USB Nervosa Thread Decrapifiers, pro interfaces, and bears oh my

Take the Wyrd out and if it doesn't sound noisey or glitchy, then there's no need for it.

 

Done.

 

With the appearance of Fulla 3, Fulla 2 became available used in Germany and I took the opportunity. Connecting it to my workstation, I was somewhat underwhelmed by the amount of digital noise coming through in the treble. Although I may be spoilt by using a Digione as regular source, I wondered how the elders could recommend this in the way of the noob. It was uncomfortable with HD598 and and unbearable with HD600.

So I connected this to a front port of my workstation and there were _seriously_ less problems. The front USB is also delivered by the motherboard. Going back I noticed that in the neighbour USB port there was my WIFI stick sitting. After unplugging the WIFI stick, things improved seriously and sounded like the front port. Not sure if that was due to RF interference or due to a shared port.

Having a Wyrd at hand, that increased things a bit further. But Wyrd was a lot bigger help when plugging it into the original port with WIFI stick setting next to it. The more f**k there is, the more can be defuckified, it seems.

Cyrus Soundkey as alternative Dac/amp was somewhat less sensitive to USB noise. Together with Wyrd it turned out to be a mixed bag. Sounded like more authority and snappier response in the bass. But somewhat overfiltered in mids and highs and loosing immediacy. Would be interesting to test this with a standard powered USB hub.

When connecting Fulla 2 via Wyrd to my cheap smartphone, the Wyrd made some difference. Switching the smartphone to airplane mode then made yet a tiny difference - not sure if I could detect this in a blind test. If yes, only with selected material.

Take-away lessons: If unhappy with your USB hole, try a different one. If using a phone, try airplane mode and see if you can spot a difference. Does not cost you a dime.

While we are at it: Took my Steinberg audio interface downstairs for some room measurements recently. Could not get it to work for half an hour. Finally put a powered USB hub inbetween, and bamm - it was working. This was on a thousand Dollar laptop with a special USB port marked with a promising power icon leading you to believe that this could deliver some power, while in fact it was not.

Take-away lesson: do not believe in every icon you see.

Looking forward to hear Schiit Unison one day. Until then, I am happy to avoid USB when possible by using SPDIF. It's a mess. Fulla 2 sounds good for the money when on a diffferent port of my workstation, so mystery solved.

To all the "It's only ones an zeroes" believers out there (I used to be one myself): Before we can even discuss ones and zeroes it turns out that even basic power delivery is often shit with USB.

To all noobs out there being noobier than myself: spending 100$ for a Wyrd decrappifying a 100$ Dac/amp doesn't make sense of course. I just have it here for playing around.

 

It is only ones and zeroes. But you are reaching the right conclusion. Not only may the port not be delivering power (and/or the cable unable to carry it) but it may also be that your audio ones and zeroes are fighting for priority with some other process using that USB port/internal hub. An ultimate example of this is when moving the mouse causes loud noises on the audio (cf stuff like DPC latency)

It is only ones and zeroes, but computers and their operating systems (and proprietary drivers) are not necessarily perfect at delivering them. Even after so many decades. Aaaargh! Pass me a wall to knock my head against. Consider: it is only recently that Microsoft actually started supporting USB Audio 2.0.

 

Not only that, but the USB audio spec doesn’t include any error correction as far as I can tell, while USB file transfers do use error correction. The ones and zeroes don’t always get interpreted correctly on the other end of the cable, and there’s nothing in the protocol to mitigate it.

 

This is also true for usb audio.

All the data fields are crc-protected, and in the rare case that you do actually get bit-rot on a cable, you don't get subtle shit like sharp treble or less stage, but clicks, pops, and in anything made in the last decade, just nothing but silence from packet drops, since usb audio devices usually operate in modes where the sender requires an ACK on receive but there is no mechanism for resend on fail.

So take your meds people, because timing jitter is one thing, and dirty power is also a thing, but anything that your DAC shits out is absolutely 100% a bit-perfect copy of whatever the sender decided to shit out, or you get silence, there is no third option.

 

This is the case with most computer-equipment audio problems. Those who are desperate to experience such need only search the world for a PC with a DPC latency problem. No need of either imagination or double-blind testing! I detest the modern-day use of binary as an adjective, but hey, either the sound is there, or it is not. Either the sound is listenable or totally lost in clicks or other distortion. Binary. Ones and zeroes

There is an intrinsic problem in that our PC operating systems are not designed to do anything in real time, but on a time-sharing basis. Mostly it works anyway. When it doesn't... Aaaargh!

 

USB audio can recognize that something went wrong when comparing checksums, but there’s no retry when they don’t match, unlike file transfers. Didn’t mean to trigger anyone, I just thought it was interesting.

 

Ah! I misunderstood.

Yes, the focus of the usb audio spec - and especially USB Audio 2.0 - is low-latency and multi-channel.

I believe only control frames - used for setting frequency, volume and such - have retry capability, as the spec was designed purely to carry as many channels as possible in with as little latency as possible, in response to the very high latencies and limited channel counts for spec 1.

USB audio over the original Full Speed USB could only do 2ch CD quality with 2ms minimum bus latency, so you can see how the rise of multichannel audio via playing movies on a PC might have shaped the second spec - a buffer for retries runs directly counter to a low-latency aim.

Spec 3.0 does nothing for us either, since it's designed around low-power mobile audio and exposing noise cancelling and such to the os.

 

Not only usb audio, all usb devices that use isochronous transmission mode exhibit this behavior. Can be seen with a naked eye when plugging a monitor using displaylink port replicator and a shitty usb 3.0 cable.

 

The main thing to understand about that is that the result is not "less black blackground" or "slower transients" or "less detailed treble" but as Taverius said, pops and clicks that are unmistakably audible. Considering how rare that actually is, most USB nervosa is total BS.

 

Thank you for the answers!

I would like to refine my post a bit, substracting some frustration and adding some justice to USB.

USB is just a connection technology. What is it supposed to do when you connect it by accident to the single most noisiest port available in your household? Clean everything up magically along the cable? Would be nice but unlikely.

Expecting total decrappification in a connected Dac/amp at 99$? That would be unfair.

Comparing it to a Digione as source? Even more unfair - a comparison to Allo USBridge would be much more appropiate.

While all of the things above would be somewhat unreal or unfair, what would be even more unreal is to claim that it is only ones and zeroes and must sound the same. Main issue here is probably that I am not adressing the typical SBAF member but rather other people on the internet.


My frustration is not originating from USB connection per se, but rather from parts of the USB universe and how different they can be. I am no way qualified to judge if USB or SPDIF is generally better. In my experience SPDIF has given me more of a consistent listening experience though.


My second point was acknowlodging how big differences can be on different USB ports. This is unfortunately true also on the lower-budget side of things. Good decrappification is somehwat unreasonable here for budget reasons. The good news is that I can get much better results in my case by changing to a different port, removing a disturbing device, or changing to an alternative source ( Laptop, phone, etc.). Cost of that is 0 cent! Just play around and try when in the same situation.

My last point was that there are some USB issues that can be solved by addition of a simple powered USB hub. That is 15€ at max. Maybe you have that already, or you can borrow it for free overnight just to have a listen.

USB itself is not to be blamed if Windows does strange things to USB power, or if there is a noisy source. But the promise that you can just connect a USB audio device to any port available and be happy, is oversimplified.

 

The reason for no retry is that the isochronous transfer mode used by USB audio and video was originally designed for live, realtime streaming. Think of the live video feed for a football game. The data will keep coming continuously even if an error is detected. The receiver tries to keep reading data at the stream's current point in time and replaces any detected bad data with zeroes.

There is no practical need to be concerned about errors in USB data transmission. The USB spec allows a maximum bit error rate of 1 bit in 10¹² bits for the hardware layer. By my calculation (somebody please verify), when playing an uncompressed 24/96 stream on USB audio, that would be 1 bit error in 60 hours of playback, and that is the maximum allowed error rate.

 

There is more than just bits to a digital audio signal over USB (or any other physical connection). Those bits are encoded on an electrical signal traveling through resistive material. Trying to digitally encode data on a stochastic phenomenon is bound to create a fair range of unpredictable outcomes. The very rudimentary and arbitrary ways in which we try to measure a restricted range of those outcomes are hardly informative of what truly goes on the electrical chain. No one has a clue.

Transparent enough audio stream over physical wire is a lost cause to me. No decrapifier is truly gonna solve it. Not with current Middle Age tech.

 

How do you think your post arrived here, and from here, on all our machines?

How do you think your photos, documents, etc, are accurately stored and reproduced.

We are back to square one of audio/data fantasy here. Suggest you stick with vinyl. Or join computeraudiophile.

Or, to be more succinct.... bullshit.

 

Indeed. And one can't decrapify data that is not there. So, if there is a clash with, say, the video controller, and it has grabbed the device, and audio is forced to wait... it's lost, and that's the end of those bits.

It is not a perfect system. As I mentioned before, the basic PC design is not intended to do anything in real time. But it is a lot more perfect than the nervosasites would have us believe. And with computers that do misbehave, simply changing ports, at zero currency units is often enough. But yep, I did once throw the computer away and build a new one.

 

If you think the excerpt you quoted from my post puts any of this into question, you didn't understand what it means. What is your objection exactly?

 

Well, this bit...

and, errmm... this bit..

I can misunderstand anything. But your post seemed fairly clear.

 

Experiments performed years ago demonstrated bit perfect transport possible over USB and SPDIF:

Modi Multibit: Multibit for the masses.

Over 85 million bits transported and not one single bit deviated from source after sending from a Digital Audio Workstation (DAW) out through the following:

DAW -> USB -> SPDIF -> loopback -> SPDIF -> USB -> DAW using a Gustard U12 and RME Fireface UC.