With pcm2dsd it comes out DSD256. And in my opinion, it can be placed behind a fifo. Try it, it's the best place for me.
And you will be able to play both pcm and dsd with HQPlayer. The pass through function will switch automatically.
that is not what i have found on my system...
surely something funky is going on.
Are you sure that with a double the MCLK feeding the pcm2dsd the output rate doesn't double ? can't be a side-effect of using a faster clock rather than the designed 22/24 of the amanero ?
Try it for a test...
with a double the MCLK feeding the pcm2dsd you will have dual DSD_CLK and think you have DSD512.
But pcm2dsd was made/calculated for MCLK 22/24MHz.
I can't predict what comes out of this with MCLK 45/49MHz.
Yes, it does happen automatically. It is Amanero that controls the DSD_on pin.
It's a miracle you could have "sound" when you applied a 45/49MHz clock on the pcm2dsd board.
The input is still the original frequency(48kHz), but the internal calculation period becomes double(96kHz). I'm sure the data buffer can't trace the right track for successful FIR. The only possibility is doubled output clock, which means 512OSR instead of 256OSR, but DSD data itself can't be guaranteed. There was a chance you had destructive noise. I agree with PJotr25. The best(only) position is where the 22/24MHz clock exists.
BTW, my previous code for the dsd2pcm board didn't activate DSDON, which means unintentional mute. I will post the revised one in a week. Mine will never have a lucky consequence if you apply a 45/49 MHz clock; you would have a mess.
The input is still the original frequency(48kHz), but the internal calculation period becomes double(96kHz). I'm sure the data buffer can't trace the right track for successful FIR. The only possibility is doubled output clock, which means 512OSR instead of 256OSR, but DSD data itself can't be guaranteed. There was a chance you had destructive noise. I agree with PJotr25. The best(only) position is where the 22/24MHz clock exists. BTW, my previous code for the dsd2pcm board didn't activate DSDON, which means unintentional mute.
I will post the revised one in a week. Mine will never have a lucky consequence if you apply a 45/49 MHz clock; you would have a mess. 
I've been trying to source a couple of the FPGAs required to realise this project but they're only available on long lead-time backorder at all the suppliers I've tried. Does anyone have any available or know of a reliable source at non-extortionate pricing.
I guess the availability issue is another symptom of the global silicon supply problem?
Cheers
Ray
I guess the availability issue is another symptom of the global silicon supply problem?
Cheers
Ray
I think so, also many "big" companies have done huge investements in inventory to secure production for the near future, so the effect has snowballed even more.
BTW, i may have found a reputable source in china for the FPGA's, ordered few samples of them. They will arrive shortly on my bench. If they prove to be legit and reliable then i will share the contact source.
I'll keep you posted.
You can check very good algorithim here below;
PCM-DSD_Converter | PCMDSD.com - Converting PCM to DSD.
Source code is here;
GitHub - serieril/PCM-DSD_Converter: PCM16bitInt~64bitFloat/44.1kHz~768kHz to DSD64-DSD2048
I hope it helps to improve the embeded firmware
PCM-DSD_Converter | PCMDSD.com - Converting PCM to DSD.
Source code is here;
GitHub - serieril/PCM-DSD_Converter: PCM16bitInt~64bitFloat/44.1kHz~768kHz to DSD64-DSD2048
I hope it helps to improve the embeded firmware
Another very nice resource is xx3stksm's Wiki page: PlayPcmWin / Wiki / PCMtoSDM
Some of the calculations appear to have been done with 'Microsoft Mathematics' which is downloadable freeware.
Also of possible interest, Chord claims to use a 17th order noise shaper in their higher end dacs.
Some of the calculations appear to have been done with 'Microsoft Mathematics' which is downloadable freeware.
Also of possible interest, Chord claims to use a 17th order noise shaper in their higher end dacs.
Last edited:
Hi, Markw4. This is not my HP. But I'm sure nothing is better than this if you design DSM regardless of software and hardware. My analysis of DSM is here.
Physical meaning of DSM
But I'm sure nothing is better than this if you design DSM regardless of software and hardware. My analysis of DSM is here.Physical meaning of DSM
Last evening, with a friend of mine with great experience in audio playback and recording, I did a listening session to compare 3 sources:
1) HQPlayer set 5.6MHz ADSM7EC running on a i3 at 3.6GHz + NAA over Ethernet + DSC 2.6
2) Foobar2000 + Amanero Combo USB + CT7302PL DSD 8x + DSC 2.6
3) Foobar2000 + Amanero Combo USB + Simple DSD modulator + DSC 2.6
As was to be expected the first is a reference such as micro detail and instrument separation without ever creating a listening fatigue.
The second system creates a good detailed sound but it sometimes feels like listening to a digital source like a normal dac.
The last source creates a beautiful warm and analog sound but with respect to the first source it loses micro details, so a very good result.
The amplifier was the my new PowerFollower 2021 SW and the my Monitor 4 loudspeakers.
1) HQPlayer set 5.6MHz ADSM7EC running on a i3 at 3.6GHz + NAA over Ethernet + DSC 2.6
2) Foobar2000 + Amanero Combo USB + CT7302PL DSD 8x + DSC 2.6
3) Foobar2000 + Amanero Combo USB + Simple DSD modulator + DSC 2.6
As was to be expected the first is a reference such as micro detail and instrument separation without ever creating a listening fatigue.
The second system creates a good detailed sound but it sometimes feels like listening to a digital source like a normal dac.
The last source creates a beautiful warm and analog sound but with respect to the first source it loses micro details, so a very good result.
The amplifier was the my new PowerFollower 2021 SW and the my Monitor 4 loudspeakers.
"2) Foobar2000 + Amanero Combo USB + CT7302PL DSD 8x + DSC 2.6"
For the listeners at my location CT7302 sounded awful. I talked to the guys at ComTrue but they had no solution.
Later a professional dac designer contacted me to ask about what I tried with CT7302. I gave him the info I had and my listening impressions. He said he wanted to try to see if it could be used as a replacement for AK4137. He also said he would let me know if he found a way to make it sound good. Never heard back. Assume the effort failed, given the results of my various efforts.
For the listeners at my location CT7302 sounded awful. I talked to the guys at ComTrue but they had no solution.
Later a professional dac designer contacted me to ask about what I tried with CT7302. I gave him the info I had and my listening impressions. He said he wanted to try to see if it could be used as a replacement for AK4137. He also said he would let me know if he found a way to make it sound good. Never heard back. Assume the effort failed, given the results of my various efforts.
I don't understand what you mean by it sounds bad, it has higher performance than the AK4137 and coupled with DSC modules the sound is pleasant compared to a normal DAC.
In a contest organized by the forum Vnav this was the winner.
Attachments
It sounded *very* jittery compared to other devices, even with an external reference clock as suggested by CT in their app note. AK4137 sounded much better for DSD and for PCM upsampling. SRC4392 sounded better for PCM upsampling (it can't do DSD conversion). DSD256 from HQ Player also sounded better. Overall, CT7302 was by far the worst sounding. This was evaluated by a panel of listeners, one at a time, alone (so no influencing each other), using AK4499 eval board along with Pass Labs HPA-1 and Audeze LCD-X headphones.
Also, CT7302 appears to be a 'system on a chip' design. It uses a separate 11MHz crystal to clock itself. An external audio clock can optionally be used as output reference. Sounds maybe like the internal PLL/DPLL/PPLL runs from the 11Mhz crystal and or it is being jittered by noise from the CPU part of chip. Could be other problems though. Its hard to know unless CT wants to reveal more about its design.
Last edited:
In order to compare one element in a chain, the chains must be the same.
PCM2DSD is also designed to make it easy to compare it with eg. HQPLayer.
For example, such a chain:
HQPlayer->Amanero->PCM2DSD->DSC
When we play DSD from HQPlayer in PCM2DSD passed thru works and we listen to the modulator + filter from HQPlayer.
When we play PCM from HQPLayer in PCM2DSD its modulator works and we listen to the modulator + filter built into PCM2DSD.
PCM2DSD is also designed to make it easy to compare it with eg. HQPLayer.
For example, such a chain:
HQPlayer->Amanero->PCM2DSD->DSC
When we play DSD from HQPlayer in PCM2DSD passed thru works and we listen to the modulator + filter from HQPlayer.
When we play PCM from HQPLayer in PCM2DSD its modulator works and we listen to the modulator + filter built into PCM2DSD.