• Raspberry Pi audio server
• Access to my 30 year music collection, primarily .MP3: < 100GB so far
  
• .WAV and .FLAC files
• A nice Android / Apple app to control it all
• Good quality audio: 16 bit, 44KHz and up
• Low noise, distortion, and jitter: high every thing good, low everything bad
  
• Streaming of Pandora and Spotify, maybe others
  
• Line-level stereo audio into my whole-house stereo system

Volumio system on Raspberry Pi

First Problem: Low Voltage Warnings

Second Problem: Audio Noise

Potential Fixes

The proper fix for CM noise is to use a power supply with low CM noise.  Grounded wall-warts such as a USB Hub or other grounded AC power supply will cause a ground loop. I  experimented with a floating Lab Power Suupply. As I expected, it was quiet. But when the power supply - was grounded, the noise returned.

Since the power requirements are  low, < 1A,  a linear power supply with a good low-common-mode transformer is a decent choice. There are a handful of low noise / linear power supplies intended for RPi audio systems.

Ian Canada LinearPi Dual has filter caps and LDO regulators for ~$90. It takes in either 6VAC or 7-9V DC. But it does not address CM noise source. You still need a low CM transformer or power supply.

A second problem is noise on the 5V and 3.3V RPi power side getting into the audio.  I see various RPI power boards designed to improve this power situtation. Some "audiophile" projects go as far as to use SuperCapacitors and Lithium batteries to power a RPi.  Allo makes a handful of power solutions in the $20-$300 range. Their UsBridge at $240 is super-quiet using a RPi Zero module. It has dozens of linear LDO regulator and filter caps for every IC and connector.

I see that Banggood and Ali Express offer RPi filter boards with the name: "DAC Audio Decoder Digital Broadcasting Power Filter Purification Module For Raspberry Pi3 3B 4B" for < $20. It consists of a heap of (supposedly) high quality (AVX) Ceramic, Tantalum, and Electrolytic capacitors. What could go wrong with a Chinese board full of capacitors? I've been an electronics nerd my whole life and this is the first time I've heard of a "Purification" circuit. Love it. But seriusly, I  like that it replaces the iffy USB Micro power jack on the RPi 3 with a nicer Coaxial power jack. With this board sandwiched between the RPi and the DAC, it should help isolate the RPi noise and help eliminate ground loops.

RPi 4 with USB C power would be a good choice too.

DAC HAT Boards

I ordered 2 different DAC HAT boards from Amazon and an Ali Express Purification board. Couldn't resist. The DACs are HiFiBerry DAC2 Pro and HiFi DAC HAT from www.inno-maker.com. First one to come up and work, wins! Here is the HiFiBerry DAC2 Pro:

And the HiFi DAC Hat:


I received the DAC boards. The DAC2 Pro worked, but Volumio had no specific DAC setting for it. After a brief internet search and some trial-and-error, I found that the Volumio "HiFiBerry DAC2 HD" setting works, but no hardware volume control. You'll need to use software volume. But software volume just multiplies the 16b digital data by a volume number of 0.0 to 1.0. This effectively reduces the bit resolution of the digital data, and is worse at low volume settings. Not great for a high-quality audio system. I do like that it has header connectors for +5V and audio output. These are handy for mounting your server in a larger enclosure (below).

The HiFi DAC HAT works with the "Allo Boss" setting. Hardware volume control work fine. For some reason it has an unreasonably tall connector, requiring 16mm spacers. The standoffs that came with the HiFi DAC HAT are pretty bad: too short for the tall connectors, and cheap plastic. The official standoffs for RPi are M2.5, but I found that #4-40 and #2-56 also work. #4-40 are a tight fit in the RPi holes but can be coaxed in with a bit of threading. They fit fine in the various HAT board holes. I used #4 7/16" for the Filter board and 5/8" for the DAC. I like male-female (M-F) aluminum, 3/16 hex standoffs for board stacks. Erickson Engineering doesn't currently stock M2.5 standoffs, just #4 and #2. Those $10 brass M2.5 kits on Amazon look good.

Sound quality is very good with both DACs. I noticed that the HiFi DAC Hat uses a 470 ohm resistor and 2200pFcap output filter. I was using this board to drive an audio transformer directly and the hum is pretty bad. This is not a good solution. Should have the output buffered.

The "Purification" board arrived and I installed it. No problems, and I can't tell if it improves noise. I'll need a proper isolated 5V power source with a coaxial power plug.

RPi Shortage 2022

Enclosure

Better Audio Testing

• 24b Signal generator: Sine, square, noise
• 24b Digitizer: Scope, spectrum analyzer, waterfall plots
  
• Frequency response
  
• Speaker impedance and response testing using a calibrated and amplified microphone.
  
• 2 channels
  

• Calibrated input input and output gain and attenuation: -40 to +40dB,  6db or 10dB steps
  
• Switch path for loop-back
• Calibrated measure circuit
• USB control
• 100+dB SNR
• Packaged to work next to (or on top of) a USB Sound Card
  
• Maybe balanced I/O
• Some type of ground loop eliminator: Big ferrite plus USB isolator?
  
• Advanced feature: Notch filter for better distortion testing?
  

Clean Isolated Power

If you are building hi-end audio equipment or precision instruments, I strongly suggest that you watch Marco Reps excellent video on AC leakage currents. He goes to great measures to measure and then design low-leakage AC powered equipment. How does this apply to Audio? With home equipment and un-balanced connections everywhere, the line leakage current flows through your un-balanced audio cable shields to find the shortest path to ground. This is true for every piece of 2 wire line cord audio equipment you  have. They all induce leakage currents into the audio grounds.

The commercial guys use well designed transformers with inner shields between the primary and secondary windings. We mere mortals, dealing with off-the-shelf transformers do not have such luxuries. I know of no well shielded, off-the-shelf transformers. But there are things we can do. Split-core transformers have physically separated primary and secondary windings, and so are quite low in leakage, typically less than 1uA, which is quite good. 1uA of 60 Hz AC current through your decent RCA cable grounds is pretty low noise: a typical cheap 6' RCA cable has about 0.15 Ohms of ground resistance, so 1uA * .1ohms = 0.15uV of 60 Hz. Compared to -10dBV line-level signal (0.316V) this is 130dB down. Multiple (stereo) cables in parallel, higher quality ones, or shorter ones will all reduce the shield resistance. The problem is that most audio equipment has much higher AC leakage, in the 10's of uA.

I just figured out a neat trick to reduce leakage on a split-core transformer even lower than the typical 1uA, by about 10x. Tried it on a Hammond 162G12, 12.6VCT, 12W. It reduces AC leakage current by about 10x. The trick only works with 120VAC US power and off-the-shelf dual-primary transformers. You connect one primary of a dual primary AC transformer to the 120V line and neutral. Then connect the other primary reversed, with only the neutral connected, so it generates the opposite phase120VAC. The other pin of the second winding is not connected, it just floats. This out-of-phase winding cancels out the electric fields from the primary, and so most of the leakage current also cancels. I get almost a 10x reduction.  Of course the transformer won't perform as well with only one of its 2 primary windings connected: Less total power, more V drop. But damn is the leakage current low!

I'm trying to build a quiet, isolated, 5V linear power supply for my Raspberry Pi music server. This trick will also work for instrumentation. I plan to use the circuit below to drive a 5V linear regulator. The transformer is 12.6VAC center tapped, 6.3VAC per side. It generates about 6.3V * 1.414 - 0.7 = 8.2V DC. Subtract a bit more for the ripple voltage, low line voltage,  and the single primary, and you should still get a bit more than the 7.0VDC needed for a LM7805 type regulator's 2.0V dropout voltage. LT1086 and others have lower dropout and higher currents.

I may add a Common-mode filter to the AC line to reduce high-frequency noise as well.
 

Clean Isolated Power

I'm aiming for a clean, linear RPi power supply, simpler, low cost and DIY friendly. Something anyone can build for themselves.

• Low common mode and output noise and ripple. Lots of filtering
  
• +5V at 1.5A plus, Enough for a light RPi system
• Very good isolation, < 1uA AC leakage
• AC powered, IEC 3 wire line cord
• Maybe a line filter for high frequency noise.
  
• Reasonably efficient: Low-drop-out voltage
  
• Low cost: ~ $40 parts including enclosure
  
• As small and light as possible
  
• Easy to build and package: Off-the-shelf transformer, thermal design, and enclosure, minimal machining, easy thru-hole and SMT soldering
  
• Safe and reliable, fused, all line components UL/CE, but no overall safety certs. Builder beware!

Good efficiency is needed so this can be powered on 24/7 and not waste power. As I pointed out, my current RPi music server draws < 300mA from +5V, even when driving an HDMI display and keyboard. That's about 250mA * 5V = 1.25W. I'll aim for 50% efficiency, so about 2.5W typical from the AC line.

The LT/LM1086 regulator looks good. Input capacitance and output capacitance is a handful of 1000uF electrolytics in parallel. A 1000uF 20V cap is about $.25. I plan to use 8-10 of them. Here is a simulation of the regulator showing 120Hz ripple at about 100uV, and load regulation < 1mV. Not bad.

Turns out I don't really need this, but I thought it would be fun to build a RPi server with high-quality power for DIY folks.

After a bit more googling for low noise audio power supplies, I found several on Ebay and Ali Express. I bought this Ebay one, couldn't resist the $16 price for the board and components. The title is "HiFi DC5V-24V Ultra Low Noise Studer900 Linear Power Supply Regulated Board/Kit". It is based on a power supply design used in the Studer 900 Swiss-made Studio mixing console. It is a decent, discrete regulator design, similar to what you'd find in an old electronics instrument.I built the board up pretty quickly. Assembly went well.

I found a similar Studer 900 schematic on-line and then reverse-engineered the board to generate a schematic. It is a low drop-out design using an NPN pass transistor MJE15030 in the negative path. So an isolated input supply such as a transformer is needed. The coarse output voltage is set by RB. Several values are provided with the kit from 5V to 24V. I use 1K for 5V. The current limit is high, about 0.6V / 0.1 ohms so about 6A Increasing the 0.10 ohm resistor would reduce the current limit. I'm thinking 0.33 to 0.47 ohms. It uses a separate negative power supply to provide V- for the TL074 dual op-amp.

The output ripple and noise are quite good, I measure under 50uV when driving a 0.5A load, with a 6.3VAC transformer, wired with a single primary and the "low leakage trick". I'll need to find a suitable enclosure to house this board, an AC input fuse and filter, and transformer.

Here is the original Studer900 schematic on which this board is based:

And here is the reverse-engineered one I made from the board.

I measured the +5V supply on my RPi audio server and found the largest contributor to +5V noise. Is it the +5V regulator? The 5.5mF filter board? Nope: it's the USB or 5.5mm coaxial power cable. This cable's + and GND resistance (R) in conjunction with the RPi's widely varying current drive cause the +5V to vary considerably, particularly in the audio frequency range. My Studer900 supply with a 24AWG 2m coaxial cable caused as much as 100mV p-p noise. I swapped the high resistance cable with a 20cm, 20AWG cable and the noise reduced considerably.

Clean, Switching, Non-Isolated Power

I want to mount the RPi server in my Whole-house Audio system, in the existing case. It needs to squeeze in along with the current 4 preamp boards and 8 channels of amplifiers. The system has +/- 30V power supplies for the amplifiers, and a +/-12V dc-dc converter for the preamp. There is a small 5V regulator, but not enough power for the RPi server. I'll need to convert the +30V to +5V at about 1A, cleanly and efficiently. I plan to use a 30V -> 6V non-isolated dc-dc switcher followed by a 5V LDO linear regulator. I only need to build 1 or 2 of these, so planned to build on a small proto-board.

For the 30V->6V switcher, I first tried using one of the cheap, Chinese "LM3596" blue, buck converter boards, but the output and input noise of these is pretty bad, about 200mV at 55KHz. I want to begin with a clean dc-dc. There are many choices, but the most convenient are 3-terminal 780x replacements, and some small boards. I like Pololu for small boards and have used their CPUs and motor drivers in the past. Their small D24V10Fx switchers use the ISL85410 synchronous buck DC-DC. I like that the output filter uses multiple (6) ceramic capacitors to lower ESL and ESR. However they do not specify output noise, so it's a bit of a crap-shoot.  I bought the 6V one for $10.  6.0V should work fine. If I need to add extra L-C filtering between the dc-dc and the LDO, I might need to increase the 6V a tad. The output is adjustable over a small range but requires changing one of the SMT voltage set resistors. For the 5V LDO I plan to use a LT1085 or equivalent on a small heat sink.

The Pololu board arrived. It measures 40mV p-p output noise at 500KHz with a 0.6A load. Using no filter, and a LM1117 LDO board, I get about 20mV of noise on the 5V. I'll need to mess with some LC filters to reduce it further. This is harder than I thought.

I'm having second thoughts about putting the RPI inside my Whole-house enclosure. The Whole-house system is already complicated enough. And removing / installing it to work on is not a simple process. Besides, I use it every day. Putting RPi outside, in a separate enclosure with its power supply makes life easier. I already have a nice clean AC power supply for it.

Final Solution

Here is the system in a Hammond clam-shell enclosure, type 1598D which is about 8"x7"x2.5". The RPi USB and ethernet connectors exit from the rear panel. So I'll need to reach around back to turn it on/off or to access the USB thumb drive. I did this partly because we have curious 1 and 3-year-olds at our house a few days a week.

The case is simple, and works perfectly but maybe not the most beautiful. Any nicer case ideas?

I use a switched and fused AC inlet from Amazon. I had to decide how to ground the rear panel. In order to avoid ground loops and to minimize the AC leakage current, I'm purposely not using the ground pin of the 3-wire AC line cord. If the transformer had a mounting strap, I'd ground it to the AC ground to reduce leakage. Also I made the opening for the RPI connectors large enough so they would make contact with the rear panel. Instead, I just used non-isolated (grounded) RCA connectors. They are the only thing connected to the rear panel and since they are what provide the external ground, it makes sense.





I decided to add a headphone jack to the box. I recently bought a nice pair of headphones, and want to be able to listen to my RPI server with minimal hardware between DAC and Ears. Unlike most DAC HATs, The HiFi DAC HAT has a decent headphone amplifier built in. All I need is a cable to connect the 3.5mm jack to a 1/4" headphone jack on the front panel. I'll also add a power ON LED.