Yes, Jobs you are doing very well! great!
*please continue to work!
Good luck!
quanghao
Ok,
I know i'm just quoting myself...
Simple answer is yes: jRiver will make first dsp changes and then upscaling.
Reference:
https://yabb.jriver.com/interact/index.php?PHPSESSID=eumeqccs8r5oov1fuleiihg8h2&topic=82074.msg562285#msg562285
Here's this weekend's experiment;
Balanced/flipflop board mounted on the JLSounds with the differential outputs going straight to Silk STA522S transformer volume controls - the TVC is the LP filter and there's no R/C filter.
Input is DSD256, upsampled from 16/44.1 PCM with HQPlayer.
As it stands this delivers the most amazing sound quality, I think perhaps the best I've yet achieved with the no-dac approach, but, frustratingly, it is also the worst I've had for noise - there's audible white noise/mush during quieter passages.
It's so good I'm going to have to persevere with it.
I'm currently powering everything via USB so the first thing I'll try is to hook up some good quality low noise power supplies, which will also re-enable the USB board's isolation.
Also, the wires connecting the flipflop outputs to the TVCs are quite long (didn't want to cut the wires short until I had an idea of the potential) and there are a lot of unscreened wires from the TVCs to the switch, which could be acting as antennas.
Thoughts?
There's info about the Silk TVCs here;
SAC Thailand
You can see that their frequency response is pretty flat out to 35KHz after which it falls away at a similar rate to a first order low pass filter.
Anyway, I need a walk now but I'll return to the fray a bit later.
Ray
An externally hosted image should be here but it was not working when we last tested it.
Balanced/flipflop board mounted on the JLSounds with the differential outputs going straight to Silk STA522S transformer volume controls - the TVC is the LP filter and there's no R/C filter.
Input is DSD256, upsampled from 16/44.1 PCM with HQPlayer.
As it stands this delivers the most amazing sound quality, I think perhaps the best I've yet achieved with the no-dac approach, but, frustratingly, it is also the worst I've had for noise - there's audible white noise/mush during quieter passages.
It's so good I'm going to have to persevere with it.
I'm currently powering everything via USB so the first thing I'll try is to hook up some good quality low noise power supplies, which will also re-enable the USB board's isolation.
Also, the wires connecting the flipflop outputs to the TVCs are quite long (didn't want to cut the wires short until I had an idea of the potential) and there are a lot of unscreened wires from the TVCs to the switch, which could be acting as antennas.
Thoughts?
There's info about the Silk TVCs here;
SAC Thailand
You can see that their frequency response is pretty flat out to 35KHz after which it falls away at a similar rate to a first order low pass filter.
Anyway, I need a walk now but I'll return to the fray a bit later.
Ray
Very interesting
It seems like tvcs aren't enough to act as lpf.
It's a shame, on paper they are the best way.
Can you monitor the waveform with an oscilloscope?
It seems that an active buffer is a must.
EDIT: too long unshielded wires and radio frequency waveform is not good for a silent audio experience.
EDIT2: you can try to close board and tvcs in a solid metal chassie.
Last edited:
Maybe u can try cascading the 2 TVC to one channel only, the input of the 2nd TVC to the output of the first TVC just to see if there is noise on one channel... or adding the RC filter in the signal path just to experiment...
white noise on the 20KHz band might not be coming from the DSD signal... @256DSD the noise is pretty far away...
I was able to play DSD(2.8Mhz) without problems using digital filter method from electrart and yanasoft...the filter cutoff is rather modest...of course there is the opamp LPF on the electrart board and the yanasof board had a Lundahl transformer to further cut off the signal...
the electrart board i got initially had static noise as well... but i solve it by removing most of the parts and resolder each of them back... maybe its pure luck i got it to work...
Pure DSD DAC (DAC Less)
yet to try the passive LPF... hopefully i get he time to experiment it myself...
I've just spent a couple of hours listening to the TVC project.
I hooked up some good quality low noise power supplies. Although they didn't fix the noise issue they did deliver a good step upwards with the sound quality.
I don't want to lose what this project is delivering, the music is so immediate but with an organic sort of quality. Everything in the music has it's own space and the details really tell you about the playing. To be honest it is really hard to put into words; perhaps real would be a good descriptor.
But then there's the noise!
With a little experimentation I have determined that the noise isn't general environmental interference being picked up as it isn't present when there is no playback. I'll still look out for a metal enclosure I can put it inside to confirm this hypothesis.
So if it is playback related it must either be the result of the filtering not working well enough or by noise being generated by the digital circuitry which is picked up via the antenna.
I'll try to record the noise tomorrow evening.
I can't look at the noise on an oscilloscope as I haven't owned one for a number of years after my last one went to oscilloscope heaven.
I'm thinking I'll make up another flipflop board with LP filters, followed by the TVCs. Perhaps that sort of compound arrangement will eliminate the noise.
Ray
I hooked up some good quality low noise power supplies. Although they didn't fix the noise issue they did deliver a good step upwards with the sound quality.
I don't want to lose what this project is delivering, the music is so immediate but with an organic sort of quality. Everything in the music has it's own space and the details really tell you about the playing. To be honest it is really hard to put into words; perhaps real would be a good descriptor.
But then there's the noise!
With a little experimentation I have determined that the noise isn't general environmental interference being picked up as it isn't present when there is no playback. I'll still look out for a metal enclosure I can put it inside to confirm this hypothesis.
So if it is playback related it must either be the result of the filtering not working well enough or by noise being generated by the digital circuitry which is picked up via the antenna.
I'll try to record the noise tomorrow evening.
I can't look at the noise on an oscilloscope as I haven't owned one for a number of years after my last one went to oscilloscope heaven.
I'm thinking I'll make up another flipflop board with LP filters, followed by the TVCs. Perhaps that sort of compound arrangement will eliminate the noise.
Ray
Well.
I still think an active buffer is a must, but it's only my though.
Dsd works like a sigma-Delta dac.
The frequency shaping push out the noise.
The noise is a white noise, white noise is common mode noise.
Balanced output make possible for us to work with the noise.
By the way, i'm planning on a valve buffer that acts as unbalancer and i'm asking myself how much exactly is the single ended rms voltage at output.
I miss that value due to the long thread.
Can you help me?
I still think an active buffer is a must, but it's only my though.
Dsd works like a sigma-Delta dac.
The frequency shaping push out the noise.
The noise is a white noise, white noise is common mode noise.
Balanced output make possible for us to work with the noise.
By the way, i'm planning on a valve buffer that acts as unbalancer and i'm asking myself how much exactly is the single ended rms voltage at output.
I miss that value due to the long thread.
Can you help me?
When there is no playback, there is no high frequency signal.
So you still don't know nothing about the 'goodness' of the devices.
Probably a set of good shielded wires will work well.
Ray, I would love to try a pair of Silk TVCs but they are a bit pricey right now, especially with AUD exchange rate. But I think that this is most elegant solution to filtering/buffering/volume control in one device, and I hope that you can make this work.
For what its worth, I think that the noise must be picked up. I say this because:
1. My initial build had lots of noise - but it was a very naked prototype just like yours. I've now enclosed everything and it sounds fine.
2. I use Lundahl transformer for filtering, and I don't have a noise problem, The -3db point on these trannies is above 150kHz!!! So I doubt that you need additional filtering with the Silks, which don't seem to extend as far.
I'm really interested to see how this goes. Good luck.
For what its worth, I think that the noise must be picked up. I say this because:
1. My initial build had lots of noise - but it was a very naked prototype just like yours. I've now enclosed everything and it sounds fine.
2. I use Lundahl transformer for filtering, and I don't have a noise problem, The -3db point on these trannies is above 150kHz!!! So I doubt that you need additional filtering with the Silks, which don't seem to extend as far.
I'm really interested to see how this goes. Good luck.
I don't believe that an active filter is required. In fact, I know that an active filter is not required. Please refer to my earlier circuit diagram.
With regards to output voltage - as i documented in an earlier post - I created a 0db test tone (440Hz) and played it back through my no-DAC. I measured 1.55volt output (balanced). But also note that supply rail for my flip flops is a 6V SLA (actual voltage under load is around 6.4V) and I assume that if use use a lower supply rail (say 5volt) then output will be less.
Update on Linux.
I have been running Ubuntu Studio 14 as this seems to be a well regarded and supported distro. But I recall in one of Ray's threads that he is using Debian Stretch. So I thought that I would give it a go. I have noticed 2 positive changes:
1. Sound appears to be better. Its hard to describe these changes when everything already sounds so good, and of course its impossible to verify with any quantitative measures but its just a feeling I get - everything is just a shade more relaxed, 'organic' is the word that Ray uses and I agree.
2. CPU load is reduced. I have done some testing to compare CPU load between 2 OS. Note that all files are upsampled to DSD256. HQPlayer settings are Filter poly sinc mp 2s. Modulator DSD7 256+fs. Output rate + auto (12.3Mhz for 48k family, 11.3Mhz for 44.1k family). I tried 4 different formats and results are:
Source 24-96k. CPU load Ubuntu 42-43% CPU Load Debian 38-40%.
Source DSD64. Ubuntu 43-44%. Debian 40-41%.
Source 16-44.1k. Ubuntu 36-38%. Debian 34-35%.
Source DSD128. Ubuntu 49-51%. Debian 45-47%.
Source 24-192k. Ubuntu 45-46%. Debian 41-43%.
So there is a consistent improvement of around 10% using Debian, for all formats.
If your PC is struggling, then I would suggest that you try Debian. My PC is a few years old, 6 core AMD CPU, and as you can see its working hard (fan noise gets annoying - maybe that's why i don't hear any DSD-related white noise
). So 10% reduction in CPU load is welcome. However I still can't playback with standard poly since filters as I get stuttering and drop outs, even with Debian. But with the -2s filter, all is good.
I wonder - is the reduced CPU load contributing to the perceived SQ improvement? If so, would SQ improve if I bought a new (more powerful CPU), Can anybody make any meaningful comparisons between CPU?
So now I have Silk TVC, and upgraded PC, on my shopping list. If this hobby doesn't kill me, my wife will.
I have been running Ubuntu Studio 14 as this seems to be a well regarded and supported distro. But I recall in one of Ray's threads that he is using Debian Stretch. So I thought that I would give it a go. I have noticed 2 positive changes:
1. Sound appears to be better. Its hard to describe these changes when everything already sounds so good, and of course its impossible to verify with any quantitative measures but its just a feeling I get - everything is just a shade more relaxed, 'organic' is the word that Ray uses and I agree.
2. CPU load is reduced. I have done some testing to compare CPU load between 2 OS. Note that all files are upsampled to DSD256. HQPlayer settings are Filter poly sinc mp 2s. Modulator DSD7 256+fs. Output rate + auto (12.3Mhz for 48k family, 11.3Mhz for 44.1k family). I tried 4 different formats and results are:
Source 24-96k. CPU load Ubuntu 42-43% CPU Load Debian 38-40%.
Source DSD64. Ubuntu 43-44%. Debian 40-41%.
Source 16-44.1k. Ubuntu 36-38%. Debian 34-35%.
Source DSD128. Ubuntu 49-51%. Debian 45-47%.
Source 24-192k. Ubuntu 45-46%. Debian 41-43%.
So there is a consistent improvement of around 10% using Debian, for all formats.
If your PC is struggling, then I would suggest that you try Debian. My PC is a few years old, 6 core AMD CPU, and as you can see its working hard (fan noise gets annoying - maybe that's why i don't hear any DSD-related white noise
). So 10% reduction in CPU load is welcome. However I still can't playback with standard poly since filters as I get stuttering and drop outs, even with Debian. But with the -2s filter, all is good.I wonder - is the reduced CPU load contributing to the perceived SQ improvement? If so, would SQ improve if I bought a new (more powerful CPU), Can anybody make any meaningful comparisons between CPU?
So now I have Silk TVC, and upgraded PC, on my shopping list. If this hobby doesn't kill me, my wife will.
Thanks for the feedback/ideas iperv
I still have that option, though I'm not sure it is necessary.
The noise I'm getting isn't simple white noise, there's a sort of variable mush, a bit like tuning a radio in the old days. I'll try to record the end of one track and the start of the next, when it is most audible.
BTW, I'm running the TVCs with balanced input to SE output at the moment - didn't make that clear before. My Broskie BCF is a Bal to SE buffer so maybe worth a try.
Aound 650-700mV as I recall.
The point I was making is just that if there is no noise when there is no playback then presumably it isn't RF noise pollution being picked up from the general environment. That still leaves several possibilities for where the noise is coming from, inadequate LP filtering or noise generated by the digital sections of the USB board and/or the flipflop?
I'm not sure what you mean by "...'goodness' of the devices"?
Possibly, but I'm still trying to get my head round how you might screen that sort of rats nest of wiring?
Ray
I still have that option, though I'm not sure it is necessary.
The noise I'm getting isn't simple white noise, there's a sort of variable mush, a bit like tuning a radio in the old days. I'll try to record the end of one track and the start of the next, when it is most audible.
BTW, I'm running the TVCs with balanced input to SE output at the moment - didn't make that clear before. My Broskie BCF is a Bal to SE buffer so maybe worth a try.
Aound 650-700mV as I recall.
The point I was making is just that if there is no noise when there is no playback then presumably it isn't RF noise pollution being picked up from the general environment. That still leaves several possibilities for where the noise is coming from, inadequate LP filtering or noise generated by the digital sections of the USB board and/or the flipflop?
I'm not sure what you mean by "...'goodness' of the devices"?
Possibly, but I'm still trying to get my head round how you might screen that sort of rats nest of wiring?
Ray
Thanks for the encouragement. Yes, the simplicity is amazing and the results very encouraging.
What sort of noise where you experiencing with your prototype? I'll try and post a sample of the noise I'm getting a bit later.
Thinking about how to break this problem down into rational, testable bites I'm incline to give the LP filter aspect the benefit of the doubt for now as there are some very obvious compromises with the wiring, particularly the long wires from the flipflop outputs (when laying out the PCBs the message from those with the knowledge/experience was keep everything as tight as possible). I'm also trying to work out how I could screen the flipflop device. Would be delighted with some input from any professionals out there?
To test the LP filter aspect I'm thinking I'll put another PCB together but with a LP filter and connect the TVCs after the filter.
Ray
Update on Linux.
I have been running Ubuntu Studio 14 as this seems to be a well regarded and supported distro. But I recall in one of Ray's threads that he is using Debian Stretch. So I thought that I would give it a go. I have noticed 2 positive changes:
1. Sound appears to be better. Its hard to describe these changes when everything already sounds so good, and of course its impossible to verify with any quantitative measures but its just a feeling I get - everything is just a shade more relaxed, 'organic' is the word that Ray uses and I agree.
2. CPU load is reduced. I have done some testing to compare CPU load between 2 OS. Note that all files are upsampled to DSD256. HQPlayer settings are Filter poly sinc mp 2s. Modulator DSD7 256+fs. Output rate + auto (12.3Mhz for 48k family, 11.3Mhz for 44.1k family). I tried 4 different formats and results are:
Source 24-96k. CPU load Ubuntu 42-43% CPU Load Debian 38-40%.
Source DSD64. Ubuntu 43-44%. Debian 40-41%.
Source 16-44.1k. Ubuntu 36-38%. Debian 34-35%.
Source DSD128. Ubuntu 49-51%. Debian 45-47%.
Source 24-192k. Ubuntu 45-46%. Debian 41-43%.
So there is a consistent improvement of around 10% using Debian, for all formats.
If your PC is struggling, then I would suggest that you try Debian. My PC is a few years old, 6 core AMD CPU, and as you can see its working hard (fan noise gets annoying - maybe that's why i don't hear any DSD-related white noise
I wonder - is the reduced CPU load contributing to the perceived SQ improvement? If so, would SQ improve if I bought a new (more powerful CPU), Can anybody make any meaningful comparisons between CPU?
So now I have Silk TVC, and upgraded PC, on my shopping list. If this hobby doesn't kill me, my wife will.
Thanks Hazard, interesting.
Just to clarify, I'm running HQPlayer on a workstation running Win10, Debian Stretch is on the HQPlayer NAA device. My NAA is completely silent and the grunt workstation is physically remote (I control playback via Muso on the workstation and a tablet with a web interface to Muso). Maybe you should add an NAA to your shopping list?
Last edited:
Thank's Hazard for your experience.
Your own post was the one that push me to the flip flop board, so I believe you of course.
High quality irons are very priced.
@nautibuoy
Thank you too for the vrms.
To shield your trafos you can simply put them in a metallic case and connect them to the board with shielded wires.
The board too in a little metallic box is a good idea.
Then put both box in a bigger metallic case.
Yes, it's easy to say but the execution is harder. I'm looking out for something to experiment with.
