I used to think the DAC was 90% of the sound. This sub-$300 DIY project proved me wrong

Tinkering
49

We’ve strayed quite far from Edison’s cylinder with stylus and horn - all effectively analog (no bits). Tape capture with silver oxide, if I’m not mistaken, is degrading the pure analog a “bit” or bits. And the way we transmit 1’s and 0’s is done over an analog wave, whose form, as you suggest, gets mangled in various ways, not least by shortening or lengthening the gaps between and in other ways. Normal data processing can deal with this, but audio is an accomodation.

I look at digital as in a still immature stage of development compared to other technologies in our hobby. Solid state still has a ways to go it’s way to maturity, it seems. So, the Diretta folks clearly have a point to their approach. Maybe designers of all components in the chain will eventually settle on simplified and better standards.

Yet when it comes to our sense of hearing, we’ve got far, far more to learn about how the ear/brain processes musical reality. Same applies to how our memories capture the visual, sonic, and olfactory sensations of our lives.

I’m interested in eventually pursuing Diretta, though I’m not eager to make it into any time consuming project (I’m not the tinkerer I used to be). I’d be more willing to get an off the shelf solution if one is or becomes available.

1 Like
50

While not entirely tinker-free, it’s pretty close. SOtM has enabled Diretta functionality in recent firmware versions (since April of this year) for their sMS-200 family of products. I know a number of folks now who are using the sMS-200 Neo for the Host and sMS-200ultra Neo for the Target. In addition to these two components, you’ll need a crossover network cable (SOtM sells one, but a crossover adapter may be less expensive).

SOtM provides one-time setup instructions here:

Considerably less effort than my 66-page document!

1 Like
51

Thanks. I’ll look into this. By the way, you might find this podcast series of interest https://podcasts.apple.com/us/podcast/unexplainable/id1554578197?i=1000735010354. A 4-part series on our perception of sound.

2 Likes
52

Hi Gents, I’ll just note that the instructions from SOtM are a little vague in places, I’ll do a write up from what I learned and offer it to May to see if she wants to use it

I’ll add it here too later

Not sure if I mentioned before on here, you need a USB/Ethernet which has an RTL8153 chip in it - very cheap ~£6/$8 on eBay

1 Like
53

Looking at what SOtM has to offer I’m not sure it’s a route I’m interested in. Since all I’d want is to use Diretta between Minimserver on my NAS and my dCS Upsampler and use JPLAY to manage the library, it would appear your original HW setup is the way to go. Not at all interested in abandoning my existing Roon Server to Grimm “Roon Ready” endpoint.

I suppose this is possible with the right setup scripts and configuration settings, right?

My current experimentations comparing Roon → Grimm → sdCS versus Minimserver → Grimm → dCS using mconnect lite suggests the latter is slightly better, though I’d be hard pressed to express in just what ways…so far. Eventually I’ll install JPLAY on my iPad to determine if it is a worthwhile rival to Roon, and especially if it sonically betters the Roon client, which I tend to load up with a few albums or set on radio and then kill the app on the iPad.

54

Hi @stevebythebay,

Ah, thank you for clarifying the “Roon Server” part. That makes the architecture much clearer!

To answer your technical question first: Yes, the AudioLinux OS on the Diretta Host does support UPnP renderers. So technically, you could configure it to accept streams from MinimServer/JPLAY just like you do now.

But before we go down that rabbit hole of configuration, I have to ask: Is that actually what you want?

You mentioned that you currently prefer MinimServer -> Grimm over Roon -> Grimm because it sounds “slightly better.” That is a very common experience with standard network setups, and it makes sense why you’d want to keep that sonic advantage.

But let’s imagine a scenario where Roon -> Diretta sounded better than your current MinimServer setup.

  • Would you still want to juggle JPLAY, MinimServer, and the iPad app?
  • Or is your goal simply to get the best possible sound, ideally using the richer Roon interface you already pay for?

The reason I ask is that the entire goal and purpose of my Diretta project is to close that sonic gap for Roon. My suggestion would be to build the Diretta path strictly for Roon first and see what happens.

You can run it right alongside your current setup (as a second Zone/Input) without changing a thing. If Diretta does its job, you might find that you don’t need the complexity of UPnP to get the sound you want.

Does that make sense as a starting point?

55

My own quest for a sonic step up over the electrically-busy Roon via a cleaner UPnP or Lyrion server was derailed when I found that David’s 3-tier setup made upstream network changes/investments moot.

I can do without my much-loved EtherREGEN or LHY/Finisar fiber segment now that Diretta delivers just the music, with unprecedented clarity, to the DDC converting Target’s USB into AES/EBU for my active speakers.

2 Likes
56

Certainly my goal is to have all my sources (local music files, Qobuz, and internet radio stations) sound the best. Today using Roon I can tap into all these and even send them about my house to various zones.

But, as I understand it, what the Diretta protocol is doing, when compared to Roon, is to severely reduce the bursty internet activity, that adversely affects a DAC’s processing of the bitstream. It’s upstream source and transport level functions aim to reduce noise and data delivery.

I’m not clear on how Minimserver is doing flow control. And since I’m currently using the free version, I know that’s not something I have any control over. I believe this is accomplished via MinimWatch. In any case, it’s likely there’s something different in just how it’s managing packets (size, flow rate, etc.). On the other hand, the Grimm MU1 is operating like a CD player, synchronously with my DAC. I know that my dCS DAC has it’s own methods for dealing with timing issues (jitter) as well as clocks and filters.

Seems the only way to test it is to build your configuration and try it. If it’s clearly better then it would prove a good investment, Otherwise, a waste of time and money.

Would love to “go to school” on anyone else’s experiences with UPnP versus Diretta, but I’ve not been successful at discovering anyone who’s made the journey.

57

Hi @Brad_Burnside,

Thank you for chiming in! That is exactly the experience I was trying to describe.

The fact that your quest for “cleaner UPnP” was “derailed” by Diretta is the key takeaway here. Why chase a different protocol (and app ecosystem) if we can just fix the sonic issues with the one we love?

Also, your mention of using a DDC to convert the Target’s USB to AES/EBU is perfectly timed. It proves that even if one requires an AES input (like for @stevebythebay’s dCS), the Diretta architecture can still deliver that “unprecedented clarity” right up to the final conversion step.

It’s fascinating (and validating) to hear that it made your etherREGEN and fiber segments moot. That “dome of silence” theory seems to hold water!

58

So I have to say that @David_Snyder invited me along on this ride early on when he was first thinking about trying it. And I didn’t go on the ride, because I was busy with other stuff. Rebuilding servers in my house, getting my daughter into college, figuring out what speakers to put in my barn (current leader is tekton double impact, but might be klipsch heresy’s, we shall see). Anyways, I didn’t dig in. And now I’m sad I didn’t because y’all have been on quite an adventure.

So what I’m hoping is that I can benefit from all the work you’ve done and I can have the “lazy cheater’s” guide to diretta so I can use it - ideally in my HQPD / Holy Cyan 2 chain. But if I have to cut out HQPD in order to try it, so be it. What’s the simplest place to start given all you have learned?

2 Likes
59

Welcome! Besides buying a pair of Raspberry Pi computers (4 and 5) and going through at least the first 29 pages of my 66 page guide, you could go the SOtM sMS-200 route. Or buy the parts and have me assemble them for you. No change. I just ask that, if you like the sound, you’ll help me to spread the word. It’s still early days for something great.

60

Is it possible to check it out without buying a license with some other linux, or is AudioLinux mandatory?

61

Both Audiolinux and GentooPlayer offers some form of trial before engaging in a Diretta license. I don’t remember the limits while trialling though? 30minutes i think in Gentoo and perhaps only RedBook resolution in Audiolinux?

62

That is a fair question. Unfortunately, Diretta is not open source, so the company only provides code to integrators under NDA. For the Raspberry Pi, the only two options are AudioLinux and GentooPlayer.

While AudioLinux does not offer a free trial, the ~$69 license covers unlimited Raspberry Pi units within your household. This is critical because my Diretta project
requires two computers (a Host and a Target). The single $69 fee covers the OS for both.

The Good News (The Diretta “Trial”):
You do not need to pay the €100 Diretta Target license to test the sound. The free/unlicensed version of the Diretta protocol works indefinitely but is simply capped at 44.1 kHz. (High-res playabck works but halts after six minutes.)

My Recommendation for Evaluation:

  1. The Entry Fee: Purchase the AudioLinux license ($69). This is your only “at risk” capital.
  2. The Build: Configure your Host and Target Pis as documented.
  3. The Config: In Roon’s MUSE DSP settings, enable “Sample Rate Conversion” to downsample everything to 44.1 kHz.

This allows you to fully evaluate the “Diretta sound” signature for as long as you like.

The “No-Regrets” Fallback:
If you decide Diretta isn’t for you, that $69 wasn’t wasted. You can simply reconfigure those two Raspberry Pis as standard Roon Bridge endpoints. In my experience, Roon Bridge running on AudioLinux sounds superior to general purpose Linux distributions anyway, so you still come out ahead.

Three Ways to Minimize the Financial Risk

If the upfront cost is still a barrier, here are three strategies to lower the effective cost of trying this out:

1. The “Resale” Safety Net
Raspberry Pi computers hold their value incredibly well. Market data shows that used Raspberry Pi 4 and 5 models frequently sell for 70% to 90% of their original retail price. If you decide to abandon the project entirely, you can sell the hardware on the used market and recoup almost all of your hardware investment. Your actual financial exposure is effectively just the software license.

2. The “Group Buy” Strategy
Since the AudioLinux license covers the entire location (household), you could split the initial hardware and OS cost with a local audio friend. You can build the ‘stack’ (Host + Target) together and pass it back and forth to evaluate. If you both like it, the second person just buys their own hardware and OS license later. If neither likes it, you sell the hardware and split the small loss on the OS.

3. The “Club Kit” Approach
If you are part of a local audio club, propose this as a club project. The club buys the AudioLinux license and the two Pis to create a “loaner kit” that members can check out to evaluate Diretta in their own systems. This spreads the $69 OS cost across the entire membership, making the individual cost to try it negligible.

Edit: AudioLinux is now $79. Prices subject to change, so check the official site for the latest.

2 Likes
64

Just wondering but are you still using your etherREGEN switch to receive music before sending it on its way via your Diretta setup? And if so, is that out from the “B” side? The underlying question, as you’d guess, is whether the etherREGEN makes any difference.

1 Like
65

Hi David. I read somewhere that the cap is 48kHz?

1 Like
66

You are technically correct! I just checked the official Diretta documentation, and it explicitly lists both 44.1 kHz and 48 kHz as the “bottom Hz” rates that work indefinitely without a license.

That said, I usually recommend locking it to 44.1 kHz for evaluation simply because the vast majority of source material is CD-quality (44.1 kHz). Keeping the conversion at 1:1 (or an even integer downsample if you have high-res files) is generally cleaner than resampling 44.1 content to 48.

But yes, if you have native 48 kHz content (like video concerts/rips), that should play fine without the time limit.

67

I bought it when first introduced in 2019, and have experimented with placement among three systems since then, along with a variety of different power supplies, cables, fiber and other doo-dads.

It was the last network device before my main endpoint (2.1 system, featuring Meridian DSP9 speakers) until I implemented the separate Host RPi scheme and discovered that the ER had become superfluous. So I moved it into my home theater (“A” RJ45 in, “B” out), just ahead of a Meridian 218 controller. Still a worthy device.

1 Like
68

I pulled EtherREGEN out of the network path because someone had said that Diretta worked best at gigabit speed, and I wanted to test that idea without ER’s 100 “B” port in the mix. Made no difference, in or out.

I can’t confirm or deny whether Diretta wants gigabit.

69

Interesting. I thought about using the etherREGEN in my video system. But since the “B” side max is 100 Mb and the NAS housing my video library has many UHD 4K titles. I’d likely run into inadequate bandwidth issues. So, I’d been awaiting the next version which replaced the “B” side with gigabit, among other things. Will be interested how my system will respond to the Diretta setup.