Pull down resistors on bus of microcontroller board

[Rene]

Hello to all!

I have a question that may sound stupid to one who knows the answer but
I am bit uncertain about whether what I would like to do would be a wise
thing. For some time now (I have not much time to spend on my hobby) I
have been building an experimentation board for AVR processors, quite a
sophisticated one if You ask me (it's my own design) . Often one sees
that output leds are coupled directly to the output ports (well, with a
resistor off course). I did not like that so I chose to use 74HCT540 bus
drivers (they sink the current through the led so the latter lights up
when there is a one on the bus). The bus is 18 lines wide (plus two for
PS) and which line has which function depends on the adapter that is
plugged in (it holds the controller)(btw for LED 17 and 18 I use two
ports from a 74HCT04 in case someone would wonder how to buffer 18 lines
with 2 octal buffers). The problem is that when a line is "not defined",
it's led turns on. And sometimes off. And on. And so forth... Point Your
finger to it and it will do something. The impedance of those drivers is
obviously extremely high. I noticed that if I use a pull-down resistor
of 1Mohm on every line, this unwanted behaviour does not occur. However,
the datasheet of several AVR-processors says that it is not allowed to
put more than one gate (a HC-buffer) on the oscillator output (XTAL2) of
the processor if one is using an external crystal. In a datasheet with
general specs of the 74HC family I found that input leakage current is
maximum 1 uA, at 25 degrees C even only 0.1 uA. They tested this with a
VCC of 6 volts. That means an impedance of 6 MOhm and that is more than
my 1 MOhm.
My question is whether anybody foresees problems when I mount these
resistors. Off course there might be a problem if I want to measure an
analogue voltage from an high-impedance source (the impedance of the ADC
is, iirc, about 32 MOhm) but in that case I will have to buffer it, no
problem. But what are Your opinions, apart from that? And when some line
of the bus will be connected to the crystal, do You think that in real
life that will cause problems? I can off course make some of these
resistors "disconnectable" by means of jumpers (I wish I had done that
with the leds, but then again, then You just wouldn't see them anymore,
the buffer might still start oscillating), but as I can connect several
diff. types of controllers through the adapters, it would mean more
jumpers than for just one type of controller.

Thank You very much in advance for Your opinions/experiences.

Yours sincerely,

Rene

P.S. I just thought of what might a smarter solution to the crystal
problem, if I want to use a controller with an external crystal I can
put the jumper on the specific controller adapter so that the line will
not be connected to the bus on the development board. I guess that would
be best. Remains the question whether there could be other problems.

P.P.S. I could have rewritten my question after my brainwave from the
first P.S. but in that case someone might have come up with a suggestion
about that the crystal might give problems, I thought it would be wiser
to just leave it like this so You might see that those specific pins
won't be a problem anyhow.

 

[John Popelish]

Rene wrote:
(snip)

I
have been building an experimentation board for AVR processors, quite a
sophisticated one if You ask me (it's my own design) . Often one sees
that output leds are coupled directly to the output ports (well, with a
resistor off course). I did not like that so I chose to use 74HCT540 bus
drivers (they sink the current through the led so the latter lights up
when there is a one on the bus). (snip)
The problem is that when a line is "not defined",
it's led turns on. And sometimes off. And on. (snip)
I noticed that if I use a pull-down resistor
of 1Mohm on every line, this unwanted behaviour does not occur. However,
the datasheet of several AVR-processors says that it is not allowed to
put more than one gate (a HC-buffer) on the oscillator output (XTAL2) of
the processor if one is using an external crystal.

(snip)

The crystal oscillator function is analog, not digital, and
quite sensitive to not only DC leakage ands loading from
external resistors, but also to capacitive loading. I am
not familiar with the AVR line, but I suspect the oscillator
is usually connected to only a few of all the possible pins.
If you can confirm this, it would be a good idea to have a
relocatable jumper that would disconnect your pulled down
buffer input from the oscillator pins and possibly free it
up for a secondary use.

 

[colin]

Hello to all!

I have a question that may sound stupid to one who knows the answer but I
am bit uncertain about whether what I would like to do would be a wise
thing. For some time now (I have not much time to spend on my hobby) I
have been building an experimentation board for AVR processors, quite a
sophisticated one if You ask me (it's my own design) . Often one sees
that output leds are coupled directly to the output ports (well, with a
resistor off course). I did not like that so I chose to use 74HCT540 bus
drivers (they sink the current through the led so the latter lights up
when there is a one on the bus). The bus is 18 lines wide (plus two for
PS) and which line has which function depends on the adapter that is
plugged in (it holds the controller)(btw for LED 17 and 18 I use two ports
from a 74HCT04 in case someone would wonder how to buffer 18 lines with 2
octal buffers). The problem is that when a line is "not defined", it's led
turns on. And sometimes off. And on. And so forth... Point Your finger to
it and it will do something. The impedance of those drivers is obviously
extremely high. I noticed that if I use a pull-down resistor of 1Mohm on
every line, this unwanted behaviour does not occur. However, the datasheet
of several AVR-processors says that it is not allowed to put more than one
gate (a HC-buffer) on the oscillator output (XTAL2) of the processor if
one is using an external crystal. In a datasheet with general specs of the
74HC family I found that input leakage current is maximum 1 uA, at 25
degrees C even only 0.1 uA. They tested this with a VCC of 6 volts. That
means an impedance of 6 MOhm and that is more than my 1 MOhm.
My question is whether anybody foresees problems when I mount these
resistors. Off course there might be a problem if I want to measure an
analogue voltage from an high-impedance source (the impedance of the ADC
is, iirc, about 32 MOhm) but in that case I will have to buffer it, no
problem. But what are Your opinions, apart from that? And when some line
of the bus will be connected to the crystal, do You think that in real
life that will cause problems? I can off course make some of these
resistors "disconnectable" by means of jumpers (I wish I had done that
with the leds, but then again, then You just wouldn't see them anymore,
the buffer might still start oscillating), but as I can connect several
diff. types of controllers through the adapters, it would mean more
jumpers than for just one type of controller.

Thank You very much in advance for Your opinions/experiences.

Yours sincerely,

Rene

P.S. I just thought of what might a smarter solution to the crystal
problem, if I want to use a controller with an external crystal I can put
the jumper on the specific controller adapter so that the line will not be
connected to the bus on the development board. I guess that would be best.
Remains the question whether there could be other problems.

P.P.S. I could have rewritten my question after my brainwave from the
first P.S. but in that case someone might have come up with a suggestion
about that the crystal might give problems, I thought it would be wiser to
just leave it like this so You might see that those specific pins won't be
a problem anyhow.

The maximum load of 1 HC gate on the osc output is probably ac limited due
to capacitance
and its effect on osc stability.
It should have no trouble driving a 1M pulldown if it is a true/typical cmos
output.

ofc you shoulnt have anything at all connected to oscillator input when used
as an xtal osc.

Colin =^.^=

 

[whit3rd]

Hello to all!

I have been building an experimentation board for AVR processors
... I chose to use 74HCT540 bus drivers [to drive LED indicators]
The problem is that when a line is "not defined",
it's led turns on.

There are several options on each pin of the processor, including
'output', 'input' and sometimes 'current-limited output'. In
particular,
it is common for 'input' to be the default, and for a weak pullup
to be programmed for that state (this is a convenience to
prevent unused pins from being in an unstable state).

By using an inverting bus driver you are allowing the input-pin
state to (through the weak pullup) light the LED. Conventional
wisdom would be to ADD WEAK PULLUPS and not pulldowns
(this is a holdover from TTL which dissipates extra power
when pulldowns are used), and to drive LEDs between the
74HCT540 output and ground (rather than between output and
V+). Because the 74HCT540 is CMOS, it can source current
for this function. And your added pullups establish the
situation 'inputs float high' which is conventional (and found in
the most unlikely places- you can't ignore this convention).

Another conventional solution would be to swap the 74HCT540
for the 74HCT541 (noninverting but otherwise similar).

In either case, the 'inputs float high' convention is satisfied, and
the polarity of your command-the-LED-to-light is inverted from
the existing situation. Sorry about that.

 

[Rene]

Dear John, Colin and With3rd,

Thank You very much for Your input!

Yours sincerely,
Rene