DS I2S input

IMG_2894.JPGMy primary intention was to see if the I2S input of the DS sounds better than the USB or AES input. So I got this little device Gustard U12 which basically is a USB converter with AES/I2S output. I am observing strange behavior when connecting the DS over I2S

  1. In Jriver, if I set output as 2 x Dop/Native, the interface seems to work fine.

  2. For PCM, there is a small icon that shows up on the DS screen at the top left below the settings icon (this icon is not seen when changing the Jriver output to Dop/Native as described above). The stereo channels seems to be reversed, presents a weird soundstage and highs are rolled off. It plays perfectly fine when connected over AES/EBU.

In both the cases, 192Khz bitperfect test fails over I2S (but works fine over AES). I have tested both the I2S input and both works the same way. My initial thought was maybe the I2S signals over HDMI interface is not compatible but then it doesn’t justify when playing using DoP/Native.

Any thoughts ?

Thanks in advance.

– Dev

The icon indicates that the current input is preemphasized. That flag is only read and displayed for 44.1k inputs. When it’s shown the deemphasis is engaged and you’ll hear a HF rolloff.

The other variables can be the polarities of the I2S connections. One indicates the channel, one the clock (which can cause a lot of noise if inverted) and the last the data which if inverted may not make much of a difference - e.g. DSD will have the opposite polarity but is otherwise fine. You can find pinouts for many I2S devices at http://www.i2s.sonore.us/

Thanks Ted. The link is very helpful.

PS Audio
PWD DSD
SD (-) / DSD DATA R (-)
GND
SD / DSD DATA R
BCK / DSD BCK
GND
BCK (-) / DSD BCK (-)
L/RCLK (-) / DSD DATA L (-)
GND
L/RCLK / DSD DATA L
NOT USED
GND
NOT USED
NOT USED
NOT USED
SCL - i2c
SDA - i2c
GND
NOT USED
GND
internal master clock
LVDS
HDMI
PS Audio Spec
Gustard
U12
SD / DSD DATA R
GND
SD (-) / DSD DATA R (-)
BCK / DSD BCK
GND
BCK (-) / DSD BCK (-)
L/RCLK / DSD DATA L
GND
L/RCLK (-) / DSD DATA L (-)
MCLK
GND
MCLK (-)
NOT USED
NOT USED
DSD detect
NOT USED
GND
5V
GND
1X or 2X
LVDS
HDMI
Gustard Spec
Looks like Gustard folks are really incompatible with the DS. Pin 1/3 and 7/9 are reversed. Also Pin 10/13 are MCLK (+/-) on the Gustard and NOT USED in PS. I am assuming 10/13 can be ignored on the DS ?

So if I am able to cook up a hdmi cable with inverted pin for 1/3 and 7/9 on one end, should that work theoretically ?

– Dev

Reversing those two pairs should be everything you need to do, but it doesn’t explain the preemphasys detect unless the “DSD detect” of the Gustard or some other of it’s “NOT USED” is being grounded or driven. That may or may not be a problem after the two swapped pairs are swapped back. I keep meaning to add an option in the DS software to allow configuration or the I2S input polarities (or using some ad hoc automatic detection) but I’ve not gotten it done yet.

Ted Smith said Reversing those two pairs should be everything you need to do, but it doesn't explain the preemphasys detect unless the "DSD detect" of the Gustard or some other of it's "NOT USED" is being grounded or driven. That may or may not be a problem after the two swapped pairs are swapped back.
I asked most of the hdmi cable manufacturer around and no one would terminate a custom one So I got a $5 radioshack cable and spliced it to revert the required pins. The channel swap and phase problem went away but the pre-emphasys still remained and highs are rolled off. As you indicated, there is some other lines driven/grounded. Is there any way to find this or a trial and error method is the only way ?
I keep meaning to add an option in the DS software to allow configuration or the I2S input polarities (or using some ad hoc automatic detection) but I've not gotten it done yet.
That would be immensely helpful and would make the DS virtually compatible with any I2S DDC.

It’s the SCL and/or SDA lines (I2C lines), but I don’t know what they should be to not indicate preemphasys - perhaps someone from PS Audio can clarify that.

Ted Smith said It's the SCL and/or SDA lines (I2C lines), but I don't know what they should be to not indicate preemphasys - perhaps someone from PS Audio can clarify that.
From the chart it looks like,

SCL (PS Audio) → DSD Detect (Gustard)

SDA (PS Audio) —> Not Used (Gustard)

I am not sure what the Pin definitions for SCL/SDA are but the Pin DSD Detect in Gustard is high whenever a DSD stream is detected. As for an experiment, maybe I can cut the Pin 15 (DSD detect) to see if that works but hdmi cabling is pretty complex and finding that pin is a pain.

What is confusing is that the preemphasys is only indicated when I am playing PCM and NOT when playing DSD. This should mean that the Gustard DSD Detect Pin is not high when playing PCM. I don’t have a way to ascertain it but that what it seems logically. Interestingly, the sonore rendu which is compatible with the PS Audio spec, both of these Pins are “NOT USED” (so how does it make it compatible ?).

Best,

– Dev

Preemphasis is only used for 44.1k. It’s ignored at any other sample rate. PS Audio DACs are designed so that an undriven I2C inputs indicates no preemphasis.

Got the pin definition for SCL/SDA. Its part of hdmi specs

SCL - DDC Clock

SDA - DDC Data

"

HDMI has three physically separate communication channels, which are the DDC, TMDS and the optional CEC. HDMI 1.4 added ARC and HEC.

DDC

The Display Data Channel (DDC) is a communication channel based on the I²C bus specification. HDMI specifically requires support for the Enhanced Display Data Channel (E-DDC), which is used by the HDMI source device to read the E-EDID data from the HDMI sink device to learn what audio/video formats it supports.

"

Is the DS expecting on SDA whenever SCL is driven ? How is this related to preemphasys ? From the definition of DDC, it looks like it could be useful for the PS Audio Transport but may not be really needed for other DDC.

Ted Smith said Preemphasis is only used for 44.1k. It's ignored at any other sample rate. PS Audio DACs are designed so that an undriven I2C inputs indicates no preemphasis.
It is also indicated with 96 and 192 samples. I just played the 192 Bitperfect test - it indicated Bitperfect as well as preemphasys. Is that even possible ?

The preemphais indicator flag shows up at times when I wouldn’t expect it to (i.e. I believe we have a bug in the display/control software), but it’s only reported to the user interface software for 44.1k PCM and the flag is explicitly ignored in the FPGA for DSD.

Nothing should be inferred from any spec’s about HDMI - this isn’t HDMI - it’s solely a way of passing I2S over a quality cable that’s widely available.

The lines labeled for I2C over HDMI are used between PS Audio products to signal preemphasis - they shouldn’t be used for any other purpose with PS Audio hardware.

Thanks Ted. I understand but can you iterate what exactly is meant by preemphasys and its effect on the sound quality ? Is it expected that the highs are rolled off whenever the preamphasys is asserted ?

I2C SCL = 0 Clock

I2C SDA = 1 Data

This is the Marker for preemphasis. This is because the Transport uses these signals to signal preemphasis.

Dennis

When CD first came out the state of the art was about 14 bits for ADC and DACs. If the source preemphsized and the player deemphasized you could get less sensitivity to quantization noise with those CDs. This hasn’t been a problem for decades.

If you have a source that is preemphasized and you don’t use a deemphasis filter on it you will accentuate the higher frequencies.

If you have a source that isn’t preemphasized and you use a deemphasis filter on it you will attenuate the higher frequencies.

These days the CD pre/de emphasis flag is often lost somewhere in the ripping, storing, playing chain and pre/de emphasis causes more headaches than it solves. With a PS Audio transport and a PS Audio DAC and any of I2S, TOSLink, S/PDIF and/or AES/EBU connecting them you get a system that handles pre/de emphasis as it was originally designed.

Ted Smith said When CD first came out the state of the art was about 14 bits for ADC and DACs. If the source preemphsized and the player deemphasized you could get less sensitivity to quantization noise with those CDs.
Similar to Dolby, both now seem quaint.
Dennis Kerrisk said I2C SCL = 0 Clock

I2C SDA = 1 Data

This is the Marker for preemphasis. This is because the Transport uses these signals to signal preemphasis.

Dennis


Thanks Dennis. Would 0 and 1 mean that the pins are not driven and driven respectively and this would be the only combination that would mark preemphasis in DS ? Would it work if I disconnect both the SCL and SDA wires in the cable ?

Ted Smith said When CD first came out the state of the art was about 14 bits for ADC and DACs. If the source preemphsized and the player deemphasized you could get less sensitivity to quantization noise with those CDs. This hasn't been a problem for decades.

If you have a source that is preemphasized and you don’t use a deemphasis filter on it you will accentuate the higher frequencies.

If you have a source that isn’t preemphasized and you use a deemphasis filter on it you will attenuate the higher frequencies.

These days the CD pre/de emphasis flag is often lost somewhere in the ripping, storing, playing chain and pre/de emphasis causes more headaches than it solves. With a PS Audio transport and a PS Audio DAC and any of I2S, TOSLink, S/PDIF and/or AES/EBU connecting them you get a system that handles pre/de emphasis as it was originally designed.


Ahh got it. Thanks for the explanation.

Dennis Kerrisk said I2C SCL = 0 Clock

I2C SDA = 1 Data

This is the Marker for preemphasis. This is because the Transport uses these signals to signal preemphasis.

Dennis

Thanks Dennis. Would 0 and 1 mean that the pins are not driven and driven respectively and this would be the _only_ combination that would mark preemphasis in DS ? Would it work if I disconnect both the SCL and SDA wires in the cable ?
1 is driven to a logic Hi.
0 is driven to a logic lo.
It should work with both pins disconnected.
Dennis

Here is the update - after the disconnecting the SCL/SCA pins, the preemphasys is gone and high are no more rolled off music-078_gifI am A/B’ing the I2S to the AES/EBU connection and apparently after a brief listening, both sounds superb with the edge over the AES/EBU. This could be contributed to the cable - the AES is Straightwire whereas the I2S is a $5 Radioshack one. I still feel that I2S has more potential and with a decent short run, it might trump the AES. Moreover, I am now inclined to return this incompatible Gustard (I unknowingly bought since it was cheap) and get a different DDC w/ compatible I2S PS audio spec. I could then have more choice to buy standard hdmi cables.

Since PS stopped making the hdmi cables, any other recommendations ?

The guys on computeraudiophile.com have mentioned this thing called Hydra-Z. Not sure if it meets your desired use case, but it looks interesting. DSD 2x over I2S. I’m curious.