On Fri, Sep 4, 2015, at 11:57 AM, Jon Elson wrote: > On 09/04/2015 03:58 AM, andy pugh wrote: > > On 4 September 2015 at 02:16, Jon Elson <[email protected]> wrote: > >> Apparently, this device can't do bidirectional translation. > >> It can do 22 channels of conversion all going to a lower > >> voltage, or all going to a higher voltage. But, when wired > >> that way, it can't flip direction > > That's not the way that I interpreted the data sheet when I made the > > board for the 3.3V to 5V conversion on the RPi, and the boards > > appeared to work. > > > Well, as a typical NXP data sheet, they leave a lot of > helpful info out. Looking at it for 5 minutes, I really > couldn't see how it worked to step up the voltage. Stepping > down seemed pretty simple. Glad it worked. >
You need pull-ups on at least the high voltage side. Simple but clever really. Pull down on either side, and the other side gets pulled down. Release both sides, and both sides are free to rise to their respective supply voltages, without the pullup on the high voltage side bleeding thru to the low voltage side. Driving low-voltage side high does NOT drive the other side high, that's why you need pull-ups. I'm not 100% clear if driving the high voltage side high drives the other side high. If it does, it only goes as high as the other side's reference voltage. If it doesn't, you need pull-ups there too. It is bidirectional in the same way that any net driven only by open- collector outputs is bidirectional - anybody can pull down, nobody can pull up. If using it for uni-directional signals, you supply a pull-up on whichever side is the output. Each channel can go either direction. -- John Kasunich [email protected] ------------------------------------------------------------------------------ _______________________________________________ Emc-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-users
