Pairs of resistors in opposite directions

[oz7aff]

I just noticed a Vandal Resistant Push Button Switches on the page, dont use it as power switch, it is only rated to 48VDC

[GoatGuy]

Oh, a pair of resistors ... to do the work of "one" ... has a lot of utility! It is certainly the cheapest way to get higher-wattage dissipation tolerance, if space and layout is of little concern. Series-connected resistors are also quite a bit more tolerant to over-voltages (which for most amplifiers isn't really a concern). Did I mention "cheaper"? Its usually true: a pair of 1/2W resistors is usually less expensive than a 1W. Oh yes ... there's also increase-in-precision: the variance from stamped-on value tends to average by the number of resistors placed in parallel (or series). The increase in precision beats specially-vetted high-precision resistors any day (at least when manufacturing the things). Its also possible to use DIFFERENT resistors - with different temperature coefficients, to cancel each other's drift. Did that all the time up at LBL. Hmm... and cheaper.

Of course, the most-usual reason of all for using a pair or trio of resistors is to obtain a harder-to-find resistance value with off-the-shelf-in-your-lab candidates. That too saves money. And time. And frustration when the wrong part arrives.

GoatGuy

[DF96]

Originally Posted by GoatGuy

there's also increase-in-precision: the variance from stamped-on value tends to average by the number of resistors placed in parallel (or series).

That would be true for resistors with random variation around the stated value. I have heard that in practice most resistors are a bit low, as that speeds up the laser spiral cutting, so they are more likely to average low too.

[cyclecamper]

Ah, but it can be very important for consistency rather than accuracy of stated value, for instance in push/pull or balanced circuits. That's why the Jensen discrete op amp uses an integrated circuit with massively paralleled transistors...the resulting characteristics are statistically averaged and achieves precision consistency so any two production items are perfectly matched. In theory...

[Robert Kesh]

Just been doing a bit of maths, combining normal distributions with the paralleling resistor formula.

I make it two 1% sd paralleled resistors of equal value become a 1.6% sd resistor, so paralleling is bad, not good. The error gets smaller, but not by as much as the resistance gets smaller, so comparatively the error is bigger.

However, combining series resistors does improve things. Two 1% resistors become a 0.7% resistor.

Of course I could have royally messed up here.

[DF96]

Massively parallel transistors are almost always for low noise, not consistency. BJTs are very consistent in following the Ebers-Moll exponential model.

Paralleling resistors should have the same effect on error as putting them in series, as you are just adding the conductances instead of the resistances.

[Robert Kesh]

Originally Posted by DF96

Paralleling resistors should have the same effect on error as putting them in series, as you are just adding the conductances instead of the resistances.

errors in conductance don't have a linear relationship with errors in resistance. for example a +/- 10% tolerance in resistance becomes (roughly) a +11%/-9% error in conductance

for adding, multiplying and dividing standard deviations see.

http://www.cartage.org.lb/en/themes/...ubtraction.htm

Multiplication and Division of Values with Standard Deviation

[DF96]

If you have 10% resistors then it is better to buy better resistors.

If you have 1% resistors then +-1% gives +1.01/-0.99% which is close enough for the Central Limit Theorem to work its magic.

[Robert Kesh]

Originally Posted by DF96

If you have 10% resistors then it is better to buy better resistors.

If you have 1% resistors then +-1% gives +1.01/-0.99% which is close enough for the Central Limit Theorem to work its magic.

10% was just an example.

CLT is for the arithmetic mean of many samples, and so works for their sum too, which is why it works for series resistors, hence the 1/sqrt(2) in my first post on this, sd will reduced by a factor of 1/sqrt(n) for n similar resistors.

Paralleling resistors is related to the harmonic mean. I am not convinced it gives a similar result to the CLT. Though I confess I did the numbers when I was very tired.

[oshifis]

I see only one reason to connect resistors parallel in opposite direction: the stamped value (or the color code) can be read from either side...