Take Me to School (Understanding Audio Better)

Technical Discussion
82

I’m surprised that people say RAAT sounds inferior.

I appreciate the clocking consideration, and hadn’t thought about that as most of my connections are USB. I use TOSLINK only on lower quality stuff, like from TV or cable box, and usually only when my more convenient connections are already used.

What I wonder is if the digital stream from my iPhone 12 differs from that of a later model iPad with different chips or the MacBook Pro, or something from Android (using UAPP or ROON ARC to fix the droid USB problem). I haven’t done any careful listening, and am not sure what I would listen for.

1 Like
83

That’s all about electical noise from the source making it to the DAC.
The portable devices likely sound batter than the Mac, just because they will generate less noise.

But most portable devices have there own issues, any Lithium battery device is going to have a switching regulator in it, so you’ll likely get switching noise, and I don’t really know how practical it is to filter that effectively inside a small device like that.

What I always tell people is if you can do the test easily, just do the test and and see what you hear, pick the one you prefer. Most of any difference if you can hear one will be in staging, mostly focus, and depth, with some effect on the treble and overall smoothmess.

In his CanJam talk Rob Watts had a whole piece on people preferring devices with specific types of common mode noise, because it gave an impression of false contrast.
It’s one of those things that makes discussing better in practical terms difficult, and isn’t the same as discussing preference.

1 Like
84

No personal experience, but RAAT is a multi-function protocol that does lots of other things in addition to delivering the data. Simply being aware of that may cause people to think a simpler protocol sounds better.

USB devices delivering the same source are delivering the same data. If they weren’t they couldn’t transfer files. Differences between them are coming from something other than faulty data.

1 Like
85

But are they delivering the same data? Is the USB from an older iPhone delivering the same data as an Android with UAPP or a Linux box or or or… Assuming they are all streaming the same 96/24 file from Qobuz.

Is the data delivered to the DAC the same? How can a lithium battery switching noise in the streaming device affect sound when it’s delivering data to a DAC?

Am I being obtuse?

1 Like
86

Yes. USB detects errors and re-sends the data if it’s messed up. It’s not like streaming a movie over Netflix, which is via a network protocol called UDP (“User Datagram Protocol” – catchy, eh?). UDP is “best effort” meaning it doesn’t even try to detect, much less correct, any transmission errors.

To get a little further into the weeds, as of 2.0 all USB packets are delivered in order without numbering. That means each packet is checked and confirmed as OK (ACK’d) by the receiver. So each Data packet (containing 0 to 8192 bits of data in multiples of 8 – I’m sure you’re just giddy over knowing that little tidbit) is checked and verified before the next packet gets sent. This works extremely well.

I haven’t done it, as I trust the protocol, but one can get a USB packet sniffer for about $50 if one really wants to verify this.

The short answer is “not via the data”. EMI/RFI or something, maybe. Ineffective galvanic isolation perhaps.

Not at all. No real reason for normal people to examine the internals of USB. Luckily, there are a few of the right kind of weirdos around here for that. :slight_smile:

3 Likes
87

Yes
Transmission errors might exist, but I’ve tested network and USB transmission before and frankly over short distances with adequate cables, it’s just a theoretical possibility rather than common reality.

Because digital signals are transmitted on analog wires, and those wires aren’t perfect conductors.

Every metal thing with length acts as an antenna, including traces on a circuit board, interconnects digital or analog.

What happens is the switching noise which is in the 100’s of KHz/ Mega Hertz range is picked up by traces on the source device, it gets passed to the DAC via the wire (This is true of any noise).

Some of it will be filtered out by isolation (if the device has any), but some of it will make it into the primary DAC circuitry, eventually your DAC has to turn an analog signal into a stream of 0’s and 1’s it does that by using a threshold. The noise on the analog line makes the transition vary slightly in time, and that is Jitter.
Our hearing is somewhat surprisingly extremely sensitive to jitter.
Rob Watts talk from Can Jam is actually very good discussion of this.

There is more to it than that, that noise is going to make it to the other analog components in the DAC, and that will have some effect as well. I’ve done a lot of schilling of Rob Wats CanJam talk, but his observations on the effects of insanely low levels of certain types of noise is at the very least interesting.

4 Likes
88

Thank you for explaining this in a way that is understandable, @Polygonhell and @Lou_Ford .
Much appreciated!

4 Likes
89

Thank you @Polygonhell and @Lou_Ford. I know much of this intellectually, but had not applied my understanding to this case.

“Not via the data” was very useful. So was reminding me about how even circuit board traces might act as antennas (antennae?)

Finally was not familiar with the software layer of protocol in USB. Shameful for someone who actually read Doug Comer’s “Internetworking with TCP/IP” for fun in my disturbed young adulthood.

4 Likes
90

We tend to think of these things as being trivial effects.

I wish I had an easy way to capture my Oscilloscope output.
I was recently testing a cheap switching mode bench supply I picked up for a project to see if it was workable.
The difference between measuring the noise at the output terminals of the PSU vs at the load, was 10x.

1 Like
91

While this is true for a file transfer, USB audio is streamed using a different protocol: USB - Wikipedia. So actually it IS like streaming a movie over Netflix. This is one of the many things I learned from @Torq when he was still active here.

However @Polygonhell is correct that transmission errors are rare as long as the USB cable is within spec. I think I read where a USB engineer calculated an average of one error within a month of continuous audio streaming, and that was for UAC 1.0.

92

Yes and no. UAC 2.0 and 3.0 still provide for error detection, but not for getting data resent (burst mode, which I overlooked). So the DAC will either mute and you’ll hear a dropout, or play the bad data and you’ll almost certainly hear something really ugly.

If a DAC did play the bad data, which is highly unlikely, the errors would have to be (a) fairly consistently distributed, and (b) always occuring in the least few significant bits for there to be a subtle but detectable difference to listener. You’re more likely to win the lottery 50 weeks straight than for that to occur.

I’d be surprised if any decent DAC didn’t mute on error, precisely because it can’t know the significance of the corrupted bits. So if you’re hearing anything at all, you’re hearing the proper data.

I’ll stick with my claim. Different USB sources don’t sound different because of the data the DAC is playing.

Is there a non-zero probability that the data are different? Sure, but that’s true even with error-detect-and-resend protocols. Nothing is ever 100 percent guaranteed.

4 Likes
93

Or it will interpolate the data surrounding the error and attempt to conceal the problem, like CD players do.

I agree. It is extremely unlikely to have enough contiguous errors to be heard as long as everything follows the USB specifications. I think we’re all in agreement on this. There is only analog noise left to consider as @Polygonhell mentioned.

1 Like
94

I’ve seen it written that IEMs with low impedance may hiss when paired with amplifiers with high output impedance. Is that really the case though? Wouldn’t the high output impedance of the amp mitigate the low impedance of the IEM?

I’ve also seen it written that IEMs with a low output impedance don’t pair well with amplifiers that also have vey low output impedance. That would seem to be my experience , but it has been a while since I had an IEM to listen to with a very low impedance spec to listen to. In those past cases my iFi amps’ IEMatch feature came to the rescue.

95

Generally you want the output impedance of the amplifier to be much less than the impedance of the device/driver it’s driving. You’ll see rules of thumb like 4x, 8x, 10x or 20x.
The primary reason for this is in effect the output impedance of the amplifier forms a divider with the input impedance of the device being driven, higher output impedance means the amplifier is less effective at driving the load.

The ratio of load impedance to output impedance is sometimes called damping factor, and is often stated to be a measure of control an amplifier can exert on a load. This is sort of true, and big mismatches can often give you a boost in the Bass frequencies with a loss of Bass control.

But it’s not really a rule. Some driver types tend to like higher output impedances, in the speaker world, Horns often prefer damping ratios at or below 1.
And for transformer coupled tube amps into speakers, it’s uncommon to see damping factors much better than 4, but they generally do not have an issue controlling an 8 or even 4 Ohm driver.

Hiss is usually a function of passive components (resistors, diodes, capacitors), it’s more obvious with IEM’s because they are so sensitive, there is an element of component quality in hiss, but amps with lower gain are going to have less audible hiss.

5 Likes
96

I’m not an expert on IEMs, but to add a bit more to the above… If you hear hiss or noise in an IEM or other lower impedance headphone, chances are it’s because the headphone is simply too loud (sensitive in audiophile jargon) for the amount of gain or noise in the amp.

If you have to set the volume on your amp to a low setting for comfortable listening, that’s a pretty sure sign the amp has too much gain for the headphones, or the headphones are too sensitive/loud for the amp. (It works either way.)

To fix this, you can either try a headphone that is quieter/lower in sensitivity (and frequently also higher in impedance). Or try using a lower gain setting on the amp, if it has that option. Either one of these could reduce the audibility of the amp’s noise floor in the headphones.

There are other ways of addressing the issue (as Polygonhell alludes to above). You could get a better amp with a lower noise floor, for example. Or simply try omitting the amp altogether, and using the headphone output on your player instead, if it has one.

Another less likely possibility is that the noise or hiss is originating in your source or DAC. Generally speaking, you want the gain controls as high as possible on these devices, without clipping the signal, to keep the noise floors on them as low as possible.

Finding a good synergy between an amp and headphone (or other transducers) can be a bit of a challenge when you first get into audio. Much of it falls within the general rubric of proper gain staging and impedance bridging though.

The bottom line is that headphones and other audio gear are generally designed with specific uses or applications in mind. And if you try to go too far outside of the general scope or parameters of those applications, the performance is likely to degrade. The degradation in performance can take a variety of forms, such as increased noise, distortion, or clipping, insufficient volume, or too much volume and not enough sensitivity in your amp’s volume control.

There are a couple different methods for comparing a headphone’s sensitivity or efficiency btw. And to compare apples to apples, you need to know which one a manufacturer is using, either dB/V or dB/mW. dB/V will almost always be the higher of the two.

2 Likes
97

An only tangential to audio question. I’m successfully using Tailscale to connect ROON Arc without port forwarding.

The Tailscale docs imply that conceptually it is creating a virtual network that connects your devices as if they are on your local network.

So why must I use ROON Arc and can’t use the regular ROON endpoint?

98

Orrr if you’re powering them single-ended and are OK with staying single-ended (and losing any mic signal), you can plug an impedance adapter between the headphones and the amp, “eating up” more of the power and requiring the amp to be turned up, which gets you away from the noise floor and possibly more adjustment precision. Of course there’s a catch with impedance adapters:

  • If you’re going buy one of those absolute cheapest ones, they’re usually simple resistors being added to each headphone channel’s circuit in series with the driver from that side. In one direction that does make the total load appear larger to the amp, requiring higher volume and fixing your noise floor and adjustment range problem. But in the other direction it also makes the amp’s output impedance (Z-out) appear to the headphones as being much larger, which can change the final frequency response in undesirable ways (AFAIK this only happens with dynamic drivers or other technologies that don’t have a flat impedance-vs-frequency curve, plus its effects have been greatly exaggerated in audiophile discourse, see Solderdude’s Impedance, Resistance and Other Issues for more on this)
  • If you’re willing to spend more, “smarter” solutions exist that are electronic neworks of multiple components (no idea how those work) that present a high resistance to the amp side but only a very small additional resistance to the headphones side, as in iFi’s very expensive IEMatch, which can show sub-1-ohm values to the headphones even as it’s showing tens or hundreds(?) of extra ohms to the amp. Luckily there are also in-between solutions like this JCally impedance adapter, which shows e.g. 150 extra ohms to the amp (I bought the 150 version and tested that one) while only showing about 15 extra ohms on the headphones side. Depending on what headphones/IEMs you’re going to be using it with, it could be a pretty good cost-performance compromise, considering Solderdude’s explanation on how much the amp Z-out really impacts the FR. (No idea why JCally aren’t clearly marketing this as a “smarter” impedance adapter that’s not just a series resistor, they could be making a tad more cash off of it but w/e.)