I want to convert a sine wave produced from AD9850 (opens *.pdf) into square wave with its inbuilt comparator. AD9850 has a fast comparator with 3ns rise/fall time and propagation delay= 5.5ns.

without accounting the propagation delay , it should work up to 1/(2x3ns)= 83MHz. If I add propagation delay, it will be 1/(2x3ns + 5.5ns)= 43.48 MHz. if I account the same time for settling and flat part of the square wave and some error tolerance, it should at least tolerate around 20MHz.

But it is not what really happens. the output wave is not a square wave higher than 5MHz ( it is more like a noise). The input signal of the comparator is well low pass filtered and is a clean sine wave with around 2v pk-pk.

Is there anything wrong in my calculations?

EDITION: I am using this module with a reliable ground plane:

  • \$\begingroup\$ Is your o-scope fast enough? I'd say 3ns rise, 3ns fall and 6ns top and bottom should give you a reasonable 80MHz signal. \$\endgroup\$
    – Andy aka
    Commented Oct 15, 2013 at 18:01
  • \$\begingroup\$ A schematic indicating what you have implemented and how you are measuring this would be helpful. \$\endgroup\$ Commented Oct 15, 2013 at 18:32
  • \$\begingroup\$ @Andyaka yes it is a 200MHz one with good quality. \$\endgroup\$
    – Aug
    Commented Oct 15, 2013 at 18:48
  • \$\begingroup\$ @Aug - does the onboard comparator require localized de-coupling capacitors? Are there any decoupling capacitors you have forgotten about. What about the signal size going into the comparator - bigger the better (within the power rails of course)! \$\endgroup\$
    – Andy aka
    Commented Oct 15, 2013 at 18:59
  • \$\begingroup\$ Spec indicates ground plane , decoupling cap close to 5V and ground and 15pF load. Use barrel of probe & tip with sleeve & gnd clip removed between Qout and Gnd with 10x calibrated probe. \$\endgroup\$
    – user30080
    Commented Oct 15, 2013 at 19:10

1 Answer 1


It appears there's a bit of circuitry between the output of the dac and the input of the comparator on that board. Connect a four channel scope to the outputs and inputs directly on the chip and view the waveforms present. I suspect you'll find your answer.

You'll notice in the datasheet that the comparator jitter graph is performed at 40MHz, so it's capable of running at that speed. Chances are good you have low signal quality at VINP, an improperly set VINN threshold, or both. As the frequency goes up, the circuit impedance reduces the signal power, and it's having a hard time seeing it.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.