Diretta is licensed software that quiets network activity by sending steady, uniform packets instead of the usual “bursty” electrical/cpu spikes. I have been using it to good effect for a couple of years, but have now taken a big sonic step up with a simple DIY addition to my system.
The idea is to implement this software using two Raspberry Pi’s in what the designer calls a three-tier system: Server > RPi Host > RPi Target > DDC or DAC with the Pi’s on their own isolated direct link. In my case, a NUC running Roon Server connects over RAAT through my final LHY switch to a Ethernet-to-USB3 dongle into a RPi4 with Diretta out to the second RPi4. USB is then converted by a Singxer DDC into AES/EBU for my Meridian DSP9 active speakers.
This scheme works just as well with Audirvana, Lyrion etc. as it does with Roon, which inspired the idea due to that player’s extensive network activity -to the detriment of sound quality according to many.
So for example, for Roon RAAT, stream goes over Ethernet to RPI Host, then it’s transformed into USB on RPi Host and then how is the RPi Host connected to the RPi Target? Then another USB to Singxer to convert to AES/EBU?
Are you sure a good streamer with AES/EBU to DAC wouldn’t trash that?
No, it is not converted to USB on the Host RPi, which is connected by a super quiet link to the Target RPi where USB output happens. Target could be fitted with SPDIF, AES or i2S instead if you prefer.
I have used a number of quality streamers (and conventional Diretta implementation) in the past, but they have never produced the kind of clarity and immediacy I enjoy today.
In the Diretta world, their Host software usually runs on your server (Roon, in my case) to send more uniform, less “bursty” packets to a Target, which is software to receive those packets on your streamer.
In this scheme, a separate RPi Host machine is connected to your server using a Ethernet-to-USB3 dongle to input the stream on the Host, and then uses the Diretta protocol over a short cable -completely isolated without any other network activity- to the streamer.
This looks interesting but, being a Naim streamer user, I’d need an Ethernet connection from the target Pi rather than USB. Is that possible? Many thanks
No, I don’t think it’s possible without the final network device (which receives the Diretta packets) running their licensed Target software. There are multiple Linux operating systems, including GentooPlayer and Audiolinux, which can incorporate this option, and run on many different transports but not Naim.
I get it now, the Host to Target leg is ethernet over direct UTP. What I’m worried about is the quality of Target RPi and final Target to DAC connection.
Depending on the DAC/DDC, the USB connection is usually the worst. It’s asynchronous and even with reclocking, susceptible to jitter. It is greatly influenced by power supply and source clocking. Even with this Diretta protocol handling Host to Target transfer, you still have this final connection as liability.
I find it hard to believe that a cheap RPi in the last leg to DAC will preserve whatever the Diretta is positively doing in the leg before. You would need a better Target with a OC clock, LPS and reclocker to make this worthwhile.
In your specific case, you have a DDC which usually reclocks inputs and provides much better final digital connection to the DAC, so I’m not sure what improvement could Diretta add to that? Maybe minor, but the heavy lifting is on the DDC anyway.
Of course, experiences may vary, but I believe you may have a false positive experience because of the DDC. Try it with direct cheap Target to DAC and see then.
Hmm, but at least when the host device is connected direct to DAC Audirvana doesn’t have network activity while playing, as I understand that it loads the entire music file into RAM before playing and if the host device is dedicated then that won’t be doing anything on the network either. This is probably one reason why Audirvana has always sounded great.
Whether Audirvana or Roon or playing from memory, we are still sending packets over our local networks, which carry traffic beyond the music itself such as iot broadcasts. Plus, the electrical activity generated by the usual send/receive process also impacts our DACs’s ability to deliver everything in the recording. When we eliminate bursts and block everything but the music, more of the fine details are revealed.
As far as the quality of RPi hardware is concerned, I have compared what can be realized using credible power and isolation etc. (Shunyata, IanCanada and Singxer, in my case) to excellent electronics from companies like dCS and Meridian to find that these cheap little SBCs can compete.
I was made aware of this new approach (a separate Diretta Host on the network with an isolated link to Target) by David Snyder, who explains how to implement an inexpensive and effective filter
What are you listening through? The DAC is the critical component in a digital front end, and some are far more susceptible to, for example, noise superimposed on the digital datastream and AFAIK all renderers, whether part of a streamer or in a separate player, fed from a directly connected store or over a lical network or internet, feed a continuous real-time digital music datastream to the DAC.
There are four DACs feeding six amplifiers inside each of my Meridian DSP9 active speakers, which accept up to 192kHz AES/EBU input. They call their new software and electronics platform Extreme Engineering (XE), and claim that it rejects noise, but sounds even better to me with the Diretta filter in place.
You clearly stated your belief that a Raspberry Pi used as Target is not up to the task:
And you challenged me to directly connect the RPi to my speakers without the DDC (which converts USB to the AES they need) for some reason, possibly to prove your point?
I have said that giving it good power etc. does, in fact, elevate this humble SBC to audiophile performance as part of a system with Diretta implemented as described and using the other equipment I am personally listening to in my own room. That’s all.
I am not selling anything. My goal was to share a discovery, developed by someone else who also has no profit motive, that makes a real difference in a hi-end system for few dollars. You can try it, or not.
Maybe the details about which cables and power supplies I’ve used to enhance SQ from a Raspberry Pi would alleviate concerns that it is too inexpensive to sound really great. (Best to try it for yourself, though…)
My RPi4 Host uses a 5v iFi iPower2 and is connected by Network Acoustics eno2 Ethernet filter to a supercapacitor-heavy IanCanada 5v powered RPi4 Target, both running Audiolinux, then LHY 7.5v LPS for a Singxer UIP-1 PRO isolator and 9v Farad Super3 with SR Purple fuse for my Singxer SU-6 DDC, both of which use Ghent star-quad umbilicals. I have a Shunyata Triton V3 power distributor for all of these with quality cords from Furutech, Shunyata and Zu.
So, I’ve spent a little more than just $50 for the RPi4 and $70 for its iPower2. But all of those other hifi products were in place before this new arrangement, and may, or may not, be essential to the success I enjoy today.
I actually removed my LHY FMC, Amphenol dac and EtherREGEN from the network upstream of the new Diretta Host because its totally isolated filtering made them superfluous.
The point is that I made a modest change to a very good-sounding setup which result is so gratifying that I wanted you to hear about it.
@MeridianDevo May I ask for some more info on your equipment setup please:
What case are you using with the RPi target given you are using the IanCanada power supply?
What usb power adapter are you using to connect to the IanCanada power supply?
What brand USB cables are you using from the RPi target to the Singer kit
Hi Jack! The IanCanada battery/supercapacitor device powers RPi4 from below using spring-loaded connectors, so a case is problematic. Because of this, I have two sheets of 3M EMI absorbing material top & bottom (as a sandwich) wrapped in Faraday RFI fabric. Ugly, but it works.
