Hermes-Lite v1.21

Revision 1.21 addresses issues found with the revision 1.2 Hermes-Lite. FPGA firmware is the same for 1.2 and 1.21.

No SATA power connector

Revision 1.2 specified the wrong gender SATA power connector. Rather than rework this, I abandoned the onboard SATA power connector for the following reasons:

• There is a dearth of inexpensive SATA power connectors of the correct gender.
• One can still use a SATA power connector by hacking and splicing in a SATA power y-splitter. This costs about the same as a connector. There are through holes where 3.3V, 5V, 12V and ground wires can be soldered easily.
• The original plan was to eventually try a SATA interface as all connectors I found included the SATA data portion. With the release of the FTDI USB 3.0 bridges, I no longer give that high priority. Instead, I want to focus on USB 3.0 and gigabit ethernet as common interface standards. The SATA data portion of the connector takes quite a bit of space. I want to free this space for other uses.
• I have had mixed results and seen lots of variation with PC power supplies. Some produce lots of noise, not through the power connection, but radiated out the AC side. Others do not produce a constant voltage at the relatively light loads of the Hermes-Lite.
• The Hermes-Lite RX can be powered directly via the BeMicro CV. The BeMicro CV has a barrel power connector and USB connector.

Because of these reasons, revision 1.2 has a more flexible and experimental power supply interface. It is pads for a board-edge mounted 0.1 inch spaced dual inline header. I hope it will be useful to experiment with what power supply is best. For the future, I am leaning towards these goals:

• I do not intend to use LDO regulators as in rev 0.9 as the power lost to heat is more than I can stomach.
• I've read some articles about using switching power supplies with high-speed ADC and the trick is to design a good filter. See this and this. I would like to experiment along those lines, especially late next year when I hope to include the FPGA on the Hermes-Lite.

Experimental 24-pin board-edge mounted 0.1 inch spaced connector

As mentioned in the SATA power connector discussion, there is a standard 24-pin connector where the power and SATA data connector used to be. Half of it is for power, and the other half has 8 FPGA signals, ground, plus 3 spare connections to wire up extra signals, power, etc. It is designed so that the 10 pins closest to the BeMicro connector can form an Alex connector (5x2) when one of the spares is connected to 12V.

Test pads increased from 2 to 6

I found the test pads useful when bringing up revision 1.20. Although they probably can't all be used, I did expose all unconnected FPGA pins as test pads. All FPGA signals from the BeMicro connector are now used or available somehow.

J6 added for BeMicro SDK half-duplex support

To try and achieve best full-duplex operation, I need to use dedicated clock input pins for the 122.44 MHz RX clock. When not in full-duplex mode, this pin must be an output to drive a 61.22 MHz RX clock. The BeMicro CV's Cyclone V clock inputs can be configured as outputs. The BeMicro SDK's Cyclone IV clock inputs can not be outputs. So, J6 is added to use another pin as an output.

With revision 1.20, I had to run jumper to duplicate this for the SDK. The jumper was to revision 1.20's TP1, which has become the alternate clock in revision 1.21 so that firmware is still compatible. More details of this mod will appear in the build instruction.

J7 added to power Hermes-Lite from BeMicro CV

Since it is possible to power the Hermes-Lite from the CV (not the SDK), this jumper has been added to enable/disable this.

J5 and J8 added to disconnect PWREN from ground, disconnect 5V supplies

Although the default is to have these jumpers always connected, the idea came up since revision 1.20 to use the built-in JTAG interface as a slow and inexpensive way to transfer the data for one 96kHz receiver to the host PC. This jumper enables that type of experimentation.

Rearranged and larger mounting holes

Since the SATA connector has been abandoned, the mounting holes have been arranged in a more regular square pattern. The spacing matches that seen on the daughter boards. Also, the holes are slightly larger to support a wider range of mounting screws.

Rounded corners abandoned

While looking into panelization for automated assembly, I decided it would be easier to panelize the Hermes-Lite if there were no rounded corners. Also, a flat edge where the SATA connetor used to be helps with panelization options.

Fiducial marks added

To allow future assembly by www.smallbatchassembly.com, I've added fiducial marks on two opposing corners to align an optical placement system.

Silkscreen includes open source hardware symbol

This project is released under the TAPR open hardware license. The license meets the requirements for using this symbol.

Frontend connector can handle 2A at 12V

To support higher power amplifiers, the 12V traces have been updated to support 2A.

Frontend connector space widened and offset reduced

The frontend connector is a bit wider to meet the specs for a wider range of PCIe connector manufacturers. Also, the frontend connector is not seated as far back as it used to be, but still well within the connector specifications. This makes it easier to solder.

AD9866 thermal pad on PCB includes large "thermal via"

This is to facilitate manual assembly and soldering. One should take care not to put the AD9866 in thermal shock when soldering this pad.

All vias are "untented"

There is no solder mask on top of vias. This allows for easier testing and modding with vias as well as more opportunity to solder the AD9866 thermal pad from the bottom side. It does increase the potential for shorts from via to pads, etc.