Eight. It's got a Class-H TEC driver (cooling only in this version), temperature controller, gain vs. voltage compensation, a Cockroft-Walton for bias generation, pHEMT bootstrap, and three-stage PGA on the output. It also speaks MODBUS over RS-485 and RS-422 for test, calibration, and gain control. (The principal gain control mechanism is a DC voltage, as in the Hamamatsu PMT modules--we're aiming to be an improved drop-in replacement with a lower price. (Not so low as to destroy the market, you understand.)
It has population options.
First is for 6 MHz with fine gain adjustment via what AFAIK is the world's fastest dpot, the AD5273BRJZ1 (1k, 64 taps, I2C). We use the top half of that to get ~0.2 dB gain steps over 6 dB. The second stage is X1-X2-X4-X8 using a TMUX1511 quad SPST to switch the input resistors, and the third stage X1/X8 using the fourth section. The first two are ADA4898s in 3x3 mm CSP and the last is an ADA4860 (SOT-23).
Second is for 50-75 MHz, which shorts out the dpot. Bandwidth is a compromise between the fast rise time (~1 ns) and slow recovery time (~20 ns) of the Hamamatsu parts, and is limited by the GBW of the ADA4899 in the second stage.
Third is 300-600 MHz (primarily for the On Semi parts, which are considerably faster than the Hamamatsu ones). That one swaps out the ADA4899s for AD8045s, which are about twice as fast (1 GHz, 1400 V/us) and runs them at fixed gains (TBD depending on gain requirement).
It fits into the smallest Hamamatsu module form factor (1 x 1 x 2 inches, end-on), which is why we needed the LFCSP op amps. The main board has to go diagonally as it is. ;)
The MPPC board goes inside the end face, and has the bootstrap, TIA, and one of those very nice Molex 5025981593 FFC connectors--0.3 mm pitch, fanned out into three rows of 1-mm pitch contacts. Magic.
What's your current gizmo?
Cheers
Phil Hobbs