussion. In contrast to Ray, I'm very interested in running NAA on the Botic image. I’m not keen on a Intel NUC, I’ve had two Intel mini boards die on be before.
I'm a bit unclear on how to interpret you comment about, “the more you tweak the poorer the result”.
First of all, the perfect fullrange 7 octave single linear driver doesn’t exists yet, (as far as I’m aware), so we’ll have to “tweak” somewhere. But for argument sake I presume you are refering to some kind of DSP tweaking in software. I do not know of a logical reason why running a convolution engine in software would be detrimental. Unless the algoritme used is not up to scratch, which is why I made the comparision between Foorbar and HQPlayer. I understand you are running ALSA cross-over on the BBB.., how does this compare?
I agree that trying to fix for example a driver issues with EQ will not work. Making it do something it doesn't want to, is a different problem all to gether.
As for acoustic correction. When I bought my current house this was an explicit consideration, I have very few room nodes already and I’ve added quite a “lot” of absorption and diffusion as well. But there is always something left to be desired for. Define the perfect acoustic properties of a room for me, please? 😉
Last edited:
Stijn, I'll still follow up and try to get BBB/Botic running as an NAA because of my current Buffalo IIIse DAC...
Ray
hi again
Try to change the driver but same results.
When I set DSD64 the noise disappears, with 128 it came back. (Hiss noise)
Still try to compare between the DSD and S-D