BUILDS / DIGITIZER / V1
digitizer
The two-channel digitizer and signal generator on module front ends (tier 2): two inputs and two outputs on an ECP5 FPGA board. Two AD9226 modules sample the inputs, a dual AD9767 module drives the outputs, and the host reaches it through the board's USB debugger.
Parts
| QTY | PART | FORM | |
|---|---|---|---|
| 1 | LFE5U-45F | Colorlight i9 v7.2 module, on the i5/i9 extension board | |
| 2 | AD9226 | module with a 2x10 male header (5 V on pins 1 and 2), 3.3 V outputs and an SMA input | |
| 1 | AD9767 | dual module with a 40-pin female header in the AN9767's order (5 V on pin 2), 3.3 V inputs and two BNC outputs | |
| 4 | pin header, 2x15, 2.54 mm, male, straight | ||
| 38 | jumper wire, female to female, 10 cm, for the ADC modules' male headers | ||
| 36 | jumper wire, male to female, 10 cm, for the DAC module's female header | ||
| 1 | cable, USB-A to USB-C, with data lines | ||
| 2 | cable, BNC male to SMA male, 50 Ohm, from each output to an input |
Walkthrough
At your own risk: not certified or calibrated test equipment. Legal
Solder a 2x15 male header to each of P2, P3, P4 and P5 on the extension board, and seat the i9 in its socket.
With no module connected, plug the extension board into a USB-A port with the USB-A to USB-C cable, press PWR, and measure P2 pin 23 and P2 pin 1 against P2 pin 2.
EXPECTAbout 5 V on pin 23 and 3.3 V on pin 1. A USB-C to USB-C cable may give nothing: the board's USB-C port has no CC resistors (the schematic marks CC1 and CC2 not connected), and a USB-C source drives no VBUS until it detects a sink's pull-down on CC1 or CC2 (USB Type-C spec 1.3, sections 4.5.2.2.7 and 4.5.2.2.8.2, as quoted in Zephyr's subsys/usb/usb_c/usbc_tc_src_states.c).
AGENTSPowering a board is a hardware action: logged.
With the board still powered, read the FPGA's JTAG IDCODE through the extension board's DAPLink debugger, with OpenOCD built for CMSIS-DAP, as Muse Lab's get-start.md (wuxx/Colorlight-FPGA-Projects at 5042201f) programs it. Write nothing to the FPGA or its flash.
EXPECT0x41112043, the LFE5U-45F's IDCODE in prjtrellis-db's devices.json (YosysHQ/prjtrellis-db at 015e0330, read 2026-10-06). The LFE5UM-45F reads 0x01112043 and the LFE5UM5G-45F 0x81112043 there.
AGENTSReading it is a hardware action: logged. If it reads anything else, stop: the module is not the LFE5U-45F this build is written for. Record the reading in the i9's incoming-QA record.
For each AD9226 module, before wiring any data line: note its package (28-lead SSOP or 48-lead LQFP), power it from P2 pin 23 to its pin 1 and P2 pin 2 to its pin 3, and measure DRVDD (SSOP pin 28, LQFP pin 29) against ground.
EXPECT3.3 V: the V1.2 schematic runs DRVDD from a 3.3 V regulator. Whether LQFP boards do the same is unknown.
AGENTSIf DRVDD reads 5 V, stop: the module's outputs would put 5 V on FPGA pins, and the ECP5 allows at most 3.63 V on an I/O pin, even briefly (Lattice FPGA-DS-02012 1.9, Table 3.1).
With the module still powered, measure the AD8138's pin 2 (VOCM) against ground (docs/scope.md, "Known part issues").
EXPECT2.0 V on a module wired as the V1.2 schematic is, with VOCM on the AD9226's reference: the variant that clips early.
AGENTSRecord the reading, the package and the seller in the module's incoming-QA record.
Power the DAC module from P4 pin 23 to its pin 2 and P4 pin 2 to its pin 1, with no data line, and measure its DVDD against ground.
EXPECT3.3 V: the AN9767 schematic runs DVDD from a 3.3 V regulator.
AGENTSIf DVDD reads 5 V, stop: the DAC would need 3.5 V for a logic 1, more than 3.3 V logic drives.
With the USB cable unplugged, wire ADC A's 14 signals to P2's I/O, ADC B's to P3's, and the DAC's 32 to all 18 of P4's I/O and 14 of P5's; each module's 5 V to pin 23 of P2, P3 and P4, and its grounds to that header's grounds. Leave P2's pins 3 (K18) and 28 (L2) and P5's pins 13 to 18 unwired: the i9 and the extension board's debugger use them. Which signal takes which pin waits for the LiteX platform file (roadmap.md, Phase 1).
With the capture gateware (roadmap.md, Phase 2), feed each input a slow ramp from the DAC through a BNC to SMA cable, and read the codes.
EXPECTCodes that rise with the ramp: the V1.2 schematic ties the SSOP's MODE pin to AVDD, which selects straight binary (AD9226 data sheet, Table I). A single jump at mid-scale is what two's complement output looks like (Table IV). If the codes jump about, the module labels its bits in the other order: sellers number D0 to D11 both ways.
Claims
| CLAIM | SIMULATION | ON THE BENCH |
|---|---|---|
| Samples each input at up to 65000000 Hz | unknown | unknown |
| Updates each output at up to 125000000 Hz | unknown | unknown |
| Every module's logic runs at 3.3 V: the AD9226 modules' output drivers and the AD9767 module's digital inputs are supplied by 3.3 V regulators on the modules | unknown | trusted aliexpress-modules |
| Each module runs from 5 V alone, which pin 23 of each of P2 to P6 carries | unknown | trusted aliexpress-modules |
| One USB port powers the i9 and all three modules | unknown | unknown |
| Reaches the host over Ethernet | unknown | unknown |