Tech Forum

Generally, if you require greater precision, better stability and/or less noise, you go with metal film resistors. In the olden days, this was an expensive option, but these days, precision metal film resistors have plummeted in cost, at least for the Asian imports. So use metal film if you're working with test equipment attenuators, voltage dividers and analog timing circuits.

Metal Film Is Better
  If you need a circuit to remain within operational specs for a long time without periodic adjustment.
  If you’re designing a low-noise audio preamp.
  If you need precise timing for your 555 circuit.

Some notes here.
  1 - The ancient trick of beginning with a carbon composition resistor of a lower value than needed and filing a notch in it until the value was right on, gave you a good, tight resistor for a few minutes. But even if that notch is sealed with clear fingernail polish, carbon resistors still have a lousy temperature coefficient. The resistor will change in value with any changes in temperature far beyond what you'd want with a precision part. It isn't the same as using a precision metal film resistor, no matter what that old timer tells you.
  2 - The second note concerns power dissipation. The old carbon comp resistors could take short spikes in power and survive even with short bursts of ten times the power rating, because the mass of that slug of a resistance element could absorb and dissipate that spike. Carbon film and metal film resistors can handle their specified power level, but can't handle spikes well, for they don't have that same mass in the resistance element. They can dissipate the average rated power, but not absorb the heat of a big power spike. So, that same spike in power can cause a carbon or metal film resistor to literally burn out.

Even with the problem of film over composition described above, carbon film resistors do seem to have tighter and more stable values than the comps. They still suffer from bad temperature coefficients and noise, but the typical new 5% carbon film resistor will often be well-within 2% of its marked value.

As a "rule of thumb" I recommend: Choose carbon film resistors when low cost (at high quantity) is the main consideration. Use metal film resistors when low noise, low temperature coefficient, and/or greater precision are important.

All types of resistors generate thermal noise, also called "Johnson noise," which increases as resistance and temperature increase. This comes directly from the laws of physics which, as Engineer Scotty famously said on Star Trek, "you canna change." However, current flowing through a resistor causes additional noise, including "shot noise," and this is greater in carbon film than in metal film types. That is why the latter are preferred for critical analog circuitry, such as audio preamplifiers. Resistor noise hardly matters in digital applications.

For more information about noise, see Joe Geller's excellent resistor noise measurement project (JCan) in the July 2007 Nuts and Volts. After building the JCan kit, I could easily measure more noise in carbon composition types compared to metal film resistors of equal value when current was applied. There was no difference with zero current — just the thermal noise "background." The noise difference between carbon film and metal film resistors under power was smaller, but detectable.

The temperature coefficient tells you how much a resistor's value can change with temperature, in units of parts per million (PPM) per oC or oK. Common axial lead (through hole) ¼-watt metal film resistors are rated 50 or 100 PPM/ oC, but their carbon film cousins are 350 to 700 PPM/ oC. Sensitivity to temperature is a consideration in many sensor and measurement applications, but it does not matter in digital circuits.

Assuming the resistor offerings at Mouser Electronics (www.mouser.com) are typical of the industry, the vast majority of ¼-watt axial units are available only at 1-percent tolerance for metal film, and 5 percent for carbon film. Even if I wanted one, I could not buy a 5% metal film or a 1% carbon film resistor. Therefore, choice of type is related to precision, for practical purposes. The cost difference is negligible for small quantities; you are mainly paying for the labor to count them out and bag them.

I'm not sure about a rule of thumb but I would not use metal film resistors in an RF circuit. It would be my luck that I create a resonate circuit and have all kinds of strange oscillators going. Good Luck