EH,
Here is a link to the D90 manufacturer with all the descriptions of the D90. How about you look through it and find what you say is there.
TOPPING D90 AK4499 AK4118 Full Balanced DAC Bluetooth 5.0 DSD512 Hi-R
Since you won't find it, stop posting your nonsense and let people discuss the subject of the thread.
Here is a link to the D90 manufacturer with all the descriptions of the D90. How about you look through it and find what you say is there.
TOPPING D90 AK4499 AK4118 Full Balanced DAC Bluetooth 5.0 DSD512 Hi-R
Since you won't find it, stop posting your nonsense and let people discuss the subject of the thread.
You can't route a two layer board with an uninterrupted ground plane unless your design has barely anything on it. If you have any reasonable number of signals or more than a couple supply voltages then it is completely out of the question to make a good two layer design.
There are modern parts (not necessarily audio) that you can't even escape on a standard 4 layer board.
Look at every high performance data converter EVB from TI, Analog Devices / LT, Maxim, etc. Most are > 4 layers, haven't seen a 2 layer one for anything over 12-bit.
Last edited:
Pretty bad at first sight: the ground plane is interrupted by long horizontal traces and there are no vias above and below those traces to short the gap via top level ground fill. In fact there is no top level ground fill anyway above a large part of those traces, because the top layer is crowded with vertical lines. Whether any of this actually matters depends on what signals those traces are for; if both the horizontal and the vertical lines are digital GPIO lines that only switch once per hour, it may not matter at all.
btw AK4490 EVB is 2-layer I suppose. of course it is far from being the best implementation of this DAC
Sure but most audio isn't that complicated and you can use zero ohm links to jump topside traces over one another.
It's worth pointing out that the ak4499 eval might used a 4 layer board but they do not use an uninterrupted ground plane on any of their layers chopping the copper up on purpose in an attempt to control ground currents. Mostly I would expect this to be unnecessary but they do it anyway.
Sure there are parts that necessitate a 4 layer board, or more, and when using them you obviously go that route.
Any high speed trace requires a ground plane beneath it. With that in mind you can see where the problems with that board are.
One ak4499 board I saw cut the ground plane directly in the path of the master clock
to me it looks like inner planes reflect their data sheet recommendations:
"10.2. Grounding and Power Supply Decoupling
To minimize coupling by digital noise, decoupling capacitors should be connected to AVDD, TVDD, DVDD and VDDL1/R1/L2/R2. AVDD and VDDL1/R1/L2/R2 are supplied from analog supply in system, and TVDD and DVDD are supplied from digital supply in system. Power lines of VDDL1/R1/L2/R2 should be distributed separately, from the point with low impedance of regulators or other parts.
When not using LDO (LDOE pin = “L”), TVDD must be powered up before DVDD is powered up or at the same time. AVSS, DVSS, VSSL1/R1 and VSSL2/R2 must be connected to the same analog ground plane. Decoupling capacitors for high frequency should be placed as near as possible to the AK4499."
VDD then is supplied by the famous NJM78M05 discussed above
it appears that power supply wise the biggest difference between AK4499 and previous velvet DACs is for Vref though. 120mA vs 2mA for AK4497
Yes but still, they've chosen to split up the ground planes rather than use one layer as a solid plane. They could quite easily attach the return path for the vref supplies to the solid plane at an advantageous point, rather than slicing it up and providing the return beneath the vref power traces. These are inherently low bandwidth, what with the 10R in series with the main vref caps, they don't need a plane directly beneath them to return to. One wonders if they've tried the alternative or not. Either way it works and I guess that's the important thing.
Of course it's not optimal as it causes an increase in low frequency distortion, but with a large vref cap the increase is pretty insignificant. You can offset the reduction in absolute output by increasing the vref voltage too if necessary.
If the preliminary performance of the SMSL M400 comes to be reality and not just a cherry picked unit then there's good enough reason to insert series resistors. As far as I can see they use them.
If the downsides to using series resistors were more significant then I can see the point in trying to remove them but they are a simple/cheap solution to a problem.
If the preliminary performance of the SMSL M400 comes to be reality and not just a cherry picked unit then there's good enough reason to insert series resistors. As far as I can see they use them.
If the downsides to using series resistors were more significant then I can see the point in trying to remove them but they are a simple/cheap solution to a problem.
Two resistors per channel, actually. It sounds good with that circuitry. Depending on capacitance used and other factors a little bass distortion may sound preferable in some implementations (not in mine
though). The foregoing assumes, of course, the result is not intended for use in an instrumentation DAC.
Last edited:
Besides factors already mentioned by others, putting the I/V opamps off-board would seem of questionable wisdom. Wonder how it sounds...
Last edited:
The extra length will compromise the performance to some degree. An active VI stage assumes zero source impedance and with 130 dB plus dynamic range it takes little resistance/inductance to compromise the performance. He got it all running with some impressive bradboarding skills but merging it all to a single PCB would improve things still.
