The goals for this project were two fold. Having been inspired by Tony's (Bemused) dual mono set up when I heard it, I wanted one too. The second was to get up to speed with SMD, after all its the future innit, and to overcome my complete inexperience with this mode of working.
The beauty of a board like this is that it allows you to experiment at a component level at a cost only marginally more than the cost of DAC chip. I paid around £125 for two boards and the retail price of just the ES9018's would be £110 in the UK. The supplier must make a living out of these so he must have been able to source the chips at a favorable price.
To help with component selection, I referred to the Application Note published by ESS which has some very useful information.
Basically they recommend all resistors be metal film or thin film. As this was going to be a SMD project, I opted for 1206 1% Melf resistors by Vishay.
They also recommend that X7R ceramic capacitors are not used in the signal path because of their inherent non-linearity and according to them result in substantial increase in THD and loss of DNR. I used COG/NPO's in the signal path. Diyers are not immune to a bit of "fooism" as there are some people who recommend that even for decoupling/bypass duties one should use COG/NPO's as X7R's smear the sound. Having tried both in this role, to my ears there was no audible difference between the two types. The final boards have a mixture with anything below 1uF being COG/NPO and the 1 and 10uF being X7R's.
The board has locations for ferrite beads in the digital supply lines (and even return for the 1.2V). I have not used these before and have not experimented with their effectiveness or audibility but have installed them. After all as DIYers we are seriously hampered by the lack of measurement gear. It doesn't have to be audible to be technically better.
There is a facility on the board to disable the onboard clock and provide an external clock via pins or ufl connectors or even power the onboard clock from a seperate external 3.3V supply. Quite a few people have reported that Epson SAW clocks are audibly superior to the Crystek ones common fare in most ES9018 implementations. The Crystek clocks have technically some of the best phase noise and jitter figures available but the SAW clocks use a completely different method of generating the clock. This apparently makes an audible difference even though its phase noise/jitter performance is inferior. I have installed a 75MHZ clock readily available from RS with a view to experimenting further with Crystek CCD950 or 575 if I can find a source for them here in the UK. Any clues anybody?
So what about SMD? Well if I can do it, anybody can as I am the worst solderer in the world. Absolutely essential are a good pair of tweezers, a magnifier, 0.3-0.4mm solder wire and flux. Anything missing from this list and it is likely to end up in tears. The benefit of some guidance (as I had from Bemused) goes a long way too. It is definitely worth watching somebody else do it as you can copy their techniques. In fact after doing these two boards (which also have a couple of positions of 0603's) I prefer SMD. I found reworking through hole incredibly difficult and was forever lifting tracks and burning boards. SMD is a lot easier to rework and less danger of lifting tracks in the flimsy boards we have to contend with.
Any views or comments on components or clocks gratefully received.
To come, my experiments of different power supply configurations, different op-amp configurations for IV including a CFB and discrete IV versus Opamps.