Power board v3

[thanasipantazides]

Overview

This issue will track changes to the v3 power board design to address FOXSI-4 hardware errata, documented here.

Fixes to make

• Connect LT8612 BST to SW via 0.1 µF capacitor.
• Connect LT8612 SYNC to INTVCC.
• Route SPI.SDO to J1.7 and SPI.SDI to J1.9 (swapping them).
• Change R58 value to 3 kΩ.
• Timepix_5V_out and Timepix_12V_out naming or placement in header should be swapped. They carry one another's voltage. (Error was introduced by module-level wiring in the hierarchical schematic).
• The MAX7317 SPI.SDO line needs external circuitry to be hi-Z when inactive (when nCS3 is high).
• Confirm feedback resistor values for all converters (measure on board).
• Something needs to be done to address the CMOS 2 always-on issue that emerged during sequence testing in Alaska. Consider new SRRs, input-side (+28V_clean) circuit protection, or output-side circuit protection. Overvoltage protection? Fuses?
• Increase the MAX7317 pullup resistor value to ~47 kΩ. The output ports are weak (max ~10 mA drive). Don't waste effort when they are sinking to ground.
• Confirm input AD7490 divider values.

[thanasipantazides]

For the MAX7317 SPI.SDO fix, can use the TI SN74AHC1G125 inline, switched by SPI.nCS3, to toggle.

[thanasipantazides]

Oct 10 2024:

Implemented these locally:

• Connect LT8612 BST to SW via 0.1 µF capacitor.
• Connect LT8612 SYNC to INTVCC.
• Route SPI.SDO to J1.7 and SPI.SDI to J1.9 (swapping them).
• Change R58 value to 3 kΩ.
• Timepix_5V_out and Timepix_12V_out naming or placement in header should be swapped. They carry one another's voltage. (Error was introduced by module-level wiring in the hierarchical schematic).
• The MAX7317 SPI.SDO line needs external circuitry to be hi-Z when inactive (when nCS3 is high).
• Confirm feedback resistor values for all converters (measure on board).
• Something needs to be done to address the CMOS 2 always-on issue that emerged during sequence testing in Alaska. Consider new SRRs, input-side (+28V_clean) circuit protection, or output-side circuit protection.

[thanasipantazides]

For voltage spike suppression, consider the LittleFuse SMCJ36CA which was used on the CMOS boards.

[thanasipantazides]

Or the SMBJ36CA, which is slightly smaller.

[thanasipantazides]

A proposed set of mitigations for the CMOS 2 issue. This avoids changes to the CdTe canister channels, for which we expect much less inrush and which performed well:

1. Add a TVS diode on the input for transient suppression. That's done above.
2. Maximize the R19/R20 resistance to maximize turn-on time. I think the largest we can do is ~1 kΩ; I will validate this in the lab.
3. Duplicate the CMOS channels' G3VM-61 chips. Run two chips in parallel per CMOS channel. This gives us much more current overhead, uses two of the same part that already flew, and minimizes board space consumption. Are there downsides to this plan?

[thanasipantazides]

I corrected the Timepix_5V_out / Timepix_12V_out schematic-side. So there is no change to the FOXSI-4 power pinout, but the schematic is also now internally consistent with the voltage you actually get out of the pins.

[thanasipantazides]

Phil made two suggestions in review:

1. Move topside copper out from under 12 V inductor. Lower layer or route around. This affects REGS_Isense and Net-(JP9-C).
2. Get more consistent voltage across each duplicated CMOS SSR LED. This applies to U19 + U33, and U20 + U32. The idea is each LED will turn on at slightly different times, based on their steep transition in response to applied voltage. To mitigate, Phil suggests either wiring them serially, or giving each LED its own resistor. Serial hookup is undesirable because of lack of redundancy, if the topside fails. Phil included a setup (for the low voltage/high current switches) with similar paired SSRs in his original v1 board, and succeeded with separate resistors per LED.

So for (2) I will plan to implement dedicated resistors for each LED.

[thanasipantazides]

Another edit before purchase: add some exposed +28/return copper on bottom side just after input Gecko. In case we need more components for any reason.

[thanasipantazides]

Re: Phil's suggestions: I've implemented this:

So for (2) I will plan to implement dedicated resistors for each LED.