TTL Differential Driver
Problem
We needed to send a TTL pullup to remote equipment for toggling an RF amplifier. We were concerned that the pullup voltage may be affected by noise or differing reference ground voltages on the source and sink.
Trade
There were a few commercial options, but they seemed few and far between, especially with some of our nice-to-have requirements (i.e. preferring a DIN mount or inline package). I found a chip online that would drive a differential signal that we could convert back on the far end. I proposed the project to colleagues, but in the end we selected a COTS product for its demonstrated reliability.
Goals
I decided to pursue the project on my own time, because I lacked experience with board layout, board CAD programs (like Eagle), and hadn't had a SMT soldering project in a while. To that end, the following were the project goals:
- Learn CAD board layout by completing a small electronics project
- Satisfy the (admittedly) stubborn desire to prove the differential driver would work
- Have a bit of fun with SMT soldering along the way
- Release the project to fill a role that I found to be sparse in my initial search for COTS options
Summary
I created a layout based on the reference design and datasheet for the TI SN65ELT22D mixed signal translator (and SN65ELT23D). We planned to use standard CAT-5E twisted pair cable between the two devices, so an RJ-45 connector was an easy choice. I chose the 1206 packages because I was soldering by hand without reflow and wanted easier parts to work with.
I created a simple 2-layer board in Eagle terminating in a terminal block on one side and the RJ-45 connector on the other. All other components were placed between or on the underside to conserve space. The transmit and receive boards were identical to reduce complexity. The only change is the TI part on each.
Each of the TI chips comes with two lines, so I gave each signal line two wires on the CAT-5E. Since the signal lines come in positive and negative, I chose to combine them in twisted pairs. One twisted pair carries both positive and negative. Another twisted pair carries identical signals. The two remaining twisted pairs similarly carry the other signal pairs
I ordered boards from OSH Park and components from DigiKey, and assembled them at a basic solder station.
BOM
| Digikey PN | QTY Rx | QTY Tx | Note |
|---|---|---|---|
| 296-20717-1-ND | 1 | 1 | TI 5V0 Linear Reg |
| 296-32152-1-ND | 1 | 1 | TI 3V0 Linear Reg |
| 296-24085-5-ND | 0 | 1 | TI TTL-DIFF Tx |
| 296-24820-5-ND | 1 | 0 | TI DIFF-TTL Rx |
| A31465-ND | 1 | 1 | Ethernet |
| 277-1860-ND | 1 | 1 | Screw Terminals |
| 478-1567-1-ND | 2 | 2 | C1, C4: 1uF |
| 490-1799-1-ND | 2 | 2 | C2, C3: 2.2uF |
| 311-100FRCT-ND | 4 | 4 | R1, R2, R3, R4: 50 Ohm |
Files
Conclusions
Basic testing suggested that the design would have worked for our needs. That said, the COTS solution is also working well, and the upfront risk with the COTS solution was less. I want to be clear that I agree with that choice; I just saw this as a fun project to take on and learn.
I could have shrunk the SMT packages a bit more, but the board size would likely have remained the same, since it is driven largely by the two thru-hole connectors (RJ-45 and terminal blocks). Overall, I'm very happy with the design and result. I accomplished everything I set out to.
There are a few improvements I would like to make, and I'll try to compile and release my notes soon. If I were to produce this for actual production use, I'd also want to test and characterize it far more than I did.
License
Please treat this as released under CC-BY-SA 4.0.
