For use with DAC. Usb vs optical, what do you use? What is better?

[Srius1]

As title states. I recently got a DAC but am wondering if I should use optical or usb as source. Usb is more convenient as 1 less cable but I’m not opposed to optical . Thoughts?

[FloRolf]

Imo it doesn't matter. Digital is digital. 

[LAwLz]

It doesn't matter. 

Both are digital signals so in terms of audio, they will be virtually identical. There might be some very small differences in jitter but that won't matter. If they aren't identical then USB will most likely have the very slight edge (if async mode USB is used). 

 

Two benefits of S/PDIF however are that since it's optical, it might fare better in some areas with a lot of electrical noise. This will probably not matter unless you run your audio cables along fluorescent lights or live in a power plant however. Besides, such high noise would probably interfere with the DAC as well. 

 

The other benefit is that you don't have to mess with the USB audio drivers. Windows has historically been pretty bad when it comes to USB audio drivers, but that has mostly been fixed now with windows 10 and above. Also, if you use S/PDIF then you might need the audio chipset drivers which may or may not be terrible as well. So pick your poison. 

[Spindel]

As @LAwLz pointed out the streanght of optical is that it will not transfere potential electrical interference from the source. 
 

As to sound quality digital is digital. 

[Srius1]

Thanks y’all! I will continue to use usb as I have zero issues and it’s 1 less wire since I use a hub 

[Heliian]

2 hours ago, Spindel said:

As to sound quality digital is digital

But is it lossless? 

[Spindel]

1 hour ago, Heliian said:

But is it lossless? 

Only if you want to waste space and bandwidth 

[H713]

Optical offers galvanic isolation, which solves a lot of problems. Good DACs also handle USB isolation properly, and they solve most of those issues. If you happen to have a DAC that doesn't do this well, then optical isolation is a good thing.

[GoldenSound]

19 hours ago, Srius1 said:

As title states. I recently got a DAC but am wondering if I should use optical or usb as source. Usb is more convenient as 1 less cable but I’m not opposed to optical . Thoughts?

 

18 hours ago, Spindel said:

As to sound quality digital is digital. 

 

19 hours ago, FloRolf said:

Imo it doesn't matter. Digital is digital. 

 

18 hours ago, LAwLz said:

It doesn't matter. 

Both are digital signals so in terms of audio, they will be virtually identical. There might be some very small differences in jitter but that won't matter

TLDR: You'll usually get better quality from USB. Optical specifically is often best avoided.
The 'bits are bits' argument is not true in audio as it ignores some aspects specific to digital audio transmission protocols.

Long Version:
So there is a lot of misunderstanding surrounding digital audio. With the 'bits are bits' argument often being used.

It's true in the sense that the actual CONTENT of the signal, the audio data itself, the 1's and 0's will either reach the DAC intact, or they will not.
There are no 'audiophile bits' and as far as data integrity goes, optical SPDIF, coaxial SPDIF, AES, USB, I2S or anything else will be able to get the exact same data to the DAC.

BUT, the issue with audio is that it is not purely about data integrity. TIMING is important too. If you have exactly the same audio content, (ie: Exactly the same PCM samples) but they are converted at different times, you'll get distortion.
This is called jitter.

 

 

USB is an 'asynchronous' protocol.
This means that the timing is actually controlled by the DAC.
A super simplified explanation is that audio data from your PC is sent to your DAC, often in chunks at a time.
This is put into a buffer on the USB receiver of the DAC, and then the DAC uses it's own internal clock (crystal oscillator) to determine when to convert these samples.

When the buffer is getting low the DAC instructs the PC to hand it some more data and it sends more to put into the buffer ready for conversion.

The DAC is dependent on its own clocks for timing accuracy, and whether you plug this DAC into a Mac, a PC, a Pi, or any other device, the jitter performance will be absolutely the same. The timing with which the source device sends data to the DAC makes absolutely no difference so long as the DAC always has the next sample in the buffer ready to convert.


SPDIF is a 'synchronous' protocol. And unlike USB, the source you use DOES have a direct impact on the quality of the output of the DAC, even if the data itself is completely identical.
Why? Because with SPDIF, the clock signal is actually sent by the source device (PC), NOT the DAC's own clocks.

Data is sent in a constant stream, accompanied by the clock signal, and so if there is additional jitter on that given clock signal, the DAC itself will be converting samples with less accurate timing and the analog output will be less accurate.

We can show this quite easily by measuring the output of the DAC.
This for example is the Schiit Yggdrasil, playing a 'J-Test' file, when fed SPDIF from a good source:

 

And then this is that same DAC, playing EXACTLY the same file, with no data integrity issues, but with a poorer quality SPDIF source that has higher jitter:

 

These are not digital simulations or readings, these are measurements of the actual XLR analog output of this DAC.

 

Many DACs will have methods of attempting to get rid of jitter from a poor source device.
One common method is a PLL, however a PLL does not eliminate jitter, it's more accurate to describe it as jitter attenuation. The level at which it can do so is dependent on both the design of the PLL itself (performance of PLLs varies MASSIVELY), and the reference clock used. As the PLL can only ever be as accurate as the clock you're using as a reference.

So unfortunately, in audio, bits are not always bits. And other aspects of the digital source can absolutely make a difference in quality even if the 1's and 0's themselves are no different.
This is why there is an entire product category called 'digital to digital converters' which take audio data from your PC, and output over I2S, AES, SPDIF etc, with exceptionally low levels of jitter. The idea being that you can use these to gain an improvement in jitter performance over your DACs internal clocks when using USB. And in many cases these work quite well.

An example being the singxer SU-6: https://shenzhenaudio.com/products/singxer-su-6-xmos-xu208-cpld-femtosecond-clock-usb-digital-interface

And the last thing to mention would be noise. As @H713 mentioned, optical offers galvanic isolation, meaning there is no actual electrical connection between the source and DAC (cause it's just light). This prevents any sort of noise being conducted from the noisy, beefy gaming PC to the DAC which can degrade performance.

Some DACs will have this internally but typically only at the higher end of things.
But DDCs like the singxer SU-6 will almost always include this feature as well so they can provide an exceptionally low noise output on SPDIF/AES/I2S even if your PC is horrifically noisy.
There are also products to do this on USB such as the Intona 7055-C https://intona.eu/en/products/7055-c

Generally though the noise advantage of optical is not worth the often drastic hit to jitter performance. ESPECIALLY if you're using a PC. Seriously most motherboard optical outs that I've tested were beyond awful it's a miracle some dacs can even successfully lock to them.

 

 

 

[H713]

14 hours ago, GoldenOne said:

So unfortunately, in audio, bits are not always bits. And other aspects of the digital source can absolutely make a difference in quality even if the 1's and 0's themselves are no different.
This is why there is an entire product category called 'digital to digital converters' which take audio data from your PC, and output over I2S, AES, SPDIF etc, with exceptionally low levels of jitter. The idea being that you can use these to gain an improvement in jitter performance over your DACs internal clocks when using USB. And in many cases these work quite well.
 

 

Excellent writeup - I never realized that people actually use the motherboard fiber output. I wouldn't expect it to be very well-behaved. I've always done my testing using the AES output on my Lynx card and an AES to fiber box I designed a few years ago.

 

With all of that said... unless you're using a really terrible source (and PC motherboards might be that bad - I can't remember the last PC motherboard I dealt with that had a fiber output), the jitter is likely to be more or less inaudible. The graph you posted shows jitter components are each about 120 dB below the fundamental. I don't know about you, but I sure as hell can't hear distortion / jitter products that tiny. 

[GoldenSound]

13 hours ago, H713 said:

Excellent writeup - I never realized that people actually use the motherboard fiber output. I wouldn't expect it to be very well-behaved. I've always done my testing using the AES output on my Lynx card and an AES to fiber box I designed a few years ago.

 

With all of that said... unless you're using a really terrible source (and PC motherboards might be that bad - I can't remember the last PC motherboard I dealt with that had a fiber output), the jitter is likely to be more or less inaudible. The graph you posted shows jitter components are each about 120 dB below the fundamental. I don't know about you, but I sure as hell can't hear distortion / jitter products that tiny. 

It's worth noting that the J-Test is a bit of a 'hack' to show jitter (which is a time domain issue) on an FFT (A frequency domain graph). The levels on a J-Test result cannot be directly compared to other things like harmonic distortion etc.

So whilst it may seem at first that -120dB would be no issue at all, actually that could be quite problematic indeed.
In fact, there's even evidence that human hearing is so sensitive to time domain that it beats the fourier uncertainty principle
https://phys.org/news/2013-02-human-fourier-uncertainty-principle.html#:~:text=The Fourier uncertainty principle states,required to represent the sound.


FFT analysis is extremely useful for many things but it's important to remember that it can't be used for looking at time domain issues. The J-Test is a useful workaround for jitter, but other things like phase can be tricky too.

For example here's a square wave:

 

Here's that exact same square wave, with absolutely no frequency domain content changed whatsoever and the FFT will be 100% identical, BUT, it's been phase shifted by 90 deg:

 

 

The interesting thing about these is that despite all frequency domain content being identical, these two signals can be audibly distinguished and are changed in different ways based on things like DAC reconstruction filter, severity of jitter, or the transducers themselves. Even though from a frequency domain standpoint they're identical.