The iADC, which can digitize either two streams of data at 1 Gsps or one stream of data at 2 Gsps, was designed to be the IBOB's input data source. It has seven SMA input ports: I+, I-, clk_i, sync, Q+, Q-, and clk_q. Each of these input ports was designed for 50-ohm termination. I- and Q- traces have been cut in all 0.2.x ADCs, and likely will be in future runs.
The ADC input circuit can be divided into three sections: SMA connector to the balun input; balun; balun output to ADC input. Each of these sections was isolated in the measurements below. Each ADC input is terminated by two 50-ohm resistors to ground; these termination resistors were changed to 25 ohms as noted below.
Strange behavior (e.g. spikes, dips) in output data was noticed during tests of various designs, including the pocket spectrometer and eight-station correlator. The following measurements of the ADC's input impedance were undertaken to determine if imperfect input impedance matching was a cause of such behavior.
Consulted for advice because of his RF experience, Doug Thornton at Campbell Hall offered several suggestions for implementation in the next revision of the ADC board:
1. Use sidelaunch SMA connectors instead of right-angle connectors.
2. Reduce the length of balun leads as much as possible (reduces inductance); use Mini-Circuits TC1-1-13M instead of current Macom TP-101.
3. Move balun pads as close to balun as possible.
4. Minimize copper pad ring around SMA input pin (reduces capacitance).
5. Terminate ADC input with 25-ohm resistors instead of 50-ohm resistors (to terminate balun properly).
Other possible improvements include enlisting an RF expert to redesign the board and using a fab that specializes in RF applications, and using a different ADC chip that digitizes a wider bandwidth.
Data were taken at the BWRC with ADC 0.2.13 using their HP 8719C Network Analyzer. The instrument was set to measure s-parameter 11 (forward reflection) with a test port power of -10 dBm. The test signal was swept from 130MHz to 2GHz, and data points were taken in 10MHz increments.
Overall, the input impedance varies +/- 4 ohms from 50 ohms at -10dBm, even when attached to a 4' cable. Measurements of other test port power levels (-30dBm to 0dBm) will be added to this memo at a later date. Any input signal power to the ADC must be no greater than -10dBm (though the input clock may be 0dBm without issue). Input signal powers greater than -10dBm experience severe degradation at the input due to imperfect termination.
All of the following plots and data may be found in the Excel spreadsheet located at http://seti.berkeley.edu/casper/memos/adcimpedance/adcimpedance.xls.
First, the input impedance of ADC 0.2.13's I+ input was measured. (ADC 0.2.12's measurement is used here because the network analyzer used for 0.2.13 was faulty, and subsequent modifications to 0.2.13 made this measurement impossible. It is assumed that all boards have similar characteristics, as they were manufactured simultaneously by the same company.) The I+ input was connected to the network analyzer via a SMA coupler.
For comparison with the above, measurements of the Q+ input and an old ADC without cut traces are provided. Both inputs were connected to the network analyzer via a SMA coupler.
.PNG)
Upon Doug Thornton's suggestion, the 50-ohm ADC input termination resistors were switched to 25-ohms. In the first plot, the ADC was connected to the network analyzer via a SMA coupler; in the second, the ADC was connected via a four-foot SMA cable. The third plot shows the original 50-ohm termination with the ADC connected via a four-foot cable.
Scraping off copper pad around the SMA input pin did little to improve the impedance. The ADC was connected to the network analyzer via a SMA coupler.
Bending down the long balun leads (moving them closer to the ground plane) also did little to improve the impedance. The ADC was connected to the network analyzer via a SMA coupler.
The following two plots show the effects of parts of the input circuit on the impedance. These measurements were taken after the copper pad was cut away and the balun leads bent. The ADC was connected to the network analyzer via a SMA coupler.
For this measurement, the balun was disconnected from the SMA connector and replaced with a 50 ohm resistor to ground, isolating the connector from the rest of the circuit.
After restoring the connection between the SMA connector and balun, the balun-ADC traces were cut at the balun pads and replaced with two 25-ohm resistors to ground, isolating the balun from the ADC.
It is apparent that with signal levels at -10dBm, the input impedance of the ADC is within 10% of 50 ohms for most of the input range. The effect of this variation may not be sufficient to explain the strange behavior observed during testing of various designs.