Design ideas for Aune M1 DAC card

I'll outline here some thinking in choosing the major building blocks (aka ICs) for this card - any comments welcome as this progresses.

First up the DAC chip will be the TDA1387 initially. I don't know for sure that the output from the ARM/Xilinx card is I2S but I'm going to verify that fairly soon. There's nowhere near enough room for the passive shunt I've adopted previously so the bass performance probably is going to have to suffer. I shall pay considerable attention to the power supply arrangements though in an attempt to make up for the LF lack.

After the DAC, passive I/V will follow and then a filter using the TDK 7mm inductors I've used previously. I've slotted them into the gap between the PCB and the case and there's just enough height available. Since space is at a premium I'll experiment with a 3 inductor design - the stop band attenuation will suffer but probably I'll add a secondary LC filter at the output to make up for that somewhat. The secondary filter will also provide NOS droop correction.

Given that the only choice is for low-valued inductors (because bigger valued ones are so bulky) the signal level at the filter's output will be around 15mV. Thus a gain of 100 stage is called for to bring us to standard CD output level. I'm veering towards using AD603s for this - they come with a DC volume control which will allow volume remoting and permit a smaller and cheaper pot to do volume control. They're also remarkably cheap on Taobao compared to the official price from ADI.

The power supply for these parts is going to be a major determinant of the subjective SQ - from a brief perusal of available options of boost regs I rather like the TPS61220 - mainly because its both efficient and cheap. Perhaps I'll do an emitter follower discrete regulator after it to keep the HF hash down to the absolute minimum.

So far then the power budget is the boost reg (90% efficiency) a TDA1387 (30mW) and two AD603s (250mW, ouch!). One or two 7660s I'll use to create the -ve rail.

My first simulated stab at a suitable filter is attached. Given the constraint on space (hence number of inductors) I've adopted a quasi-elliptic design. The addition of stop-band zeros (the caps shunting the series inductors) gives a steeper slope than a pure Chebyshev response and I've not so far been able to find a downside The 75uH coils will probably be hand-wound ferrite beads to save some space.

Update1 - starting the first build, I measured some inductors (around a dozen) and they all but one turned out below 1mH, around 950uH was the most common value, so I've rejigged the filter component values to suit. Plot attached.

Attached Thumbnails