Hermes-Lite v1.20

Modular RF Front-End

My biggest complaint with the Hermes-Lite prototype has been the limited things I can do with it. I receive and can transmit QRPp, but not at the same time. I want an inexpensive modular front-end so that I can easily add TR switches, amps, preselectors, antenna tuners, etc. This is the biggest new feature in 1.2. On the side of the board where the RF connectors would be, there is a straddle-mount PCIe female connector. This has two TX balanced lines, and one RX balanced line, plus some digital I/O (SPI bus, PTT, 3 GPIO). I was very careful to separate the digital and analog domains. The general idea is that various inexpensive daughter cards can be made and will plug into the PCIe connector.

This approach to modularity is inexpensive. I chose PCIe because it is ubiquitous and hence inexpensive. The main cost is in the 2-3 dollar straddle-mount connector. PCB daughter cards only have to have the proper edge connector and no expense for a connector. Daughter cards can be 2-layer and through-hole. Three 2-layer RF daughter cards from OshPark will cost ~$10.

This approach is mechanically sound. There are now 4 mounting holes on the Hermes-Lite main board. Daughter cards will have 2 mounting holes. There are no right-angles as the connector is straddle mount. See the picture at https://www.xmos.com/products/xkits/slicekit for a better idea of how this works. One can bolt the main board and daughter card to a substrate for a permanent setup. Unfortunately, there are no mounting holes on the BeMicro. This remains the weakest mechanical point in the system. One can use a square wooden dowel of the approximate height (from one of the big home improvement stores) to support the BeMicro when connected.

This approach does not sacrifice signal integrity. I am more worried about a sharp edges on the 122 MHz digital signals then I am about integrity or noise here. If one inspects what has been done on the AD9866 eval board and the Hermes with respect to RX and TX routing, this Hermes-Lite design is much more conservative. PCIe connectors support signaling in the gigahertz range. All RX and TX signals are surrounded by analog ground.

I am not getting into the business of selling RF daughter cards. There will be templates on the main Hermes-Lite github for people to easily layout their ideas. I will pull the most popular designs back into the main Hermes-Lite repository. Each project will include a BOM and a zipped package that can be sent to OshPark, iTeadStudio or Hackvanna to fabricate the boards. People can organize small group buys if they like.

There are many possibilities here that I am quite excited about. First on my list is to shamelessly lift the front-end from the SoftRock RXTX Ensemble and put it on a 2-layer through-hole board. (Tony responded to me on the SoftRock list that the designs are open source.) Imagine buying 3 boards from OshPark for ~$10 to build 3 "superband" front-ends 40/30, 20/17, 15/12/10, etc. If someone else wants to take on this project, I'm glad to see them do it. I will post the templates with instructions later this week. What about a bridge for a VNA? What about a matching network so that the Hermes-Lite can be placed right at the end of a long-wire but operate on all bands? With an inexpensive modular front-end, this type of experimentation is made easy.

There is no backup option on the Hermes-Lite 1.2 for onboard SMA connectors. I actually spent quite a bit of time trying to fit this in at the top of the board, but never liked the resulting crossing of signals so threw it out.

There is no Hermes compatible IO on the Hermes-Lite. If one wants an alex header, they can design (or use) a daughter board for this purpose. After some investigation, I think the existing Hermes peripherals are too expensive and not easy to build or connect.

Standard Power Supply

This design uses a standard SATA power connector to supply power. This includes 3.3, 5 and 12V which are also supplied to the daughter board. There are standard size through-holes if one wants to not stuff the SATA connector and use their own power supply. One can also cut an inexpensive SATA power adapter cable to connect an arbitrary power supply to the Hermes-Lite. One can also build a daughter card that has the power supply circuitry they desire and feed power through the PCIe.

Miscellaneous

• Footprints for termination resistors are included for both the lower power and high power output paths from the AD9866. These are optional.
• Footprints to connect the IOUT_G/N lines to the high power output path form the AD9866 are included. This allows power output to be increased by 3 dB. These are optional.
• Footprints for decoupling capacitors on the RX path are included. These can be jumpered if no capacitors are desired.
• To save space and expense, all 3 config bits are tied low. This was tested in v1.0. Different configuration can still be selected via software.
• The clock is now 61.440 MHz for compatibility with the latest Hermes efforts.
• A bit of the wiring has changed between the AD9866 and BeMicro connector. This was just to ease routing and implies a different FPGA pin assignment file.
• RXCLK, which in full duplex must run at 122 MHz, is a dedicated route to a dedicated clock pin. As recommended in the AD9866 datasheet, there is also a footprint for an optional termination resistor here.
• CLKOUT1 includes a selection matrix so that it can connect to one of two pins on the BeMicro connector. This is to support driving clock inputs on the SDK and CV. It also includes a footprint for an optional termination resistor.
• J5 is permanently wired so that the Hermes-Lite supplies power to the BeMicro. The intent is that one never uses the USB when connected to the Hermes-Lite. To program, one will disconnect the BeMicro SDK, program via USB, then program nonvolatile via standard HPSDR ethernet programming, then reconnect. Once programmed, one will only have to disconnect the BeMicro and reprogram in this manner if the firmware is corrupted. There is no ground under the J5 jumper trace so that it can be cut if desired.
• FPGA 3V (F3V) is run to a connector for experiments with the BeMicro CV driving DVDD as the regulators on the CV should handle this.
• A serial eeprom for IP and MAC address is included. This is a backup option if changes to the firmware can not use the existing BeMicro eeprom to store this information. If not used, the pads are extra general purpose connections to the FPGA.
• RESET is connected to the BeMicro's RESET_EXP#. The BeMicro has a pull-up resistor on this line. It connects to the FPGA so that the FPGA can control reset.
• EXP_PRESENT is wired to DVDD. This is so that the BeMicro can detect if the Hermes-Lite card is connected or not. This will allow running in test-mode for programming when no Hermes-Lite is connected.
• There are 2 test pads for general purpose use around the the connector to the BeMicro.
• The SATA data connector has 4 connections to the BeMicro. This is for experimental use only. They are not LVDS. Eventually I'd like this SATA data connector to actual support SATA data.
• There are quite a few filter caps and ferrite beads on the power distribution. Some of this may be left off to reduce cost.
• The SATA connector is placed on the bottom side to avoid mechanical conflict with the BeMicro CV, and also to place the SATA data connector near the FPGA. The SATA connector is not much thicker than some of the filter caps on the bottom.
• Components and mounting holes are kept away from the BeMicro CV to accomodate the overhang of the BeMicro CV.