Selecting passive components

Question

I am in the case where I read a lot of datasheet and I quite often see some circuit where the chip manufacturer indicate some value for resistors and capacitor but by not indicating any part number, no tolerances, no voltage etc...

How do I select components in this case ?

for example if we look at this simple I2C line:

(source)

We see two pull up resistor of 4.7k. How do I select the tolerance of those resistors, how do i select the part number etc... ?

My question is rather about how to deal with general case rather than with pull up or pull down and it concern indeed the choice of a component who may have different tolerance, package size, power I am looking at rule of thumb for R, C, L and D :) regarding tolerance % , X5R, 7R, Voltage . Thanks to Peter for his first proposal, if other have their own rule of thumb to propose, I am happy to hear them :)

Thats the point where you apply all the stuff you studied for your EE degree and all the experience you gathered so far. — PlasmaHH, Jun 02 '16 at 11:16
This is two questions: one about I2C resistors and another, too vague, one about all passive components. There are long app notes on general applications of *just* ceramic capacitors. This doesn't get into RF capacitors/inductors/resistors, high power, precision, temperature conditions, etc. — user2943160, Jun 02 '16 at 15:53
[This answer in a earlier thread](http://electronics.stackexchange.com/a/30792/7036) describes how to calculate values [specifically] for the I2C pull-up resistors. — Nick Alexeev, Jun 02 '16 at 16:29

dim · Answer 1 · 2016-06-02T13:59:58.423

The unmentioned parameters are always either implicit, or not critical.

In your example:

The voltages the resistors must withstand is implicit: it is the supply voltage.
Their tolerance is not critical: what will happen even if the resistor is 4.2k or 5.3k ? Nothing really wrong. No part of the circuit will misbehave. So you may well choose the worst possible tolerance from any manufacturer.
And actually, often, a parameter that is unmentioned is both implicit and non-critical: Here, the resistor power is implicit: it is U²/R, with U being the supply voltage. This gives a few mW, which clearly makes it non-critical.

Regarding the implicit information: It forces you to have a good understanding of the overall circuit, that is right. But it simplifies the drawing and prevents redundant information, which can be more confusing than useful.

Regarding their non-criticality: Often, on a schematic, the engineer will only put information that is important for the device to work. And that is actually important not to put unimportant information. Because, if you think an information is critical when it's not, it may lead you to select a component that is over-specified.

Now, in a datasheet, they will never go as far as providing a part number (espacially for things as uncritical as pull-up resistors), unless there is actually only one part in the world that work for this specific schematic (that may be the case, for example, in SMPS supplies where the transformer is often made specifically, or, of course, for the other ICs of the circuit). The part number is always up to you to choose, because suppliers catalogs are always evolving, and the pricing is often the main driver in selecting components.

However, you can often find one or several reference designs when looking for documentation on a given IC, and these reference designs often come with a complete schematic and bill of materials, which provide full part numbers, even for mildly-critical components (transistor, mosfets, diodes, ...).

+1 for "non-criticity". Although I believe the technical term is "non-criticalnessitudinal". — Dan Laks, Jun 02 '16 at 11:54
@DanLaks Dammit, I knew it wasn't a word... So what's the real term, then ? Oh, anyway, I'll leave it like that, it's funny. — dim, Jun 02 '16 at 12:00
@vicatcu Thanks. Then, I guess I have to edit now. I have to say I'm doing it reluctantly. I liked "criticity" so much... — dim, Jun 02 '16 at 13:12
Oh, gosh, my post is boring now, without non-existanting words... — dim, Jun 02 '16 at 13:15
For part stocking reasons, as mentioned by Peter Green's answer, it's best to stock 1% conventional tolerance SMD resistors, usually. Tolerance with through-hole resistors or any capacitors will require more specific judgement on price versus more parts required in inventory. — user2943160, Jun 02 '16 at 15:51
@user Yes. But my post is more about "why aren't all parameters mentioned" and, to some extent, "what would work in the circuit" rather than "what you should stock". — dim, Jun 03 '16 at 05:13

score 6 · Answer 2 · edited Apr 13 '17 at 12:32

We see two pull up resistor of 4.7k. How do I select the tolerance of those resistors?

As with any part of designing a circuit, you have to actually stop and understand the circuit. Those are just pullup resistor, so have a reasonably wide latitude. They are a tradeoff between requiring too much current to pull the line low, versus the line not floating high fast enough when let go, or not overcoming enough coupled noise.

I go into much more detail on what pullup resistors do and how to decide their values here.

score 6 · Accepted Answer · answered Jun 02 '16 at 13:15

Some tips (this is not exhaustive, it always pays to think about your circuit and what is important in it)

My rule of thumb with resistor tolerance is to go for 1% unless specified otherwise. In some cases 5% would be adequate but the price difference (at least in SMT) is small and stocking multiple different tolerances has costs of it's own.

Resistor power rating is not normally an issue for resistors in the kilohm range in circuits running at normal logic supply voltages but if your voltages are a bit higher or your resistances lower or your resistors are especially tiny then you should work out a worst case voltage and hence a worst case power.

Resistor voltage rating is normally only an issue once you get up to circuits operating at mains voltages.

For capacitors I usually use X5R or X7R for power supply decoupling and charge pumps. Maybe with a large electrolytic or tantalum for "bulk" capacitance (this is less nesacery nowadays though as big ceramics have become more readially available). For very small capacitors C0G/NPO ceramics are the norm. For capacitors in an analog signal path I would normally look at film capacitors.

thank you very much for your rule of thumb this is what I was looking for :) It helps. — chris, Jun 02 '16 at 14:05
Definitely try to only use/stock 1% SMD resistors, particularly now that it's not 10+ years ago. For lower cost, the cheaper ceramic capacitor dielectrics are appropriate (but, like all ceramic capacitors, come with caveats and negative characteristics). — user2943160, Jun 02 '16 at 15:49
Thank you for the advice. I take note for the resistor. What rule of thumb you use for selecting capacitor face to value of % and V ?Same question for L and D. How you prefer to select 0603 or 0402 or 0805 ? You have a rule of thumb for stocking according to size ? — chris, Jun 02 '16 at 16:17

score 1 · Answer 4 · answered Jun 02 '16 at 12:54

There are simple cases, and more complex cases.

In your example, you can use Ohm's Law to calculate the "worst-case" power dissipation. 4K7 @ 3V3 is 2mW so even a tiny 1/10 watt resistor would be OK. And the resistance tolerance for that particular application is pretty broad. They could have easily selected a resistance value higher or lower without significantly changing the performance.

Of course some component values are more critical than that and if you aren't sure of the circuit design, it is safer to just go with the specified values.

As to exactly what kind of component, that depends on how you are constructing it. Do you need conventional resistors with leads out each end for breadboarding or through-hole PC boards? Or do you need to make it compact with surface mount devices (SMD)?

Selecting passive components

4 Answers4