What do you mean by “above”? Shouldn’t it be at USB layer? I’m not sure if there’s any kind of logging around USB buffer size adjustments, but if there is, you will likely see adjustments there, with and without Diretta, so you would need to show that the patterns are different.
1 Like
It was obvious. Please don’t.
1 Like
… as it’s against forum policies. This is a place for human discussions.
1 Like
This, and most of it was incorrect (or irrelevant) hallucinations
3 Likes
Hello Matthew, interesting to discuss the ideal settings for the pc Host.
I tried for my part as well and I almost obtained an ideal one under a Win10 host PC (and therefore ASIO).. but, the target blocks at one point and I have a continuous whistling sound.
I have to go back to the start.
Is a version of this (unofficial) doc downloadable somewhere?
1 Like
As I mentioned previously I am not affiliated with the Diretta. I do not have access to their SDK. I’m just a hobbiest who set out to understand how to configure Diretta so that audio improved in my system. Without specialized USB or DAC-side measurement equipment, I cannot measure electrical jitter or noise at the DAC. What I can observe are changes in host/target feedback behavior, scheduling stability, and buffer correction in logs and packet timing. Those changes result in audible differences, but they don’t constitute proof of changes inside USB Audio itself.
I’ve explained, at a system level, how Diretta is intended to work and documented how its configuration affects behavior and sound. Some of that was shared here. My guide focuses on repeatable setup, observable logs, and controlled listening outcomes.
I responded to this thread because there were comments that Diretta had been debunked, was vaporware, and that users reporting positive results were essentially imagining things. My original intent was not to argue measurements or redefine USB Audio, but to point out that the tested configuration does not represent how Diretta is designed or typically deployed when it works well. I’m comfortable standing behind my configuration guidance and tuning methodology, but not making assertions that require proof at the USB or DAC hardware level.
And I don’t dispute Marian’s results given the configuration that was tested. Based on that setup, I agree it’s entirely likely Diretta sounded worse than RAAT. Marian and I also agree that the line noise claims on Diretta’s site are fantasy, which I wasn’t aware they claimed that until he pointed that out.
Hi Stef, I have not tested the ASIO version. My use case was using Audiolinux Diretta-ALSA.
My conclusion after my listening tests is that it can’t be said it made any perceptual difference, good or bad. Considering the fact that I followed David’s instructions ad litteram and double-checked with him directly, it can be said that it should have made a significant difference. And what you are saying suggests that whatever improvements people hear with David’s setup are imagined.
3 Likes
There is a specific setting : fixASIO otherwise it’s identical to the Linux version.
And about your guide ? A link ?
But none of that could affect anything related to audio output.
How did you establish that there is an audible difference? @Marian had performed a proper test and the result was that multiple people could not tell Diretta and RAAT apart at all. Did you do any convincing testing that would show otherwise?
But… that’s not what @Marian had found. Is there a straw man here somewhere?
So we can at least agree that Diretta’s designers are peddling some nonsensical claims. Is there any reason to believe that their other claims are any more valid, when what they describe (inasmuch as one can find from their descriptions that consist of 100% pure technobabble that contradicts laws of physics and principles of computer science) does not make any more sense and has been shown to not affect anything tangible.
2 Likes
Can you PM me?
I don’t think there are any other claims.
2 Likes
That’s rather hard to say, since nothing they have posted on their site making any sense at all, but being charitable I allow that they might make some other ones.
I do have to admire your willingness to put to test something that makes somewhat less sense than a magic fuse though.
Or just predictive text and probabilities. Though even these sound these better using Diretta.
4 Likes
I have attempted to answer all questions, but no one has stepped up to actually test my configuration. The reason I first replied to the thread. I am confident when Diretta is correctly configured results are repeatable.
So I challenge someone to actually invalidate my Diretta claims with Roon Core at the heart. I will give you three options to test, each with its benefits over David Snyder’s base configuration.
This must be a fair assessment, and approached as Marina did with his sons, using third-parties with no real knowledge of the test. Commit to the work and follow the steps. They should be straightforward, and the first two options should take less than 30 minutes if you have basic familiarity with Audiolinux. If you want to run pcap captures and noise analysis that’s fine but my focus has been on my systems sound. Let’s agree to begin there.
Option 1: Diretta Host and Target routing through a switch.
Option 2: Diretta Host and Target direct connect, IPv4.
Option 3: Diretta Host and Target direct connect, IPv6, with full CPU isolation and IRQ pinning.
All three options are valid configurations, but I encourage option 3 for a real sense of Diretta as this configuration offer’s the most impactful tuning options. Or, go through the options as the changes are minimal from 1 → 2.
To support my reasoning for an IPv6 configuration, local IPv6 eliminates entire classes of background activity. With IPv4 or global IPv6 enabled, Linux runs background mechanisms that are irrelevant to real-time streaming. With IPv4 and global IPv6 disabled, IPv6 link-local only:
• No routing table lookups
• No ARP
• No RS/RA traffic
• No address renewal
• No DNS queries
The NIC and kernel do less work, more consistently.
Who will step up?
Not me.
- I’ve already spent enough time on it.
- I don’t want to look for other people with no knowledge of the experiment and waste their time also.
- If what I’ve done so far didn’t matter at all, I can’t see how what you’re proposing will.
I’d probably have more fun dropping weights from the Leaning Tower of Pisa.
4 Likes
Dude, that’s on you. Why do you think anyone is obligated to do your work?
1 Like
Marian, here’s what you need to change.
- Set the end0 network interface on both the Host and Target
- Change the Audiolinux cpu isolation on both the Host and Target
- Apply changes to the Host setting.inf. Reboot
That should take 5 minutes
No, its not all on me to give you all everything you ask for without anyone willing to verify what I’ve suggested. No matter what I show its about how it sounds in your personal system, and I don’t believe based on this thread that you’ll accept it.
And half an hour for each participant. No thanks. The person who makes claims has the burden to prove them; the rest of the world has no obligation to prove them wrong, as you ask. I made an exception, to show that I’m willing to “step across the aisle” and give it a try, and encourage anyone to try that once in a while. My job is done here.
It is on you. Neither @Suedkiez nor I asked anything of you. I’m sure people who believe these tweaks make a difference will hear a difference and won’t even bother to look at my tests. My job here is done.
2 Likes