I am using AD9547 to generate two clock signals -> a 325 MHz LVDS signal on out0 pin and a 81 MHz CMOS signal on out1 pin. I see that at -40 degrees, the out0 pin has amplitude surges, but the out1 is stable. Is this normal?
I moved your question to the Clock and Timing Community.
Someone here will be able to help you.
On 12 January 2017 at 02:41, mcee <email@example.com>
While there is some expected variation of most parameters over temperature, without seeing what this looks like it is hard to determine if this is normal behavior or not. The levels should still be within the spec of LVDS, as the output driver of the AD9547 is LVDS compliant.
The signal in light green is the 81 MHz signal - CMOS
The signal in cyan is the 325 MHz signal - LVDS with "gaps" in between
The numbers in the spec table are guaranteed over process, temp, and voltage. In the case of amplitude, they are measured with the output driver static (not toggling), and yes, they are compliant with the LVDS spec. Refer to Figure 13 in the datasheet in order to understand the expected loss in amplitude as the frequency goes up.
If you are experiencing amplitude variation at low temperature, I recommend that you change the output frequency to something much lower (like 10 MHz or hold the driver static) in order to at least remove any frequency dependent factors.
I have referred to Fig 13. My question is, if there should be say 0.37 V @ 325 MHz, would I not expect this amplitude as a constant, and not with breaks or gaps as seen in the image above?
In your picture, the cyan trace appears to be very low amplitude and on top of a green trace. If this is the correct trace, the 325 MHz clock appears to be sub-sampled by the scope, or at a minimum, the screen resolution is limiting the ability to accurately display the waveform. I recommend zooming in on both axes (so that we can see about 5 periods of the clock), triggering on the channel, and setting infinite persistence. Also, in order to have an accurate waveform image, you'll need a scope with a 1 GHz analog BW.
The gaps appear to be an artifact of the scope sampling.
i have the following image -
where the signal in cyan can be seen better (325 MHz output)
The signal in light green is the reference input clock of 13 MHz just before pin 44.
It still appears to me that there is some sub-sampling on the scope. Can you set the zoom on the scope to 2.5ns/div? You may need to reduce the total number of samples, as it appears you are gathering data at 20us/div. Since the scope uses the lowest reference rate as the trigger, sub-sampling is a concern when sampling such disparate frequencies.
I do not understand how there can be sub-sampling. The scope samples at 2.5GSps as shown in the image, and we are sampling a 325 MHz signal, so roughly 10 sample points per period. This matches sample points seen in the image as well...
Before going further, have you designed the analog reconstruction filter on the AD9547 with a >325 MHz cutoff frequency? The default one on the evaluation board is 240 MHz.
The waveform anomalies that you're seeing are something that we commonly see in our lab, and our experience has been that it's been related to how the scope samples the waveform. We would like to be sure that we don't have the same issue here before investigating any potential issue with the AD9547 LVDS driver.
Also, is this 325 MHz clock signal causing any issues in your system, or is the distorted waveform your primary concern?
Would it be too much trouble to ask if you could make the following plot just to be sure that we're not seeing a scope issue?
Turn off all other channels.
Trigger the scope of the 325 MHz waveform with the highest possible sampling rate.
Configure the timebase and vertical scale with settings similar to Figure 23 in the datasheet, except may catch three cycles instead of two.
Enable infinite persistence.
Retrieving data ...