Further discussion about configuring Diretta

Tinkering
23

Analogies are almost always bad, and this one is no exception, but I’ll play along. A carburetor - or a fuel injector - is an essential part of an engine, without which it simply won’t work. Digital audio works just fine without Diretta, so Diretta can’t be a carburetor. What you’re actually doing is more like attaching magnets to the fuel lines with millimeter precision.

You keep throwing David and all the people who were impressed by his setup under the bus.

That’s a big red flag. How can you expect reliable results from it if they don’t bother to provide detailed documentation?

You are contradicting Diretta themselves. Their home page talks about controlling power consumption and shows graphs of electrical currents, which happen very much at the physical layer (highlights are mine):

There is only one way to achieve this goal: averaging of processing and reduction of fluctuations in power consumption.
…
The transmission is controlled by forecasting Target buffers.
In this way, fluctuations in power consumption of a player as the Target is averaged as much as possible.

You may want to sync up with them on this.

I measured the power rail noise on the Diretta target, not the DAC, because that’s what they claim they are improving. Please make sure you understand what was measured and how - and let me know if you have any questions. Regarding the DAC’s output spectra, that’s precisely where you need to look to see how much of the target’s noise seeps into its analog output.

The jitter test output, which is very sensitive to jitter - a.k.a. timing errors - at the D/A stage, shows absolutely no difference with and without Diretta. So no, the test doesn’t fall short there.

Correction: no analog difference whatsoever were observed, with an accuracy that greatly exceeds the abilities of human auditory system.

I have no doubt Diretta changes all that you mention here. What nobody demonstrated so far is why and how those changes are impacting the analog output of a DAC, which is what eventually becomes sound waves. Trying ad nauseam to “educate” people what Diretta is or does in the digital domain serves no purpose.

7 Likes
24

Look at Uptone Audio - Svenson Explains:

[John Swenson's Tech Corner – UpTone Audio

Diretta relieves the PHY of the USB DDC.

25

When I woke up at 2.30AM this morning, I read the post from @Matthew_Arndt1 and was trying to compose a post along exactly these lines but, in my sleep deprived state I couldn’t put in the effort to make the point without the potential to cause offence.

This statement, as far as I’m concerned, hits the nail on the head.

No one has disputed what Diretta does with respect to network activity, timing and associated buffering strategies. The point of contention is that Diretta claim that these changes improve audio quality. I.e. change some aspect of what comes out of the DAC. The measurements performed refute that particular claim without needing to measure and validate that the host/target network activity itself is what Diretta describes.

With respect to the ‘what it does’, I am perfectly happy to take Diretta at their word.

The claim that it improves audio quality and the suggested mechanism by which it does it, however, is a different thing entirely an that is what the measurements and tests performed by @Marian address.

For my part, I am very glad to see these tests done even though, for me, they produced no surprises.

3 Likes
26

For me i don’t care how it works. I hear a massive improvement in my system. bigger than changing DAC. And everyone that came to listen said the same.

2 Likes
27

Note the timeframe of those posts and look at the jitter performance of recent USB DACs. It looks like these old issues have been solved.

1 Like
28

It’s placebo, that’s how it works.

4 Likes
29

Hi Marian — you’re right, and I owe you an apology. I wasn’t aware that Diretta’s own site makes explicit claims about reducing power-supply noise at the DAC, and I shouldn’t have stated that it never made those claims. I personally think that aspect is marketing overreach and not something Diretta can literally do at the DAC level. My intent in the thread was to focus on correct configuration and what Diretta actually changes in system behavior when set up properly, but in doing so I overstated my position. Thanks for calling it out.

5 Likes
30

We encountered this problem with jitter measurements years ago when comparing Linn streamers with those from LaRosita.

The often-modified Linns (improved voltage, better clocks) showed very good jitter measurements, but the LaRositas sounded clearer and more open. For a long time, the cause puzzled us; no experienced electronics engineer could imagine that the network transmission of the audio data was to blame. After all, this data is buffered, etc.

Jitter measurements likely refer to the quality of the actual music timing, not to the temporal composition of the “fluctuating” data.

31

What data “fluctuates”, how does it do that and what does “temporal composition” mean?

2 Likes
32

@Matthew_Arndt1, no need to apologize. Admitting mistakes is not easy, I appreciate that.

I agree, and since that’s the core premise of Diretta, everything about it is marketing. And the fact that they don’t have clear documentation and leave it to other people like you or David to figure out how it works - or how it shoud work - and come up with practical solutions and tweaks doesn’t do them any favors.

2 Likes
33

Take a PC x and connect a USB DDC.

Then change the configuration by splitting the load using Diretta. PC x will become the host.

The asynchronous USB DDC will show you the same jitter values ​​in both configurations because the clock speed within the device itself doesn’t change.

But it sounds completely different. Why?

34

Because you think it does. I don’t think it does, or that it can, so it doesn’t. And btw, you’re not answering my questions; you’re using terms you don’t bother to explain and then deflect with questions like “why is the sky green?”. And btw, “DDC” usually means “digital-to-digital converter”; I think you mean “DAC”. This is a technical discussion of my tests, so I insist on using accurate technical terms.

2 Likes
35

Diretta is an enhanced form of USB audio. USB audio stands for Digital-to-Digital Conversion (DDC). Digital audio data is converted into digital music signals.

A DAC converts digital music signals into analog music signals.

These are two technically distinct processes with different requirements.

36

It converts digital data to audio.

It stands for Universal Serial Bus.

Doesn’t have to have anything to do with USB connection. A DDC converts one digital in format to another. Ie. Toslink in and coax out. Or toslink in and i2s out. one if the inputs or outputs could be usb, but doesnt have to be.

And it has nothing to do with the dac stage.

6 Likes
37

I agree with Marian that Diretta’s closed business model, and lack of any meaningful documentation, has caused the community to distrust their claims and question users who have found Diretta to be a great solution in their system. Especially when many cannot reproduce their results.

Here’s what I can say about why Diretta can sound different

Diretta does not change the audio data or the DAC’s clock; it changes how and when data arrives at the DAC. Unlike similar streaming solutions that prioritize robustness and abstraction across diverse networks and devices, Diretta uses a tight, feedback-driven delivery loop between Host and Target. This loop continuously adjusts send timing based on the Target’s buffer state rather than relying on larger buffers to absorb timing variation.

As a result, timing errors caused by OS scheduling, IRQ contention, and network jitter are not masked but actively managed closer to the DAC’s consumption point. When the system is properly configured, this can reduce low-level timing variance at the DAC input, which listeners often perceive as improved image solidity, vocal density, and coherence. When the system is not properly configured, Diretta exposes instability rather than hiding it.

In short, most solutions are designed to work everywhere reliably; Diretta is designed to work optimally when the system itself is made deterministic. The audible difference is not about format or bandwidth, but about delivery discipline at audio-relevant timescales.

Diretta is based on documented standards for Real-Time Systems & Scheduling

Foundational sources
• Jane W. S. Liu — Real-Time Systems (Prentice Hall)
• Hermann Kopetz — Real-Time Systems: Design Principles for Distributed Embedded Applications
• Linux Foundation — Real-Time Linux (PREEMPT_RT) Documentation
• Ingo Molnár et al. — Linux scheduler and threaded IRQ design
Relevant concepts
• Deterministic scheduling
• Jitter minimization
• Priority inversion
• Threaded interrupts
• CPU isolation (isolcpus, nohz_full, rcu_nocbs)

And Closed-Loop Feedback & Control Systems

Foundational sources
• Franklin, Powell, Emami-Naeini — Feedback Control of Dynamic Systems
• Åström & Murray — Feedback Systems: An Introduction for Scientists and Engineers
• IEEE Control Systems publications
Relevant concepts
• Control loops
• Feedback delay vs loop stability
• Overshoot / oscillation
• Control authority windows

We can derive this from the parameters in setting.inf as well as the log structure. ThreadMode, InfoCycle, FlexCycle, CycleTime/CycleMinTime, periodMax/periodMin, periodSizeMax/periodSizeMin, syncBufferCount, Fragment are all based on documented standards. What I don’t know is what the system math is to pull this together as a cohesive system.

I’m also very aware that what I say about Diretta is doubted as I’m not a known entity nor affiliated with Diretta.

38

That’s an impressive collection of terminology and references! Alas, it does not address the fundamental problem: does this matter to the DAC? There are empirical results above that show it does not. There are also good reasons given for why it simply can’t.

3 Likes
39

The first statement is correct. The second is false.

With asynchronous USB, the DAC determines when the data arrives. And the how is not fungible. I am not sure everyone is aware of this, but with USB Audio a frame of data is always sent every 125us (micro-seconds), regardless of the output rate and sending device. The only thing that really changes is the size of the frame, depending upon the output rate. So that’s 8000 frames per second no matter what. Eg with 96/24, the 24 bits are padded to 32 bits and you need 96 bytes per packet.

Diretta cannot change either of these and has nothing to do with USB Audio per se.

Here’s a good primer on USB Audio with some clear examples. https://thewelltemperedcomputer.com/Lib/Fundamentals-of-USB-Audio_1.0.pdf

9 Likes
40

Those are fair questions. Because the DAC does not just convert data — it converts data arriving in time. Technically, this matters for three reasons:

The DAC input stage is time-sensitive even if the clock is internal. While the DAC’s master clock ultimately governs conversion, the input receiver (USB, Ethernet, I²S, etc.) must still handle packet arrival, buffering, and flow control. Variability in arrival timing changes how often buffers are filled, drained, or corrected. Those micro-corrections alter internal activity (buffer management, PLL interaction, interpolation timing), which can modulate noise and phase behavior upstream of the conversion stage.

Lower delivery variance reduces corrective work inside the DAC. When data arrives more evenly, the DAC’s input logic does less catch-up, throttling, or interpolation adjustment. That does not change the clock rate, but it reduces the number of timing-related housekeeping events happening near the conversion process. Less internal correction correlates with more stable reconstruction conditions.

Timing stability affects analog behavior, not just digital correctness. Even when bit-perfect, fluctuating digital activity changes current draw, ground reference stability, and local noise inside the DAC. A steadier delivery pattern produces a steadier internal operating state. Listeners perceive this not as “more detail,” but as improved image focus, vocal body, and decay continuity.

So Diretta doesn’t improve the DAC clock. It reduces how much the DAC has to react to upstream timing irregularities. That distinction is important — and it’s why Diretta only shows benefits when the Host, OS scheduling, IRQs, and network path are already under control.

41

With USB asynchronous audio the DAC does not react to the upstream timing irregularities.

It observes these irregularities (likely by monitoring internal buffer levels) and tells the upstream device to modify it’s behaviour by increasing or decreasing the number of samples in the USB frame which is regularly spaced in time.

This is the strength of USB asynchronous audio. The DAC is in control and it makes sure that it gets what it needs to keep its buffers filled to levels that it (the DAC) determines such that there is always a sample available to be presented for conversion when required by the DACs sample clock. In this way, the absolute clock accuracy and even the clock stability of the upstream device are isolated from the digital to analogue conversation and become irrelevant.

Diretta does nothing to change this. Indeed it can’t because to do so would require a different USB audio protocol and your USB DAC would have to support that protocol.

5 Likes
42

Are you generating this stuff with an LLM? I don’t believe a human can produce these amounts of impressive-sounding text while not having any actual understanding of what’s being said.

6 Likes