Options for the "Big" resistor (i.e. pulldown)

[Youkai]

Whenever I've seen a "big" resistor used or been instructed to use one it seem to always be a 10k resistor. I assume that is because 10k is very common and also our monkey brains like round numbers.

Is there anything special about 10k? A while ago I bought a variety pack of resistors and it came with a bunch of different sizes. I'm running short on my 10k resistors. I also have 8k, 12k, 15k, and 18k, etc,. Would any of those be a suitable replacement for a 10k resistor?

My understanding is that the point of these resistors it to have a very hard but technically there path for the electricity to flow; that way we can get a solid ground reading (in the case of a pulldown) without dumping too much energy through that path. That being the case it seems like, within reason, any large resistor value should work. Am I correct in that thinking? If so what is the practical limit on how large a resistor you could use in a 5v circuit?

[Zero999]

I don't know what you mean by big resistor? Assuming a 5V supply, the power dissipation in a 10k resistor would only be 2.5mW, so a small resistor could be used.

The value isn't critical, just use 8k2, 9k1, 11k, 12k etc.

8k is an usual value, so it's unlikely your pack of resistors will have that. Resistors are mostly produced in standard values, known as the E series of preferred numbers, which doesn't include 8. It's possible to get other values, but they're uncommon.
https://en.wikipedia.org/wiki/E_series_of_preferred_numbers

[Youkai]

I don't know what you mean by big resistor?

I just mean a resistor that's vastly larger than the other ones I'm using in resistance. E.g. I use a 470 on my LED but a 10k for my pulldown. That's 21 times "bigger".

8k is an usual value, so it's unlikely your pack of resistors will have that.

It's 8.2k, I rounded for the OP.

[atmfjstc]

I'd advise you against thinking of resistors as "big" or "small". What matters most of all is their function. If they're pullup or pulldown resistors, yeah, you can almost always replace them with reasonably close values (+-25%). In fact it's recommended to use values from the E-series instead of nice, round didactic values like 10K, since as mentioned before, the former are standard, much more plentiful and cheaper.

If the resistors are part of a filter, voltage divider, setting the OP for a transistor, etc., you need to have an understanding of the circuit to know how far you can stray from the schematic.

[Ian.M]

Lets consider pullups or pulldowns for a typical MCU, e.g. the ATmega328P used on many Arduino boards.

Refer to its full datasheet:  The input leakage current for an I/O pin is +/-1uA.  V_IL(max) is 0.3*Vcc, and V_IH(min) is 0.6*Vcc, which at Vcc=5V are 1.5V and 3V respectively.  3V is 2V below Vcc. 

Assuming there is no other significant leakage in the input circuit, in theory, the maximum resistor values that could be tolerated for a pullup and pulldown are 2 Meg and 1.5Meg respectively, as if you used greater value resistors, the leakage current through them would lead to invalid logic levels.

In practice, there are always other components, surface leakage between board tracks etc. so a conservative design where there are no other known sources of large leakage currents would use at least an order of magnitude smaller resistors e.g. max. 200K for pullups and max. 150K for pulldowns.

If you use a different MCU or logic chip, you need to re-do the worst case calculation with *YOUR* MCU's (or logic's) specified worst-case input leakage (or bias) current and logic thresholds.  E.g. for original 7400 series bipolar TTL logic it turns out that you cant tolerate more than 500R for a pulldown and still meet the logic '0' threshold,  so if you are dumb enough to design a 'classic' TTL circuit that actually requires a pulldown, it had better be 470R or lower!

Note that high value pullups (or pulldowns) increase sensitivity to external EMI, so in electrically noisy environments, or with long input wiring, you will have to choose between using much lower value resistors, or fully screening the input circuit.

There may be other considerations like switch wetting current that prevent the use of high value pullup/pulldown resistors. Edit: specifically see T3sl4co1l's comments on speed below.

[T3sl4co1l]

Right, "10k" is often code for "who cares".  In a low current application it might be 100k, or even more.  Sucks for logic speed -- you do need to make sure none of these signals are rising too slowly for their own good.  Logic gates can chatter when the input spends too long (more than fractional microseconds?) in the transition region, or say, transistors used to switch power domains on and off, can spend too much time in the linear range, dissipating excessive heat, not to mention the potential for oscillation.

On the low end, pull(up|down)s down to 50 ohms are typical for terminated signals, like Ethernet (which is 100 ohms nominal; coupled through a CT transformer, two 50 ohms are used in series across the CT winding -- same thing).

Even lower can be used for driving MOSFET gates (large load capacitances) at high speeds, or specialty purposes (e.g., a low impedance attenuator for signal injection).

There aren't many use cases these days for unipolar, logic level, terminated signals, but there were some families for that purpose, back in the day.  A relative of Ethernet used 50 ohm terminated coax cable.  A typical use case for 74/LS TTL was outputs driving ribbon cable, with a termination at the far end consisting of 330 ohms pulling up and 150 ohms pulling down: the parallel (Thevenin) equivalent is ~100 ohms, matching the ribbon cable (alternating signals and grounds gives about 100 ohms characteristic impedance)

Intermediate values, say 200 to 4.7k, are used to varying effect.  Such a pullup on a TTL output will (eventually) reach the +5V rail, a handy hack to interface to CD4000 and 74HC CMOS.  If you need the speed, you might choose a lower value; if not, a larger, generic value will be fine.

A very typical modern application is I2C, an open-collector serial bus.  Performance is limited by the rising edge rate, which is typically not very important (if you need high bandwidth, you don't choose I2C!), so a relatively large resistance can be used (10k or more?).  When a lot of devices are connected, or long traces or even cables get involved, the high load capacitance can require smaller resistors to maintain performance, down to 1k or so, maybe a bit less.  Beyond which you really should consider multiple buses, or a more powerful networking standard!

tl;dr the motivation is largely risetime, and if it's noncritical, well, it can be very noncritical.  Transmission line termination is another motivation, in which case a value similar to or slightly larger than Zo is desirable.

Tim

[Zero999]

I don't know what you mean by big resistor?

I just mean a resistor that's vastly larger than the other ones I'm using in resistance. E.g. I use a 470 on my LED but a 10k for my pulldown. That's 21 times "bigger".

But it will draw around 21 times less current, thus dissipate 21 times less power.

I = V/R
P = V²/R
P = I²R

[rstofer]

There may be a reason for using a low value pull-up when working with switch inputs.  Switch contacts are sometimes 'nasty' in the sense that the surfaces are oxidized.  You need a higher voltage to break through the oxide layer (and that usually isn't possible) and a  significant current flow to maintain the contact.  This would be more problematic for toggle switches which maintain a position over a long period of time than with a pushbutton that only needs to make contact momentarily.

In the world of TTL, it wasn't unusual to use 1K pull-ups or even 470 Ohm.  If it's a momentary condition, it might not matter how much current is flowing through the switch.  For a toggle switch, this needs engineering.  That's why they invented toggle switches with bifurcated gold plated contacts with a wiping action.  They're a lot more expensive than the bargain bin switch.

Very high value pull-ups are susceptible to noise and even though a 1M pull-up may technically work, it would probably never be used.

[Zero999]

There may be a reason for using a low value pull-up when working with switch inputs.  Switch contacts are sometimes 'nasty' in the sense that the surfaces are oxidized.  You need a higher voltage to break through the oxide layer (and that usually isn't possible) and a  significant current flow to maintain the contact.  This would be more problematic for toggle switches which maintain a position over a long period of time than with a pushbutton that only needs to make contact momentarily.

In the world of TTL, it wasn't unusual to use 1K pull-ups or even 470 Ohm.  If it's a momentary condition, it might not matter how much current is flowing through the switch.  For a toggle switch, this needs engineering.  That's why they invented toggle switches with bifurcated gold plated contacts with a wiping action.  They're a lot more expensive than the bargain bin switch.

Very high value pull-ups are susceptible to noise and even though a 1M pull-up may technically work, it would probably never be used.

Putting a small capacitor, around 100nF in parallel with the switch, can alleviate the oxidation problem by providing a high current spike, when the switch is closed and suppress noise pick-up.

[rstofer]

There may be a reason for using a low value pull-up when working with switch inputs.  Switch contacts are sometimes 'nasty' in the sense that the surfaces are oxidized.  You need a higher voltage to break through the oxide layer (and that usually isn't possible) and a  significant current flow to maintain the contact.  This would be more problematic for toggle switches which maintain a position over a long period of time than with a pushbutton that only needs to make contact momentarily.

In the world of TTL, it wasn't unusual to use 1K pull-ups or even 470 Ohm.  If it's a momentary condition, it might not matter how much current is flowing through the switch.  For a toggle switch, this needs engineering.  That's why they invented toggle switches with bifurcated gold plated contacts with a wiping action.  They're a lot more expensive than the bargain bin switch.

Very high value pull-ups are susceptible to noise and even though a 1M pull-up may technically work, it would probably never be used.

Putting a small capacitor, around 100nF in parallel with the switch, can alleviate the oxidation problem by providing a high current spike, when the switch is closed and suppress noise pick-up.

That's a nice solution!