After completing four I2S projects on the BIII I still prefer SPDIF. How can I improve the SPDIF transmission?
System context: RedBook CDs burned onto a CAPS-style PC with SoTM filters (on fan and HDD). Logitech Media Server runs on the PC and provides uncompressed FLAC files over wifi to a Squeezebox Touch with Enhanced Digital Output (EDO) app. EDO buffer tuning is set to ‘Very Large + Randomize’.
The Buffalo III: The 4 Channel-SPDIF input card is configured for RCA SPDIF input. It is directly soldered to the Sidecar which is connected to the BIII via a very short ribbon cable. I2S is routed through the Sidecar. In Sidecar mode the SPDIF auto-detect is set to BYPASS. The BIII outputs to a Legato output stage. The BIII has the latest versions of the Tridents/AVCC. The FIR rolloff is set to SLOW. The DPLL bandwidth is set to LOWEST (best jitter rejection).
I have done four major experiments using I2S with the Buffalo III:
1. DIYINHK XMOS USB to I2S PCB (purchased Oct 2013) direct to BIII Sidecar
-PCB was placed in the BIII enclosure right against Sidecar
-PCB was powered from USB source
2. DIYINHK XMOS USB to I2S PCB via Teleporters to BIII Sidecar
-The PCB and Teleporter transmitter were powered from the USB source
-The Teleporter receiver was powered from BIII Sidecar
3. Raspberry Pi I2S output to BIII Sidecar
-Raspberry Pi was placed inside BIII enclosure, but configured so the USB wireless modem was sticking out of the back of the chassis.
-Picoreplayer was used as the Logitech Media Server client player.
4. DIYINHK Isolated XMOS USB to I2S PCB (purchased Feb 2015)
-An I2S port was built on the chassis top of the BIII enclosure, right above the Sidecar, with short I2S lines, to allow plug-ability for future external modules.
-The DIYINHK PCB was connected to the BIII in this way, on top of, not inside the enclosure. This was to minimize potential USB interference with the BIII (which might have negated the isolation feature of the PCB).
-The dirty side of the PCB was powered by the USB source.
-The clean side of the PCB was powered by an older Trident, which in turn was connected to the BIII 5.25v digital power supply.
In all the experiments above the sound quality was very, very good, with the best of the group being #4. But I always compared these back to SPDIF. With the Squeezebox Touch (in experiments 1, 2 and 4) it was very easy to switch back and forth and compare USB vs SPDIF. In the case of the Raspberry Pi, it was a little more complicated, because I always took the modem out or completely powered down the Pi when comparing to SPDIF.
In all cases I preferred the SPDIF. While the I2S connections may have provided just a touch more clarity, I always went back to SPDIF as it provided a more open soundstage and more engaging presentation. I must say I was surprised by this unexpected result.
So, my first question is: Has anyone experienced this?
My second question is: What can I do to maximize the quality of my SPDIF signal chain?
The obvious one for me would be a mod on the Squeezebox Touch to bring out a proper 75ohm cable terminated at the BIII end with a ‘real 75ohm’ BNC plug into a ‘real 75ohm’ BNC jack on the BIII.
Other ideas that have occurred to me:
- remove the 4-channel SPDIF card and the Sidecar, and use a single-input SPDIF card straight into the BIII. (reducing the number of connections, junctions, etc)
- Remove just the Sidecar (same reasoning as above)
- Upgrade the signal isolation transformers in the 4-channel SPDIF with the very latest, premium quality transformers. (do these exist?)
- Other potential mods on the 4-channel SPDIF card (??)
- Modify the Squeezebox SPDIF output to provide an AES/EBU signal, and set up the 4-channel SPDIF card to receive that. (A lot of work! Would this provide such a big jump over properly implemented BNC?)
I would appreciate any comments on these ideas, or other suggestions to improve the SPDIF input on my Buffalo III.
- Steve
System context: RedBook CDs burned onto a CAPS-style PC with SoTM filters (on fan and HDD). Logitech Media Server runs on the PC and provides uncompressed FLAC files over wifi to a Squeezebox Touch with Enhanced Digital Output (EDO) app. EDO buffer tuning is set to ‘Very Large + Randomize’.
The Buffalo III: The 4 Channel-SPDIF input card is configured for RCA SPDIF input. It is directly soldered to the Sidecar which is connected to the BIII via a very short ribbon cable. I2S is routed through the Sidecar. In Sidecar mode the SPDIF auto-detect is set to BYPASS. The BIII outputs to a Legato output stage. The BIII has the latest versions of the Tridents/AVCC. The FIR rolloff is set to SLOW. The DPLL bandwidth is set to LOWEST (best jitter rejection).
I have done four major experiments using I2S with the Buffalo III:
1. DIYINHK XMOS USB to I2S PCB (purchased Oct 2013) direct to BIII Sidecar
-PCB was placed in the BIII enclosure right against Sidecar
-PCB was powered from USB source
2. DIYINHK XMOS USB to I2S PCB via Teleporters to BIII Sidecar
-The PCB and Teleporter transmitter were powered from the USB source
-The Teleporter receiver was powered from BIII Sidecar
3. Raspberry Pi I2S output to BIII Sidecar
-Raspberry Pi was placed inside BIII enclosure, but configured so the USB wireless modem was sticking out of the back of the chassis.
-Picoreplayer was used as the Logitech Media Server client player.
4. DIYINHK Isolated XMOS USB to I2S PCB (purchased Feb 2015)
-An I2S port was built on the chassis top of the BIII enclosure, right above the Sidecar, with short I2S lines, to allow plug-ability for future external modules.
-The DIYINHK PCB was connected to the BIII in this way, on top of, not inside the enclosure. This was to minimize potential USB interference with the BIII (which might have negated the isolation feature of the PCB).
-The dirty side of the PCB was powered by the USB source.
-The clean side of the PCB was powered by an older Trident, which in turn was connected to the BIII 5.25v digital power supply.
In all the experiments above the sound quality was very, very good, with the best of the group being #4. But I always compared these back to SPDIF. With the Squeezebox Touch (in experiments 1, 2 and 4) it was very easy to switch back and forth and compare USB vs SPDIF. In the case of the Raspberry Pi, it was a little more complicated, because I always took the modem out or completely powered down the Pi when comparing to SPDIF.
In all cases I preferred the SPDIF. While the I2S connections may have provided just a touch more clarity, I always went back to SPDIF as it provided a more open soundstage and more engaging presentation. I must say I was surprised by this unexpected result.
So, my first question is: Has anyone experienced this?
My second question is: What can I do to maximize the quality of my SPDIF signal chain?
The obvious one for me would be a mod on the Squeezebox Touch to bring out a proper 75ohm cable terminated at the BIII end with a ‘real 75ohm’ BNC plug into a ‘real 75ohm’ BNC jack on the BIII.
Other ideas that have occurred to me:
- remove the 4-channel SPDIF card and the Sidecar, and use a single-input SPDIF card straight into the BIII. (reducing the number of connections, junctions, etc)
- Remove just the Sidecar (same reasoning as above)
- Upgrade the signal isolation transformers in the 4-channel SPDIF with the very latest, premium quality transformers. (do these exist?)
- Other potential mods on the 4-channel SPDIF card (??)
- Modify the Squeezebox SPDIF output to provide an AES/EBU signal, and set up the 4-channel SPDIF card to receive that. (A lot of work! Would this provide such a big jump over properly implemented BNC?)
I would appreciate any comments on these ideas, or other suggestions to improve the SPDIF input on my Buffalo III.
- Steve
I don't have much experience with AES/EBU to form an opinion of my own, only what I've read. Do the BNC connector on the Touch.
Myself, I prefer I2S over SPDIF with my dual mono BIII. I get more clarity, better bass without losing soundstaging. It all comes to down to personal preference, there are no wrong answers when you pick what you like.
I modded my Touch for I2S out using Teleporters to send/recieve. Did the same with a Denon 3910 to output DSD and I2S. There are some decent tutorials out there for opening up the Touch.
Myself, I prefer I2S over SPDIF with my dual mono BIII. I get more clarity, better bass without losing soundstaging. It all comes to down to personal preference, there are no wrong answers when you pick what you like.
I modded my Touch for I2S out using Teleporters to send/recieve. Did the same with a Denon 3910 to output DSD and I2S. There are some decent tutorials out there for opening up the Touch.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
Thanks, Rich. I once made an attempt at tapping I2S out of the Touch, but it was beyond my skill level. I can see that would be a good way to go.
The DIYINHK pcbs are of very high quality, and I would certainly be happy over the long term with their latest. Via I2S I agree on the clarity and the bass. But there's just something about that SPDIF sound that intrigues me. This is getting over my head, but as I understand it the SPDIF going into the Sabre is decoded directly, without conversion to I2S. See the following:
http://www.esstech.com/PDF/sabrewp.pdf
"A. SPDIF Interface
The SPDIF interface is more complex than the DSD and I2
S since it must first derive the embedded clock in the bi-phase
encoded data. In fact, experience with many forms of SPDIF
decoder suggest that most fail in the presence of high jitter due
to the lack of robustness in the clock recovery process. To
avoid this potential problem the Sabre SPDIF interface avoids
having to extract the clock at all: decoding is done using a
method that does not require an explicit measure of the clock
frequency. Specifically, the digital input is first corrected for
50% duty cycle by means of a discrete digital delay line that is
able to delay either the positive edge or the negative edge of
the signal such that after this delay line the signal is at 50%
duty cycle. Thereafter an assessment is made of the width of
each pulse based on its relation to recently seen pulse widths
and a decision circuit assigns each a width of 1, 2, or 3 units.
A state machine then operates on the assigned widths in
succession; this state machine is searching for the block
boundaries and the bit states. The state machine makes no
attempt to re-time or otherwise decode the clock – it simply
“time stamps” the event and passes it to the downstream
processor. Using this method the SPDIF interface is able to
accommodate 50nS of random jitter and 200nS of sinusoidal
jitter in the incoming data."
I don't have the technical background to know whether this is a unique Sabre thing, or whether many DACs do it this way.
What ever it is, I find the sound laid back and lovely.
The DIYINHK pcbs are of very high quality, and I would certainly be happy over the long term with their latest. Via I2S I agree on the clarity and the bass. But there's just something about that SPDIF sound that intrigues me. This is getting over my head, but as I understand it the SPDIF going into the Sabre is decoded directly, without conversion to I2S. See the following:
http://www.esstech.com/PDF/sabrewp.pdf
"A. SPDIF Interface
The SPDIF interface is more complex than the DSD and I2
S since it must first derive the embedded clock in the bi-phase
encoded data. In fact, experience with many forms of SPDIF
decoder suggest that most fail in the presence of high jitter due
to the lack of robustness in the clock recovery process. To
avoid this potential problem the Sabre SPDIF interface avoids
having to extract the clock at all: decoding is done using a
method that does not require an explicit measure of the clock
frequency. Specifically, the digital input is first corrected for
50% duty cycle by means of a discrete digital delay line that is
able to delay either the positive edge or the negative edge of
the signal such that after this delay line the signal is at 50%
duty cycle. Thereafter an assessment is made of the width of
each pulse based on its relation to recently seen pulse widths
and a decision circuit assigns each a width of 1, 2, or 3 units.
A state machine then operates on the assigned widths in
succession; this state machine is searching for the block
boundaries and the bit states. The state machine makes no
attempt to re-time or otherwise decode the clock – it simply
“time stamps” the event and passes it to the downstream
processor. Using this method the SPDIF interface is able to
accommodate 50nS of random jitter and 200nS of sinusoidal
jitter in the incoming data."
I don't have the technical background to know whether this is a unique Sabre thing, or whether many DACs do it this way.
What ever it is, I find the sound laid back and lovely.
It's proprietary to ESS Sabre. What it basically means is you don't get a huge improvement with I2S versus SPDIF with the Sabre like with some DAC chips.
I'm getting old so a lighted magnifier helped. A steady hand with small tip iron is essential too.
I use the exaU2I with the pc-part of the usb reciever powered and grounded to its own 5v PSU and disconnected from PC usb, but connected to the star ground I made for the dac implementation. This proved quite a substantial upgrade for me. I guess the PC having a different ground reference than the DAC might have been the biggest culprit.
Just my 2cents but moving from SPDIF to I2S and incorporating the chronus reclocking has resulted in the best digital sound ever. Adding teleporters allows locational flexibility without harm to the sound.
As to your question try adding a Bybee bullet at the receiving end of the spdif. i also prefer true 75 ohm bnc connections instead of rca or even balanced aesebu.
As to your question try adding a Bybee bullet at the receiving end of the spdif. i also prefer true 75 ohm bnc connections instead of rca or even balanced aesebu.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.