Advice on a simple DAC board

[juretrn]

Hi folks, my first time posting here!
I need some advice on a circuit that's misbehaving.
I am trying to make a DDS function generator on a Terasic DE0-Nano board. The digital stuff I believe is acting well.
I am generating a sine wave, triangular wave and a square wave. The square wave is driven by a SN74LVC2G04, and is completely separate from the DAC.
I've limited the frequencies to max. of 2 MHz.
The board itself is just a single layer, milled PCB with some of the bigger components on the other side.
Problems lie in my own DAC board (that's using the AD9760ARZ). I needed to use an "I-V" converter to create my output signal, and I used one directly from the AD9760 datasheet, and stuck a Sallen-Key filter at the end.


However, above 1.5 MHz - for the sine wave - strange things start to happen. I start getting really bad distortions on my wave.

For the triangular wave, things look bad even earlier, at about 1 MHz.


Note: the connection between R3 and pin 2 of AD8041 has been bodged, because apparently reading datasheets is hard.
Any ideas what might be going on?

[awallin]

That R10/R11 DC-biasing of U2 looks suspicious, especially if the DAC output is also DC-coupled?
If you don't need very low frequencies, maybe try AC-coupling after R8 and R9 with 100n or 1u caps?

also you haven't drawn decoupling capacitors on DVDD, close to the op-amp power pins 10u and 100n or similar...

[capt bullshot]

Did you follow the layout consideration of the AD8041 datasheet (page 14), same for the ADA4891 (page 21)?
These are quite high speed amplifiers, power supply decoupling and low stray capacitance layout is important here

[juretrn]

The output of the DAC is not really DC coupled:

I didn't use the Copt.
What looks suspicious about the R10/R11 biasing?

DVDD is decoupled, but not close to the AD8041:


As for the layout of the board, I wouldn't really brag. Besides, it's a single layer and milled PCB, and it's a one-off. So no ground planes and power planes.
...
After reading the AD8041 datasheet layout spec, the layout could be one of the issues.
I shoud also mention the fact the distortion appears at the junction between R3 and R4, but not earlier, that is between R4 and R9. There is no distortion present on the IOUTB side, but I guess that's mostly because of the fact that one doesn't have a feedback loop?

[capt bullshot]

The circuit itself looks OK for me.
You'll need the decoupling caps very close to the supply pins of the Op-Amp (no headers or jumpers in between), you should have a ground plane and you probably want low stray capacitance on the "-" inputs of the amplifiers.
I guess your layout would work if you replace the amplifiers with slower ones (<= 20MHz GBW)

Use a square wave excitation (at the output of the DAC) to watch the response of the first amplifier for clean steps (no ringing) and smooth edges at the output of the filter.

[juretrn]

I guess the first priority will then be decoupling DVDD as close to the chip as possible, and trying to minimize stray C's.

[capt bullshot]

Still guessing:
to me, your waveform look most likely a decoupling issue. If this problem goes away with better decoupling, look at the square wave signal: excessive ringing may be a result of too large stray capacitance.
Connecting the AD4891 (filter) output directly to a coax cable or other kinds of reactive load may also cause instabilities and ringing. Better fit some kind of isolation resistor (>=10 Ohm) or back termination resistor (50 Ohm) between the Op-Amp output and the connector.

[juretrn]

OK, I've ran a square wave over the circuit, I get this on the output - before the filter.
EDIT: will it suffice for now if I just bodge in a few caps, just to see if it has any effect?

[capt bullshot]

The waveform really looks interesting. I still guess missing decoupling, stray C doesn't appear to be an issue.
Yes, bodging a cap directly to the IC may help, it's worth trying.
 

[juretrn]

Ha, it actually worked.
Used a 100nF cap to decouple the 3.3V power; seems to do the trick.

These are the unfiltered signals now:

[capt bullshot]

Looks good, not perfect. There's still some noise or spiking or whatever to see, but you are on the right way. I cannot tell what the reason might be, I'd want to see your test setup now. Maybe still not enough decoupling, I doubt more caps will help, I think you'll need the ground plane to get it better.
Stray capacitance isn't a issue here, you'd see ringing on the square wave if this was an issue.

[juretrn]

I talked to my prof (this is my master's thesis project  ), and I've been promised to get a 2-sided board in a while. My guess is also that getting a board that's not as beat up as this one will make things much better.