I2C master

Good question -- you use the diode drop to steer to neighbors:

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com

the

point

asynchronous

just the

it... ;-)

networking issue.

None of the USB to I2C gizmos I've seen follow the specification. For example, the FTDI ones configure the clock line (SCL) as a push-pull output rather than open drain.

I contacted FTDI support about that once and was told "it works fine for other customers". Grrr.

Allan

On Saturday, September 23, 2017 at 8:30:06 PM UTC-7, Allan Herriman wrote: ..

There is an FTDI API library call to enable clock stretching - maybe the default is totem pole to allow faster operation to support high-speed I2C

FT4222_STATUS FT4222_I2CSlave_SetClockStretch(FT_HANDLE ftHandle, BOOL enable)

kevin

You have two separate wired-OR gates, one for the inverted and one for the non-inverted side.

On the inverted side, connect the port to the diode anode, connect the cathodes together and to pulldown resistor to V-

On the non-inverted side, connect the diode cathode to port, connect anodes together and a pullup to V+

Thanks. I was using an FT2232, which didn't seem to have that ability. This was in a design from ~2009, when the '2232 was new.

Allan

a) We have 2 boards, both Jedi ^H^H^H I2C masters, talking over RS232

b) We use I2C from one board to talk to I2C accessory boards.

c) I wondered "Can we skip 232 and add the other master to the I2C bus?"

Clearly the answer is "stick to what works...."

-- A host is a host from coast to snipped-for-privacy@nrk.com & no one will talk to a host that's close.......................... Unless the host (that isn't close).........................pob 1433 is busy, hung or dead....................................20915-1433

On a sunny day (Sat, 23 Sep 2017 20:06:06 -0500) it happened "Tim Williams" wrote in :

That is a very clever circuit, had not seen it before, no power source needed. OTOH economics: I bought 2 good 8 port switches that each came with a power adaptor on ebay for about 12 $ each...

To make a PCB, buy diodes and connectors, housing, time.. would be more expensive.

Or you can use CAN controllers in microcontrollers, but with a simple open collector bus. You get the best of all worlds then - a convenient and reliable multi-master bus without much extra hardware (no components at all for your bus, if your microcontroller has flexible enough pin drivers) and with simple software.

Well, not ALL worlds. Ethernet has better distance, and RS-423 with terminators (sold widely as LocalTalk by Apple, with third parties making ports that work for Windows) does good fast serial with (nominally) about 500m wire limits.

Open collector is kinda... IEEE-488. Short distance and low bitrate.

OK, not /all/ worlds. It only makes sense for local connections, such as between devices on a board or on closely-coupled boards. But the

CAN with an open collector bus would also work - just with much easier multi-mastering.

10Base2/5 and 10BaseT with hubs had only CD (Collision Detect) arbitration. On 10/100/1000BaseT with some extra hardware (switches) is needed to queue the messages to other ports.

Is that RS-423 (a.k.a DEC-423) shit still around ? That was supposed to be RS-232 compliant. Of the first batch of VT-220 terminals we received, the transceiver chips were replaced all but one during the warranty period.

Anyway, claiming 500 m for any single-ended non-isolated system is really stupid in the real world. A balanced system like RS-422/485 could work 1000 m at 115k2 bits/s provided that the connection is isolated (no ground potential issues).

TTL Open collector systems, such as Unibus can do about 15 m in a flat cable. However, the real show stopper is the 0.4 V noise margin in the TTL "0" state. We had to connect the CPU and each disk unit drive frames with 5 cm wide copper strips in an old building with TN-C electrification to keep the CPU/disk communication error rate at a reasonable level.

The RS-423 serial ports did take a transformer to network. The drivers, and receivers, had virtually NO problems, they were AMD's AM26LS30 and AM26LS32. Some connector wizardry was employed, and two-port backbone connectors, so any unoccupied connector got the right termination resistor applied (until you plugged something into it).

The biggest problem: everyone who wired to an RS-232 plug forgot that they wanted the negative output signal.

Transformer WTF ?

That would require Manchester coding or similar to get rid of the DC component, not very usable for traditional NRZ signaling.

Sure did. The serial communications controller mode was called 'FM0'. It wasn't just a UART, it was kinda... smart. Zilog 8530. That was in all the Macintosh models roughly 1984-2000, so even the oldest Mac128 would do networking.

On a sunny day (Wed, 27 Sep 2017 03:16:12 -0700 (PDT)) it happened whit3rd wrote in :

Ah yes, good old 8530, did a Z80 based baycom packet radio modem with it

AX25 etc ... wrote all that code in Z80 asm, 27 MHz packet radio. And I still use the TCM3105 in an RC airplane com link... was posted here years ago. SDLC HDLC...

And Linux still has the drivers and docs: /usr/src/linux-2.6.37.6-p1_small/Documentation/networking/z8530drv.txt or use locate z8530drv.txt

Do not see the datasheet on this PC anymore. Things change... Threw away the whole huge z80 system one day. Too much stuff...