On Monday, February 17, 2014 11:38:46 AM UTC-6,
Fabio Battaglia wrote:
I don't have 1K/1.2K SIP-10 resistors, but I do
have enough 1.5K, do you think those would be OK too?
I obtained a copy of the Wilcox book to go with my 68000 board
build and this kind of thing is discussed at length (pages and
pages of equations) early in the book, from memory less than
500 ohms could hit the overcurrent limit for his design and
more than 8500 ohms would undercurrent as a pullup depending
how thirsty the inputs were for current. Lower switches
faster but wastes power, higher uses less power but switches
slower.
I ran the numbers myself using a real live Motorola 74LS682
datasheet and you've got two criteria:
1) Too low of a resistance to Vcc and a TTL L output will
overcurrent and catch fire.
So figure the data sheet Vol = 0.5V and Vcc at 5 V, thats 4.5
volts across the resistor. And the 74LS682 (or most any LS)
will catch fire around 24 mA so 4.5V/0.024A = 187 so less than
190 ohms to Vcc could burn out a LS output. Thats assuming
there's no load on the TTL output other than the resistor,
which would be pretty pointless. So I'd keep it WELL above
200 ohms.
2) Too high a resistance to Vcc and the input current of 0.4
mA will pull a false low. So figure 5 volts Vcc, half a volt
for ground bounce and half a volt for noise, that leaves 4
volts, then Vih for 74LS is 2.0 volts, so the resistor must
drop less than 4-2 = 2 volts. At the previously defined 0.4
mA. So 2/.0004 = 5000 ohms. Thats assuming it only drives
one input. So I'd keep it well under 5000 ohms.
So, well over 190 ohms and well under 5000 ohms. How bout a
geometric mean? (5000*190)**0.5 = 975 ohms so ideally it
would be around 1K which is probably where the designers are
getting 1K from.
How bad is 1.5 K? Well, lets get a percentage of the range,
from at the low end where it catches fire and the high end
where it eternally outputs a false Low level. and call that
full range 100%. That 1500 ohm pack is (1500-975)/(5000-190)
= about 11 percent high of center of the safe range. Now you
were probably building stuff with 10% tolerance resistors so
the range of plus or minus 5 percent higher or lower means the
possible ranges just barely don't overlap by about a percent,
so there should be a measurable effect.
But will the measurable effect matter? Well, worst case is
your 10% resistor could be 5% high (or 5% low, which doesn't
matter) and you've decided to use something that centers 11%
high, for a theoretical peak of 16% high of perfectly centered
on a 100% scale. 16 outta 100 is a long, long way to go until
it fails in eternal false L level. So you're all good, by
far.
What if you got the idea of throwing two 1500 resistor packs
in parallel? Well that would be 750 ohms.
(750-975)/(5000-190) is 5 percent low, so its technically
better, however, look at the failure mode... Too low means it
burns out permanently, but too high doesn't hurt anything. So
you're closer to parallel two, but in a dangerous direction,
and further away but in a safer direction to just use one.
Combined with two of them costing twice as much, I'm thinking,
don't parallel them. Also the component you never install can
never fail or have a soldering issue or add to capacitive
load, so its twice as reliable, sorta, not to install parallel
resistors.
TLDR don't worry about 1500 ohms, on the other hand, I
wouldn't go trying 4700 ohm packs..