AD9780 / AD9781 / AD9783 Specs: Current noise, Settling Time, Rise Time and Glitch Energy

Hi Ben3 ,

Thank you for your interest in the AD978x family.

• For Noise Spectral Density, there are data under Table 3, Figure 25-28, Figure 43, Figure 49 and all measured at FS = 20mA.
• For Glitch Impulse, typically negligible because communication transmit chains use reconstruction filters that strongly suppress out‑of‑band spurs (2 × fDAC) far from the desired signal.

Thanks for the quick response. I'm mostly interested in the broadband noise of the DAC in the DC-10MHz regime. I also want a fast response time (low latency as well) so I don't want to put too aggressive an output filter. So I'd like to understand the glitch size to inform the filter design. After the filter, I'd like the glitch to buried in the broadband noise of the DAC.

Thanks for pointing me to the NSD Table and figures. I'm confused about the units in table 3. It reports the NSD in dBc, but I'd expect a density with per Hz units. The plots in Fig 24 & 27 are show dBm/Hz which makes more sense. Also, I'd like to convert the dBm reported there into current -- what is the load impedance used for that data?

Thanks,

Hello,

Perhaps you may want to consider the AD9114-9117 family if your application is more time domain orientated with focus on settling-time, rise/fall time, ext.  Suggest that you operate at 20 mA to achieve the lowest current noise density specification.  Time domain specs are typically measured with each current output terminated with 50 ohms on the PCB (prior to connector) with a co-ax cable feeding both outputs into a wide-band, high-speed digital o'scope capable of measuring settling time to 0.1 % to within 2 nsec (i.e 5x better than device under test).  Ideally the 50 ohm load on PCB is as close as possible to the output pin using 0402 resistor type with 50 ohm trace leading to SMA connector since accurate settling time measurement needs the best possible 50 ohm environment.

Anyways.............just a thought if you application seeks best possible time domain performance.

For the AD9783 the flat part of the PSD plot (at higher frequencies) is at -156dBm/Hz. I'm assuming that's into two 50 ohm resistors that terminate the differential outputs. So that's 100Ohms total impedance. So -156dBm is 50pA rms. So that would be 50pA/rtHz which is worse than the AD9117. Is that calculation correct? 

With regard to the AD9117, its NSD is not very flat, which makes it hard to trust the NSD -> current noise calculation since the 37pA/rtHz spec implies flat frequency noise.  The AD9117 is -142dBm/Hz at 25MHz on its NSD plot while the AD9783 is -164dBm/Hz. The AD9117 is running at lower sample rate and has lower noise at lower frequencies, but even at 5 MHz it is -154dBm/Hz, so that says that the AD9117 has substantially higher broadband noise than the AD9783? But that goes against the 37pA/rtHz vs 50pA/rtHz calculation from above. Do you know how to reconcile this and compare the broadband noise on these to DACs?

Thanks,

The wideband noise specification is likely performed using the evaluation board with the double balun configuration with an IOUTFS=20 mA at FCLK=500MSPS as shown below. Also included is the frequency response (normal mode)that shows measured output power in dBm of the CW while noting the reduction in amplitude out to 240 MHz is most attributed to the DAC's inherent Sinc response.

A full-scale digital sine-wave is swept over frequency with the wideband NSD measured in the "flat region".  The noise measured in this area includes effects of wideband phase (PM) noise,  thermal  noise (AM noise from current source array and bias generator) as well as quantization noise.   The wideband PM noise (i.e. jitter) causes the noise floor to raise with frequency. 

Since the AD9783 is 16-bits and operating at 500 MSPS, the quantization noise is around -182 dBFS/Hz (=6.02*N+1.76+10*log10(250e6))  thus are well below the PM and AM wideband noise.  If the clock rate is reduced to 125 MSPS (factor of 4), the noise floor would increase by 6 dB for -176 dBFS/Hz.


The AD9117 is 14-bit resolution (vs 16-bit)  so its quantization noise spectral density is - 162 dBFS/Hz at clock rate of 125 MSPS.  A plot comparing the different resolution devices over frequency is shown below.  The quantization noise effects vs device resolution are quite evident.

From a broadband noise comparison which removes the effects of clock phase noise (jitter), one can compare the noise measurements at very low frequencies (<5 MHz).   Since the AD9117 is also limited by quantization noise at this low frequencies, its difficult to ascertain what its thermal AM noise contribution is by just looking at datasheet characterization curve but I would surmise that it would be similar to the AD9783 since both DAC cores are operating at same 20 mA full-scale.



  




-----------------------------AD9717-----------------------------------------