Shuttler analog bandwidth

[jmatyas]

According to wiki, Shuttler's analog bandwidth is supposed to be around ~15 MHz, however after some testing on Shuttler v1.0 it seems to be closer to 6 MHz :

To achieve those results:

• a sinusoidal signal with a given frequency was generated on DAC channel
• peak-peak amplitude on one of the DAC channel's analog output (after AFE) was measured
• signal frequency was changed and the measurement was repeated

[kaolpr]

@dhslichter @adbr8389 @gkasprow How do you find current bandwidth? Do we need to fix AFE to follow statement in Wiki (~15 MHz)?

[dhslichter]

@kaolpr what is the gain in the AFE currently? What is the bandwidth for signals measured before the AFE? There are several places that could be causing this issue with the bandwidth.

[kaolpr]

AFAIK this was measured before AFE. Tomorrow I'll ask @jmatyas to provide more details.

[kaolpr]

But the main question is - was this intended? The simulation @gkasprow provided in #28 seems to confirm that DAC_AFE (the one on FMC) was designed for ~6 MHz.

[dhslichter]

the simulations in #28 appear to be with an ADA4896? The schematic calls for LM6172 so I am confused. EDIT: I see now that you are talking about the "AFE" on the Shuttler card itself, and not the remote AFE.

[dhslichter]

@kaolpr looks like this is just an effect of C40/C41 giving a lowpass rolloff. We should change those capacitance values to 33 pF from 100 pF and will get back the ~15 MHz bandwidth. @gkasprow was there a reason these capacitors needed to be 100 pF?

For our purposes we would rather have the output bandwidth of the system limited by user-implemented passive filters placed after the remote AFE outputs, and have the FMC and remote AFE boards keep as much of the available bandwidth as possible. The FMC DAC AFEs should have a 3dB bandwidth of ~30 MHz or higher per ADA4896 datasheet, with no filter capacitor in the feedback loop (confirmed with simulations). Cutting this down to ~15 MHz with a small filter capacitor (22 or 33 pF) in C40/C41 is acceptable, but let's not go to stronger filtering than this on the FMC card.

[kaolpr]

We did a test with 33 pF. Now bandwidth is around 12 MHz. The measurement was done on AOUT_N net (after R36).

We'll investigate further next week.

[adbr8389]

What is the IC used for the differential amp after the ADA4896's? If it has some smallish capacitive input, it could explain the discrepancy between Daniel's estimation of a 33 pF cap for a 15 MHz bandwidth.

[dhslichter]

The 33 pF = 15 MHz was only a coarse estimate, I think the measured data seem reasonable. Glad to see that this increased the bandwidth @kaolpr. Next question is whether we should increase further (e.g. 10 pF, or just DNP these capacitors). I would be in favor of testing with capacitors removed and ensuring the op amp outputs are stable (they should be, since we are at G=5). This gives maximum bandwidth out of the FMC card. One can always use following components to reduce the bandwidth as desired for applications, but this would make the FMC card maximally useful for different tasks.