Secret project (Svart AD/DA converters)

well i think the least you could do is quit your day job and get crackin'


......


8)

Just kidding ... of course..

I am really interested in this...

cheers

Wiz

I don't know what TI was thinking, but If you look into the pins of the adc like you were the signal, the left input is on the right and the right is on the left.

Wtf?

Either i can swap the inputs so that the physical connectors are left on right, and right on left, or i can cross the traces on the pcb, which has its own issues.

I'll see what i can do.

But on a side note, i hooked up the demo board from TI tonight and shot it out with my sound card inputs.

Let's just say that the difference wasnt subtle in the least.

I'm as excited as a little school girl!

when life hands you lemons, make ADDA converters, everyone knows that! 8)

nice, your not kidding around!

Its very cool dude.... very cool.

I will say it a third time

very cool


cheers

Wiz

Oh, well,
doing a 312-style pre is simple.
svart's new baby needs knowledge about digital and analog electronics, at least a bit of programming of microcontrollers, board layout tricks for mixed analog and digital signals etc. pp.....
I wouldn't call that particularly simple....

Low speed digital like this isn't too bad.  I follow the Dr. Howard Johnson way of doing mixed signal design.  I don't separate digital and analog grounds, I actually add them together and strengthen the ground system.

Also of importance is using physics to our advantage.  SMD devices have properties that lend themselves to fixing issues that used to plague board designs.

One such issue is transient response and ground bounce.

Digital circuits don't really care which source is + and which is ground.  You could essentially power something with 0V and -3V3 just as well as you could 3V3 and 0V.  As long as the difference is 3V3 and you have no I/O issues, then it's doable.

Anyway, as the IC attempts to direct current to an output, that current must reference to ground, but as the IC switches fast and hard, the ground potential will actually start to bounce up and down.  0V will be -0.5V or +1V or whatever it happens to be for very short periods of time.

This causes the IC to not only have issues with input and output signals, but it causes ground to have currents that can cause all kinds of issues with other chips, most notably in audio, digital hash.  In this case, ground separation doesn't matter because you only confine the issue, rather than solving it.  Having much more robust grounding gets away from band-aiding the problem, however it requires that the grounding be VERY low impedance.  Full ground planes are the norm here.

Decoupling and stiff grounding is the answer to our problem, but..

With leaded components, the leads themselves act as low value inductors.  As you know, inductors impede high frequencies, so they will slow down fast decoupling currents from capacitors.  The higher the operating frequency of the digital part, the worse the effect, to the point where leaded parts are actually useless to use for decoupling no matter what the value!

So SMD parts technically have no leads, so they have less parasitic inductance when use in higher speed devices, to a point.  Even then, the bonding between the solder cap and the dielectric material can have inductance at REALLY high speeds.  What Dr. Johnson has found is that there is correlation between capacitance value and physical package size.

Essentially it boils down to using the largest capacitance value possible with the smallest package size, as close to the power pins as possible.  Each power pin gets it's own decoupling capacitor as well.  You must use NO vias on the board between the power pin and the capacitor, as vias are essentially parasitic entities on their own.

This means that the optimum decoupling capacitor is 1uF in 0402 SMD package(currently, but I'm sure that will become 0201 in the near future..).  This is what I use into the 4-5GHz range daily, and it'll work right on down to DC.

But I digress..

So I'll be using a (mostly) solid ground plane, very thick power traces and lots of 1uF 0402 caps.  That should be sufficient.  I'm also waffling on the idea of using a solid inner ground plane layer, but we'll see how much of the bottom layer of a 2 layer board I use before going up to 4 layers.  4 layers costs a bit more..

That's step two.  The plan is to release the A/D first because that's what most people seem to want, and I've been told many times that they want it NOW!  Unfortunately it's just me doing this, and I only have so much time to do it in.

But yeah, the plan is to have an installable card that will allow expansion.

For example, there will be an expansion port that has unregulated power and I2C communication bus so that another card can be hooked up via a cable, or hard connector.  The DAC card will also allow stand alone functions or slave functionality.

I'm also kinda developing this on the fly, so it's evolving as I go.  

^ This!
The ADA4898-2 made a clear difference in the DAC, compared to the LM4562 in the standard Ross Martin, i heard both side by side. I really thought about asking RM if he could build it into the ADC, too, so it's on par with the Superbeast DAC!
Simply excellent audio opamp. Clearly preferred.