What is the advantage of I2S over USB?


#1

I thought USB had the advantage of the computer being clock locked with the dac. I2S can be set up that way too (I believe) but if the player’s clock is controlling the dac isn’t that a lesser desired situation?


#2

The DS controls the clock in such a way as to make all inputs look like asynchronous USB. The real differences when using the DS are groundloops and RFI/EMI, etc. These are quite system specific. I2S cables often have more solid grounds and better shielding than USB cables which can help both noise from groundloops and RFI… USB also has the downside of the signals being right next to the power and ground lines for a long distance which pollutes the USB power connection.

Still, like most things in audio, the real differences will depend on the components in question and the system they are put in.


#3
Ted Smith said The DS controls the clock in such a way as to make all inputs look like asynchronous USB. The real differences when using the DS are groundloops and RFI/EMI, etc. These are quite system specific. I2S cables often have more solid grounds and better shielding than USB cables which can help both noise from groundloops and RFI... USB also has the downside of the signals being right next to the power and ground lines for a long distance which pollutes the USB power connection.

Still, like most things in audio, the real differences will depend on the components in question and the system they are put in.


Ted,

Would the AES connection have the same or similar advantages as I2S?

Karl


#4

AES is balanced which can be an advantage over S/PDIF. I2S is also balanced in that it has differential signals. In general I2S was only designed for short runs and has a high bit rate. AES can be reliable over long cables. I2S can do, say, 352.8k PCM or double rate DSD and AES/EBU will probably only do 192k or single rate DSD.

Still no other input beats the universality of USB or it’s bandwidth. In some systems it’s just as good audio quality as the other inputs.

I should also mention that TOSLink can beat them all just because it’s galvanically isolated and it can also at times do 192k and single rate DSD with good cables and good quality transceivers.

When jitter is less of a problem (as it is with the DS) the worst evils of TOSLink, AES/EBU and S/PDIF are ameliorated and the potential improvements of I2S are smaller.


#5
Ted Smith said The DS controls the clock in such a way as to make all inputs look like asynchronous USB. The real differences when using the DS are groundloops and RFI/EMI, etc. These are quite system specific. I2S cables often have more solid grounds and better shielding than USB cables which can help both noise from groundloops and RFI... USB also has the downside of the signals being right next to the power and ground lines for a long distance which pollutes the USB power connection.

Still, like most things in audio, the real differences will depend on the components in question and the system they are put in.


Actually this brings up a question: I’ve found an unpowered usb hub that lets you switch off the power. If I use this with the DS, wouldn’t there be a benefit because you can cut the power line?

#6

If it works :slight_smile: Whether such devices work depends not only on the DS, but on the specific HUBs, PS hardware, OS and drivers. It’s worth trying: things like this work for some systems and not for others.


#7

Seems like the bottom line is, roll the dice and listen what works best - if you can even tell. Precisely what I am planning to do with my I2S source currently on order.


#8
Ted Smith said USB also has the downside of the signals being right next to the power and ground lines for a long distance which pollutes the USB power connection.
So I can assume that if my server has a USB out that is power free and only includes the file information, then there is no issue in relation to interference from power or ground lines.

John


#9

There has to be a ground signal. The power signal is also used to tell when a device (vs, say a hub) is connected - if your system works without the USB power connected then you may be avoiding some noise pollution at the USB device - but you will almost certainly suffer some reliability problems, especially when booting either device or plugging or unplugging a USB cable. Get your system working reliably before experimenting with cables, or other devices that disconnect the USB power.


#10
Ted Smith said There has to be a ground signal. The power signal is also used to tell when a device (vs, say a hub) is connected - if your system works without the USB power connected then you may be avoiding some noise pollution at the USB device - but you will almost certainly suffer some reliability problems, especially when booting either device or plugging or unplugging a USB cable. Get your system working reliably before experimenting with cables, or other devices that disconnect the USB power.
Thanks Ted for the clarification re the ground signal not that I know what it means. My server has 2 X USB only out. One powered and one not. Supposedly the non powered out is better for the SQ. I cannot give a definitive comment as to which is better as both seem excellent to me. That being said when I got the server 6 months ago I initially had some reliability problems. The DS stopped playing whereas the server was still going. I had to re-intialise the DS and reconnect via a PC. For some unknown reason that problem has now disappeared entirely. The only two changes that happened that may have contributed to the change was that I put in a much better spike in the ground for my system earth and the release of the wonderful PP. The USB cable that I use now is the same as when thing were unreliable. I did have one cheap printer cable that never failed. To me the improvement is a mystery.

I use the non powered USB and have not had any operational problems for a some time. If I unplugg I have to reconnect via a PC. I do not do that often.

John


#11

Ted … would a powered USB hub connected to a linear power supply address the USB noise issue?


#12

In fact I’ve been advocating that for a while here :)

It will help compared to the typical noisy USB supplies in computers and the wall wart USB hub supplies. You only want the DS on the USB hub with a linear supply and you want to use a shorter USB cable there (6’ or whatever, not 15’.)


#13

Finally got my HDPLEX LPS. Connected the 12V output to my Intel NUC and the 5V output to a powered USB hub. JCAT USB cable from hub to Directstream.

Got a significant improvement in SQ: the system is now very quiet, blacker background and enhanced sound stage. Thanks to Ted for suggesting the powered USB hub.

My question: would the Pink Faun I2S card be an improvement on my current setup?


#14

Only way to find out is to try. The card is not very expensive (euro exchange rate is helping), and Jord @Pink Faun will probably ship you a card with money back option if you don’t like it.


#15

Hi Ted,

My understand is that DS uses ASRC (or a similar algorithm by PS Audio, less likely) to reject jitter? Is it correct? If not, then what is the purpose of 10 times upsampling?

re. RFI/EMI, am I right if I understand that they contribute to bit error reading (DAC side) only, not contributing to jitter? DS uses its own clock to clock data in (2 stages of jitter rejection for Async USB and 1 stage for others) using possiblly the algorithm mentioned above and is not dependent on the source clock in this aspect so personally I don’t think RFI/EMI can affect jitter. However, from the practical view with measurement in your lab, it is true?

Besides, It is digital world (at least before Digital to Ananlog conversion implemented in your FPGA) , if RFI/EMI is not very bad, it would not affect data too much to the point where the receiver, DS in this case, cannot differentiate 1 and 0. Is it true in practice? Even it is too bad, high voltage balanced port like AES would be helpful?

Is it very safe to state from the practical angle that with DS in place, I don’t really need to buy very expensive sources? It is like the expensive music servers, if they really do their job (sending data out with no voltage drop, less noise, good square shape, timing precision), so no garbage; the DS does not care garbage in or not, DS cleans garbage anyway. In that sense, these expensive music servers are nice to have than must have because I have a DS? Personally, Mac Mini is enough for me when DS in place. Is that very true in practice? To me, looks like people bring the bad impact of noise in analog world and apply to digital world. It does not make any sense to me but just ask you here because you have the insight at theory and in the lab as well. Appreciate your truthful answers. Thanks


#16

Another question is: re. I2s vs Async USB, the bit outcome can be same and you mentioned the good side of I2S cable and downside of USB (“USB also has the downside of the signals being right next to the power and ground lines for a long distance which pollutes the USB power connection”).

I think of galvanization of Async USB to I2S converter at DS which may help to reject some of the bad RFI/EMI noises from source and along the way picked up on cable? Does I2S interface at DS has the same galvanization? If not, then this one gives +1 point to Async USB implementation which can benefit sounding more?


#17
tamnguyennb said My understand is that DS uses ASRC (or a similar algorithm by PS Audio, less likely) to reject jitter? Is it correct? If not, then what is the purpose of 10 times upsampling?
No ASRC, only Synchronous Upsampling (always integer ratios): 10 x DSD rate upsampling is the minimum rate that multiples of 176.4k and 192k have in common. (LCM(176.4kHz,192kHz) = 28.224MHz = 10* DSD rate)
tamnguyennb said re. RFI/EMI, am I right if I understand that they contribute to bit error reading (DAC side) only, not contributing to jitter? DS uses its own clock to clock data in (2 stages of jitter rejection for Async USB and 1 stage for others) using possiblly the algorithm mentioned above and is not dependent on the source clock in this aspect so personally I don't think RFI/EMI can affect jitter. However, from the practical view with measurement in your lab, it is true?
No bit errors with any reasonable cables. RFI/EMI are picked up by induction in any loop, the amount of current injected into the loop is proportional to the loop's area and hence big ol' groundloops formed between power cords and interconnects is often the biggest receiver of RFI/EMI. That current can cause noise directly in the power supplies, it's frequencies can be aliased down in the audio band by sum and difference frequencies with any clock in the system, etc. You can never completely get rid of jitter, ASRC actually just encoded the incoming jitter into the signal (it changes the signal based on the relative timings of the incoming and outgoing clock.) Anyway power supply (and other) noise induces jitter and jitter induces noise... The best you can do is low pass filter the jitter with reclockers, etc. In practice the noise on the 5V power in the USB cable is the biggest source of noise with USB, tho suppressing the horrible jitter caused by the 8k USB block rate (which doesn't evenly divide 44.1lk ...) is a good thing.
tamnguyennb said Besides, It is digital world (at least before Digital to Ananlog conversion implemented in your FPGA) , if RFI/EMI is not very bad, it would not affect data too much to the point where the receiver, DS in this case, cannot differentiate 1 and 0. Is it true in practice? Even it is too bad, high voltage balanced port like AES would be helpful?
As I said bit errors are essentially non-existent with reasonable cables, etc. (Just run the bit-perfect test as long as you want - it will flag any bit errors.) AES/EBU and I2S (at least the way we do it) have the advantage of begin differential signals that reject most of the common mode noise. The higher voltage helps just a little, but the DS rejects enough jitter that if you hook I2S, AES/EBU, S/PDIF and TOSLink between a SINGLE transport and the DS you won't be able to tell them apart (or even get clicks when you change inputs.) The key point is that if all of the sources share the same ground and are sending the same data then the groundloop noise is the same when the DS changes inputs so most of what's left is jitter related (and the DS rejects most of the jitter.)
tamnguyennb said Is it very safe to state from the practical angle that with DS in place, I don't really need to buy very expensive sources? It is like the expensive music servers, if they really do their job (sending data out with no voltage drop, less noise, good square shape, timing precision), so no garbage; the DS does not care garbage in or not, DS cleans garbage anyway. In that sense, these expensive music servers are nice to have than must have because I have a DS? Personally, Mac Mini is enough for me when DS in place. Is that very true in practice? To me, looks like people bring the bad impact of noise in analog world and apply to digital world. It does not make any sense to me but just ask Ted here because you have the insight at theory and in the lab as well. Appreciate your truthful answers. Thanks
The biggest different will be good hygiene in your whole system - solid grounds, minimized ground loops, power cords away from interconnects, away from speaker cables, etc. Running a signal from a PC that's on a different outlet than the audio system adds a big ground loop but keeps some of the PC's noise away from the audio system - whether it's better to plug the PC into the same outlets as the audio system is an experiment that everyone should try.

Balanced cables can help a lot to avoid noise in a system, but some people’s systems just don’t have enough noise to worry about them.

That said, everything matters and cables all sound different (tho perhaps just microscopically so) and the same for sources. Whether this difference is significant, better or worse or whether you care about the changes you do hear depends on you.

Most people suggest getting used to the DS before you start tweaking other things, then you’ll have a base line to compare any changes… The DS reacts differently to power cords, interconnects and various sources that almost any other DAC so optimizations you might have done for another DAC may actually be hurting the sound quality from the DS (and also optimizations for the DS may hurt the sound quality when used on another DAC.) For example: many people found improvement with the Empirical Audio Off-Ramp in their previous systems, at least some have reported it hurts the sound from the DS - to be honest I’m not sure how, but I suspect another ground loop caused by the Off-Ramp’s power supply was the culprit, not a failing of the Off-Ramp proper or of the DS.


#18

I guess one have to distinct between emitted noise vs. degree of immunity (e.g. balanced noise suppression) and noise originating from the power supply, propagating along with the audio signal.

Also, every time there exist a difference in electrical potential on your ground you might get yourselves a ground problem. If you have several places on a PCB where you ground different items you create ground loops simply because you cannot make a copper layer with a big enough sectional area to completely eliminate the electrical resistance between the connection points. That’s also why grounding cables have a huge AWG to lower the wire resistance. In the design one have to remedy this to the degree possible.


#19

Thanks, All.

@Ted: interesting answers. Personally I don’t believe in bit error but just asked to confirm. I did run bit perfect test before. Good to have the confirmation from you to get rid of this doubt from my mind. From your answers, I can say much of everything boils down to GROUND LOOPS. Totally agreed that optimization on top of what I have in the system is more important (I don’t use expensive power cables and don’t use those with shields; I just arrange them properly; I also use some good ferrite clamps for each power cables, believing they can help reduce some RFI/EMI going to power cables if any just to make sure but I believe the real problem is RFI/EMI going out from these power cables to ICs and speaker cables so distance should be maintained; my speaker cables with proper twisted pairs and analog interconnection cables are decent ones all from Robert Lee, Zen Acoustics).

I have another 2 questions based on your answers:

1/ I had an P10 and all audio gears are fed with power from P10 so the ground loops are kept small. I have maintained only one source in my system to keep ground loops simple. From your answers, I understand that as long as ground loops are broken, DS can handle jitter well. I have tried to break a ground loop in my system before by using a good TVC with transformer and used balanced cables with best Neutrick connectors so I believe ground loop is broken from the DAC to the amps mainly due to the transformers within TVC in between.

If I forget about Async USB, I2S from computer audio (planned to add one more source) and just use optical out (i did) from my existing source and optical in of the DAC. With TVC and the this optical link, virtually I have no ground loop in my system which can add noise to power supply which affects clock and contributes jitter in this sense. Is it correct? I maintain only one source in my system.

2/ I don’t go into the details of USB protocol and have little knowledge on that, but if 5V power pin of USB is that bad like you mentioned, is it possible to break that pin (in the cable; use customized cables with that wire cut off) and DS uses its power supply to feed USB instead of using +5V from source via USB cable? DS USB has galvanization which can reject some of these noise already?

Thanks.


#20

Ferrite beads are a mixed blessing on power cords or interconnects - they can take the life out of things if overused. In particular I sometimes use them on power to digital equipment, but I haven’t found hardly anywhere else where they help. I find that power cords make more of a difference in my system than interconnects or speaker wire. After your system is working the way you want and you are used to it you might borrow and try some different power cords from a dealer or a friend… You might be surprised (or not.)

Don’t get too paranoid - it’s easy to try S/PDIF (or AES/EBU) vs TOSLink, most devices have both. For some devices TOSLink adds so much jitter that it’s worse than S/PDIF, but certainly not always: sometimes it’s just the ticket. With a good TOSLink cable you may well be able to do 24/176.4k (and therefor DSD via DoP) and maybe even 24/196k.

The problem with the 5V on the USB cable is that the amount of current during connection is used to signal when a device is connected and whether it’s a hub… Depending on the specific models of hubs, USB controllers, and even the drivers in the computer and OS a given connection may work or not with the 5V disconnected. Some people have cables with a jumper or switch and they close it when they power the system up and then open it when playing. Sometimes that works…