AD9910 output not activated when using "set_frequency"

steine01

I use an Urukul board to generate RF with an AD9910 at 80 MHz and found an interesting behaviour. I would like to know, if it is by design or not intended.
When I use the following code to start the AD9910 after disconnecting ARTIQ from the power supply, there is no output measurable at any frequency even though the LED panel is green:

def build(self):
    self.setattr_device("core")
    self.dev1 = self.get_device("urukul1_cpld")
    self.urk1_ch0 = self.get_device("urukul1_ch0")

@kernel
def run(self):
    self.dev1.init()
    self.dev1.cfg_switches(0b1111)
    self.urk1_ch0.init()
    delay(1000*us)
    self.dev1.get_att_mu()
    delay(1000*us)
    self.urk1_ch0.set_att(0*dB)
    delay(100*us)
    self.urk1_ch0.set_frequency(80*MHz)
    delay(100*us)
    self.urk1_ch0.set_amplitude(1.0)
    delay(100*us)
    self.core.wait_until_mu(now_mu())

If I use the set function instead of set_frequency, I measure the expected output at 80 MHz. Also afterwards I can change the frequency by using set_frequency:

def build(self):
    self.setattr_device("core")
    self.dev1 = self.get_device("urukul1_cpld")
    self.urk1_ch0 = self.get_device("urukul1_ch0")

@kernel
def run(self):
    self.dev1.init()
    self.dev1.cfg_switches(0b1111)
    self.urk1_ch0.init()
    delay(1000*us)
    self.dev1.get_att_mu()
    delay(1000*us)
    self.urk1_ch0.set_att(0*dB)
    delay(100*us)
    self.urk1_ch0.set(80*MHz)
    delay(100*us)
    self.urk1_ch0.set_amplitude(1.0)
    delay(100*us)
    self.core.wait_until_mu(now_mu())

I checked the documentation, but no such behaviour is mentioned in set_frequency or set_ftw. When looking at the source code, there is a comment in set_mu (which is called from set), that mentions the shared IO update pin is pulsed to activate the data. No such comment is found in set_ftw and also I do not see any code, that would pulse the IO pin.

Is this behaviour by design? If yes, why is it not mentioned in the documentation itself?

rjo

That's by design. Documentation does typically not describe things that don't happen.
Also, the "FTW register" (see the docs for set_frequency(), set_ftw()) and the "profile register" (see set_mu(), set()) have very different functionality. Even with an IO update pulse, you won't get output from in the first snippet.

steine01

I checked the documentation of the AD9910 and now see, that the FTW register as well as the ASF register and the POW register (set by the set_x functions. x not mu) are different from the Profile 0 data. But what do they do? The only one I understood from the documentation is ASF, because it is involved in ramp generation.
And I see, that it is not a good idea to document, that something does not happen, but from the perspective of the Artiq documentation, the functions look very similiar to me. I assumed, that the set_x functions are just the set function split up into its components, so that one would not have to specify every parameter when changing one of them.

rjo

The ARTIQ documentation will not document the AD9910. And it doesn't and can't abstract over its functionality and expose a different one since that would limit the features and restrict the use cases. Documentation can certainly be improved in some areas. But there is no market and insufficient demand for it compared to "new features". If you want this to change, we can talk about options.