I am playing with a MiniSharc board as a benchmark for my own DSP processor.
When a run SPDIF from my server to the Minisharc with no processing and back out to SPDIF to my DAC I think I hear a subtle degradation in sound, especially in vocals. I will confirm with a blind test as I am doing the switching myself but assuming this result holds, why?
1. Could it be jitter introduced in the Minisharc?
2. Is it the ASRC step from Redbook to 48Khz?
3. Is it my DAC (Denafrips AresII) is not as good with 48khz?
Is it all of the above?
Is any of the above?
Is the answer knowable?
Thanks, jd
When a run SPDIF from my server to the Minisharc with no processing and back out to SPDIF to my DAC I think I hear a subtle degradation in sound, especially in vocals. I will confirm with a blind test as I am doing the switching myself but assuming this result holds, why?
1. Could it be jitter introduced in the Minisharc?
2. Is it the ASRC step from Redbook to 48Khz?
3. Is it my DAC (Denafrips AresII) is not as good with 48khz?
Is it all of the above?
Is any of the above?
Is the answer knowable?
Thanks, jd
Jitter seems like a pretty good guess. Could be more than one problem.
in the past I looked at the specs for ASRC in a few different ADI DSP chips, including some Sharc. IIRC, those ASRC are rated around -120dB distortion. By way of comparison the best hardware ASRC I am aware of (AK4137, SRC4392) offer distortion levels down around -140dB or lower. In addition, my own listening tests show subtle yet audible differences in the sound of ASRC processing by AK4137 and SRC4392. Since I therefore do not consider their ASRC to be completely inaudible (under my listening test conditions), I most likely would not want to use a lesser ASRC at all (including Sharc's).
in the past I looked at the specs for ASRC in a few different ADI DSP chips, including some Sharc. IIRC, those ASRC are rated around -120dB distortion. By way of comparison the best hardware ASRC I am aware of (AK4137, SRC4392) offer distortion levels down around -140dB or lower. In addition, my own listening tests show subtle yet audible differences in the sound of ASRC processing by AK4137 and SRC4392. Since I therefore do not consider their ASRC to be completely inaudible (under my listening test conditions), I most likely would not want to use a lesser ASRC at all (including Sharc's).
Last edited:
Sorry Mark but jitter seems like a stab in the dark to me.
@JD: Have your tried as an experiment any HiRes material of known provenance so you are not upsampling. There would still be the ASRC in the way, but it would be interesting to see if what you think you hear is still there when 48 or 96 all the way through.
@JD: Have your tried as an experiment any HiRes material of known provenance so you are not upsampling. There would still be the ASRC in the way, but it would be interesting to see if what you think you hear is still there when 48 or 96 all the way through.
That is a great idea! Let me try that this weekend. I would probably need to download something. I don't have any 48 material currently.
I suspect this is a product of nonsyncronous ASRC inside SPDIF receivers combined with relatively bad phase stability of adjustable PLL-generators.
Two steps of SPDIF mostly anyway will kill bitperfect transmission or perfect timing even if one step is a PLL-locked receiver and second is a ASRC-receiver.
If my DAC only accepts SPDIF or USB would I2S all the way through DSP and one final SPDIF conversion be the way to go?
Heh, it's all very complicated.
Your USB-converter, DSP and SPDIF'ed DAC must be inside one masterclock domain.
So, the best way if DAC can output masterclock to external devices and each previous devices will be slave in timing domain while computer will be even asyncronous.
I doubt in this.
So, check which device in what time-domain around masterclock are.
I will be able to get my source and dsp on the same mclk but not the DAC. But the current source and DAC are not either and I am happy with sound.
So if I can get my source and DSP on the same mclk but let the DAC recover the clock from the final SPDIF signal would that get my DSP block "transparent"? I guess it all depends on the final SPDIF implementation too.
Yes, two time domains would be better than three time domains.
Clearly!
There are very much different SPDIF types with very different time and bits accuracy.
So if you can exclude some of them in a syncronous island in the chain - do this.
4. Possibly listening position change unless headphone was used.
5. Possibly listening level unmatched unless it was verified at the speaker terminal for each setup.
jderimig,
May I ask what dac you are using?
Also, depending on various factors using SPDIF could cause some audible degradation of sound quality. If you are using a computer as a USB source for audio playback then it should be possible to avoid any ASRC by using basically using the computer as a FIFO to bridge clock domains. Diyinhk makes a 4 channel playback and 4 channel record USB board for which an ASIO driver is available. However, haven't tested it to see if the ASIO driver works for recording too. Using it as an I2S source and destination for the DSP chip could allow processed audio to be returned to the computer where it could then be routed to the USB dac. There are now a few different audio routing programs for Windows that should make it easy to do. Obviously though, if going to all that trouble it would probably be easier and higher quality to do the DSP in the computer in the first place. Many people do that now if they have a music computer. There is lot more computing power in modern PC than in a typical audio DSP chip.
May I ask what dac you are using?
Also, depending on various factors using SPDIF could cause some audible degradation of sound quality. If you are using a computer as a USB source for audio playback then it should be possible to avoid any ASRC by using basically using the computer as a FIFO to bridge clock domains. Diyinhk makes a 4 channel playback and 4 channel record USB board for which an ASIO driver is available. However, haven't tested it to see if the ASIO driver works for recording too. Using it as an I2S source and destination for the DSP chip could allow processed audio to be returned to the computer where it could then be routed to the USB dac. There are now a few different audio routing programs for Windows that should make it easy to do. Obviously though, if going to all that trouble it would probably be easier and higher quality to do the DSP in the computer in the first place. Many people do that now if they have a music computer. There is lot more computing power in modern PC than in a typical audio DSP chip.
Last edited:
Okay, a stereo dac. Should we assume then the DSP you want to do is for something like room correction, and not for doing digital speaker crossovers?
Also, I am not aware of an Allo product with that name. Could it be a USBridge Sig? USBridge Sig - Ultra low noise RPI
Anyway, I believe you that bypassing mini Sharc sounds better. Not surprised, and wouldn't use it myself. Depending on what you are trying to accomplish, there is probably a better way.
EDIT: Almost forgot: is the RPi the source device for playback, or is it streaming content from some other system?
Also, I am not aware of an Allo product with that name. Could it be a USBridge Sig? USBridge Sig - Ultra low noise RPI
Anyway, I believe you that bypassing mini Sharc sounds better. Not surprised, and wouldn't use it myself. Depending on what you are trying to accomplish, there is probably a better way.
EDIT: Almost forgot: is the RPi the source device for playback, or is it streaming content from some other system?
Last edited:
Allo DigiOne (sorry left off a word) which is the source for playback.
My end goal is to use a ADAU1466 as my DSP engine but initially using the Minisharc as a benchmark. The first application will be Room Eq + experimentation with speaker crosstalk cancellation. A later project would be speaker crossover, probably an LXMini project.
I was using the Minisharc for very subtle room eq. I didn't like what I was hearing compared to the Minisharc bypassed. So I was wondering if it was the EQ I didn't like or something else. So I removed the EQ filters and played straight through and recognized what I didn't like WASN"T the EQ filters.
My end goal is to use a ADAU1466 as my DSP engine but initially using the Minisharc as a benchmark. The first application will be Room Eq + experimentation with speaker crosstalk cancellation. A later project would be speaker crossover, probably an LXMini project.
I was using the Minisharc for very subtle room eq. I didn't like what I was hearing compared to the Minisharc bypassed. So I was wondering if it was the EQ I didn't like or something else. So I removed the EQ filters and played straight through and recognized what I didn't like WASN"T the EQ filters.
Last edited:
The input was Redbook, Minisharc resamples to 48K to process and send to the DAC.
Currently I cannot send USB and SPDIF to the DAC using the same source. The Minisharc DigiFP only accepts SPDIF so I cannot compare A/B the Minisharc versus direct with anything other than SPDIF (now).
I am going to try higher def FLAC files as the source and see what the difference that has in the "difference".
Regarding USB, I was just trying to see if you ever determined the best your dac can sound under any circumstances. I would expect that might be with high quality DSD256 (or higher) from a PC over USB. If you can hear the sound of the minisharc and don't like it, can you also determine if you do or don't like SPDIF verses USB, and whether you do or don't like high sample rate DSD verses CD audio?
If you like all sorts of junk except you don't like the junk from minisharc, then that seems a little incongruous, but hey its for your listening pleasure I'm just trying to help. Sorry, don't mean to be offensive.
If you like all sorts of junk except you don't like the junk from minisharc, then that seems a little incongruous, but hey its for your listening pleasure I'm just trying to help. Sorry, don't mean to be offensive.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.