Study: Is I²S interface better for DACs than S/PDIF or USB?

[asrUser]

Some motherboards have a pin header for spdif output (optical and/or coax, depending on the output adapter). If you don't have that there are inexpensive USB to toslink options like the Hifime UT23 doing up to 24 bit 96 kHz.

I'd be ready to pay more such as the https://hifimediy.com/product/s2-digi/
However I don't know if Hifime makes good quality. See amirm's review of another device: https://www.audiosciencereview.com/...and-measurements-of-hifime-uda38pro-dac.4064/

I found something like this from Thomann https://www.thomannmusic.com/swissonic_hdmi_2.0_audio_extractor.htm?reload=1 which uses HDMI as source. Is there some other higher quality alternative to the Hifime?

 

[somebodyelse]

Minidsp do things like the USBStreamer. Thomman's UA-202 is a cheap option with transformer isolation on its coax output.

 

[CleanSound]

Today, the most prevalent USB bridge chip for DACs are the XMOS chips. They have a limitation on up to 512 DSD, I am not sure if that is the limitation of USB or the XMOS. But I^2S enables you to transfer up to 1024 DSD. But in all practicality, there aren't any 1024 DSD content and even if there are, I don't believe it's audible.

 

[asrUser]

Minidsp do things like the USBStreamer. Thomman's UA-202 is a cheap option with transformer isolation on its coax output.

USBStreamer isn't in stock anymore and not in stock elsewhere. The US-202 got a poor internal DAC. Will this be bypassed, if I only use the optical output?

 

[Roland68]

Is the optical input better than USB in terms of Galvanic isolation? Though I would need an additional DAC just for the optical output, because my PC got no optical output.

What do you need galvanic isolation for?
With the newer DACs with XMOS chips, such as the XU216 and XU316, this is usually unnecessary.

The only USB isolator that really works for me is the Intona USB 2.0 2.5kV. Alternatively, DDCs, e.g. from Xingser or Gustard (U18).
But as I said, unnecessary with the newer DACs.

 

[somebodyelse]

With the newer DACs with XMOS chips, such as the XU216 and XU316, this is usually unnecessary.

The usual reason for needing galvanic isolation is to avoid ground-related noise such as ground loops or modulated leakage currents (eg. when moving the mouse). Usually this is with single ended outputs - balanced outputs should be immune if correctly designed. This is an analogue phenomenon, unrelated to the USB receiver model. Having said that, if there is no audible issue there's no need for isolation.

 

[somebodyelse]

USBStreamer isn't in stock anymore and not in stock elsewhere. The US-202 got a poor internal DAC. Will this be bypassed, if I only use the optical output?

IIRC there's also the MCHstreamer and more expensive UDIO. The UA-202 has a transformer isolated coax output, not an optical output. It outputs the digital data as received via USB. From that review:

SPDIF output works fine.

You could also look at the Topping D10s but I'm not sure if its coax output is isolated.

 

[ThatGuyYouKnow]

Seeing a DAC I otherwise like with an I2S input sort of angers me. It is just taking up space and development cost that could have went towards adding another useful USB input. USB audio is plenty matured and flawless with most modern DACs.

 

[antcollinet]

Will this be bypassed, if I only use the optical output?

yes - if you are outputting digital, the DAC (whose function is to convert from digital to analogue) is not used.

 

[Roland68]

The usual reason for needing galvanic isolation is to avoid ground-related noise such as ground loops or modulated leakage currents (eg. when moving the mouse). Usually this is with single ended outputs - balanced outputs should be immune if correctly designed. This is an analogue phenomenon, unrelated to the USB receiver model. Having said that, if there is no audible issue there's no need for isolation.

This has nothing to do with single-ended or balanced outputs. Balanced/XLR connections only help with interference in the cable. If the interference is already present in the DAC, it will be transmitted in the same way as the audio signal.

I have never encountered such problems with MACs and high-quality PCs, such as HP business devices.
To me this always indicates poorly designed or configured hardware. In such a case, it always makes more sense to eliminate the cause than to tamper with the effects.

However, for a DAC developer, the isolation of the 5 volt USB voltage in front of the XMOS chip and the galvanic isolation afterwards should be part of the basics.

 

[antcollinet]

This has nothing to do with single-ended or balanced outputs. Balanced/XLR connections only help with interference in the cable. If the interference is already present in the DAC, it will be transmitted in the same way as the audio signal.

I have never encountered such problems with MACs and high-quality PCs, such as HP business devices.
To me this always indicates poorly designed or configured hardware. In such a case, it always makes more sense to eliminate the cause than to tamper with the effects.

However, for a DAC developer, the isolation of the 5 volt USB voltage in front of the XMOS chip and the galvanic isolation afterwards should be part of the basics.

I'm not sure what you are saying here.

With a ground loop, the noise is not already on the signal in the DAC. It gets added to the signal at the receiving end of an unbalanced analogue interconnect due to the ground potential difference between the sending device and the receiving device, which is caused by the ground current and the impedance of the ground connection.

With a balanced interconnect this noise is still added to both signals but cancels out if the impedances are balanced.

However, breaking of the ground loop using galvanic isolation somewhere downstream of the ground current source is a far better solution. Toslink is a very effective and easy way to achieve this.

Ground loops are not a component problem, they are a system problem. Though some components are better at creating ground currents than others.

 

[asrUser]

What do you need galvanic isolation for?

To get rid of ground loops. Once I recognized some ground loop when using minidsp Flex.

You could also look at the Topping D10s but I'm not sure if its coax output is isolated.

I just want to use Toslink and will probably get one of those because they look good.

 

[Roland68]

I'm not sure what you are saying here.

With a ground loop, the noise is not already on the signal in the DAC. It gets added to the signal at the receiving end of an unbalanced analogue interconnect due to the ground potential difference between the sending device and the receiving device, which is caused by the ground current and the impedance of the ground connection.

With a balanced interconnect this noise is still added to both signals but cancels out if the impedances are balanced.

However, breaking of the ground loop using galvanic isolation somewhere downstream of the ground current source is a far better solution. Toslink is a very effective and easy way to achieve this.

Ground loops are not a component problem, they are a system problem. Though some components are better at creating ground currents than others.

I only reported on experiences with the appearance of USB 2.0. This only affects problems with hum loops that exist between the computer and the DAC and could also be identified as the cause there.
In most cases, both outputs, SE and Balanced, were affected. In these cases, the PCs were always cheap consumer or self-built devices.
However, these problems became less common with DACs with XMOS chips after the XU208 generation. This may be mainly due to the developers' better support for power supply and galvanic isolation. XU208 and other USB solutions for audio were often powered from the 5 volt USB cable.
With MACs and business devices in the pro audio and hi-fi sector, I only had the problem with a hum loop in one of well over 50 different DACs.
This is striking to me, but it doesn't necessarily correspond to other experiences.
But I haven't had these problems with ground loops in my hi-fi system (ok, twice in 30 years), despite many different devices.

However, this experience corresponds to that from my IT job in the area of high availability hardware. When I started, this employer used inexpensive PCs with many special devices for development, measurements, calculations, print systems, etc., most of which were connected via USB. Lots of hardware issues and failures, lots of USB issues and lots of Windows/OS issues. After replacing all devices with HP business devices, all of these problems, especially with USB, were eliminated and were now the exception and not the rule. An IT team of 12 people for around 1000 devices could be reduced to 4 and concentrate on important tasks.

 

[wgh52]

Hello folks,

not sure this is the appropriate place to get advice for my case, but please let know

OK. I'm building a system with a DSP board and DAC board, located in the same system box next to each other.

Now I have the choice of either using CMOS Level SPDIF or I²S to connect the two. Connection length is probably 12cm or 5 inch max.

To implement the I²S I'd have to make a modification to the DAC to accept the I²S signal, the connection with the SPDIF coded CMOS level signal woould be simpler to implement.

So is it worth the hassle to change the DAC from async clock mode for SPDIF to sync and slave clock mode?

Thanks and regards,
Winfried

 

[Roland68]

Seeing a DAC I otherwise like with an I2S input sort of angers me. It is just taking up space and development cost that could have went towards adding another useful USB input. USB audio is plenty matured and flawless with most modern DACs.

i2s is already in the DAC anyway. Connecting it externally costs less than 5% of the development time for a USB interface and a few cents in components. No comparison with the implementation and programming of a USB interface with XMOS.
But you can always add another USB interface with the i2s connection, e.g. with the SMSL PO100 Pro, with an Amanero, Xing Audio or XMOS board, other DDCs, etc.

 

[Roland68]

Hello folks,

not sure this is the appropriate place to get advice for my case, but please let know

OK. I'm building a system with a DSP board and DAC board, located in the same system box next to each other.

Now I have the choice of either using CMOS Level SPDIF or I²S to connect the two. Connection length is probably 12cm or 5 inch max.

To implement the I²S I'd have to make a modification to the DAC to accept the I²S signal, the connection with the SPDIF coded CMOS level signal woould be simpler to implement.

So is it worth the hassle to change the DAC from async clock mode for SPDIF to sync and slave clock mode?

Thanks and regards,
Winfried

12 cm can be limiting for i2s, especially if they still have interference from various components.

 

[wgh52]

Hello Roland,

thanks for responding! Yes, I am aware of the potential signal line length issue, just never implemented I²S so far. Which also is the reason I'm asking for advice. My actual plan was (and is) to try the SPDIF CMOS Level connection first, just wanted to evaluate if I should even attempt to try out the internal I²S.

Kind greetings,
Winfried

 

[Roland68]

To get rid of ground loops. Once I recognized some ground loop when using minidsp Flex.

As I said, I am always in favor of eliminating the basic problem rather than creating an elaborate workaround to mitigate or eliminate a problem that previously arose.
Take a closer look at the device to which you connected the DAC or minidsp Flex.
In my circle of friends, the minidsp Flex is often used and such problems are unknown.

 

[antcollinet]

In my circle of friends, the minidsp Flex is often used and such problems are unknown.

As I pointed out, ground loops are a system problem. If a system incorporating a Flex creates a ground loop. (Grounded kit used both upstream and downstream of the flex - for example, a PC upstream, and a Sub or grounded amp downstream) AND there is a source of ground current (eg one of the bits of kit filtering interference to ground, or magnetic coupling of mains frequency into the loop) then the ground current will flow around the loop, including through ground on the flex inputs and outputs.

This has nothing to do with weakness in the flex itself (any more than any other kit which doesn't provide galvanic isolation).

I'd agree that solving this at source is the correct solution. However the source of a ground loop problem is first the loop itself, then the source of the ground currents. Eliminating the ground loop (by introducing galvanic isolation) is often easier than preventing the ground currents in a loop that exists. If neither of these is possible then a workaround for the ground loop is balanced interconnect.

 

[Jimster480]

Optical is the only real isolation that I know of.
The rest of them require expensive connectors.