DIY Streamer: In search of one ultimate streamer...

In principle a standalone streamer(with no DAC or Amp) is a box, takes digital signal in some form and outputs a digital signal for an external DAC. But IME, that one box is anything but simple. The challenge with that one box has two parts
1. The hardware itself: The types of power supply, the type of clock, the type of power filtering, the different types of inputs supports, the different types of outputs it supports, then crucially the frequencies it supports. Does it support upto 768K PCM and DSD1024 (for the folks who are upsampling or would at least like to have that option for future).
2. The bloody and messy software ecosystem. Roon does as OK job of remote playing and network support, HQPlayer with NAA supports upsampling. But what if you want to play Apple Music. The whole ecosystem is a mess, there is just no single software answer for all use cases. In the end many people choose to stick with a workflow that works but if you want to expand your use cases of playing music the combination of hardware and software becomes severely limited.

I imposed many of these limitations and was just left with a single choice Holo Red Steramer. I sometimes use HQPlayer for DSD1024 (i have Holo May lvl3). I also use Apple Music because Qobuz's library sucks. Holo Red is a RPi based streamer which works wonderfully well and sounds pretty neat. I used it for 6 months without a hitch. It has multiple inputs/outputs. Can act as DDC, Roon Endpoint and HQPlayer NAA endpoint. Supports DSD1024 via USB and all PCM frequency ranges. I also optimized its ethernet input with optical isolation. In my journey to Holo Red I passed on vinyl and dante protocol setup that @atomicbob suggested. These two options sounded great but for me these two mediums were limited for my streaming use case. I don't want to own music files either. That doesn't suit the experience i'm looking for.

Then in DIY community i read few things
1. RPi introduces a lot of unwanted noise to the signal. People have gone to great lengths to shield the RPis in their setup. Those hacks are almost comical to look at even for audiophiles. Although in Holo Red they don't use the standard RPi outputs, the RPi board itself is a great source of EMI. But with the casing of Holo Red this should not be an issue.
2. The re-clocker circuit and the clocks used have a very huge impact. and Holo Red uses a cheap clock and their re-clocker circuit is not SOTA (according to DIY community).
3. Power supply have equally big impact on the output jitter o the streamer circuit. For the money Holo Red sells for it won't be possible to have truly great power supplies to all different modules.

This led me invest in a DIY streamer. My build goals were
1. Ability to try different clocks (to get some understanding of how clocks make a difference in digital sound).
2. Ability to upgrade power supplies to various modules.
3. Support frequency ranges I wanted for both PCM and DSD and be able to change. Lower frequency clocks tend to have lower jitter. So if let's say just 44.1K then you could prolly get better sound with a lower frequency clock.
4. Keep RPi away from the streamer. But sill have the ability to use RPi. Basically use Holo Red for the software support part.

This led me to various DIY components from IanCanada [1][2]. My reasons were
1. I did not have any prior DIY experience and not any EE background. I don't even quiet know soldering.
2. His modules are well documented and support from DIY community is really great.

NOTE: From this point on, my words may seem like advertisement of https://iancanada.ca. My Intention however is to share my learnings from last 9(6 with Holo Red and 3 with DIY streamer) months. I don't yet have experience beyond his modules. I hope someday to build more complex projects on the digital side.
Don't for a second think this is cheap. When you stack of lot his components it will become expensive.

I have built the streamer in multiple phases. So that i can understand the differences better.
1. Decent power supply with stock clocks. Apparently this clock in the stock DIY streamer is the same quality as Holo Red and is meant to be replaced. The power for DIY streamer supply according to DIYers is better than Holo Red.
2. Decent power supply with great clocks.
3. Great power supply with great clocks.

Subjective Impressions of DIY streamer vs Holo Red
I use two Dacs
1. ESS Dac with tube output stage for PCM.
2. Holo May lvl3 for DSD (mainly DSD512, I still don't quite like DSD1024).

Sources for Holo Red
1. IPad for streaming. USB C -> USB B Holo Red.
2. HQPlayer NAA via ethernet for DSD.
Source for DIY streamer
Holo Red's Coax RCA output (as DDC) to DIY streamer's RCA input. As I still need to add USB board to my DIY streamer. Holo May Dac is NOS mode with no upsampling. HQPlayer not used for this comparison yet.

I still have not added HDMI output to my streamer, so I currently am not able to play DSD out. So all my comparisons are currently for PCM only in HiRes from Qobuz and Apple music.
Headphones used : Focal Utopia and ZMF Atrium Open, The impressions below are bit of an average from both headphones because their very different strengths.

1. Decent power supply with stock clocks vs Holo Red
The DIY streamer is slightly more of everything. More micro dynamics. Slightly better imaging and very slightly wider sound stage. I spent 2 weeks on this setup. The BOM here is 599$ plus shipping. Slightly better performance than Holo Red. Some key notes:
- The holo red seemed more sibilant on Utopia after listening to DIY streamer.
- The background of Holo was slightly darker compared to DIY streamer.

2. Decent power supply with great clocks vs Decent power supply with stock clocks
- Upper frequencies rolled off a little or sounded so.
- The resolution of the music improved a lot. The micro dynamics popped out more.
- Vocals were presented a little a little upfront.
- Imaging became much sharper, especially with Utopia
- The sound very organic and for the lack of better term very analog like.

3. Great power supply with great clocks vs Decent power supply with great clocks
- Sound has more weight and denser. Reverbs are more tangible.
- The low level and bass especially is very addicting. I have heard Atrium on a lot of systems but I don't quite remember it doing this well in the bass and sub base region.
- I can play at higher volume with my ESS dac without feeling exhausted.

 

Basic design principles
- Power supply is critical. Moreover, passive power supplies (LiFePO4 batteries and UltraCapacitors) are considered superior to LPSUs connected to mains. There are many reasons that are documented in the DIY community. I'm not an expert yet, so i'm taking it as the truth for the moment. And my listening also backs this claim.
- Re-clocking circuit and Clocks are equally important. There have been some recent advancements made in this domain. In fact there are highly revered clocks and ppl in DIY community have favorite clocks! There is this fabled Pulsar clocks from a genius name Andrea Mori that are now unobtanium and sound very analog (maybe someday i'lll hear them). Then there are more commonly available Accusilicon clocks that are supposed to sound cleaner but has digital and analytical sound and so on.
- Most streamers these days have the concept of dirty side and clean side. Where the input signal from the source is considered dirty and the output of the re-clocker is considered clean. So the power supply on the cleaner supply should be as high quality as possible.
- Optimum performance can be reached by providing individual power supplies to various modules. This can be pretty expensive and can take a lot of space for a streamer only box.

Cons of DIY streamer
- Deals with very high capacity capacitors for passive power supplies. 400F-3000F (yes not millifarads but farads). For novice DIYers this means more caution is needed.
- No off the shelf enclosure available.
- Some things can be tweaked and improved as reported by DIY community. So one can become pretty obsessed
- If you want to go for highest quality power supply it take a lot of money and space. There is really no need for this. You can stop with what I have currently and call it a day (seriously). The DIY streamer I have can rival almost any other streamer solution (without DAC). I still don't know where the diminishing returns start. But I do want to see what higher quality power supply would unlock or if its just voodoo.
- You still need some source that generates the music. Be it ipad/iphone or laptop. This is a DDC only platform without RPi. This is my design choice though. You can add RPi to this setup if you want.

Various Modules Used
- PurePi II 5V+3.3V ultracapacitor/LifePO4 battery power supply combo - Provides 5V to the dirty side using ultra capacitors for the dirty side and 3.3V using LiFePO4 batteries for the clean side.
- UCConditioner 3.3V : Stabilize the 3.3V output of the PurePi II for the clean side. Improves the tonality and bass and sub-bass response quite a bit.
- ReceiverPi DDC (RPi free mode) - This module accepts various digital inputs. HDMI, USB (optional and I will add this later), Coax (RCA and optical).
- FifoPi Q7II (has upgraded clock)- The all critical reclocker circuit. Also houses new upgraded high quality clocks from Ian. So if you buy these now, you might not even have to upgrade the clocks. But the SCPure clocks are measured to be of lower phase noise.
- SC-Pure Extremely-low Phase Noise Audio Clock - I had the old version of FifoPi Q7 in which the clocks were not of great quality. The SCPure is objectively second best clocks measured, behind the fabled Pulsar clock. If you just want a streamer which is above Holo Red quality you may want to skip this clock. The default clocks with the new FiFoPi Q7II should outperform Holo Red.
- TransportPi AES, Flagship AES/SPDIF transport - The output stage after re-clocker phase. The master clock signal MCLK is slaved to the output of FiFoPi Q7II.

BOM
- Decent power supply with default clocks (PurePi II 5V+3.3V ultracapacitor/LifePO4 battery power supply combo , ReceiverPi DDC (RPi free mode), FifoPi Q7II (has upgraded clock), TransportPi AES, Flagship AES/SPDIF transport) - Project - 599$ plus shipping.
- Decent power supply with great clocks (Add SC-Pure 45.15K and 49.15K clocks to the above setup) - Add $358
- Great power supply with great clocks (Add UCConditioner 3.3V to the above setup) - Add 108$

Add shipping.

Build Instructions
- This guy Gabster has a great youtube channel. For this particular build see this video.
- The DIY thread.

 

Is this the best way to do streaming? Not quite sure yet. But I'm realizing DIY route might be more fruitful in return for a lot of hard work to realize the sound signature i'm looking for. And allows me to experiment with what else is out there. I can't imagine the sound I have never heard. Who knows what all exotic sound signatures exist out there.

Why advocate this build? The DIY community is more interested in objective discussions. They do make subjective notes on the sound but that reads a bit superficial. Maybe with this kind of lego like assembly (no soldering needed), we may get more audiophile review/take on DIY components.

Any more changes planned for the streamer? Yes
- Upgrade the clean side 3.3V power supply to UCPure MKIII with balancing board.
- Upgrade 5V dirty side power supply with LinearPi 5v and UCConditioner 5V.
- Add USB input.
- Add support for I2S output with HDMI for DSD using HdmiPi Pro II Flagship HDMI transport interface


But this is good starting point. I think a DAC will benefit more with clock and power supply improvements. So eventually I want to build the full stack dddac . According to DIY community this dac with fully maxed features (power supply, clocks and output stage transformers) pips Chord Dave. The build is challenging so this is good learning bed for me. I first would like to build a high quality TDA1541A DAC.

 

This is an interesting thread.

In some ways this kind of thing has been around quite a while. I have an old XMOS DDC from Breeze audio (the DU-U8) that I used with Theta DACs that don't have USB input. That was almost ten years ago now (jeez, that's crazy). I actually bought the Theta Pro Basic from Zach at ZMF when he was just modding Fostex cans. Anyway, I spent a few hundred on nicer clocks, and I remember it sounded really good. I'm sure there's a lot to be gained from upgrading the power supplies. But I'm curious if there are other major improvements from overall topology, etc, or if this is fundamentally the same thing just better components.

I don't really understand why someone hasn't yet made an "end-game" streamer/DDC with TOTL clocks and PSUs with the software all worked out so someone can stream tidal, etc, and control it with their phone via bluetooth. Even if it's 2 separate cases: one for DDC, one for the streamer side with wifi/bluetooth, etc.

Regardless, it will be interesting to see how things change for your sound if you're able to get the USB board added to your stack and get the Holo Red out of the chain entirely.

 

My knowledge is still limited as I'm learning so take my words with a grain of salt.

If we remove the clocks and the power supplies out of the picture, then in terms of topology we have the following choices -
- The interesting part is the output I/V stage of the DAC (that have current mode output). The choices are quite vast here - active vs passive, discrete vs opamp etc. Most opamps are not really well suited to be in the I/V stage. This article from Nelson Pass goes a bit more deep into the weeds. Then I/V stages with tubes etc. Unfortunately the I/V stage would really depend on the end user, what his sonic preferences are.
- Discrete R2R topologies - Most tested discrete R2R dacs are linear upto 14 bits. Exceptional R2R dacs can stay linear to upto about 16 bits. So there is some work needed here. As I understand this is still an open problem (but I could be wrong or have limited knowledge here.)
- For chip based DACs like ESS, AKM etc, the DAC board itself seems like a solved problem. Throw good components at it to get better sound.
- Digital volume control might be another area where different topologies may play some role.

 

First of all, congratulations on your streamer. Jumping into a project like this and making it work is a real accomplishment. Also, I am not an expert either, so my words may be wrong too.

My understanding was the Pulsar clock was not by Andrea Mori, but Andrea Mori and his clocks, dacs, and other solutions are alive and available at

https://www.thewellaudio.com/

I have a dac project from Andrea that I haven't put together yet using his Driscoll oscillator masterclock where the "5" MHz clock has a measured phase noise of -150 dB at 10 Hz. And he is always working on something new. That's not to take anything away from IanCanada. They both make great stuff.

As you state, a discrete resistor R2R dac is typically limited to 13 or 14 bits of linearity. This is entirely based on the manufacturing tolerance of the resistors. Mori addresses that with a calibration technique that measures the tolerance error of the completed dac board and adjusts for the error. He claims 23 bit linearity with calibration for his Dac Lite dac. I'll start a thread on the Mori dac once I get some time to put it together.

Your project brings up a question that I don't know how to answer. It's a little like putting optical isolation between the router and streamer. From what I understand, the streamer decodes the clock signal and doesn't rely on the router's clock for anything so noise on the router shouldn't matter as long as the data can be read. That said, lots of people have put optical isolation (read fiber) between their router and streamer and heard worthwhile improvement in sound.

The other question is how can a better streamer clock improve sound if the dac strips the clock channel and reclocks the data (not all dacs do this). I have experienced a streamer improving sound that I assume is due to a better clock but I can't explain why. That brings up the question, is a high quality master clock best used at the streamer/DDC or at the dac?

This is fun. Keep up the good work.

 

This was my question, which leads to:

Based on the assumptions (about clocks/asymetrical-data-streams/dacs) above the answer would be dac, and that any improvement from different/betterer streamer must be based on less EMI/"noise" being passed to the dac and thus not interfering with it.

What am I missing?

 

As far as I'm concerned, you are not missing anything. But then I'm more mechanical than electrical so I don't know enough. If you are a network expert, then cool, you answered my questions. If not, I hope someone else joins in.

@neo_the_one - Sorry to have sidetracked your thread but these seem to be part of your quest, so I don't feel too bad. Ihope you come back to this thread when you try other options.

 

Based on my rudimentary understanding, I was always kind of confused by transports having an impact on sound. The explanation of jitter is probably the prevailing one, but I was still always confused about why transport jitter mattered to the dac. it should be possible to completely fix transport jitter. But alas, for many years I also heard differences between different transports with DACs.

Enter my purchase of a Chord DAC, the Dave, where jitter is irrelevant, Rob Watts actually encourages people to use toslink. because if jitter doesn't matter, which is the primary knock on toslink, than what you are getting is the main upside of toslink, which is perfect galvanic isolation. This is born out in my own listening with Chord DACs, using toslink is the best the DAC can offer, with other inputs only potentially being as good as toslink.

This gets to the next way that transports can impact the sound, and that is noise from the transport getting into the DAC. This can also still happen with Chord DACs. As an aside, Rob Watts is always going on about the importance of blocking RF noise, even up into the GHz region, I can't say how true or not this is, but the explanation does make sense to me that the RF noise can have an outsized impact on the analog electronics in the device.


So all that said, I think both jitter and electrical noise can impact the sound with transports in most DACs. most Dacs do not "solve" transport jitter. And whether or not I think they all should be doing so, now that I have heard some that do, doesn't really matter as not everyone likes Chord DACs(some others may also do this).

 

Thanks for the encouragement. I indeed want to build a maxed out dddac where the clocks would be part of the dac and power supplies directly powering the said dac.
But before that I need to learn more with simpler systems and get an understand of various components. In the coming months I will make multiple changes/upgrades to this streamer and will eventually also add a ESS DAC for comparison. Hopefully might get to try out a few different I/V stages as well.

 

Cables?

 

For interconnects i use Nanotec. It smooths the sound and takes some edge out on my system. But recently they have become super expensive. I will learn to build these interconnects myself from bulk cable. That should lower the cost a bit.

 

How about from the streamers to DAC?

 

For now i am using one the RCA interconnects for Nanotec as digital coax cable . I do like how it sounds. Might continue with it or get a different coax RCA cable. But I'll do that once i have more clarity on what the final system would look like.

 

How are you handling streaming with Apple Music?

 

For now iPad USB C -> Holo Red USB B. This setup is ok for now but if ever I have to go remote I will most likely have to go to Tidal Connect since Apple will never release similar feature.

 

Cool thread, thanks. I vaguely remember when the IanCanada stuff started becoming available and folks were pointing to it on other forums, but completely forgot about it. If the boards were already totally populated with no soldering required I would probably give some of this a go, but as it is I'd need to improve my soldering skills before taking a crack at a build.

The real attraction with this stuff is being able to get features that match kilobuck (or, really, multi-kilobuck) pre-built transports/servers/streamers for less. The market for transports/servers/streamers upstream of a DAC has gotten significantly more flush with options than, say, ten years ago, but as it is if you really want to play with esoteric power supply and clocking stuff integrated into a streamer, this is probably among the cheapest ways to go, otherwise you get into nutty costs for a pre-built solution.

 

Regarding clocks in a streamer/DDC:

They shouldn’t matter at all if you’re using USB to connect the DAC to the streamer. All modern USB is asynchronous, meaning that the data is sent in bursts, put into a buffer at the receiving end and will be sent from there to the DAC chip using the DAC’s clock. So all current USB implementations are actually reclocking the signal. (Electrical noise entering the DAC via the USB cable might still play a role, so a streamer/DDC can help, anyway. Also, there theoretically is synchronous USB audio, where a continuous stream at a defined data rate is used, and there the signal clock could be used by the DAC. This is a very old standard, though, and isn’t used in the real world anymore, AFAIK.)

With S/PDIF and AES, it depends on the DAC. My understanding is that most affordable DACs will use the clock that comes with the signal, so the streamer might make a big difference. Which could actually be an advantage: With USB, it’s not possible to fix a shitty DAC clock using a streamer.

 

Yes purposefully not using USB. I am using SPDIF and my current ESS DAC works in sync mode which is slaved by incoming signal. For Holo May you can turn the PLL off.
But eventually the goal is to build 2 DACs that will use the clocks directly in Sync mode.

 

Couple of more changes
1. Added Amanero USB module
2. Upgraded FifoQ7 -> FifoQ7 MkII

Previous chain
iPad -> USB C -> Holo Red as DDC -> Coax RCA DIY Streamer -> Coax RCA ESS Dac.

New Chain
iPad -> USB C -> Coax RCA DIY Streamer -> Coax RCA ESS Dac.

So the Holo Red is free for its NAA duties. I now also have the HDMI Pro MKII board, so i can switch it with TransportPi AES Pro board to stream DSD (upto 512) from HQPlayer.

Any changes to the sound?
I the system running for 24 hours so the clocks should have settled down. The main difference I hear after removing Holo Red from the current chain is slightly lesser digital artifacts. Its so crazy, till you hear the alternate option I could have sworn there are no digital artifacts coming from Holo Red. The background on Holo Red is a bit darker though. The outline of imaging on Holo Red is a little sharper. I wonder if psychologically that darker background and sharper imaging is what my mind is interpreting as digital artifact.

Are the goals still the same for the system?
When I first started on this route the ambition was to create a streamer that was "BETTER" than Holo Red. But now my goal has changed with the learnings. I now wish to build a streamer, a ESS Dac and a NOS Dac that I would enjoy listening.
This building experience has humbled me in many ways and now have a mad respect Holo Audio. To be able to sell the quality of products at the price point they manage to do is really commendable. Not just Holo but all audio hardware engineers out there I understand a bit now, how difficult it is to come up with something differentiating in the market. I'm just assembling stuff and still I have to make so many engineering decisions.

What's next?
1, Maybe test out DSD with Holo May. I say maybe because the PCM is sounding lovely and that too on an ESS dac.
2. Build the ESS Dac first. This will be a grand project in terms of power supply and would directly use the SC-Pure clocks. It will be built in two version (a). Utilizing the active OPA861 Active I/V converter stage (b). Using the highly regarded passive I/V stage from Ivan.
3. If my soldering skill improve then I might take on the DDDAC NOS Dac project this year or that will have to wait for next year till I develop enough skills.

Pics or it didn't happen
1. Original configuration with Coax input from Holo Red