Building a DIY Music Server
Building a DIY Music Server
When everyone was crazy about low-powered CPUs, I tested some high-powered CPUs (with Euphony) and realized there is something special about them. That was before I knew about any high-powered CPU audio servers on the market. I reached out to several manufacturers to ask about a high current LPS (something in the 10-20A range). The only person who got back to me with a positive response was Sean Jacobs. He told me he had been working on a design but does not really know how It sounded yet. I was familiar with Sean’s designs having built a DIY version of his DC3 LPS for my computer and jumped on the opportunity. I bought his high current modules (paid full price for them) from Sean and started playing with them. They were feeding the 12V EPS on my computer and were good for about 10A of continuous current (as long as you have a massive heatsink as they were dissipating about 80W of heat at 10A and even more in the later revisions) and could handle 20A of peak current. Good enough to power many decent CPUs. That is how my journey with high-CPUs started. I tried several Intel i7/i9 CPUs and some AMD Ryzen CPUs. A lot of these experiments have been documented later here and in the novel thread.
In the meantime, Sean suggested that there are some parts of the design that can be voiced. He gave me a few ideas, and I started experimenting with different things, just to realize how important the LPS feeding the 12V EPS rail was. I changed some parts and some small sections of the design. And I realized I could tweak the sound of my server to my liking. I got actively engaged with Sean on this project, and at one point I seriously considered starting my own business and making the first high-end high-current LPS designed for audio. Sean was okay with this idea, and we were bouncing around some ideas of the terms of the relationship if we were to go that route. But the main thing was the product. It had to be the best in the world. In fact, I remember telling Sean that I would only go ahead with this if he can design the best high-current power supply in the world. Something that no one else can compete with. Not even the Paul Hynes custom dual regulated SR7 XLR (now called DR7). Sean accepted the challenge, and we started working on it. He sent me different prototype boards to listen to and kept improving and tweaking the design based on my feedback. He implemented parts of his DC4 design and continued to refine this high current LPS. The biggest problem was the heat dissipation but with good heatsinks, that wasn’t really a problem. The prototype was a beast that could do 20-30A continuously and 60A peak if the heatsink allowed.
Sean then released his DC4 LPS and he got a bunch of pre-orders. He got super busy. And his orders queue kept growing and growing. I think at some point it was getting closer to 6 months wait time on new orders. Our high-current power supply product was 99% ready, but Sean all of a sudden did not have time to work with me and finish the last 1% so we could go to production. This was an absolute cost no object LPS we were working on. It was much more expensive than the DC4. We considered everything available including massive Duelund caps that cost over $2K each, massive Bocchino connectors and other things I’ve never seen done before.
I could not believe how good the sound in my system was at the time, but I could not share much with the world and I could not finish the product I wanted to release.
Two things happened while I was waiting for Sean to free up some of his time:
1. Someone gave me a hint that Taiko might be developing a groundbreaking power supply that would change the market.
2. I offered Sean to help him with some of his DC4 orders, so he can clear up his queue and continue working on our design. Sean gladly accepted my offer and that is how my affiliation with Sean started.
In the meantime, I built my dual ASUS Sage motherboard and powered it with the latest prototype of the high current LPS Sean and I nearly perfected.
I was following the Taiko thread on WBF, and one day I decided to send a message to Emile showing him what I have done. I expected my message to be ignored in a typical big manufacturer manner, but to my surprise Emile welcomed my DIY build in a very warm way. He told me that he has been sharing a lot of details about the Extreme design in the hopes that people would try these parts and follow this path. He actually genuinely seemed happy to finally see someone in the DIY world doing it. He told me my passive cooling solution looked better than his first mockup, asked me how I liked it, and even gave me some hints of what to try. All that was a big surprise to me, and I tried to open a communication channel with Emile, while trying to also be respectful of his time and IP. He was open about some aspects of the Extreme, and he was completely closed about some of the secrets he did not want to share.
What surprised me even more was how passionate Emile was. He is the biggest DIY-er I have met up to date… He told me that the main reason he started Taiko Audio was to experiment with things that are not available on the market and require solid investment to make.
We started chatting from time to time and sharing ideas. Let me explain what “sharing ideas” with Emile means… I would suggest a new idea I would like to try to Emile, and he would respond with when he had tried this, what the results were, and add a bunch of other similar things he has tried (basically taking the idea into extremes I could not even imagine). That’s typically how those conversations went. But I’ve learned a lot from that!
At some point, while I was helping Sean clear up his queue with DC4 orders, Emile reached out to me and told me about the DC to DC ATX he has developed. He explained some of the design concepts. I gave my word to keep those confidential, so you will not hear about them from me. The main thing is that this SMPS DC to DC ATX is super transparent and doesn’t switch at KHz but at MHz frequencies. He thought it might be good for DIY and asked me if I am interested to evaluate it. Of course I accepted right away!
That’s something that Emile would learn from this post, but when he sent me his Taiko ATX last year, I was still thinking about releasing the high-end high-current LPS with Sean. I was absolutely convinced that regulated Linear power supplies is the way to go and had what I believe was (and perhaps still is) the best high-current LPS in the world. I was convinced in that just like some of the people who posted earlier here.
I received the Taiko ATX, and the first thing I did was swapping out the HDPlex 800W DC to DC ATX. Big improvement right away just like @Exocer reported. It was a big eye opener (or a reminder of) how noisy that HDPlex ATX was.
Note: Let me clarify something here. I didn’t always use the HDPlex DC to DC ATX to power my computer. The HDPlex powered up only the 3.3V and 5V rails on the ASUS SAGE motherboard. I used to power those up with separate rails of DC3/DC4 LPS. But since I was constantly tinkering with this build, I ran a few times into BIOS resets and other issues where an ATX-speced power supply was needed. So, I accepted some level of sound quality degradation with the HDPlex in favor of the convenience of the ATX controller.
Replacing the HDPlex (which was only feeding the 3.3V and 5V) with the Taiko ATX was a big improvement. But Emile told me the strength of the Taiko ATX is in powering the 12V EPS rail. I put together a higher current higher voltage LPS (mine was 12V which was not enough for the Taiko ATX) and while experimenting I managed to destroy one of the rails on the Taiko ATX. Oops! That happened on the day I received it. Emile added protection in the design, so all the units he is selling have that now. But it took a while to send me a new unit. However, I managed to bypass the ATX on the Taiko ATX and just use the 12V EPS rail on the Taiko ATX. Did that sound better than my pimped out high-current LPS? No, it did not! At least not at the time. I rolled up my sleeves and started working on this unregulated LPS with some help from Emile. I quickly realized how transparent the Taiko ATX was. Any change you make to the unregulated LPS, and you can immediately hear. I played with different transformers, different rectifiers (including mosfet rectifiers), different capacitors, bypass capacitors, inductors, etc. I have been doing this for about 6 months.
This is becoming a very long post already, so let me shorten the story a little bit and get straight to the point. What I learned from this experience is how noisy linear regulators become under high current. No matter what regulators you use… and I’ve tried many – Belleson, paralleled LT3045s, Sean’s regulators, etc. It’s that type of noise that you never thought it was there until it was gone. I can see a bunch of LPS manufacturers attacking me on this and telling me “my design does not have any noise”. Yeah, right! I thought the same. Let’s not argue about that until you try the stuff I am referring to.
To me, based on experience, a rule of thumb is: if it draws more than 5A, LPS is NOT the best way to go. But the most important thing is that I called Sean, explained what I heard with the Taiko ATX, and we decided not to release the special high-current LPS. In my point of view our design (as good as it is) was already obsolete. In fact, I consider myself lucky that the Taiko ATX came in my hands before licensing Sean’s high-current LPS technology and starting my own business.
Now that you have the full “behind the scenes” story, let’s talk about my affiliation. I have been completely transparent about that.
- I do make power supplies for Sean Jacobs. Whatever I make from this is mainly invested in my system and my DIY experiments. It’s a way for me to support this expensive hobby with my hobby (and not drain my bank account).
- I also have a full-time job that pays my bills.
- And I have zero financial affiliation with Taiko. I haven’t paid the Taiko ATX that was sent to me last year to test, and that is the only benefit I have received by Taiko. I did however send a DAVE DC4 to Taiko, because we have a big DAVE community and Extreme servers, and Emile thought the DAVE is a horrible DAC. The DAVE DC4 is a game changer, and I wanted him to hear that. Shortly after I sent the DAVE DC4 to Emile, I asked him if I can buy one of the Taiko USB cards from him. He refused and told me that the USB card would only be available to Extreme users. I will not hide that I was disappointed from that message, but it clearly shows that I don’t get special treatment for spreading the word about the Taiko ATX. As mentioned, the Taiko USB card I have in my system was loaned to me by a friend who has an Extreme server.
Although I decided not to start my own company and release a high current LPS, Sean has been offering high current LPS for quite some time, both as finished units and for DIY. They don't have a buil-in ATX controller, but 99% of the time you don't need the ATX section.
I work with Sean. If anything, I should be hiding the Taiko ATX and pushing the Sean Jacobs power supplies instead. It makes sense, right? I don’t make anything from Taiko’s sells. And I do make some profit from the Sean Jacobs power supplies I build. Make your own conclusions from that.
All I can say is the Taiko ATX is a game changer. I am not saying it’s the best way to power any computer on the market. But I am not going back to regulated linear power supplies for this dual CPU build. And this dual CPU build I am using is way better than any other source I have tried.
I can see how my posts can disappoint some manufacturers. I am sure I have pissed off many manufacturers before. That's not the first time and won't be the last.
JCAT is releasing a new fully linear ATX LPS. I am sure there will be plenty of customers for this product. I haven't tried it and I can't comment on it.
Sean Jacobs offers a high current LPS. Farad is coming up with an ATX product. HDplex has one. Paul Hynes has some good products. It's a competitive market, and that is good for the consumers.
I have no desire to try other lower power CPUs at the moment. And I certainly have no desire to try other regulated linear power supplies and compare them with what I use. Other people would need to do that part and share. I also don't have any desire to argue objectively or subjectively and try to prove anything to anyone. I don't make claims what's best, just share my experience. I've spent more than 2 years of my life trying all kinds of power supplies and a lot more years doing all kinds of DIY projects.
My system has never sounded better and my digital source has never been better. My digital source consists of:
- a DIY server using the same motherboard, CPUs, OS, and internal storage as the Taiko Extreme. The server is powered by the unregulated LPS I am sharing for DIY (around $1000 in parts) and the Taiko ATX (1250 Euro).
- I have a JCAT XE in the server that I recommend powering with the best linear power supply you can get. In my case this the Sean Jacobs DC4. Please note, the Taiko ATX does NOT work best here!
- A Chord DAVE DAC powered by DC4 using offline upsampled files with a program called PGGB. BTW, I can't stand the stock DAVE DAC.
I am sorry if sharing my findings upset manufacturers. It's hard to be honest and transparent and not upset someone. But I am committed to remain transparent and honest to the DIY crowd. The good news is there are plenty of new products for people to test. I encourage people to test, compare, share, etc. That's what these forums are about.
Building a DIY Music Server
Building a DIY Music Server
The lower phase ITX could have performed better due to not using doublers and or parallel connections. Often times when there are too many that are not in use, the PWM Controller (voltage regulator) turns off some of them. Which lanes open/close at times could be at random similarly to how CPU cores work. What is available happens at varied intervals. The MOSFET's that are sitting further away from the controller could be the ones that are in use. It may not look like a long distance to us but there is a latency lag when gates open and close. Along the path there are different traces the signal must travel through, each attributing to crosstalk and a potential capacitance created between them. The Aorus I mentioned features the newer server grade PCB weaving for less signal loss. The 6 or 8 layers PCB (depending on model) improves crosstalk since these traces no longer have to inter-lap one another so closely. The 2X copper increases signal fidelity and improves thermal dissipation. In the newer X570 designs there is also an impedance improvement for the traces that go from the CPU to the PCI-E and M.2 slots. This benefits the add-on cards and Intel Optane storage. All that I labeled could be a potential link to why those transient attacks and timing quality didn't meet your expectations. Other reasons not yet listed could be difference in inductor value (higher is better), capacitor value, trace/signal arrangement, or natural noise created by the voltage regulator. There are probably even more that are outside of my area of expertise (I'm an audio mixing/mastering engineer).
The feedback you received from the linked post, I would say that I agree with everything the writer says.
First I would recommend starting with a clean linear power supply, then pick out a motherboard that has a capable VRM with the benefits posed above, then everything else from add-on card to CPU to Industrial/ECC RAM are of smaller importance. The same can be said for the endpoint. The only time the CPU is as important as the motherboard or clean power is when you are deciding how that CPU will function. For example, If you need it to do heavy DSD upsampling then the motherboard VRM is vital and the linear PSU has to be able to supply the large amount of current required. These three components affect one another which is why they are of similar importance. If the device you are building is an endpoint then the CPU is less important than the motherboard while clean power is more important than both. Although the unit will perform better with more CPU cores even at a lighter workload, the motherboard is more important because of the VRM's, board layout, possibility of handling ECC RAM, slots availability for add-on cards, and it's functionality plays a bigger role. Lastly clean power triumphs both. This last mention can be seen by those who have upgraded their power supplies for their endpoints and have found it to be a drastic change. In short, a CPU just processes information, the motherboard should be more important since it is the technical grid but most motherboards are made pretty similarly nowadays whereas clean power supplies are not, especially for audiophile. This is why the priority is set this way.
For power I do not recommend SMPS's aka conventional power supplies (AX1600i, SuperFlower). Although they are good, they are only good for the average consumer. Their quality is comparable to high quality Seasonic which have under 20mV (20,000uV) of ripple noise, but compared to an HDPlex 400W (3mV), it is 17mV higher. A Paul Hynes have noise in the single digit uV's (probably). It's a big difference. So when it comes to endpoint power, purchasable brands like Paul Hynes, Sean Jacobs, Farad is probably is good as you'll get. There may be some just as good out there waiting to be discovered, but just without a reputation worthy enough for any of us to take a risk. One thing worth mentioning is although Paul Hynes SR7 can do 12A at 12V, it may still be a little low for a music server that wishes to use some of the higher filters and modulators that is provided on HQPlayer, as doing so will require a CPU with a higher TDP. These TDP numbers also lie since my 105TDP CPU actually does 150W during full throttle so if 14A is cutting it close for me then 12A might be a little under but that heavily depends on what CPU you are using. I remember running Poly-Sinc-XTR-LP with ASDM7EC at DSD256 with an Intel i7-9700K before which is also rated 95W but I never measured how much it consumed before switching over to AMD for ECC compatibility.
There are not a lot of build-able linear power supplies other than using Belleson or Sean Jacobs (albeit having to get familiar with current boosters). I went with the more simpler path. For a computer, 3.3V, 5V, and 12V voltages are required. With the Belleson I am able to achieve the following:
- Belleson SPHP 12V (14A)
- Belleson SPX 5.0V (7.2A)
- Belleson SPX 3.3V (10.91A)
The reason for SPX is that it has less noise and lower impedance than the SPHP but cannot handle as much current. Since the lower voltage rails are not as important, this drops the price of voltage regulator down while offering better performance (-$20). On a side note, Belleson has not tested their SPHP to perform over 10A, but given their datasheet if you provide it with a 5V headroom, meaning the transformer that supplies voltage to the regulator is at least 5V higher, it can potentially do 14A. This much amps at 12V is 168W. The 3900X I use on my music server draws 150W at full 100% load, but during DSD upsampling it is only around 90W. The idea of having the Belleson be able to handle that high of a current is for the sake of headroom. You never want to run a power supply at full throttle.
The standard orientation of the build:
IEC > Transformer > Rectifier Bridge > Filtering Capacitors> Belleson SPHP/SPX > DC Output
For my more advanced build: (Balanced, Pi FIlter, Dual Regulation)
IEC > Balanced Transformer > Rectifier Diode > 700,000uF Pi Filter > Belleson SPX > Belleson SPX > DC Output
Hopefully that was all understandable.
Building a DIY Music Server
Building a DIY Music ServerOn 3/25/2020 at 3:30 AM, Gavin1977 said:
Did you manage to find a price for the DFI GH171?
Not anymore. I've decided to leave the NUC route and make an endpoint out of the following:
- HDPlex H3 V3 Fanless Chassis
- X570 Micro-ATX Motherboard
- AMD Ryzen 7 3700X
- Innodisk ECC/Industrial SODIMM DDR4 RAM
- JCAT NET Card FEMTO
- JCAT USB Card FEMTO
- Clocked with sCLK-EX (from SOtM tX-USBultra)
- AudioLinux Operating System (Headless)
Power Supply: (dual regulated)
- Belleson SPHP 12V (14A)
- Belleson SPX 5.0V (7.2A)
- Belleson SPX 3.3V (10.91A)
Only this way can I have two add-on cards with the addition of them being directly connected to the CPU.
I just need to find a suitable X570 motherboard first. Since the job of the endpoint isn't as severe as the DSD upsampling music server, I've decided on a Micro-ATX form factor despite it offering less VRM's.
For my main system I went with Gigabyte X570 Aorus Master (and Xtreme) because they are the first motherboards in existence to offer TRUE 16 PHASES. What this is is that the PWM controller (voltage regulator) has 16 lanes that connect to the MOSFET's that are used to feed the CPU. Normally in any advertised 16 phase, there is only actually 8 lanes that have doubler MOSFET's on them. Doubler's take one frequency voltage and divides it into two working frequencies. The other method is using two MOSFET's in parallel to achieve these two frequencies required by the CPU. These methods come with problems. A single device outputting two signals does not perform that good and two in parallel sharing one lane comes with it's own problems as well. This motherboard that I am speaking of however has 16 lanes with each an independent MOSFET handling only one job. There is no overworking of a single transistor or two that is bottle-necking one lane and having poor communication with the PWN controller. This improvement leads to higher efficiency, lower voltage tolerance, lower latency, and perfect load balance. To put it short, on paper it should be the best sounding motherboard for audio use.
After the PWN controller and MOSFET, there is an inductor and solid polymer capacitor that filters out the high frequency noise away from the CPU circuit. I plan to increase the capacitance of the solid polymer capacitor. Increasing it cost money but as an audiophile I am not limited to what the manufacturer's BOM was limited to. Increasing the capacitance helps lower the voltage ripple noise and gives the CPU a higher capacity reservoir for pulse currents. I'll then bypass/shunt each of them with ceramic and film/foil capacitors to lower the ESR further so that more high frequency noise can be removed from the system. This further attributes to a lower ripple although most of it is highly dependent on the power supplies's rail.
The reason for using different types of capacitors is that their ESR depends on capacitance and frequency.
Brief: There are various areas that can be improved on the motherboard. It's just not since they're marketed towards the average consumer hence having to fit tightly to meet the bill of material.
Building a DIY Music Server
Building a DIY Music Server2 hours ago, RickyV said:
Have you compared the NUC7I7DN.. and the iBox v1000 and which did you prefer?
Yes I have. I haven't shared the news with anyone else other than with friends but it does sound better to me. I've got some of my audiophile friends to switch over to it. Some from NUC7I7DNHE's while other's from UltraRendu. To start off, aside from the sound quality, it has some good hardware. It's CPU and GPU (it's an APU thus combined) is a little faster than the Intel while it it's memory supports ECC and has a higher limit (3200MHz vs 2400MHz). It comes with a display port in case you ever want to use it as a normal computer and wish to pair it with higher refresh rate monitors. For the same price as the Intel offering you get better performance which when running it as an endpoint, the performance affects the sound quite a lot.
The setup I have is Music Server > EtherRegen > Endpoint
The endpoint runs AudioLinux via ramboot. The memory modules are 4GB x 2, ECC enabled, Industrial Grade. The unit runs headless after the initial setup was done to disable all unnecessary services. It occupies just a little over 3.4GB. I have it powered with a linear power supply that can handle 4A (AMB Laboratories σ11 with custom cooling) and it takes a set voltage of 12V. All superfluous (excessive) partitions in BIOS is disabled such as the WIFI card since it is not used. Now... for the sound.
I had both the NUC7I7 and this unit for about nearly 2 weeks before I decided on one and sold the other one off. It was easy to tell from the begging that I liked the iBOX-V1000 much more. It wasn't a SUPER DRAMATIC difference but I would say a 10% gain is quite significant for a product that has the same price and comes in a passively cooled fan-less case. The first thing I noticed was the imaging, detail, sound stage, and void of digital artifacts. The bass was somewhat more articulate and each sonic timbre more real with the same said about the decay. There was more edge to the sound yet it wasn't sharp. It's like re-polishing the corner of table, re-staining it, and putting a new layer of polyurethane over. It was the same sound I had liked or was familiar with but it just sounded more real yet edgy. Digital, yet analog. It only got better in time. At the 2 weeks mark I sold the Intel NUC off.
Having said all this, I could come with some guesses on why it sounded better. I'm not a computer techie, but the improvements could be related to it's low TDP of 15W yet it has faster processing power. It's an APU so it's latency between CPU/GPU processes could be lower since they're both located on the same die. Since the Intel and AMD unit are both small form factor 4 x 4's, that department ends up in a tie. Since the Intel has a wide input range of 12-24 VDC, perhaps a set 12V has less DC-DC conversion going on. The higher frequency RAM meant nothing since I am using 2400MHz sticks so it didn't matter. ECC enabled for the memory does sound a little better (3%) whereas when used on the music server is less profound (1-2%). Lastly I am using the same power supply for them both as I have two of them so it couldn't be because of PSU differences. I even swapped them but they produced the same results time and time again.
So all in all, I highly recommend the iBOX-V1000 for those wanting to upgrade from their NUC7i7. The ECC didn't play as huge of a role as I thought but little did I know, every other component was what lead to the rest of the benefits. In regards to components, I advise you to have at least 3A of current for your linear PSU. I found that it sounded better with a 4A supply and the case is probably true when going even higher. I was going to add an MPAudio HPULN (5A) after the Linear PSU I'm using now for extra PSRR and lower output noise to see if it helps with sound (which im sure it does), but in that process I had a different idea. I looked at the unit and hold myself.. "having the endpoint attached to sCLK-EX to replace it's clocks wasn't good enough". I didn't want to improve the clocks of an area that could have been better isolated. For that I felt like the best course of action was to find a way to install both an Ethernet card and USB card into the unit to further isolate the Ethernet and USB controller so that even less jitter could be achieved. This way when the sCLK-EX connects to it, the clock signals are received in optimal fashion since the receiving circuit is now much more quiet. Then comes the discussion on the new USB output's ability creating a more ideal USB signal for the DAC. Even though I have a SOtM tX-USBultra after the endpoint that does some reconditioning to the USB signal, maybe it would do a better job if the signal arrived in better condition than it had before. Or maybe Ethernet entering the endpoint through the isolated card can now be processed more efficiently since there is a direct path to the CPU rather than running it through noisy motherboard components that are all sandwiched in the one area.
For this reason I came to the conclusion that despite it sounding better, I am going to go further. For those of you who want the iBOX-V1000, don't let this discourage you. It does sound better, but I think more can be squeezed from an endpoint.
I did a search and found that I could purchase a motherboard with the same CPU that also has a set 12V input and all the goodies that was spoken about previously. The unfortunate thing here is that the motherboard is no a longer small form factor 4" x 4". It was bigger, possibly require more power, and having more components on-board for processes that aren't used may create more noise. BUT what it did come with was a PCI-E socket for an add on card. The chassis I found does have a slot for an add on card and passive cooling.
This is currently what I have found. I am trying to find a motherboard with the V1605B APU but with 2 PCI-E slots as well as a chassis that can hold both cards while offering passive cooling.
So there you have it. This endpoint is currently satisfying me but for the future I would like to have a similar one but housed in a larger aluminum case (black) with two JCAT add-on cards for Ethernet and USB. This however would require a larger linear PSU that would have to output at least 5A so I am hoping Farad will build something that can accommodate.
Building a DIY Music Server
Building a DIY Music Server
It was quiet here for a moment, but @Energy dropped a bomb.
We can debate forever on some of that stuff. It's very easy to be convinced in a specific technology or a specific way of doing this and scratch every other option available. I haven't found this to be the right approach for me. I prefer to be open minded. Typically I would pick the technology that makes sense to me, implement it to the best of my ability and then go back to the other alternatives to try them and move on after I find what sounds best in my system.
Let me comment on some of the topics @Energy mentioned - Intel vs. AMD, ECC vs. non-ECC RAM, OCXO clocks, and the HDPlex DC to DC convertors.
Let's start with RAM and Intel vs. AMD.9 hours ago, Energy said:
Furthermore it is known that ECC sounds better as it corrects on the soft/hard bits that computers tend to make. Some people on the forum has found that to be the case including myself. So between Intel and AMD, I am more lenient of the later.On 2/18/2020 at 10:43 PM, Dev said:
@Nenon and others,
have any of you guys compared Intel vs AMD ? I think @Nenon has built both, so he must have some idea. Are there much differences between the two ? If one is starting a fresh build (and leaving the behind the price for now), is there a consensus which one sounds better ?
I've had an Intel i9-9900K and AMD Ryzen 7 3700x builds side by side in my system and might have been the first who wrote about this in the Novel thread. I can tell you that they sound different. But I cannot tell you which one is better. I think it's a matter of personal preference. I did not like the motherboard USB output on the AMD as much as I did on the Intel build. Once you take the digital output from the PCIe, it's hard to say which one is better.
Then, I did a lot of RAM chip comparisons. About a year ago I had no idea that RAM could make a difference. No, that's not the right statement. I was absolutely convinced that RAM could not make any difference at all. Silly me.
However, curiosity made me check that, and I was shocked to discover that RAM makes a significant difference. I managed to obtain some Apacer RAM for myself. And I could not believe my ears. One of the guys here convinced me to do a group RAM order, which I did, and I am very thankful that almost everybody who participated in that group buy gave me some feedback. That helped me tremendously to understand how the Apacer RAM changed the sound in different systems. There was an improvement in every system. Some systems benefited a lot more than others. There were two common denominators I noticed. First, people with more resolving systems experienced bigger impact (improvement). And second, the people who had better power supplies (cleaner power) benefitted much more from the Apacer RAM. For those who had custom double regulated SR7s or custom double regulated Sean Jacobs DC3 power supplies, it was a match in heaven. They would typically hear the improvement from the first note after swapping the RAM. The other interesting thing was that on average it took two weeks to break in and even after the break in, if you just take the RAM out and put it back in, it resets some of the settled effects and needs more break in time (not weeks but at least 2 hours or maybe a full day). Those are statistics I have collected over time thanks to all of you.
Having access to all kinds of different RAM modules, I did a lot of testing myself. I discovered two things that are important - quality and speed. For completeness I will repeat what I have said in other posts:
1. Non-ECC Apacer 2400MHz was better than the Non-ECC G.Skill 3200Mhz.
2. Non-ECC Apacer 2666MHz was better than the Non-ECC Apacer 2400MHz.
3. ECC Apacer 2666MHz was better than Non-ECC Apacer 2666MHz.
But that does not mean that the ECC RAM sounds better, because "it corrects on the soft/hard bits that computers tend to make". No, it's not better for that, even if that contributes to some degree. And here is why I am saying that. I recently discovered that the ECC RAM can work on motherboards that don't support ECC RAM, such as the Intel Z390 gaming motherboards that are popular here and I used in the build in this thread. If you install ECC RAM in that motherboard, it would work as Non-ECC RAM. I compared the same version of ECC RAM and Non-ECC RAM in the build in this thread. And the ECC RAM sounded better. Not because "it corrects on the soft/hard bits that computers tend to make". It does not correct anything, because it is working as Non-ECC RAM. My only explanation is that the ECC RAM was built better / better quality.
So for those who think the ECC RAM (working as ECC and correcting some errors that makes the sound so much better) is the holy grail and that determines what motherboard and CPU to take, please think twice. You can probably use the ECC RAM as Non-ECC, and it still sounds really good and better than anything else I have tried. So, I would not prefer AMD Ryzen just for that. Motherboards make a big difference. I have compared quite a few. The RAM is not the only factor, and I feel like it is a bad decision to plan your build around ECC RAM because it corrects something. I am not saying that correction does not contribute to the sound, but that's hard to test, and there is obviously a lot more than that for RAM to sound good=.
Power supplies.9 hours ago, Energy said:
Lastly, for those who are using HDPLEX 200W + HDPLEX 400W/800W HiFi DC-ATX or 800W. I think the idea is ridonculous. You are essentially using a 3mV (3,000µV) device and adding it with a 10mV (10,000µV) device to have a max output noise of 13mV (13,000µV). The HDPLEX 400W ATX Linear Power Supply may be a failure in it's independent output but it's ATX modular Output is a heck of a lot better:
Just saying but 3mV beats 13mV. And look at that DC current headroom!
Well, again, none of these numbers tell us anything about how these things sound in our systems, right? They definitely don't tell ME anything about the sound. What about output impedance? What cables are you using? How long are they? How much headroom it has? How fast it can handle transients? I can keep going on and on. The only thing I agree with is that 10 + 3 = 13. Can't argue with that. And I also prefer full linear power supply with separate rails for everything as long as they are good quality - which means the HDPLEX 200W does not qualify, but more on that later.
Let's forget about the numbers for a moment and talk about actual experience comparing these products. I have done different builds. The one in this thread is one of the best power supply arrangements I have done. There is actually only one that was better than this, but I can't talk about it. I have also used a whole range or power supplies, DC to DC regulators, ATX convertors, nanoATX, etc. etc.
Generally speaking I try to stay away from the HDPlex linear power supplies. I have also done some builds with the HDPLEX 400W HiFi DC-ATX. And I was not impressed. But for someone who already had only one rail of decent power supply and did not want to spend more, that was a good budget option with possibility for upgrades in the future. So here is my experience.
1. The most important point - not everything works equally in all builds. What worked well in my low powered systems did not work well in my high powered CPU systems. Keep that in mind. Most of my experiments have been with 65W-95W TDP CPUs like the Intel i9 and AMD Ryzen 7. I will focus on those below.
2. For some reason the HDPLEX 800W DC-ATX sounds better than the HDPLEX 400W HiFi DC-ATX. I don't know why, but it does. It wasn't until I tried the HDPLEX 800W DC-ATX that I considered using a SMPS DC to DC ATX devices. Basically the 800W has been a drop-in replacement upgrade (for the HDPLEX 400W HiFi DC-ATX) in most cases.
3. No matter what the numbers say, an HDPLEX 800W DC-ATX powered by a really good quality power supply (i.e. Sean Jacobs DC3 in my case) sounds a lot better in my system than a directly connected HDPLEX 200W.
4. More power of the same quality sounds better and improves dynamics. My CPU is running fine with [email protected] But feed it with a LPS of the same quality that can deliver [email protected] and it sounds even better. Feed it with a LPS of the same quality that can deliver [email protected] and it sounds even better. Same is true for 15A. And that's all I have tried so far. Someone has a really good one that can deliver 30A? I would love to try it :). Why that extra power helps? I don't know. I can't catch peaks much higher than 4A with the measurement devices I have. I guess that dynamic headroom is quite important here with linear power supplies.
Based on #3 and #4, would connecting an HDPLEX 200W via it's [email protected] (190W) output to a HDPLEX 800W DC-ATX sound better than connecting the [email protected] (120W) + [email protected] (10W) + [email protected] (6.6W)? I don't know. I kind of doubt it. But 190W is more than 136.6W... and read #4 above. I am still kind of sceptical to be honest, but is it worth a try? Why not. You would never know for sure until you try it. Would I consider trying this with the HDPLEX 200W + HDPLEX 400W HiFi DC-ATX? No, I don't think so. Would I consider trying this with the HDPLEX 200W + HDPLEX 800W DC-ATX? Maybe... if I had the time.
I left that as the last topic. And this is a hot topic in my mind currently. I am not an expert here, but I am applying a similar approach as with power supplies, RAM, motherboards, etc. To repeat what I said in the beginning of my post:Quote
Typically I would pick the technology that makes sense to me, implement it to the best of my ability and then go back to the other alternatives to try them and move on after I find what sounds best in my system.
As far as my understanding about OCXO clocks goes, it's easier to make a low phase noise OCXO clock at 10MHz than to achieve the same at 25MHz. The phase noise at specific frequency is not the only important variable and it's probably not even measured consistently between all manufacturers. The implementation of the clock is arguably even more important than the quality of the clock itself. Short path to the chip is important. I studied how PLLs are implemented recently, and they are not perfect. I have many questions and not a lot of answers yet. While I am doing studies on this, I am trying to identify what is a sales or marketing BS and what actually makes sense. SOTM for example are saying that you can replace ANY clock with theirs and it would be an improvement. They have repeatedly said that if we replace the Crystek clock on the JCAT USB card (and NET Femto), it would sound better. I wonder if @Marcin_gps knew that. They said that we should start replacing the OCXO clocks with their clocks. Well, I am not buying into all these things unfortunately.
I can't imagine that on something that sensitive as the clock, we can add connectors, long cables, an external reference clock, introduce PLL that converts the master clock frequency, and it would be all much much better than a well implemented good quality clock a fraction of an inch away of the chip with clean power and done with good care. And most of these clocks we are replacing have single ended signal, not differential. What happens with the noise those long cables catch from the outside. But hey, we are open minded people, so I am willing to give it a try. The ultimate test would be how it sounds.
But before I do that, I would start with the technology that makes more sense to me. And that is not the SOTM clocks, connectors, cables, and external master clock. My first preference is the best quality OCXO I can find, implemented with the shortest possible wires (the best silver I have found), powered by the best power supply (again with the shortest possible DC cables and regulators sitting right next to the clock), sitting on a RMI/RFI and vibration isolation. I am also a truly believer that every clock benefits tremendously from vibration isolation. Every time I posted that last year, I was shot by objective engineers, but since this is my thread and now they have their own forum, I hope that won't happen again.
And here is what I am doing.
Clock removed. Silver wires soldered.
Carbon fiber plate installed. To be replaced with much thicker one. More holes for ventilation would be added. I believe the carbon fiber has some EMI/RFI properties and also vibration treating... but even if it did not, I like how it looks. The white standoff to be replaced with black vibration isolation ones.
The clock installed with a temporary DC connector for testing purposes.
I am testing every clock on the motherboard to find out where it makes sense to have one and where it does not. A lot of listening tests ahead of me. Once I settled on all that, I would not mind trying the SOTM clock with external reference clock to see which one is better. I think that is the right way to test these things, rather than looking at the spec sheets.
Building a DIY Music Server
Building a DIY Music Server
The prior digital source going into my TotalDAC D1-Direct DAC was new on 10/13/19. So the technology being replaced is almost brand new. It is a two-box setup built for me by a CA contributor. Power to the prior server was an HDPlex 200-watt LPS. I will be putting this prior setup up for sale soon. It is no longer in use, though it was quite good.
The prior server and endpoint were designed to sound best with lower-power-use, with the philosophy that lower power consumption means less noise which means better sound. The prior server had an 8 core AMD Ryzen 2 processor and a mini ITX mother board. The prior endpoint had an Intel Core i7 8650U Quad-Core processor and motherboard. The prior endpoint was powered by my Sean Jacobs supply at 19v. It now powers Nenon’s server’s CPU, at 12v (voltage is easily adjustable on the SJ PS)
For a setup with high power consumption, power supplies are ostensibly in shorter supply with longer waiting times. I don’t think the Sean Jacobs PS is in short supply and it is very, very good especially with the dual-regulation.
Roon software was moved to Nenon’s server. I had the full Euphony software on the endpoint of the prior two boxes; it was also moved to Nenon’s server. So Nenon’s server has both Roon and Euphony, and it is simple to switch back and forth.
My listening comparison is mainly with Roon playing on the prior server and Euphony on its endpoint, versus Euphony Stylus playing on Nenon’s server. My friend Bob and I also listened to Nenon’s server with Roon Core and StylusEP for an apples-to-apples comparison (same software on different hardware), which I’ll summarize later.
The comparison was not a quick A/B, because the Nenon’s server and power supply needed break-in, as did all of the cables. They probably have 300 hours+ of burn-in now. The Pink Faun USB card with the ultra OCXO clock has considerably less break-in than this, as it had a soldering issue near the DC jack that Nenon easily fixed.
The comparison is not a straight two-box to one-box comparison. For example, Nenon’s server has Mundorf silver/gold wire inside the server, in the DC cables, and in Nenon’s power supply, built from Sean-Jacobs-supplied components. Nenon’s server has six DC inputs and wires, five coming from
Sean Jacobs / Nenon equipment, and they were all JSSG360’d, while the prior server was supplied by an HDPlex via stock DC cables, not JSSG360’d. I don’t think the HDPlex PS is dual regulated (not sure), but the four outputs from Nenon’s / Sean Jacobs PS are dual regulated, as is the SJ PS that powered my endpoint which is now powering the Intel i9-9900K. My network card is powered by an LPS 1.2.
My prior server had a 19v input, plus a 12v input for the CPU. The 19v input went inside the prior server to the HDPlex DC-ATX, which then converted the 19v to 3.3, 5 and 12v, I believe. This converter is a $62.50 part. Soon after the prior server was installed, I contacted Nenon about his surplus power supply that he’d built from SJ components (he had advertised it for sale), because I thought it made more sense to have the conversion done outside of the server, which then led me even further than that to having Nenon replace everything I had just bought. Crazy stuff, but it worked out extraordinarily well in the end, and I am extraordinarily pleased with the final result.
Between my two prior boxes, I had a long run of Cardas ethernet cable. That is gone in a one-box setup. Between the prior endpoint and my DAC, I had two Lush^2 cables, an Uptone adapter, an ISO Regen, and an upgraded SOTM Tx-USB Ultra (and two LPS 1.2’s). The SOTM was being master clocked by a Mutec Ref 10. Now the only thing between my server and DAC is a 0.3 meter Lush^2.
With Nenon’s server, the Pink Faun USB card with the ultra OCXO clock and the SJ PS makes the aforementioned unnecessary, and I will be selling all but one of the two Lush^2 cables.
I am somewhat hearing impaired, so I can’t hear high frequencies well. I asked my friend Bob over, so the below is mainly his description that I agree with. We heard the same things, but it seems appropriate to have a second pair of ears given my hearing.
When Bob listened a few weeks ago, he said the system now sounded like vinyl but maybe needed to burn in as it was lacking in the bass region (which I agreed with). It turned out to indeed need to burn in.
Yesterday, Bob and I listened again. We didn’t listen for long, and only listened to two parts of two tracks. The first was the first couple of minutes of Hotel California on Hell Freezes Over (live) with much of the listening focus on the bass drums. The second was Warren Zevon’s Please Stay, from his last album/CD, The Wind.
These were Bob’s words, to the extent I could keep up with my pen.
Sounds really nice.
Really sounds lovely.
Sounds more fleshed out.
More micro detail.
More texture of the drums.
Non-digital. You do not get that flatness. It sounds like real, live music.
I mean, you got a winner.
The bass does not have as much weight or emphasis, but it sounds more accurate.
Then we played the same portion of the same track using Roon Core + StylusEP. Again, Bob’s words, and again, which I again agree with.
More digital / harsher / more etched.
With Euphony Stylus, it sounds like somebody singing.
With Euphony Stylus, you were just listening, you weren’t coming up with stuff.
Roon is a little more forward with the bass drums.
The drums are louder, and heavier, but you are losing the micro detail.
Roon does have a little bit more of an impact.
But it doesn’t have the same depth, resonance or decay.
Roon is flatter, it doesn’t have the fullness. It’s flatter sounding.
It is my understanding, in speaking with Nenon, that Euphony Stylus performs better in a high-power situation, i.e. a processor that uses a lot of power. I have read some posts where listeners like the way Stylus sounds, but they stick with Roon because it is a better user interface. On Nenon’s server, you can throw the ‘user interface’ preference wherever. It just sounds far better on Euphony Stylus, far far better. It is not close.
Then we listened to Please Stay, as discussed above. Again Bob’s words, which again I agree with.
Everything sounds natural especially the voices.
I then discussed the parts of the song when Emmylou Harris comes in to join Warren for the chorus. On the prior setup, it was still mostly Warren singing those parts; you could barely make out Emmylou’s voice even being present. Now, her presence is not only distinct from his, it is just as loud as his – there is no difficulty hearing her whatsoever. And it makes the song much more beautiful (not just that item, but that item really stood out versus the prior setup). This was my observation, with Bob agreeing with me. Bob replied that before, you could barely distinguish the background singer, now there is a distinct female background singer. It sounds totally different, much more like it should sound.
I live in the NY area between the Bronx and Connecticut, and have hosted a couple of Westchester Audiophile Society gatherings. Perhaps in connection with one of those, I can do another hosting which would include some folks from this site. I am not sure how to do that safely (strangers and all), but feel free to PM me if you might be interested. I am close to an I95 exit.
Below is a picture of my setup. Nenon's builds are shoved toward the back of the rack because the DC cables are short. The Sean Jacobs power supply for the CPU is next to Nenon's server on the second shelf with the blue light. Both Sean Jacobs power supplies are plugged into a Topaz isolation transformer with .0005 pf interwinding capacitance, and that isolation transformer is plugged into a long Tripp Lite power strip that everything else is plugged into, and that power strip draws power from another .0005 pf Elgin transformer. As I understand it from the mains isolation transformer thread, each of these has 146 dB of common mode noise reduction, so the SJ power supplies are getting clean power, and they cost only a few hundred (or less) on eBay. The Tripp Lite also has a High Fidelity Cables MC-1 Pro Helix Plus plugged into it, and a Prana Wire Ruby grounding plane hard wired to it..
SACD Ripping using an Oppo or Pioneer? Yes, it's true!
To users of sacd_extract using the server method:
I have been working on improving the sacd_extract client program from the currently inactive original sacd-ripper repository past few months. I thought I would share that in case someone is interested. Here is the git repository: https://github.com/setmind/sacd-ripper/ (Sorry, you need to compile the code to use it).
In addition to performance improvement and bug fixes, I have added features like concurrent extraction of ISO and DSF/DSDIFF in a single scan of a disc, padding-less DSF generation for some players that cannot handle DSF tail padding properly, extraction of multi-channel and single-channel tracks in one shot, and addition of more ID3v2 tags from TOC to DSF. Details with compilation instruction are described in https://github.com/setmind/sacd-ripper/blob/master/readme.rst and performance comparison with the original is here https://github.com/setmind/sacd-ripper/wiki
It seems to work fine on Linux and Windows, but unfortunately is unverified for macOS since I do not have a Mac. If anybody has a chance to try this on Mac, please report back. I welcome any feedback.