attitube: Dual AK4399 / AK4118 DAC

Dual AK4399 / AK4118 DAC

After years of research, I cannot find a CD player or a DAC within my budget to successfully replace my digital source: Audio Refinement CD Complete build by YBA in the far east. I have build two DDAC multi-TDA1543 in the past, first with 8 DACs in parallel and second with 12. The first one, with 8 dac chips and the receiver CS 8412, performed really good on battery, being brilliant on classical chamber and acoustic music. It never rocks 🙁 on other kind of music. Despite its lovely mid-range, with amazing air and detail, the bass section was its week point, whatever I tried. The second I have build (12 chips in parallel and the receiver CS8414) did not approached the sound of the first, and I let it down.

A topic on DIY-audio flashed my eyes: WEILIANG Dual X2 AK4399 DAC with LCD

After reading all the interesting posts inside the DIY-AUDIO thread, I ordered the board from e-bay, buying the version II, equipped with the receiver AKM4118. After the kit came, I bought two PCB transformers from a local store, and I powered-up the device: bare naked, this DAC showed an amazing potential, so I go further in upgrading/tuning this kit.

I have used a separate transformer for digital section, and one transformer for the audio section, placed apart in the box. I also used a IEC mother plug containing a line filter.

I decided to exchange all the electrolytic capacitors from the digital section with OS-CON, SEPF Series from Panasonic. I replaced all 10 uF decoupling caps around AK4399/4118 with 22uF/35 Volt, the rest of them being replaced with the same values. For the PSU of the audio section, I exchange the originals on board ELNA 220 uF with real ELNA Silmic II 470uF, and the decoupling electrolytic caps from OPAMP power rails (originals BC47uF) to Panasonic FC 150uF/25 volt.

I had replaced the original rectifier diodes with Phillips BYV27/200 only for audio section (makes a big difference). I replaced also the Zenner diodes, small and big transistors for the stabilized PSU that gives +/- 12 Volt for the output buffer, as the Zenner diode for the positive rail was burned from factory!!! (I was having +18/-13.8 Volt to feed the operational IC, which was not so happy with it ) . For the sake of safety, I replaced the caps near the burned Zenner diode with remaining os-cons, because polymer caps behave better at high temperatures (they live near the heatsink series regulator transistor).

Regarding the output style: in my experience, the transformer method could give the best results, if one could afford a really good one, or wounded on request ;]

The rest of the methods depends on the sound you are looking for, hmmm. I tried a E188CC Cathode follower valve output, carefully simulated /designed by one of my friends, but the sound obtained does not fit my system as I wanted; the sound-stage was incredible, with a perfect tone, but the dynamics and the signal level were not suited for my class A single ended valve powered system 🙁 With the topology I tried, the output signal coming from the DAC chip was reduced by a 0.75 factor, being too low for exciting my input triode stage of the final amplifier.

Before to get crazy, I decided to give a chance to the OPAMP at the exit stage.
The results of OPAM rolling comes like this :

– BB OPA134 PA – bold sound, highs a little veiled, decent dynamics, dark character, sound is polite and far ; good for pop/rock medium quality recordings/ original is quite expensive

– LT1028 – cannot comment too much the sound, as this beast oscillates; no matter what measures I take, the capsule gets warm in few minutes, and the medium/ highs suffers in quality. The bass was deep, controlled, textured and clean, while lower mediums were thick, natural, with a good tone. But the medium highs and highs sounds weird, suffering from a kind of compression (does not sound ‘free’ so to speak). This chip makes a cult for some, but did not worked well on my setup.

– AD797 (soic capsule on a brow-dog adapter) Sound was natural but, Jee, dry. Really dry and soul-less, with lots of details on medium and highs.. Everything was in the place, but the bass was weak. If one likes the laser-cut details, this could the the contender. For my taste, it’s a bit. Some opera voices sound interesting with this chip.

– AD8065A (soic on DIP8 adapter) This one came close to perfection! Good wide and deep stage, detailed and extended highs, decent lower register, thick and natural mediums. It does not exceeds in dynamics like some other chips. For my DAC/system, this chip came on second place with brilliance. Nice textured highs, maybe a bit too much ( high end sound pattern, if we joke a little), with nice micro-dynamics on high-hats.

– LT1357/ LT1360 Well, these are my favorites, depending on the budget (between the two of them). Compared to AD8065A, they sound a bit clear and with better dynamics. They don’t lack almost nothing, the sound they produce being fluent, powerful and agile, being so natural at the same time. The depth of field is a bit better than AD8065, as well as the lower bass: even at the lower volume settings, one could hear the lower pipe-organ registers; the double bass, cello or electric bass notes are fast, accurate and textured, without being muddy, heavy or bloated ( typical BB sound, somehow). Sound-stage is wide enough, without being artificially large. Lt1357 is cheaper, while LT1360 has better bandwidth and better speed, having a different topology (single stage). Also, there is another chip from 1360 family, called LT1362, but the difference is not big among the characteristics (band-width and speed are 10 percent better than LT1360).

Being a Valve addict, I never believed that I could say such beautiful words about a chip, but it really deserves :)) It makes dual AK4399 DAC to shine.

I ordered a beautiful crafted chassis from audiophonics (France), specially made for this project, to put the things together. Attention: the chassis is too small (depth wise) for the main board (by 2.5 mm !!!). The part of the board with the rectifier diodes hits the U-processor /LCD assembly.
So, despite the chassis is perfectly crafted, to get the things inside the box is a pain in the a*** :((

One simple solution is to replace all the PCB connectors with chassis ones (except the TOSLINK module), and put the board oriented at 90 degrees.

Regarding the connectors: the coaxial PCB connector su**s. Before to change it, my DAC sounded better on optical SPDIF entrance that on coaxial one! Not to mention that the DAC’s receiver synchronizes with the digital input signal, when using TOSLINK INPUT, until 24/196 served with a short cheap optical cable
sent by a GUSTARD-U12-XMOS-USB-to-Spdif-Converter !

I changed the flimsy PCB coaxial connector with an isolated 75 Ohm BNC panel connector; and I did that to the CD player too. I build a 1.90 meters SPDIF digital cable by using CANARE L-4.5CHWS – 75 ohm Coaxial Cable Low Loss Coax Formed PE Insulation, terminated with BNC fathers at both end. Now, the sound of coaxial approaches that of Toslink :))

I also removed the PCB RCA plugs and use good solid ones for panel mount. The PCB ones were kind of decent, but once the board was fitted to the chassis, there was a little clearance between the RCA cylinder ad the surrounding hole, that not allowed the cable RCA connector to properly connect (until the end) with the mother socket. The two RCA connectors (mother/father) overlap only for 2.5 mm instead of 5mm.

CONCLUSION:

The Weiliang kit DUAL AK4399 – Ak4118 dac has an incredible value if one wants to play PCM files only from a separate CD transport or to feed it from Computer Audio
devices via SPDIF (Toslink or Coaxial). It lacks I2S, indeed, and tha’s sad for those who needed, I don’t needed yet.

In my case, by

-using separate power supply transformers (digital/audio)

– a line filter was installed between IEC plug and supply transformers

-soft recovery diodes for audio PSU

– exchanging all caps in digital section

– all fixed capacitors around the OPAMP output buffer (those caps form the output filter were changed to Siver Mica type from Farnell )

– decoupling caps from the +/_ rails of the OPAMP changed to Panasonic FC

– coaxial PCB connector replaced with BNC chassis one

– the output stage operational amplifier changed with high speed, single stage bipolar chip LT1360

With all this modifications, I find the dual AKM 4399 (by Weiliang) stellar as price/performance ratio.

Also, I want to thank to the guys of diyaudio dedicated thread: terranigma, valerii, nino-simona and others, who share all their results with the rest of us.