Now that FFE Rev 1 was built, I had to test it to see if it met the lofty goals I set for it in my front end specs - 350MHz bandwidth, 1MΩ//15pF input impedance, 4 selectable attenuation ranges (0x,10x,100x,1000x), AC/DC coupling and switchable 50Ω termination.
*Spoiler Alert* It did not.
And that's fine, it was my first revision and that was pretty much to be expected. This log details how I tested it and the things I learned along the way!
After annexing a corner of the living room (with apologies to my housemates), I set up a power supply, oscilloscope, and RF signal generator to start frequency response testing on the FFE board. I swept frequencies with the signal generator and recorded the average amplitude off of the oscilloscope. This got tedious very quickly, but I eventually automated it with a USB GPIB controller and PyVISA. GPIB is just about as old as dirt, but if you happen to have equally as old equipment, it's an absolute lifesaver.
Here's the frequency response of FFE Rev 1 with the attenuators off, AC coupling on, and 50Ω termination enabled. The blue line is the frequency response of the oscilloscope doing the measurements to provide some sort of baseline for relative comparison (my equipment wasn't calibrated, so technically relative measurements are all they're capable of). Although the -3dB bandwidth was over 400MHz, there was massive (3db!) peaking past 200MHz. This wouldn't work for a front end. On top of that, the output oscillated from rail to rail when any of the the attenuators were enabled. The only good news was that it worked with regular scope probes.
The reason for the peaking eluded me at the time, though now I believe it was due to either the opamp's marginal stability at a gain of ten or the parasitic inductance and capacitance in the layout, perhaps even both. When filtered by the PGA's adjustable bandwidth limit (I would set it to 200MHz), I thought that the peaking would be acceptable enough to test the rest of the system until I made a new revision. As for the attenuators, switching them in massively increased the capacitance at the opamp input, which was probably the cause of the oscillation. I'd have to avoid using them whenever testing the system, which limited me to a select few voltage ranges.
It really bugged me to move on to designing the rest of the system while the FFE was still an absolute mess! Instead of scrapping the design and figuring out the fundamentals behind why it didn't work, I came up with a solution in a single afternoon (this had to be good right?). The idea was to take some of the pressure off of the main opamp, and add another one in front of it to better handle the changing capacitance when the attenuators switch.
I used the same ADA4817-1 opamp that [bunnie & xobs'] used in their oscilloscope probe, alongside the damping circuit (which they DID label as damping 1GHz peaks, so I figured tweaking the values would help me tame the mountainous peak at 340MHz). So in the same afternoon that I thought of this fix, I layed it out and sent it off to fab!
The only thing that's peaking here is the BOM cost with the extra opamp! Kidding aside, it was clear that the FFE was going to be very difficult to do right, and would need a complete overhaul for Rev 3. Until then, I would continue to build the rest of the system so that work on software could start, and then loop back and give the FFE the respect it's due.
Thanks for giving this post a read, and feel free to write a comment if anything was unclear or explained poorly, so I can edit and improve the post to make things clearer!