Go Back   Home > Forums > Blogs > wlowes

Rate this Entry

Reference Music Server Finally Making Music

Posted 13th February 2015 at 05:16 PM by wlowes
Updated 4th May 2016 at 12:44 AM by wlowes (Marconi 5U4GB NOS pair of rectifiers-awesome!)

Disclaimer that this entry is just a scrap book of my personal journey with a build. No fresh engineering here, so those seeking engineering insights can save the read.

That said, my new music server is now fully burned in, and for me its a delight. It is a collage of ideas developed by diyaudio members and by Lukas Fikus, aka Lampizator. Thanks to all who contributed. I hope these notes pass on some ideas for others.

Basic Approach
My approach is heavily influenced by Lukas of Lampizator fame. I have enjoyed his tube output amp design in my previous DAC, and several cd players. He got me interested in the TDA1541a, and I have had many years of happlily tweaking it for best sound. This server has a clone of his level 6 DAC with tube rectifiers, dual mono supply and film caps in CLCLC configuration.

Also, my digital evolution came via CICS CMP which taught me to appreciate a minimalist PC processor with network attached storage for the music library. This server uses a Beagle Bone Black processor and a WD digital NAS connected via Ethernet to a WIFI router. Linux with MPD is on the server with MPDROID on my android phone as the user interface.

I started with USB to a WaveIO, to Ian's I2S 2 PCM driving simultaneous L/R mode to the TDA1541a. I am a great fan of WaveIO. However I have moved on to Acko S03 providing master clock and isolation. I built the S03 with NDK 45/50mhz clocks. So now the chain is BBB running Miero's excellent distribution to S03 followed by Ian I2S to PCM card to Ryanj's Distinction 1541a board. The S03 finally removed the processor and the software from the sound equation.

I believe in a few key constructs for great sound:
-individual PS with very good filtering for every consumer
-separation of power supply areas from consumers. The layout is basically separate boxes for digital PS, Tube PS, digital processing and DAC sections all mounted in a nice Walnut frame. All are mechanically and EMI isolated. Sections are boxed off with either brass or copper sheets which are grounded. The sections all physically float on their mounts often with a silicon barrier to dampen vibration.
-good layout keeping power lines and signal separate
-short distance between regulators and consumers
-if PS is important, then the grounding path is even more important. There is a star grounding point near the tube amplifier with 4 gauge copper taking current to the outlet. All PCB boards are all mounted on dual sided copper PCB stock which acts as a ground plane to the star ground. Another important tweak is the use of a ferrite bead on the ground bus. Something I picked up from building power cords is that a ferrite clamp on the ground wire at the female end of the power cord is great for sound. So I took it further and put one on the ground wire close to the star ground. It had an immediate and positive effect lowering the sound floor and increasing bass quality and quantity.

The Tube Section
Again, the design is taken shamelessly from Lampizator (plagerism is a complement!). The PS is dual mono. There is a split bobbin transformer for each B+. Another for the tube rectifier filaments and another for the signal tube filaments. All the transformers are mounted with silicone in soft wood which as a unit floats on its rails and is surrounded by brass plate. I paid attention to wire routing, and used tightly twisted pair of 20 gauge Teflon wire. 14 gauge ground path is terminated only at the star ground.

I have used Lampizator tube output now on several CD players and my previous DAC. This is just a cleaner build of the same using better parts and for first time his best PS. Below are pics of circuit diagram. I believe he credits the tube amplifier design, (he claims it is not a buffer) to his friend and mentor at ARG amplifiers(Ewgenniy Kreminsky).

I do not follow all his advice. I use an I/V resistor of 60R. Best quality here is high payback. I tried many different resistors and hand wound Manganin wire used here. Alternative would be Rhopoint.

For cathode resistors I started with NOS Shinkoh Tantalum made famous by AudioNote. The whole sounded good, but I was itching to hear it with more DIY Manganin. I installed the manganin and was rewarded with immediate jump in sound quality.

The coupling cap is Russian MGBO-2 4u bypassed with 1500pF silver mica. It is great value. Upgrade might be Duelund RS.

On PS, I use all Hammand iron for trans and chokes. Values are mentioned. I use separate trans for each channel and wire for 115v output. My B+ measures 195v. The film caps are 100u 400V polyprop caps used for solar panels. I like the convenient square shape. First caps after diode is 2u Obligato PIO I had left over. If I did it again I might build 2u, 50u then 100u.

The Diodes started with Russian 5C3S. Very nice. They died early. Replaced with NOS 5U4GB Marconi.

The the tube section is dead quiet. With it attached to my OTL tube amps there NO sound with ear directly on the tweeter! Very low sound floor is letting lots of micro details shine through.

The Digital Section
Inspired by Oliver's red baron project, I used a large number of filter caps per rail. Each rail has a Hammond 229 transformer. They are similar to R Core giving good isolation, but I got them locally in surplus. They do have a tendency to have some mechanical hum. I mount them with silicone in soft wood for dampening. Filter caps are also mounted with silicone in the soft wood base. The caps in the critical -15v rail are actually set in silicon in holes in the wood for best possible dampening. I use super soft recovery diodes to build my bridges. Each rail got a CRCRC filter with around 9mf of low ESR filter caps. The last caps are black gate. The PS for the +5 -5 and -15v to the TDA are all on top. They then feed Salas shunts by Oliver Mai to feed the DAC PCB. On the bottom are the PS for S03 clock, and BBB. You may notice that these are repurposed from a Peter Daniel USB DAC. The +5v supply to the I2S 2 PCM board and the S03 isolator and digital sections is another linear supply with a small Hammond 229 transformer and 9mf of filter caps. This supply has a Salas shunt. The S03 was great as supplied. But it too benefited from PS tweaks. The Salas shunt became a 5v prereg. Each consumer on the S03 got its own tps4700 based discrete reg by Ian Canada. Ian's regs got BG NX HiQ caps and BG N caps. Every power consumer on the So3 got its own BG NX HiQ cap right on the ps pin. The Ian I2S board got more of the same. These mods took the system to another level.

The TDA board is the very nice Distinction group buy (by RyanJ). It is truly a work of art. I use 6k8 resistors to a 100pf silver mica cap on the DEM. Sounds good. Hard to know if this DEM bests the Grundig circuit I used previously. All coupling caps are .68u film right on the pins. The + and - 5v supplies have BG HiQ 33uf caps. The -15v has a BG N 33u. At the input to the pcb there is a BG FK 470uf on the digital PS and a BG N pair of 1000u caps in Super E config. The TDA chip is an S2 double crown from Taiwan. Thanks Roger!

One small touch that I like is that I used 2 power switches.. one for the DAC and one for the tube output stage. I like to leave the digital DAC on 7x24. With this arrangement you can just flick a switch and the tube section soft starts as the rectifiers come up to speed and you are ready to play. The DAC is always warmed up. I think this is always good, but with Blackgates even more important.

How Does it Sound?
Pretty darned good! It is very clear. Lots of micro detail. Very relaxed. Layered 3d sound stage. Overwhelming adjective is natural.

With new caps and USB it started out cramped and closed without much bass. What follows is the evolution
7 Days - 170 hrs
OK, its been running non stop for a week. 2 tweaks.
1. the USB cable is home built and only about an inch. Earlier I took a bunch of heated advice about how important impedance is from the usual anti cable crowd. I did some digging and found cable impedance calculators online and crafted a short cable with what should be around 90 ohms using tight twisted pair teflon 24guage silver plated wire. the signal pair is shielded with alu foil while the power pair is separated. I have to say the sound did improve.. somehow fuller.
2. I ran Rune Audio distro to make this easy for now. As good as it is I brought in some parameters I had fine tuned in my previous MPDPUP ALIX setup. For my taste, big step forward. My favorites are documented at Computer Audio Asylum

So with the maturing burnin and these 2 tweaks, I sat down for some serious enjoyment. As before, the overwhelming impression is very relaxed and natural vs artificial fireworks. Lots of detail but very clean. With my parameters changed, (nrpacks=20 is key) the big change was spontaneaous toe tapping. Yup, when my system is working right there is something about the music that just pulls you in and makes you need to move to the beat. I can't describe what it is. Could be what some people call musical. For me it is I can't stop moving to the sound.
14 Days/340 hrs
It continued to mature. Just continues to settle into a relaxed clean and natural sound scape. A little more separation around instruments.

First Tweak completed. I was intrigued to know if there would be any benefit from my beloved Blackgate caps on the Distinction TDA board when I already had Salas shunts and really well filtered supplies. In my last DAC I had a nice little Blackgate N soldered right on the #15 -15v pin. I can't bring myself to solder the pins of the S2 double crown, so I put BG's on the supply terminals. Using what I have available, I put 2x BG N 1000uf reds wired in super E configuration on the -15v supply and a 100u BG F on each of the + 5V / -5V supplies. I must say, to my ear there is nothing that matches better with the 1541a Chip than BG caps. The -15v made an obvious improvement. The BG F on the 2 digital supplies made an even greater impact. There is just a continuation of silence between notes and instruments. Elimination of any digital edge on high frequencies and with the addition to the +-5v supplies a nice turn up in bass response. Yes, I am completely biased to BG, but for me they remain a great addition if you can come by them at any kind of reasonable price. I wish I had accumulated more of them when they were easy to find. Currently there are 18 in this DAC, and I would not part with one of them. See last pic.

3 weeks burning and 2 more tweaks

Next optimization was to correct the grounding on the BBB and processor side of WaveIO. I realized that I had attached these both the main ground plane, effectively nullifying the isolation at the output of my WaveIO. I put both the BBB processor and the WaveIO on nylon standoffs to isolate them from the ground plane, and grounded their PS back to the IEC outlet ground. This bypassed the rest of the ground plane and the noise from the digital processors is isolated by the WaveIO isolator. The large ferrite clamp on the ground bus at the star ground will also prevent noise from polluting the ground plane. This change was not dramatic, but quite perceptible as a nice little bump forward and away from digital hash.

The second tweak was MUCH more impactful. If you look at the schematic for the tube output amplifier, you see the cathode resistors are really in the signal path. As good as the Shinkoh Tantalum resistors are, I was intrigued to see the impact of DYI manganin wire wound resistors in this position.

I replaced the 220r cathode resistors in the output amplifier. All I can say is WOW! As it was in tuning my amps, upgrading the resistors in the signal path is the single biggest impact. Even more important this early in the signal chain. As good as it was before, this simple upgrade was like the difference between foggy glass and no glass at all in the window. Everything clearer and more dramatic. Particularly important in the higher registers. Bells and cymbals shimmer forever. Air around everything. Even the resistors take time to break in. The surprising outcome of this change is how much better classical music sounds. I always thought my speakers were not up to the challenges of classical music and opera. They tended to get muddled in complex pieces. Now, no problem and I find myself more interested in this music.

Anyone building this circuit is well advised to skip the experiment and go straight to best possible resistors in all 3 positions. Naked foil would be the minimum. [edit] after a week of continuous play, the manganin resistors did continue to burn in and the sound just opened up.. micro detail, no digital sound, highs/lows air.. pick your audiofool adjective it is all there. Sound is now entirely dependent on quality of the recording not the DAC. Well recorded music without the heavy hand of an audio engineer is delightful.

In the pic, the 2 larger white blobs are my I/V resistors built some time ago. The 4 vertical blobs are the new cathode resistors. Basically a piece of bamboo wood with copper wire on each end.. double sided tape around the wood, and Manganin wire bifilar wrapped on the tape and soldered onto the copper leadout. The soldering the wire is the tricky part as it is almost impossible to see even with a magnifying glass. Mostly these things are build by sense of touch! Or just buy the Rhopoint!

Planned or Done
1. Blackgate N. I have several of these to place in critical locations. In the past it has been particularly good to put one right on the -15v pin. Also interested in trying them on the power pin of the clock. Done As usual I use what I have available. Put a 4.7u Blackgate N right on pin 4 of the clock on my WaveIO. Big hit. Another big jump. Like hitting the loudness switch and Holograph switch. Way more separation. Everything distinct in 3d space. Also deep extension of bass. Tight controlled and 3d. spooky. This tweak done at the same time as vibration dampening. Vibration first made an improvement in detail and bass but still bleeding together. The cap snapped it into focus. Recommended if you can find one.

2. Ground isolation. I realize that the BBB and the processor side of the WaveIO share the same ground plane as the rest of the unit. I am thinking of isolating these grounds and take them directly to the ground exit at the IEC plug. Done recommended
3. Bud Purvine's Ground Control thingy on the ground bus right at the tubes. Sorry guys, it really does work Done recommended
4. Modelling clay blob for vibration dampening on each of the clocks Done recommended
5. After the clay on the clock, added a Blackgate N directly on pin 4 of the clock. Done massively recommended
6. Next I changed the power supply for the -15v 1541a pin 15. Done It was a CRCRC followed by salas shunt. The DDAC guys inspired me to swap resistors for filament chokes. First I changed out the first resistor for a DIY filament choke. I made it with 18awe magnet wire on a 2.5" ferrite core from a torroid. The improvement was surprising. Immediate improvement in base and general separation of instruments in the sound stage. I ordered a Hammond 159y choke for the next section. It is .65mH at 7 ohms. When installed, it initially was a disappointment, collapsing sound stage and removing air and toe tap involvement. After 5 days of 7/24 run in, it has opened up again and yields even more natural sound. Very relaxed and real. So now with CLCLC and about 9mF of low esr filter caps the sound is at another level. Well worth $30 for the choke.
7. DC Block circuit Done May not belong in this blog as it was in the power distribution/conditioner. I always have struggled with some level of transformer hum. Always suspected there is DC offset in my mains supply. I decided to implement a DC blocker circuit as documented in DIYAUDIO. The circuit consists of large ripple current caps in series with the line power. These are bypassed by a set of diodes for protection. In my case I needed a lot of caps. Each pair of caps are 12,000uF 8A caps with the + to + in series with the hot AC lead. My system required 3 pairs of these large caps to filter out the DC and provide unrestricted current to the system. The result was dead quiet transformers in my entire audio chain and an increased level of dynamics. Clearly transformers were saturating and this reduced the efficiency of my otherwise over sized power supplies. Results are just stunning. This is not for everyone. If you don't have hum or buzz in your torroids, then you don't have to cure it. But if you do, then clearly it can be fixed and the results are well worth it.
8. PS optimizations ..CLC filters everywhere. Done
First, I replaced a standard silicon bridge rectifier on the PS for the WaveIO isolator and the I2S to PCM brd with another DIY bridge made from super slow recovery diodes. Obvious improvement. This simple little PS is clearly key to good sound.
So inspired again by ecdesigns I experimented with higher order filters on most of the digital and dac PS's.
First up was the the PS just mentioned. Initially added a small inductor to change it from CRC to CLCRC. POW! A big step in SQ. Improvements to this PS adds bass, texture, 3d separation and more of that relaxed precision.
Next up was the 1541a -15v supply. It was already CLCx6LCx6. Recall the first inductor is DYI 18guage on a large ferrit ring. Second inductor is Hammond Choke. So I did not expect alot... I had aquired a dozen 12mh high amp inductors in surplus. So I added 2 to this supply resulting in CLCCCLCCCLCCCCCCC. Caps are in order Panasonic FC 1800 x1; Panasonic FC 1500x12 then BlackGate F 100. This add of 2 inductors resulted in a surprising improvement.
So next I went to the 1541a +-5v supplies. These started as CCCCCCRCCCCCC. I changed them to CCCLCCCLCCCCCC. The first 6 caps are Panasonic FC 1500u followed by 5 BG 250uf +1BG 1000uf. Improvement worth the effort but far less impact.
Seems there is no end of sound if you are willing to produce excellent PS.
9. Eliminated USB. Done Brought in S03 master clock and isolator. BIG step forward. Then optimized its PS with discrete tpa4700 regs and BG HiQ caps. This is the real deal. I seriously doubt there is much to be gained from vinyl. Done and no turning back
10. Silver Mica Coupling caps Done The output caps were 4uf Russian MGBO-2 foil & waxed paper crossed with 1500pf silver mica. Found some Russian 0.1uf Silver Mica which replaced the smaller values. Surprising improvement in detail. Very smooth, clean neutral sound. At about $6 on ebay, a no brainer upgrade. Removed the metal tabs and soldered silver wire right into little tubes. Nice!
11. Tube rolling. Done 5U3C rectifier swapped for Marconi NOS 5U4GB. Awesome. Warm clear natural. Deeper into the music. Better bass. It has it all.

For now, I just need to build a top plate and listen to my music collection. This really is providing a new insight into well recorded music. Once it broke in the magic came out. You just want to listen to the music and forget about the hardware. This was my basic objective and it seems to be achieved.
Attached Thumbnails
Click image for larger version

Name:	reference music server top side.jpg
Views:	1446
Size:	291.9 KB
ID:	1568   Click image for larger version

Name:	reference music server build bottom.jpg
Views:	718
Size:	356.6 KB
ID:	1569   Click image for larger version

Name:	bird cage test.jpg
Views:	376
Size:	430.1 KB
ID:	1570   Click image for larger version

Name:	Tube PS build 0.jpg
Views:	268
Size:	183.8 KB
ID:	1571   Click image for larger version

Name:	Tube PS build 1.jpg
Views:	332
Size:	216.5 KB
ID:	1572  

Click image for larger version

Name:	Tube PS build 3.jpg
Views:	327
Size:	183.9 KB
ID:	1573   Click image for larger version

Name:	Tube power supply.jpg
Views:	651
Size:	239.1 KB
ID:	1574   Click image for larger version

Name:	Blackgates on TDA board 2.jpg
Views:	661
Size:	784.7 KB
ID:	1588   Click image for larger version

Name:	Lampizator 6n2p circuit.jpg
Views:	1998
Size:	59.2 KB
ID:	1591   Click image for larger version

Name:	DIY Manganin resistors.jpg
Views:	660
Size:	211.0 KB
ID:	1592  

Posted in Uncategorized
Views 2175 Comments 0
Total Comments 0



New To Site? Need Help?
Copyright 1999-2017 diyAudio