scan80269

MEMS microphones with PDM or I2S output should technically count as digital microphones. These are very popular for modern laptop PC designs since MEMS mics are typically quite small in size, interface easily to main boards, and can have superior specs due to lack of need for separate analog microphone amplifier and A/D conversion. SiSonic™ Surface Mount MEMS Microphones (knowles.com)

1000Base-T gigabit and faster use all 4 twisted pairs for transmit and receive simultaneously with full duplex. This differs from USB 3.x SuperSpeed where the TX & RX pairs are always unidirectional. I doubt if USB 3.x would be beneficial to DACs. The much faster switching speed makes the receiver circuits more power hungry and thus electrically noisier. This is a possibly a case where faster is not better for audio, unless the noise spectrum is easier to filter/block than that of USB 2.0 HighSpeed.

I'm extremely late to the EtherREGEN party, and have not posted here for a very long time. Long story, but it had much to do with the loss of interest in audio after the passing of a dear colleague/mentor/friend and a family member in 2019. Recently, an EtherREGEN has been introduced into my headphones system, to my major delight. Both an ISO REGEN & EtherREGEN each delivered very noticeable SQ improvements. There are now a total of 6 UpTone Audio devices serving this rig (not sure if this is a record? 🤣) Attached is a block diagram of this headphones system. It is heavy into isolation, with multiple REGENs, LPS-1/1.2 supplies and isolation transformers. China built LPS from eBay were modified with Schottky rectifiers and better electrolytic caps. This headphones rig has never sounded better. I've been listening critically to a couple of tracks recommended by audiophile buddies. There is very significant increase in piano heft, midrange fullness and treble smoothness, but the most stunning effect has to be dramatic increase in reverb effect. It sounds like someone turned up the reverb during track mastering. Reverb tails are much more easily discerned and sound longer. Many tracks now have voices & instruments enveloped in a halo of reverb, something that I have not always noticed before. There are also other improvements. Imaging of voices & instruments got more focused and sharp, while the sound stage expanded. I didn't even realize those two could be accomplished simultaneously, especially with headphones. Anyway, I'm back to doing a good deal of listening. scan80269_Headphones_System_2021-0314.pdf

Kryonaut is less sticky than Arctic Silver 5, though neither is difficult to work with. I started with AS5 a few years back and discovered Kryonaut last year. AS5 is much cheaper per gram than Kryonaut. I didn't clean the two thermal pads but instead removed them altogether as they interfered with CPU cooling drastically. There was no notification from Akasa of this issue and I discovered it the hard way. To clean the exposed surfaces of the 3 die (CPU, chipset, eDRAM) within the processor package I used cotton swabs. The amount of thermal compound needed for Akasa Turing, including the replacement of the white grease between the aluminum block and the chassis, is quite small, and I happen to have bought a rather large tube of Kryonaut some time ago, so... Between Plato X8 and Turing I believe the latter has higher cooling capability. The surface area and mass of Turing serving as CPU heatsink dwarf those of Plato X8. I saw Plato X8 released by Akasa first but became skeptical of its ability to cool a 28W TDP CPU, as it seems similar in size to earlier Plato models that only had to cool 15W TDP CPUs. I'm glad Akasa did Turing. Also see the following thread on avforums regarding a NUC8i7BEH / Plato X8 build: https://www.avforums.com/threads/intel-8i7beh-8th-gen-nuc-akasa-plato-x8-fanless-pc-build.2225432/ One can expect somewhat higher CPU core temperatures with Plato X8 compared to Turing. I'm really pleased with Turing performance with my i7 Bean Canyon. I'm contemplating building a second unit for a different purpose...

I just put my NUC8i7BEH into an Akasa Turing chassis. It was a pretty standard build except for two areas: 1. Akasa Turing uses an aluminum block to transfer heat from CPU to the case acting as heatsink. I found standard zinc-oxide type thermal grease between the aluminum block and the case. I replaced this grease with some Kryonaut compound. 2. There are two small thermal pads covering the chipset & eDRAM die on the i7-8559U substrate. Originally I left these pads in place after detaching the Intel thermal solution, but the CPU quickly overheated and reached THERMTRIP (abrupt system shutdown), so I ended up removing those pads and just put Kryonaut over all the 3 die. Those thermal pads apparently propped the CPU die up from the aluminum block surface and compromised the heat transfer. I ran the most stressful test of Prime95 for an hour, and CPU cores reached only 75C with an ambient of ~20C. CPU package power stabilized to 29-30W during this test, after several seconds of Turbo Boost with elevated frequencies. The Akasa Turing is more than capable of keeping the 28W i7-8559U CPU of Bean Canyon NUC8i7BEH well within operating limits even under extreme workloads. I have only tested with the case in horizontal orientation and the top surface got quite warm but not hot to the touch. Turing really works like a gigantic heatsink and it's very effective.

Intel's NUC7i3BNH with i3-7100U processor has an M.2 2280 slot and can support Optane as SSD. This NUC is cheaper than NUC7i5BNH.

Your assessment is correct. FUN01 is USB bus powered on the "upstream" side of the board (with USB - the power comes from the USB source device, e.g. streamer, PC, and runs through the USB cable), and externally powered on the "downstream" side of the board (with the various digital outputs). The galvanic isolation feature requires both sides of the board (separated by the isolation barrier "moat") to be separately powered. The USB side is typically USB bus powered to eliminate the need for two separate power supplies just to run the box. Singxer SU-1 also has this same power architecture by the way. SBooster VBus2 isolator cuts off +5V VBus power from the upstream USB source, so it cannot be used with FUN01 or SU-1.

An un-programmed serial flash chip will not work. If that circuit board is really missing the serial flash for the Xilinx chip, you'll need an exact replacement chip with the correct programming. Best way to DIY is to track down another FUN01 unit and hook up an external programmer to read out the code from the serial flash chip, then program a blank chip and solder it onto your board. I recommend soldering on an 8-pin SOIC chip socket at the location to allow easy drop-in and removal of the SOIC8 serial flash chip.

That may be because DACs can have much more complex circuitry than DDCs. For example, a DAC has an analog output section that a DDC does not have. Fancier DACs have dedicated oversampling circuitry like a DSP in front of the D-to-A conversion. Modding a DAC for better sound is often more challenging than modding a DDC. With that said, I have successfully modified the power section of three Auralic Vega DACs and reaped some benefits in SQ.

Holo Audio Spring DAC uses two SMD crystals, not oscillators, so Crystek CCHD-575 would not work, nor would they fit. I did replace those two crystals in my Spring last year, but don't recall offhand what brand/model I ended up with, nor did I manage to listen critically for any SQ difference after the mods.

Last year I identified the two crystals in the Spring as belonging to the Harmony Electronics HSX531S series. I have no idea what phase noise level these crystals have, whether Holo Audio uses sorted devices from the supplier, or how well designed the two crystal oscillator circuits are in the Spring. Alex @UpTone thinks these crystals are rather mediocre and I tend to agree. If unsorted Crystek CCHD-575 oscillators have lower phase noise than these crystals with their circuits in the Spring, then it can explain why the sound is better with the SU-1 driving the Spring over I2S vs. USB feeding directly into Spring. I do agree there are many factors besides crystal/oscillator performance that can affect the sound.

I believe the answer is yes, but the reason for the Singxer SU-1 route sounding better has to do with the Crystek CCHD-575 audio reference clock oscillators in the SU-1 performing better than their crystal (not oscillator) counterparts sitting on the Holo Spring DAC main board. Both the SU-1 and Spring USB inputs use XMOS XU208 USB transceiver plus galvanic isolation with I2S output, so they are quite similar. The sonic difference correlates to audio reference clock performance differences.

It looks like the "IN" pad is an exposed section of a larger top-side plane connected to the +5V coming into the circuit board. Unless that area got butchered during the rework, it should not be easy to damage that pad, as it is not a standalone pad that can de-laminate. There should be no need to bypass that pad. You are unlikely to find a wiring diagram on the net for SU-1. Most folks doing rework on SU-1 figured things out empirically.

I assume you installed a Sparkos Labs SS1117-3.3 regulator into your SU-1? The outer leg should be the regulator input. I'm not sure I understand how you can be getting correct voltages through the Sparkos and have a damaged pad for the regulator input at the same time. Why don't you post a close-up pic of your SU-1 showing the Sparkos regulator soldered to the board?

Your SU-1 may have sustained damage while your VR was wired wrong. One or more of the damaged chips may be drawing excessive current which causes the DC output of the linear power supply (encompassing the VR) to get dragged down and out of regulation. A working stock SU-1 consumes fairly little power in both upstream (USB) and downstream (digital outputs) sections. Any chip that runs very warm or hot is likely damaged.