Doing things with MODBUS should be easy but using ultra-cheap Chinese stuff for automation surely is anyting but. It's as much "hacking" as "reverse hacking" (engineering is to big a word) to figure out where those addressing a hobby and kludge-engineering market made a joke of specifications and protocols, even the most simple ones. In this project I'm having fun with these things by using them with STM8 eForth. Why not innovate a bit while we're at it?
MODBUS, especially the binary serial RTU protocol through RS485 interface is still popular - sufficiently popular to create a market for ultra-cheap (hobby-grade) components, e.g. "STM8S103 Relay Control Board" with barely usable MODBUS firmware
or the infamous Chinese "USB RS485 Adapter Black+Green" for less than $1 (including shipping).
For some time now I had fun with implementing MODBUS, testing it against frameworks and getting things running (or not) using Chinese gadgets. It turned out that finding the culprit sometimes takes time.
Lately, I played with KiCad, more for occupational stress relief than for creating something useful. But what the heck, why should this small STM8S001J3 MODBUS node disappear in the stream of the @oshpark shared projects list, eceiving no love?
This project shall make some attempts more visible and it I hope that it will lead to a usable platform soonish.
When doing hand soldering I learned "first hand" that the orientation of 0603 components matters, and consequently I worked a bit on the component placement of the tiny STM8S001J3RS485 PCB. Using 8 mil signal track width, 6 mil clearance and 10 mil via drills made it a lot easier compared to sticking with the KiCad defaults. The power supply now forms a rail from left to right, switching the side of the PCB about in he middle.
The text on the silcscreen is friendlier, too (even if they make good use of the 400DPI silcscreen that @oshpark provides (i.e. the print is tiny):
The 0603 components on the back side now are also easier to populate.
The new board is available on the @oshpark shared project pages, and the GitHub repo has been updated.
I found the first version of Mini-MODBUS PCB a bit difficult to populate by hand (using a fine tip shouldn't be necessary). The component placement around the power supply was all but ugly, and pin description and other information were missing on the silkscreen.
The reworked design is a lot nicer (and also easier to assemble). Making better use of the limits of @oshpark design rules (vias, drill diameter, clearance and track witdh) was helpful. Schematics and size remained unchanged.
Free delivery from the USA is fast enough for people with a daytime job - the PCBs lingered for 5 days in the German customs, so in a world without trade barriers it would have taken just 12 days from uploading the KiCad files to delivery!
The PCBs look quite good. Fortunately the 4k7 0603 SMD resistors did also arrive today. I hope to get the first tiny MODBUS node working by the end of the comming weekend.
My MODUS implementation on the STM8 eForth platform works. It was the RS485 communication that left me frustrated for a couple of months.
Since recently "designed" a small PCB to bring SO8 STM8S001J3M3 to good use, and now RS485 testing should better work.
The problem is that the one USB to RS485 adapter that I own, a Green+Black $0.80 gadget, doesn't play nice with my Ubuntu PC. In fact, it's able to receive things but transmitting leads nowhere. Besides, it makes as well Picocom, e4thcom and my JAMOD test hang on exit.
The PCB of the board has YYH-256 stamped on the back. The circuit appears to be simple enough:
It's based on a CH340G, the venerable Chinese USB RS232 interface chip. Has a full set of hardware handshake signals that potentially could be used for switching between Rx and Tx mode. Potentially, because there is no documentation whatsoever.
Next I started to think how cool it would be to have a circuit diagram of that simple thing. I fired up KiCad, and with the help of my trusty multimeter, I came up with this:
There is no DTR, no RTS, and thus no elaborate RS232 hardware handshake. Instead I find a *really optimistic* Tx based direction control. I would at least have expected a capacitor to prevent direction switching during transmission of a single character (!). No wonder that RS485-A is biased GND and RS485-B to 5V: the RS485 driver is only driving one polarity! The other one is just the 2.2k resistors doing their work (good luck with the promised 1 km) Is this how the CAN bus was invented?
I don't mean to say that such a circuit won't work under certain circumstances, but mine never did.
The next attempt was checking TxD for signs of life (pin2 of the CH340G), but again nothing.
A quick comparison of the circuit with a CH340G based standard USB-to-TTL adapter didn't reveal any difference: the inputs, besides RxD are N.C. as well (besides, I figured out why the same USB-to-TTL board never worked with the STM8 eForth two-wire-interface - some docs will have to be changed).
There can be just one conclusion: the CH340G on the RS485 Black+Green adapter must be broken.
EDIT: Today a new "RS485 Green and Black" dongle for $0.82 (including postage!) arrived. It worked as expected, but of course because of the design it's only usable for testing or configuring a device on the bench.