I am testing a 10-bit segmented current steering dac, 4-binary, 6-thermometer.  I have tested the Static parameters and get an INL of 0.5LSB and DNL 0.4LSB with no filter.  Now I am testing the Dynamic parameters, and it seems pretty bad.  The initial specs were for the dac to sample at 800Msps up to about 1/2 the nyquist.  At 800Msps with a 10Mhz input sine, I can only an SFDR of 33dB. At higher input frequencies the SFDR goes down to below 10dB.  I also checked the output on an oscillascope and starting from about 50Mhz the output is not a good sine wave, and by 200 it resembles a pulse.  10Mhz sine has quite a bit of glitch.  Below is the screen shot.

Could it be a clock problem?  Jitter, crosstalk?
Or a design flaw.

Equipment:
Agilent PPG
Osc
voltage generator