I've been working on a resistor substitution box and found myself curious for real data on the distribution of resistance values across large samples. There have been various other posts on this, and Dave did a couple videos a while back ("Gaussian Resistor") using through hole resistors. Still, for SMD I only saw posts with personal experience or a small amount of data.
I was specifically interested in seeing how a low tolerance part compared with a higher tolerance part by the same manufacturer and in the same series. Would I see bimodal distributions on the low tolerance parts which might suggest binning? (Answer: no, with one possible exception.) Is there actually more variance with 5% resistors than 1%, or does the manufacturer basically just make 1% resistors and sell them at two prices? (Answer: 1% often has lower variance, but even when it doesn't the mean will still be closer.) Is the average value across a sample typically lower than the nominal value, as some people have suggested? (Answer: generally yes, and the amount by which it's lower may depend on the tolerance.)
I did the measurements on a 34465A meter with Pomona 5143 tweezers. Measurements were two wire with the residual nulled out, and each data point is the average of four measurements taken with the aperture at 10 PLC. I got pretty quick at it, going through ~1 resistor per second. It took a few hours over a week or so.
Hopefully someone else finds this interesting.
Sorry the data is pretty much just about one manufacturer. The raw data sets are also attached.
Yageo RC series (0805) - 5% vs. 1%It's interesting that the main difference is a shift in the mean. The standard deviation doesn't change nearly as much as it does on the 0603 version below.
Yageo RC series (0603) - 5% vs. 1%Yageo AC series (automotive version of RC series) - 5% vs. 1%Here we're still on 0603, yet the main difference is another simple shift in the mean. The standard deviation remains roughly the same.
Yageo RT series (precision thin film) - 1% vs 0.5%The 1% actually outperformed the 0.5% since it had fewer outliers, though both are easily within spec. I double checked to confirm I didn't mix up the two, though this was cut tape so I could only check the digikey labeling. I suppose it's possible digikey mixed up the reels on their end.
I also measured a couple other brands, though I just did one sample per brand.
The TE sample was the one example I saw where the distribution appears somewhat bimodal, though I can't explain it. It doesn't make sense to me that one of the peaks would be directly over the nominal value. Perhaps there's a partial batch of higher precision resistors thrown in with these 1%?
TE CFP series - 1%
 | Stackpole RNCF series - 1%
 |
Home
Help
Search
Login
Register
