Looking more like FPGA DSP
I am leaning towards upsampling and transcoding at the point of consumption so the network infrastructure will only need to deal with lower bandwidth e.g. 44.1k music. So existing home wifi and a low power atom PC media player connected direct to DAC is all that are needed.
I am leaning towards upsampling and transcoding at the point of consumption so the network infrastructure will only need to deal with lower bandwidth e.g. 44.1k music. So existing home wifi and a low power atom PC media player connected direct to DAC is all that are needed.
Going well, I have got FPGA DSP embedded upsampler done that will convert from 44.1k PCM to DSD256. If native DSD input then all goes direct automatically. So options for low power media player or HQPlayer on a hefty PC. All being well, expected release this Spring.
This can be done by replicating the 2ch design. Best to wait until the 2ch has been fully evaluated.
Yes, it is possible to connect SDTrans (or similar I2S sources) direct to this DSD DAC, PCM or DSD input. No S03 required as re-clocker and isolation already built into the DSD DAC.
Hi Acko, you seem to have been quietly plugging away with your DSD project and it seems to becoming more sophisticated with upsampling, isolation and reclocking and, if I'm correct, digital volume?
How is the filtering developing and is it still by-passable? A block diagram f your architecture would be usefu?
I encountered white-noise on the output when I experimented with a LP filter at 80K and used HQP for upsampling to DSD256 with the volume control enabled, no problem with HQP's volume control disabled.
Do you have an ETA and price sketched in yet?
Ray
Hi Ray, I have been quite busy with this DSD DAC and there is still more work to do 
One thing that is becoming apparent now is that there are two types of DSD:
1. Natively recorded Hi-res material (Pure DSD)
2. Upsampled DSD (PCM-DSD)
DSD by nature is very noisy (generated noise) and the noise energy far exceeds that of the base signal. Upsampled ones are even worse with weird artifacts that can be difficult to remove. From the looks of it Upsampled "on-the-fly" DSD will the common use-case for a typical DSD DAC.
A simple LPF "NoDAC" may not be enough to remove ultrasonics to safe levels and sometimes may cause instability to amplifiers. So you may be experiencing these effects with some combinations of HQPlayer. I have also looked at Andrea Ciuffoli's work that is very similar and if you scroll down to the performance graphs, the THD+N for this simple setup is -68dB! (0.038%) ->AudioDesignGuide
I thought rather than rushing out to roll out boards it is best to see what sort of performance that can be realistically achieved.
So the figures stated earlier are for PCM 44.1k/16-bit to DSD
and this just in: for PureDSD direct: THD+N=-120dB and DNR=-130dB
These are based on theoretical analysis only and may be different in the final product but at least some indication that is it worth the candle to pursue further.
Things are still in a state of flux so I am afraid I cannot tell you at this stage how the final product and prices will look like. Most likely a complete DAC system first with everything integrated so that the performance can be fully evaluated. Yes, there will be Digital Volume control, filters etc all tuned to match so a full featured DAC. Some sections like Vol can be bypassed but not things like safety filters.
One thing that is becoming apparent now is that there are two types of DSD:
1. Natively recorded Hi-res material (Pure DSD)
2. Upsampled DSD (PCM-DSD)
DSD by nature is very noisy (generated noise) and the noise energy far exceeds that of the base signal. Upsampled ones are even worse with weird artifacts that can be difficult to remove. From the looks of it Upsampled "on-the-fly" DSD will the common use-case for a typical DSD DAC.
A simple LPF "NoDAC" may not be enough to remove ultrasonics to safe levels and sometimes may cause instability to amplifiers. So you may be experiencing these effects with some combinations of HQPlayer. I have also looked at Andrea Ciuffoli's work that is very similar and if you scroll down to the performance graphs, the THD+N for this simple setup is -68dB! (0.038%) ->AudioDesignGuide
I thought rather than rushing out to roll out boards it is best to see what sort of performance that can be realistically achieved.
So the figures stated earlier are for PCM 44.1k/16-bit to DSD
and this just in: for PureDSD direct: THD+N=-120dB and DNR=-130dB
These are based on theoretical analysis only and may be different in the final product but at least some indication that is it worth the candle to pursue further.
Things are still in a state of flux so I am afraid I cannot tell you at this stage how the final product and prices will look like. Most likely a complete DAC system first with everything integrated so that the performance can be fully evaluated. Yes, there will be Digital Volume control, filters etc all tuned to match so a full featured DAC. Some sections like Vol can be bypassed but not things like safety filters.
Hi Ray, I have been quite busy with this DSD DAC and there is still more work to do
One thing that is becoming apparent now is that there are two types of DSD:
1. Natively recorded Hi-res material (Pure DSD)
2. Upsampled DSD (PCM-DSD)
DSD by nature is very noisy (generated noise) and the noise energy far exceeds that of the base signal. Upsampled ones are even worse with weird artifacts that can be difficult to remove. From the looks of it Upsampled "on-the-fly" DSD will the common use-case for a typical DSD DAC.
A simple LPF "NoDAC" may not be enough to remove ultrasonics to safe levels and sometimes may cause instability to amplifiers. So you may be experiencing these effects with some combinations of HQPlayer. I have also looked at Andrea Ciuffoli's work that is very similar and if you scroll down to the performance graphs, the THD+N for this simple setup is -68dB! (0.038%) ->AudioDesignGuide
I thought rather than rushing out to roll out boards it is best to see what sort of performance that can be realistically achieved.
So the figures stated earlier are for PCM 44.1k/16-bit to DSD
and this just in: for PureDSD direct: THD+N=-120dB and DNR=-130dB
These are based on theoretical analysis only and may be different in the final product but at least some indication that is it worth the candle to pursue further.
Things are still in a state of flux so I am afraid I cannot tell you at this stage how the final product and prices will look like. Most likely a complete DAC system first with everything integrated so that the performance can be fully evaluated. Yes, there will be Digital Volume control, filters etc all tuned to match so a full featured DAC. Some sections like Vol can be bypassed but not things like safety filters.
Right on Acko. Any rough ETA of when your thinking you can expect a finished product?
Thanks for the update Acko. Based on the evidence of the LP Filter PCB buy I'm running there is a lot of interest in DSD playback and a more 'sorted' project is going to be a natural follow on from the very simple approach.
To be honest Acko, I don't have the technical savvy to fully understand the performance graph on the link you posted and I rely on standing on the shoulders of giants, such as yourself. I do like the sounds produced though! I am familiar with the material on the link. One thing I did notice is, referring to the 'schematic' at the vey bottom of the page, the LP filter specified (R=3K3, C=3.3nF) will have a -3dB point at only about 15KHz - does that matter in this context?
Regarding the white noise I was experiencing with the HQP volume control engaged, here's what Jussi had to say;
http://www.computeraudiophile.com/f11-software/hq-player-20293/index190.html#post510464
http://www.computeraudiophile.com/f11-software/hq-player-20293/index190.html#post510578
On the simple LP filter front, I've acquired a pair of Silk STA522A TVCs. I plan to experiment with using them as a combined LP filter/volume control. Their bandwidth falls away quickly above around 35KHz. Thoughts?
SAC Thailand
Cheers
Ray
Last edited:
Cheers Ray, I will have a look the details when I get a bit more time.
Question: How do plan to address the high energy noise that you can't hear?
STA522As are fantastic but not cheap so it adds to the overall costs. For all you know user may just use a simple LPF instead to save on costs and use headphones instead and in the process create a "dangerous prototype"
@acko: .... this sounds indeed very interesting Acko ... looking forward to hearing more about this in due time ;-)
One comment in relation to the link to Andrea Ciuffoli's webpages that you refer to. It looks as if the DSD rate is DSD128 and I reckon that both the ultra sonic noise and the distortion could drop if e.g. DSD512 were the playback rate ... Dont know how much, though ... (Maybe others have experiences here? My computer unfortunately barely manages DSD128 from HQPlayer and much more noise on JRiver and Foobar)
Also, recently I made a similar DSD "DAC" with two potsemi flip-flops as the active devices and a very solid PSU. Here I noticed that the distortion is heavily output current dependent. If I remember correctly (and I am not sure my memory is entirely correct here) the sweet spot was about 1 mA - below that the THD could actually be quite low.
Cheers,
Jesper
One comment in relation to the link to Andrea Ciuffoli's webpages that you refer to. It looks as if the DSD rate is DSD128 and I reckon that both the ultra sonic noise and the distortion could drop if e.g. DSD512 were the playback rate ... Dont know how much, though ... (Maybe others have experiences here? My computer unfortunately barely manages DSD128 from HQPlayer and much more noise on JRiver and Foobar)
Also, recently I made a similar DSD "DAC" with two potsemi flip-flops as the active devices and a very solid PSU. Here I noticed that the distortion is heavily output current dependent. If I remember correctly (and I am not sure my memory is entirely correct here) the sweet spot was about 1 mA - below that the THD could actually be quite low.
Cheers,
Jesper