Hi Greg,

On 04-Feb-04, Greg Mayman wrote to Roy J. Tellason:

 RJT>> Much went on with numbers back when.

 GM> Sure did. And there were some great confusions. I remember seeing
 GM> one diode number that had FIVE different types, depending on
 GM> which manufacturer it came from.

And one manufacturer who would supply 5 different diodes for the same
number, depending on which factory it came from...
They ranged from the very good (glass encapsulated) through the usual
to the rubbish (plastic which wasn't properly protected).
They couldn't guarentee supply from the good factory, or even the
acceptable factorys, so we wouldn't use them (we were talking
reasonably large volumes, >100k parts/year).

 GM>> And you could use level translator chips - the 4049 and 4050 will
 GM>> do that for you. Or a two transistor buffer will also do it.

 RJT>> If I'm remembering right they'll take a higher-range logic level swing
 RJT>> and translate it down to somewhat lower a range,  but that's not what I

 GM> Yes you're right. The input can go more positive than the + rail.
 GM> I was thinking of some other chip, but I can't find it in my data
 GM> book.

 RJT>> need here, I'd need something that would take a zero-based logic swing
 RJT>> and translate it to a bipolar one.  There's probably some simple way to
 RJT>> do this but it's not apparent to me at the moment.

 GM> .  +5v rail------------+-----------
 GM> .                      |           |
 GM> .                 ------------    4700
 GM> .  +2.5v ref ----| -          |    |
 GM> .                |  1/4 LM339 |----+---- out
 GM> .    in ---------| +          |
 GM> .                 ------------
 GM> .                      |
 GM> .  -5v rail------------

It really needs some +ve feedback to make it into a schmitt trigger
circuit to avoid instability around the switching point on slow
signals. That needs 2 more resistors, some series input impedance (say
10k) and one from "out" to 339 input (say 100k).

 GM> That's the simplest way to do it with the least number of
 GM> components. The ref source can be a voltage divider (2 x 10k)
 GM> between +5v and 0v, and the same ref can feed all four
 GM> comparators.

I'll disagree with that! :-) The simplest way is with an RS-232 line
receiver like the 1489/1489A. Quad inputs, _NO_ external components
needed, only uses the logic +V rail. Spec to +- 30 V on the input.
Available to suit most system logic types (there are CMOS ones if
needed, and Schottky ones for speed).
External components are only needed to vary the switch point.

 GM> The only problem with this is that the 339 isn't very fast, I
 GM> think it's about 1.5uS to swing the output.

My NS LM1489 data sheet gives a rise time of 175 nS max, fall time is
much less at 20 nS max.

 GM> This is a much faster level converter but with more components...

 GM> +5v rail-------------+----------------------
 GM> .                    |                      |
 GM> .                   22k                     |
 GM> .                    |                      e
 GM> .                    +-----100k-----------b   TR2-PNP
 GM> .                    |                      c
 GM> .                    c                      |
 GM> in -------100k-----b   TR1-NPN              +----- out
 GM> .                    e                      |
 GM> .                    |                    4700
 GM> .                   gnd                     |
 GM> .                                        -5v rail

 GM> Both transistors are GP types. This circuit is very fast, and I
 GM> have tested it to frequencies above 10MHz. With RF transistors it
 GM> should go even higher.

 GM> In theory, if you use MOS-FETs you can leave out the 2x100k
 GM> resistors for an even lower parts count and higher speed, but I
 GM> haven't tried it myself.



George

--- Terminate 5.00/Pro