I do have some 47-Ohm 5%'s I can try on there next. The 39s I had are only 5% tolerance, so this isn't exact, but It looks like it's converging. A little more algebra shows that the output impedance is now 41.6 Ohms, so I should use a 47-ohm resistor instead. The output swing went from 2.0V to 1.8V into a 50-ohm termination. I swapped those out for the 27's I had originally to try to make a better match to 50-Ohms. I didn't have any 43-Ohm 1206 resistors, but I had some 39's. I'd probably learn a whole bunch of stuff. ECL is a little bit of a pain to work with, but it might be interesting to make a version of this board that would go that high. While I'm thinking about next spins, there are a bunch of ECL flip-flops that could easily generate the I/Q outputs up to the maximum 1.5 GHz output of the tracking generator. The easiest way to do this would be to bring 12V onto the board from somewhere, but it's kind of a big change.
They need 5V at 50 mA, which has to be regulated with a resistor from a higher supply (at least a few volts higher). They're almost perfectly suited to this application: DC - 3 GHz, with 8dB nominal gain and a +12 dBm P1dB. I also have a bunch of MSA0486 MMICs I could use for a preamp stage on a next PCB spin. I'm not even sure I need the diodes with the 560-ohm termination there. Now that I think about it, removing the diodes (a single package) is easier than swapping the transformer, so I'll try that first.
They add a decent amount of capacitance, which could be an issue. Transformers on this core have worked very well for me in the past, but that has all been in the HF range or 50 MHz. I think the first thing to try is a different input transformer. The magnitude drops off precipitously between 90 and 100 MHz, right where the input stage stops working. I think this is due to reflection from the transformer performing poorly at these frequencies. Although the input level drops off slightly up to 90 MHz, it picks back up afterwards, becoming very large between 140 and 160 MHz. But, the shape of the traces shows the real problem. The difference in levels between the two is in part due to the transformer boosting the voltage. The yellow trace is the signal right at the input jack, while the magenta is after the transformer (across R1), shown here for reference: The vertical is set to linear mode to show the problems more clearly. This is always a little scary since the SA input is a bit delicate, but I was very careful. To capture these traces, I plugged a 10:1 oscilloscope probe into the input of the SA. I was poking around hopelessly at the front end with the oscilloscope when I realized I had a spectrum analyzer in front of me! The first find took much longer than it should have. I spent a little quality time at the bench this morning, trying to diagnose the issues with the first prototype board. I have 3 copies of the PCB from OSH Park, so I'll make one of them for I/Q outputs up to 70 MHz, one of them for a single clock up to 296 MHz, and maybe one that skips the PLL to go below 300 kHz.Ī DC-296 MHz logic clock will do for the immediate future. I think based on this experiment, I won't re-spin the board.
With lower frequencies, the scope can render the true shape more accurately, as shown here for 56 MHz (14 MHz input).
#RIGOL DSA815 TG HACK GENERATOR#
The input frequency from the tracking generator was 74 MHz (=296/4). That's what you get when you look at a signal near the scope's bandwidth - everything looks like a sine wave, since none of the harmonics make it through. The highest controllable frequency I was able to get was 296 MHz, shown here on a 300 MHz scope: I cranked the TG frequency up until the PLL stopped following at 4x (at that point the PLL's VCO just stuck). With this mod, the input signal is buffered, the frequency is multiplied by 4, and the result output on one of the ports. It only took a few mm of wire and a steady hand. I finally got around to reworking the PCB to route the 4x PLL directly to one of the output drivers.