Lynx PCIe AES interfaces: AES16 and E22

This week I have been refining my AES setup into Yggdrasil and thought I would report back on what I found. There were some noticeable improvements to be had.

Disconnecting ground
My general setup is pretty simple: PC -> soundcard -> AES cable -> Yggdrasil. My speakers are pretty sensitive to weird harsh stuff in the treble and I was getting a lot of it. Turns out that this was mostly caused by grounding issues. Under advice from @Torq it turned out that the solution was to disconnect pin 1 of the male XLR feeding Yggdrasil. This disconnects the ground of Yggdrasil from whatever the hell it's connected to at the PC end. AES is transformer isolated but that's clearly not enough to stop a ground loop from screwing up the sound.

I tried four setups with significant improvements between each:

Yggdrasil and PC on separate power circuits, ground pin connected on AES cable
--worse than--
Yggdrasil and PC on same circuit, ground pin connected
--worse than--
Yggdrasil and PC on same circuit, ground pin disconnected
--worse than--
Yggdrasil and PC on separate circuits, ground pin disconnected

Basically, each step removed another layer of nasty distortion and harshness in the treble (sort of 6khz upwards). Disconnecting pin 1 is pretty easy even for a DIY klutz like me, I just unscrewed the shell, sawed through the wire from the shield to pin 1 with (um) a scalpel and then covered the exposed end with electrical tape before screwing everything back up.

RME vs Lynx
I have been using an RME 9632 as my soundcard for a while. I am in the process of building a mini-ITX streaming box and bought a second soundcard for that which happened to be a Lynx AES16. So I got the chance to do a head to head. I never spend too long on these because the differences are usually apparent very quickly. My verdict: the Lynx is a bit better. It's marginal, but it's definitely there. I pretty much hear the differences between different digital sources as lying on a single axis of variation, and what I hear here accords with that: the Lynx is more effortless and natural, with better separation, and smoother and less tizzy/etched treble. How big is the difference? I'd say a little bit less than the difference between SPDIF and AES. The RME is excellent in its own right, but if you are looking for one of the two, I would suggest marginally preferring the Lynx.

ASIO vs WASAPI
@k4rstar reported earlier that the ASIO drivers for the RME9632 sound different from WASAPI. It's slight, but noticeable; I heard more in the lower mids and less in the upper mids and marginally better microdynamics. The Lynx also sounds different between ASIO and WASAPI, but the difference is larger. Actually with WASAPI the Lynx sounds pretty much on a par with the RME. It's only with ASIO that it moves significantly ahead. The WASAPI sounds more digital with slightly fake detail in the treble. The ASIO feels warmer and more natural.

 

@Garns great analysis! What program were you using as a media player to switch between ASIO and wasapi?

 

Thunderbolt 1 has far more than enough bandwidth for any audio application. So long as it is compatible with the software and hardware don't sweat it.

 

Thanks! I use Foobar. You need to get the ASIO output plugin.

 

USB also has plenty of bandwidth, and we all know how great it is for audio.

 

Thunderbolts 1 & 2 are similar and use the same connector, the difference is in how they packet data (1 splits the 10 Gb of bandwidth into two channels, 2 splits it into twenty which means you can allocate it more efficiently).

http://www.test-and-measurement-wor...bolt1-vs-Thunderbolt2-vs-vs-Thunderbolt3.html

 

I knew about the higher bandwidth but didn't know about the difference in allocation. Thanks for that!

 

I disliked you to get your attention. A ground loop is not the issue as ground loops are a 50 or 60hz hum depending on where you live not a 6khz one. Most computer chassis are grounded. Furthermore AES/EBU has to be transformer coupled on both ends. AES3 and S/PDIF do not but you'll find a ton of transformer-coupled consumer S/PDIF devices. Slicing off the ground of an AES XLR cable will not help you and can only hurt you here. Your feud is probably with the Yggdrasil since you mention 6 khz treble artifacts.

The Gungnir Multibit and Yggdrasil and the other Schiit DACs (even the DS ones) not the Modi multibit (which might have neutral volume treble but its softened and veiled to shit so I don't care if it does) have a neutral amount of treble. However, the Yggdrasil (and Gungnir AK4399) have an artificial, ESS Sabre-like presentation (Sabres do this outside of the glare) of lower level treble sounds like tape hiss which is probably the 6-7 khz weirdness you mention. The Yggdrasil will throw the tape noisefloor of a recording in your face instead of having it be the tape noise floor. You can tell easily with multitracked recordings where a worse grade of brand of tape was used for one instrument. Rather than that instrumental track having a slightly higher noisefloor, the Yggdrasil will throw that tape noisefloor in your face rather than just having them there as a higher noisefloor for that instrument, e.g. the cash register on "Money" off Dark Side of the Moon. Many demo anthology CDs collecting a bunch of tempos recorded on cheap tape are unlistenable on the Yggdrasil due to this while sounding like filthily recorded demo collections on everything else, including better done ESS Sabre implementations and shitty, glare-filled poverty audio interfaces like my Focusrite that picks up all the noise from desktop PC USB power lines.

The solution to your problem isn't interface and cable rolling but more caveman: ditch the Yggdrasil until it gets upgraded. I recommend the Gungnir Multi-bit. While the separation doesn't floor me like the Yggdrasil does, and it has more of the Schiit Multi-bit "bloom" (it's more bass ghosting than actual bloom or a bass boost or distortion), the treble presentation is more natural and less artificial and Sabrey minus the Sabre glare. The Gungnir Multibit only has S/PDIF connections, not AES/EBU but you can just get a cheap PCI-E soundcard from someone like ASUS that has a transformer-coupled, coaxial S/PDIF output.

 

Either they're broken, or you're imagining it. If they're working properly, you shouldn't need tasting notes.

 

Why not? Taking that assumption of yours further, then USB should sound the same as well. It's just 1s and 0s, right?

 

No, wrong. A sound driver is very different to a physical electrical connection plus screwy receiver hardware. It's not magic, it's engineering.

 

People argue over the sound of foobar versions. And if JRiver and foobar2000 sound different with the same output mode, 0dB output level, playing the same file. Dig deep enough into engineering and it's actually magic. We don't fully understand what goes on yet. Genetic evolution algorithms produce FPGA code that does literal magic. Engineering, right?

 

They do, and it's no more useful. Again, unless foobar is broken, in bit perfect mode, there is no difference. Not understanding how computers work is not a good foundation for an argument .

I'd suggest that you got and read @Torq 's rebuttal of that particular idiocy- he is a nicer and more patient person that I'll ever be, and makes an attempt to explain rather than just trying not to grind his teeth in frustration at the trope.

 

My job is involved in the process of making computers.

 

The guy who claimed to hear differences later retracted them after more critical listening.

Or maybe he didn't, because blind testing is the embarrassing audiophile boogeyman.

 

There's more than one outbreak- there's a lot of it about:

http://www.superbestaudiofriends.org/index.php?threads/computer-audio-players.134/page-13#post-95321

I sort of wish that I'd been able to express it more succinctly, though, like this:

(Assuming a bit perfect player that does no processing)

 

One wonders what the point of bit-perfect playback is if it can still sound different across implementations.

 

Why would anything be "mixed"?